Compare commits

...
22 Commits
Author SHA1 Message Date
coissac d78a6b4501 Merge pull request 'Push vpnykuuvnrnx' (#82) from push-vpnykuuvnrnx into main
Reviewed-on: #82
2026-11-19 06:17:43 +00:00
Eric Coissac 5db7b440da refactor: treat absence as state unless --free-loss and update quartet annexes
ci.yml / build (pull_request) Successful in 4m33s
Release / create-release (push) Successful in 1m56s
Release / build-linux-x86_64 (push) Successful in 11m18s
Release / build-macos-arm64 (push) Successful in 2m27s
Refactor the handling of genome absence to be a distinct state unless `--free-loss` is specified. This change introduces `QuartetConvention` to pair ambiguity policies with the free-loss flag, updating file naming conventions and logic across sampling, masking, and stats modules.

- Update `--exclude-genome` description in `pair_coverage_sampling.md` to clarify absence as a state unless `--free-loss` is specified.
- Move multi-member genome handling convention from `pair_coverage_sampling.md` to file name conventions.
- Clarify persisted annex structure in `pair_coverage_sampling.md` as `PersistentBitVec` per non-monomorphic family with counting convention carried by file name.
- Fix initial version's absence handling in `has_bipartition`, the annex, and `--sample-stats` to correctly count index 0 when absence is a state.
- Replace entropy-only builder with `AnnexRequest` and `ensure_layer_family_annexes`.
- Consolidate sampling parameters into a single `SamplingSpec`.
- Update family constraint description in `sampling_design.md`.
- Modify uniform draws and `--quartet` removal text in `sampling_design.md`.
- Clarify state counting in tests for `--free-loss` option in `sampling_design.md`.
- State that absence forms a pair without `--free-loss` but never with it in `sampling_design.md`.
- Update `--quartet` option to require at least two distinct states, each carried by at least two genomes.
- Define absence as a state unless `--free-loss` is given for family carrier definition.
- Clarify `--pair-balance` favors families covering pairs with fewest shared families.
- Modify logic for families fitting every tree topology equally well in `phylo.md`.
- Change `--quartet` restriction logic to apply after pre-checks involving `--no-ambiguity` and `--exclude-genome`.
- Update `obikmer` dependency version from 1.4.4 to 1.4.5 in `Cargo.lock` and `Cargo.toml`.
- Modify `annex_selection` in `args/mod.rs` to derive `free_loss`, `entropy`, `quartet`, and `overlap` from `SamplingSpec`.
- Update `SamplingArgs::quartet` documentation in `sampling.rs` to describe state determination based on `--free-loss`.
- Add `free_loss` field to `AnnexSelection` struct in `family_annexes.rs`.
- Change `AnnexRequest` to use `Vec<QuartetConvention>` instead of `Vec<AmbiguityPolicy>`.
- Update `request` logic in `family_annexes.rs` to conditionally create `QuartetConvention` based on `self.quartet`.
- Modify `dedup` signature in `family_annexes.rs` to accept iterator of type `T: Ord`.
- Update `ensure_layer_family_annexes` in `family_annexes.rs` to use `missing_files` and `QuartetConvention`.
- Add `missing_files` definition to `family_annexes.rs`.
- Change `has_bipartition` in `masking.rs` to filter counts where count is `>= 2`.
- Update `is_quartet_informative` in `masking.rs` to accept `free_loss` boolean parameter.
- Modify `two_pairs_of_distinct_states_is_informative` and `singleton_variant_is_not_informative` in `masking.rs` to test both `free_loss` policies.
- Add `under_free_loss_absence_is_missing_and_forms_no_pair` and `without_free_loss_absence_is_a_state_that_can_form_a_pair` definitions to `masking.rs`.
- Update `no_ambiguity_drops_multi_member_genomes_from_the_count` in `masking.rs` to check `is_quartet_informative` under `free_loss=true`.
- Add `the_annex_rule_and_the_exact_check_agree_on_the_same_family` definition to `masking.rs`.
- Remove `absent_genomes_are_not_a_state` definition from `masking.rs`.
- Update `restrict_to_quartet` in `quartet.rs` to accept `QuartetConvention` instead of `AmbiquityPolicy`.
- Modify `add_site` in `sample_stats.rs` to handle state 0 as a state participating in bipartition.
- Add `absence_as_a_state_makes_a_site_informative` definition to `sample_stats.rs`.
- Add `quartet_convention` function to `SamplingSpec` in `subsample.rs`.
- Update `sample_index` in `subsample.rs` to use `spec.quartet_convention()` for `AnnexRequest`.
- Update `restrict_to_quartet` call in `subsample.rs` to use `spec.quartet_convention()`.
- Add comment regarding `--no-ambiguity` counting convention in `helpers.rs`.
- Import `QUARTET_ANNEX_FILE_NAMES` and `QuartetConvention` from `quartet_annex` in `mod.rs`.
- Add `QuartetConvention` struct with `policy` and `free_loss` fields to `quartet_annex.rs`.
- Modify `quartet_annex_file_name` in `quartet_annex.rs` to take `QuartetConvention`.
- Add `the_conventions_never_share_a_file` function to `quartet_annex.rs`.
- Remove `the_policies_never_share_a_file` definition from `quartet_annex.rs`.
2026-11-19 07:12:36 +01:00
Eric Coissac c93616b432 feat: add threshold-based Hamming distance and refine phylo documentation
Add `threshold_hamming_dist_matrix` to calculate k-mers present in exactly one genome using union minus intersection. Update `DistanceArgs` and `distance.rs` to support this metric on count indexes via `--presence-threshold`, removing the previous restriction that hamming was only for presence/absence indexes.

- Add `threshold_hamming_dist_matrix` method to `CountPartials` in `src/obikmer/src/traits.rs`
- Implement new test `threshold_hamming_counts_kmers_present_in_one_genome_only` in `src/obicompactvec/src/tests/intmatrix.rs`
- Update `DistanceArgs` in `src/obikmer/src/cmd/phylo/args/distance.rs` to clarify threshold effects on Jaccard/Mash/Hamming
- Modify `distance` function in `src/obikphylo/src/distance.rs` to support Hamming on count indexes and use pre-computed bit structures
- Update `obicompactvec.md` documentation with new function details and partial result explanations
- Refine `index.md` descriptions for `phylo`, `unitig`, and `select` options
2026-11-19 07:11:57 +01:00
Eric Coissac c4eacb3f72 feat(phylo): expand genome distance metrics and refactor CLI arguments
Refactor the `phylo` command to support explicit distance metrics (Jaccard, Hamming) and SNP-based pseudo-alignments. Replace implicit metric selection with explicit requirements, such as requiring `--n-sites` for SNP analysis. Add a new `shared-kmers` output option and update documentation for performance characteristics of sparse vs dense formats.

- Update `phylo` CLI to support genome-vs-genome distances using k-mer content or SNPs, including options for NJ/UPGMA tree building and data export
- Replace `reindex` command with `convert` for in-place index evidence conversion (exact/approx/hybrid) and add `--upgrade-index` to `utils` for missing layer metadata
- Refactor `distance_to_compute` logic to remove implicit Jaccard implications for `--nj`/`--upgma` unless `--distance` is specified, and enforce `--n-sites` for `--pseudo-alignment`/`--sankoff`
- Add `shared_kmers` method to `Metrics` trait in `src/obikphylo/src/distance.rs` and new `write_shared_kmers` function in `src/obikmer/src/cmd/phylo/shared.rs` for CSV export
- Update `PackArgs::dense` documentation to clarify that dense format is faster for column-oriented distance matrices while sparse is better for single-row access
- Remove `requires_counts` definition and update assertions in `values.rs` to reflect new distance computation signatures and metric requirements
2026-11-19 07:11:44 +01:00
Eric Coissac 14433e80a5 refactor: restructure phylo CLI arguments into submodules
Refactor the `phylo` command's argument handling by splitting the monolithic
`args.rs` file into specialized submodules (`distance`, `outputs`, `sampling`,
`select`) and a central `mod.rs`. This improves modularity and separates
concerns for distance metrics, sampling options, and output configurations.

- Move `alignment_output`, `ambiguity`, and `annex_selection` from `args.rs` to
  `args/mod.rs` in the `PhyloArgs` struct.
- Relocate distance parsing logic (`parse_gamma_shape`, `parse_n_sites`, `target`)
  and helper functions (`distance_to_compute`, `as_classic`, `as_snp`, etc.) from
  `args.rs` to `values.rs`.
- Add new submodule files: `args/distance.rs` for `DistanceArgs`, `args/outputs.rs`
  for `AlignArgs` and `CostsArgs`, `args/sampling.rs` for `SamplingArgs`, and
  `args/select.rs` for `SelectArgs` and `AnnexArgs`.
- Update `main.rs` to register the new `Phylo` command using `cmd::phylo::PhyloArgs`.
- Refactor `mod.rs` (`run`) to use nested argument structures (e.g., `args.dist`,
  `args.sample`) instead of flat fields.
- Update `alignment_out.rs` to reference the new nested paths (`args.align.fasta`,
  `args.sample.no_ambiguity`).
2026-11-19 07:11:26 +01:00
Eric Coissac 7f829ea557 docs: clarify merge re-merge behavior and input file handling
Update documentation to explain that merged indices can serve as sources if at least one source is unmerged, and that the resulting column order follows the base index first. Clarify that explicitly named files are always read regardless of extension, with format detection based on content.

- refactor: prevent I/O errors when re-merging packed outputs by restricting `choose_base` to valid destinations via new `can_be_destination` check
- fix: handle zero-layer partitions in `build_index_layer` and `remove_build_artifacts` to skip unnecessary operations
- feat: allow `PathIter` to treat unrecognized extensions as explicit files, ensuring they are not silently dropped during directory expansion
- test: add comprehensive tests for `PathIter` including extension recognition, explicit file handling, and stdin sentinel behavior
- test: update smoke tests to verify empty partitions, `.fna.gz` indexing, and refusal of merging only packed indexes
- chore: remove obsolete cargo build environment variables from `run.sh` script
2026-11-19 07:10:58 +01:00
Eric Coissac 07c0a54e04 feat: add compatibility testing scripts and update dependencies
Add `compat_ptrhash.sh` and `compat_ptrhash2.sh` to facilitate determinism control and incremental merge testing between obikmer binaries. Update `tikv-jemallocator` dependency version in `Cargo.toml` and refresh `Cargo.lock` with new crates and checksums.

- Add scripts/docker_debug/compat_ptrhash.sh for compatibility tests (T0-T4) comparing ptr_hash versions using Docker and md5sum
- Add scripts/docker_debug/compat_ptrhash2.sh for determinism control and incremental merge operations between binaries A and B
- Update src/obikmer/Cargo.toml to change tikv-jemallocator version from 0.7.0 to 0.5.4 under musl target cfg
- Update src/Cargo.lock by adding new crates (anes, block-pseudorand, criterion, plotters, etc.) and updating existing package versions/checksums
2026-11-19 07:10:37 +01:00
Eric Coissac af525cd268 feat: switch musl builds to jemalloc allocator for stability
Update Cargo.lock to remove unused dependencies and update tikv-jemallocator. Configure src/obikmer/Cargo.toml to use jemalloc for static musl builds, addressing fragmentation issues with the default multi-threaded allocator. Update main.rs to conditionally set Jemalloc as the global allocator when targeting musl.

- Remove unused dependencies from Cargo.lock: anes, block-pseudorand, chiapos-chacha8, criterion, criterion-plot, itertools (and its either dep), nanorand, hermit-abi, is-terminal, plotters ecosystem, tinytemplate deps, voracious_radix_sort rayon dep, and web-sys deps.
- Update tikv-jemallocator version from 0.5.4 to 0.7.0 in Cargo.lock.
- Add tikv-jemallocator = "0.7.0" to musl-specific dependencies in src/obikmer/Cargo.toml.
- Set Jemalloc as the global allocator in main.rs for static musl builds.
2026-11-19 07:10:20 +01:00
Eric Coissac d0668eb167 feat: add columnar sparse matrix support and windowed transposition
Introduce `sparse_columnar` and `sparse_intmatrix_columnar` modules to enable
building sparse matrices directly from columnar data without dense intermediates.
Add a `transpose` module with `RowBatch` and `transpose_windows` for efficient
windowed column-to-row transposition based on density checks.

- Add `src/obicompactvec/src/bitmatrix/sparse_columnar.rs` with `build_from_columnar`,
  `build_from_columnar_with`, and `read_bit_window` to build sparse bit matrices
  from columnar files using windowed reads.
- Add `src/obicompactvec/src/sparse_intmatrix_columnar.rs` with `build_from_columnar`,
  `build_from_columnar_with`, `invalid`, and `read_int_window` for integer matrix
  construction from PCIV columns, handling sentinel values and overflow.
- Add `src/obicompactvec/src/transpose.rs` containing `RowBatch<V>` struct and
  `transpose_windows` function for dynamic window sizing and memory-bounded
  column-to-row transposition.
- Update `src/obicompactvec/src/bitmatrix/sparse.rs` to use `packed_bit_matrix_n_cols`
  and `MatrixMeta` for validation, ensuring stale columnar data is transposed directly
  into a single `matrix.pbmx` file while removing old column files.
- Update `src/obicompactvec/src/sparse_intmatrix.rs` to use `packed_int_matrix_n_cols`
  for column count validation, making `pack_sparse_compact_int_matrix` idempotent and
  avoiding dense intermediates by transposing columnar data directly.
- Expose `packed_bit_matrix_n_cols` in `src/obicompactvec/src/bitmatrix/packed.rs` and
  make `packed_int_matrix_n_cols` public in `src/obicompactvec/src/intmatrix.rs`.
- Add `sparse_columnar` module to `src/obicompactvec/src/bitmatrix/mod.rs` and export
  `PersistentBitMatrixBuilder` and `batch_presence_counts`.
- Add `sparse_intmatrix_columnar` and `transpose` modules to `src/obicompactvec/src/lib.rs`.
- Add `transpose_columnar` test module in `src/obicompactvec/src/tests/mod.rs` with
  tests verifying columnar transpose matches dense route output.
2026-11-19 07:08:41 +01:00
Eric Coissac 4e98de4b77 feat: add BitSliceView::for_each_set_in and optimize build_from_dense
Introduce a new `views_scan` module containing `BitSliceView::for_each_set_in`, which iterates over set bits in a specified range using optimized word traversal. Update `build_from_dense` to use this method for batched, ordered column access, ensuring sorted row output. Add comprehensive tests for the new iteration logic and boundary handling.

- Add `src/obicompactvec/src/views_scan.rs` with `BitSliceView::for_each_set_in`, which iterates over a `Range<usize>` calling a closure for each set bit while optimizing sparse traversal via `word.trailing_zeros()`.
- Add `src/obicompactvec/src/tests/views_scan.rs` with tests `words`, `naive`, `scanned`, and helpers verifying that `for_each_set_in` matches naive iteration, handles empty/reversed ranges correctly, and rejects out-of-bounds ranges.
- Modify `build_from_dense` in `src/obicompactvec/src/bitmatrix/sparse.rs` to replace `nonzero_triples` with `dense.col_view(c).for_each_set_in`, ensuring column visits are batched and ordered for sorted row processing.
- Update module structure in `src/obicompactvec/src/lib.rs` and `src/obicompactvec/src/tests/mod.rs` to include the new `views_scan` module.
2026-11-19 07:08:07 +01:00
Eric Coissac 03eaa4544a feat: add pack optimization prompt, debug infrastructure, and monitoring scripts
Add a prompt template for optimizing the `pack` step and a suite of Docker-based
debug tools including build, run, and monitoring scripts. Enhance `pack_matrices`
with LFD partition sizing and detailed timing reports, and update `stop` with
additional system statistics.

- Add `prompts/pack_optimization.md` to guide optimization of the `pack` step,
  specifying context, performance measurements, and deliverable structures.
- Add `scripts/docker_debug/` infrastructure: Dockerfile for musl targets,
  `build.sh` for image creation, `run.sh` for parallel indexing and merging,
  `monitor.sh` for PID tracking, `gen_fasta.py` for synthetic genomes, and
  `gen_fasta_real.py` for real genome generation with `read_fna`.
- Add `scripts/monitor_pack.sh` as a 1 Hz sampler for I/O and memory metrics,
  writing samples to CSV and summary reports.
- Add `scripts/watch_merge.sh` to monitor process resource usage, enforcing
  memory limits via SIGTERM/SIGKILL guards based on VmRSS and VmSize.
- Modify `src/obikindex/src/index/kmer_index.rs`: add `pending_matrix_bytes` and
  `report_pack_timings`; update `pack_matrices` to use LFD strategy for partition
  sizing and clarify documentation regarding `stat()` proxy bytes.
- Modify `src/obisys/src/stage.rs`: update `stop` definition to include new
  variables (`user_secs`, `sys_secs`, faults, context switches, I/O blocks) and
  logic checking for swap usage.
2026-11-19 07:07:50 +01:00
Eric Coissac 24291a4484 refactor: remove unused dependencies and simplify logic
Remove obsolete dependencies across multiple crates to clean up the build graph and update `obikdump` to use `try_update`. Add dev-dependencies for tracing in `obiread` and clarify memoization logic in `obiksession`.

- Remove `obifastwrite`, `tracing`, `obidebruinj`, `obikentropy`, `obipipeline`, `obidex`, and `obiskbuilder` from `Cargo.lock`.
- Remove `obifastwrite` and `tracing` dependencies from `src/obidebruinj/Cargo.toml`.
- Update `dump` function in `src/obikdump/src/dump.rs` to use `try_update` instead of `fetch_update` for updating remaining counts.
- Remove `obitaxonomy`, `obidebruinj`, `obipipeline`, and `obikentropy` dependencies from `src/obikindex/Cargo.toml`.
- Remove `obiksession` dependency from `src/obikmer/Cargo.toml`.
- Remove `obiskbuilder` dependency from `src/obikphylo/Cargo.toml` and update `kodama` version to "0.3.0".
- Remove `bytes` dependency from `src/obikrope/Cargo.toml` and set `criterion2` version to "3" with `cargo_bench_support`.
- Add `tracing` and `tracing-subscriber` as dev-dependencies in `src/obiread/Cargo.toml`.
- Remove `lazy_static` dependency from `src/obiskbuilder/Cargo.toml` and add `obikseq` as a test-utils dev-dependency.
- Clarify `memo_key_from_debug` in `src/obiksession/src/memoize.rs` with comments explaining Debug-derived field usage.
2026-11-19 07:07:16 +01:00
coissac cbc344e4b4 Merge pull request 'Push txuyrtlmyoyl' (#81) from push-txuyrtlmyoyl into main
Reviewed-on: #81
2026-11-17 14:38:33 +00:00
Eric Coissac 07bc5fc634 docs: update sparse packing strategy and benchmark scripts
ci.yml / build (pull_request) Successful in 4m53s
Release / create-release (push) Successful in 28s
Release / build-linux-x86_64 (push) Successful in 6m46s
Release / build-macos-arm64 (push) Successful in 4m8s
Update documentation to reflect that sparse packing now maps all columns of a layer at once and packs layers sequentially, reversing the previous choice to avoid materializing a packed copy. This change introduces an extra write/read cycle for the dense matrix, increasing pack benchmark time from 1.11s to 1.60s.

- Modify `pack_sparse_bit_matrix` in `src/obicompactvec/src/bitmatrix/sparse.rs` to always convert via the dense packed format, simplifying logic by removing conditional handling of existing files and ensuring a single mapping.
- Update `pack_sparse_compact_int_matrix` in `src/obicompactvec/src/sparse_intmatrix.rs` to exclusively handle conversion from the dense packed format (`matrix.pcmx`) to sparse, removing columnar format handling and simplifying file cleanup.
- Adjust `scripts/bench_merge_oracle.sh` to extract `PACK_WALL` metrics via `stage_field` for merge operations, updating grep patterns and output messages to report pack wall time.
- Revise `DevDocMD/implementation/merge_mmap_audit.md` and `merge_scaling.md` to document the new sequential packing strategy, noting that `N_part` scaling applies only to incremental merges and that `obisys::MemoryBudget` is no longer used by merge.
- Bump package version in `src/obikmer/Cargo.toml` from 1.4.3 to 1.4.4.
2026-11-17 15:35:41 +01:00
Eric Coissac 1f199c69db feat: implement in-memory column builders and refactor merge partitioning
Introduce `MemBitVecBuilder` and `MemCompactIntVecBuilder` to construct columns in RAM before writing, reducing the number of open file descriptors during merge operations. Refactor `partition_merge.rs` to use these new builders via `add_mem_col`, decoupling live memory mappings from source counts to prevent `SIGBUS` errors.

- Add `mem_builder` module with `MemBitVecBuilder` and `MemCompactIntVecBuilder` structs for in-memory column construction and PCIV file writing.
- Extend `MatrixBuilder` with `add_mem_col` and `n_cols` methods, updating `ColBuilder` enum to support `MemBit` and `MemInt` variants.
- Refactor `partition_merge.rs` to use `add_mem_col` instead of `add_col`, introducing `run_pass2` for batch processing and optimizing destination layer handling.
- Update `format.rs` with `pciv_index_layout`, `write_pciv_header`, and `write_pciv_tail` functions to manage PCIV index layout and header/tail sections.
- Modify `mmap_file.rs` to use `write_std_header` and introduce `write_built` for standardized file writing and flushing logic.
- Make `MAGIC` constant public in `bitvec.rs` and update documentation for memory mapping alignment and structure.
- Add test modules `tests/matrix_builder.rs` and `tests/mem_builder.rs` to verify column counts and byte-identical output of new builders against persistent versions.
- Update `mkdocs.yml` navigation to include "Merge mmap audit (2026-11)" documentation entry.
- Add implementation audit documentation in `DevDocMD/implementation/merge_mmap_audit.md` detailing memory mapping failures and refactoring plans.
- Add merge testing scripts `scripts/smoke_test_merge.sh` and `scripts/bench_merge_oracle.sh` for validation and benchmarking.
- Bump `obikmer` package version from 1.4.2 to 1.4.3 in `Cargo.toml`.
2026-11-17 15:16:41 +01:00
coissac b35127e4e3 Merge pull request 'Push smsmsoyssmnr' (#80) from push-smsmsoyssmnr into main
Reviewed-on: #80
2026-11-09 10:01:21 +00:00
Eric Coissac 0134c69cf4 docs: clarify VCF pseudo-alignment format and update obikmer dependency
ci.yml / build (pull_request) Successful in 3m40s
Release / create-release (push) Successful in 1m7s
Release / build-macos-arm64 (push) Successful in 1m52s
Release / build-linux-x86_64 (push) Successful in 6m41s
Update the VCF export documentation to reflect that `-` represents an overlapping deletion (extra allele `*`) and introduce the `<ABS>` symbolic allele for absent families. Add a new dedicated documentation file for the VCF pseudo-alignment format, remove obsolete descriptions from phylo.md, and fix benchmark scripts to handle errors and directory creation correctly.

- Update `obikmer` dependency from 1.4.1 to 1.4.2 in Cargo.toml and Cargo.lock
- Add `UserDocMD/formats/vcf.md` describing VCF pseudo-alignment structure, alleles, and genotypes
- Remove obsolete VCF descriptions from `UserDocMD/usage/phylo.md` and update ploidy/free-loss notes
- Update `DevDocMD/theory/evolutionary_distances.md` to clarify `-` as overlapping deletion and `<ABS>` for others
- Fix `benchmark/index_one_count.sh` to create index directory and check read command errors
- Fix `benchmark/index_one_presence.sh` to create index directory and check simulation errors
- Update `src/obikphylo/src/siblings/algorithms/alignment_vcf.rs` to use `<ABS>` for absent families instead of `*`
2026-11-09 20:59:28 +11:00
Eric Coissac 5368c98875 refactor: replace no_ambiguity bool with AmbiguityPolicy enum
Replace the binary `no_ambiguity` flag with an `AmbiguityPolicy` enum (`Keep`, `Forbid`, `OnlyDiploid`) to support multiple ambiguity conventions. This change updates CLI arguments, internal logic, and persistent storage formats to distinguish between keeping all genomes, forbidding ambiguous ones, or restricting analysis to diploid sites only.

- Add `src/obikphylo/src/siblings/ambiguity.rs` defining the `AmbiguityPolicy` enum with variants `Keep`, `Forbid`, and `OnlyDiploid`, including methods for header flags and determining if a genome is dropped.
- Update `PhyloArgs` in `args.rs` to derive `AmbiquityPolicy` from CLI flags and construct `AnnexSelection` for sibling annexes.
- Modify `SamplingSpec` in `subsample.rs` to store `ambiguity: AmbiguityPolicy` instead of `no_ambiguity`, updating masking and scalar value logic accordingly.
- Update `SiblingExt` in `sibling_ext.rs` so `build_sibling_annex` accepts an `AnnexSelection` to build only requested derived annexes, and `family_overlap` uses the specific policy.
- Refactor `FamilyOverlap` in `family_overlap.rs` to store and use `AmbiguityPolicy`, replacing boolean checks with policy-based carrier determination.
- Update `FamilyScalarAnnex` in `family_scalar_annex.rs` to store `AmbiguityPolicy` in headers, enabling distinct file names for each policy variant.
- Modify `quartet_annex.rs` to generate distinct file names based on `AmbiguityPolicy`, ensuring no overlap between Keep, Forbid, and OnlyDiploid annexes.
- Update `family_annexes.rs` to use `AnnexSelection` for building entropy, quartet, and pair weight annexes under specific policies instead of a single boolean convention.
- Refactor algorithm modules (`entropy.rs`, `masking.rs`, `pair_weight.rs`, `quartet.rs`, `snp_distance.rs`) to accept `AmbiguityPolicy` arguments, replacing boolean checks with policy-specific logic (e.g., `OnlyDiploid` masking).
- Update documentation in `phylo.md` and `evolutionary_distances.md` to describe the new `--only-diploid` option, the behavior of `--sibling-annex` with specific policies, and the separate storage of entropies per convention.
2026-11-09 20:59:09 +11:00
Eric Coissac b55123a376 feat: add --only-diploid flag to filter for diploid genotypes only
Introduce a new `--only-diploid` option that restricts analysis to genomes with one or two family members (base or two-base IUPAC codes), treating others as missing data. This complements the existing `--no-ambiguity` behavior by providing a stricter filter for diploid-compatible sites, while ensuring conflict detection between the two flags.

- Add `only_diploid` boolean argument to `PhyloArgs` in `src/obikmer/src/cmd/phylo/args.rs`, documenting its effect on pseudo-alignment and its conflict with `--no-ambiguity`.
- Propagate `only_diploid` through `SamplingSpec` in `src/obikmer/src/cmd/phylo/mod.rs` and update the `run` function signature to accept the new argument.
- Implement masking logic in `src/obikphylo/src/siblings/algorithms/masking.rs`: modify `masked_state` to accept `only_diploid`, add `only_diploid_keeps_up_to_two_members` helper, and update `genome_mask` documentation.
- Update `sample_layer` in `src/obikphylo/src/siblings/algorithms/subsample.rs` to use `spec.only_diploid` for masking decisions and update `SamplingSpec` documentation.
- Adjust `tally_2genomes` in `src/obikphylo/src/siblings/algorithms/snp_distance.rs` to pass the new `only_diploid` parameter to `masked_state`.
- Update `DevDocMD/theory/evolutionary_distances.md` to describe the new `--only-diploid` option, the `masked_state` rule (`count_ones() > 2`), and clarify toy index behavior.
- Update `UserDocMD/usage/phylo.md` to document `--only-diploid`, `--quartet`, and `--pair-balance` behaviors, clarifying that diploid choice is a convention and explaining how three/four-allele genomes are treated as missing.
2026-11-09 20:58:21 +11:00
Eric Coissac b132f0847f feat: add VCF export and refine alignment output infrastructure
Introduce `--vcf` and `--fasta` options for phylo commands, allowing users to
export pseudo-alignment data as VCF or FASTA. Implement comprehensive VCF logic
including stable virtual coordinates, canonical ID generation, and flexible
genotype representation based on ambiguity settings. Refactor internal alignment
structures to track k-mer families and site loci, ensuring consistent data flow
from scanning through to output generation.

- Add `src/obikmer/src/cmd/phylo/alignment_out.rs` with `create` and
  `write_alignment_outputs` to handle file creation and dispatching of FASTA/VCF
  formats based on `PhyloArgs`.
- Add `src/obikphylo/src/siblings/algorithms/alignment_vcf.rs` implementing
  VCF-specific logic: `write_vcf`, `alignment`, `cell_alleles`, `majority_base`,
  and site construction via `site` using fixed $K=11$. Includes helpers for
  IUPAC state inversion, genotype ordering, and stable coordinate generation.
- Update `src/obikmer/src/cmd/phylo/args.rs` to add `fasta` and `vcf` flags
  and a new `alignment_output()` method that returns true if any output format
  is requested. Add `pseudo_alignment` field documentation.
- Modify `src/obikmer/src/cmd/phylo/mod.rs` to use `alignment_output()` for
  determining alignment needs, update error messages for `--n-sites`, and import
  `write_alignment_outputs` from the new module.
- Refactor `src/obikphylo/src/siblings/algorithms/alignment.rs` by adding
  `SiteLocus` struct (partition, layer, family_idx, family_id) and updating
  `SnpAlignment` and `AlignmentReducer` to track sites. Update `new`, `finish`,
  and `reduce` to populate site information from `SurvivingFamily`.
- Update `src/obikphylo/src/siblings/algorithms/iqtree.rs`: `alignment` now
  returns `SnpAlignment` with empty sites; update `drop_ascertainment_noninformative`
  and `recode_symbols_as_absent` to handle the new `sites` field.
- Update `src/obikphylo/src/siblings/algorithms/family_scan.rs`: `GeneratedBatch`
  now includes a `kmers` field; `scan_layer_families` accepts `CanonicalKmer`
  and passes it to closures instead of masks.
- Update `src/obikphylo/src/siblings/algorithms/subsample.rs`: `SurvivingFamily`
  gains a `kmer` field; `sample_layer` closure accepts and uses this k-mer.
- Update `src/obikphylo/src/siblings/algorithms/snp_distance.rs`: `tally_2genomes`
  initializes `SurvivingFamily` with explicit `CanonicalKmer` values.
- Update `src/obikphylo/src/siblings/algorithms/stats.rs`: `sibling_annex_stats`
  signature adds `_kmer` parameter.
- Update `src/obikphylo/src/siblings/algorithms/family_annexes.rs`: closures in
  `ensure_layer_family_annexes` and `ensure_layer_pair_weight_annexes` now accept
  `_kmer`.
- Update `src/obikphylo/src/siblings/algorithms/pairwise.rs`: `ambiguous_vs_single`
  initializes k-mer using `CanonicalKmer::from_raw_unchecked(0)`.
- Add `iupac_state` function to `src/obikphylo/src/siblings/algorithms/masking.rs`
  mapping DNA characters to binary codes.
- Update `src/obikphylo/src/siblings/algorithms/mod.rs` to export `SiteLocus`
  and include new `alignment_vcf` module.
- Add `family_id` and `write_family_id_ascii` methods to
  `src/obikseq/src/kmer.rs` for stable k-mer family identification.
- Add tests `family_id_partitions_kmers_into_families` and
  `family_id_ascii_masks_centre` to `src/obikseq/src/tests/kmer.rs`.
- Update `src/obikphylo/src/siblings/mod.rs` imports to include new algorithm
  exports.
2026-11-09 20:58:05 +11:00
coissac ee62b131f9 Merge pull request 'Push ltpzonnnkpqs' (#79) from push-ltpzonnnkpqs into main
Reviewed-on: #79
2026-10-16 14:43:50 +00:00
coissac c6226dd40c Merge pull request 'Push wpvskkyozppn' (#78) from push-wpvskkyozppn into main
Reviewed-on: #78
2026-10-15 10:38:58 +00:00
114 changed files with 5552 additions and 1802 deletions

No files matched your search

+18 -32
View File
@@ -62,52 +62,38 @@ fn validate_evidence_compat(sources: &[&KmerIndex]) -> OKIResult<EvidenceKind>
Check all sources against the constraints above. Abort on any mismatch. Check all sources against the constraints above. Abort on any mismatch.
### 2. Bootstrap output from first source ### 2. Bootstrap output from the base source
Recursive file copy of `sources[0]` → `output`. Immediately after the copy: The base is chosen by `choose_base` (`merge.rs`): the source with the largest total `unitigs.bin` size (restricted to approximate-evidence sources when the output must be approximate). It is not necessarily `sources[0]`, and the genome columns follow the base first. Because `obikmer index` is not byte-reproducible (unitig segmentation varies), two near-equal sources can swap roles between runs. Its directory is copied to `output` (hard links for files `merge_partition` never rewrites, see below). Immediately after the copy:
- `index.meta` is rewritten with the final genome list (all sources, possibly renamed) and the effective evidence kind. - `index.meta` is rewritten with the final genome list (all sources, possibly renamed) and the effective evidence kind.
- In `Presence` mode, any `counts/` directories inherited from source_0 are removed. - In `Presence` mode, any `counts/` directories inherited from source_0 are removed.
- `spectrums/` from source_0 is removed and rebuilt from scratch across all sources, applying the (possibly renamed) labels. - `spectrums/` from source_0 is removed and rebuilt from scratch across all sources, applying the (possibly renamed) labels.
This establishes the partition layout, all existing MPHFs, unitigs, and evidence files. The first source's genomes occupy columns 0 … `n_dst_genomes - 1` in the destination. This establishes the partition layout, all existing MPHFs, unitigs, and evidence files. The base source's genomes occupy columns 0 … `n_dst_genomes - 1` in the destination.
### 3. For each subsequent source (parallel across partitions) ### 3. For each subsequent source (parallel across partitions)
`KmerPartition::merge_partition(i, sources, mode, n_dst_genomes, block_bits)` is called for each partition index `i`. `block_bits` is taken from `dst.meta.config.block_bits`. `obikmerge::partition_merge::merge_partition(dst, i, sources, mode, n_dst_genomes, block_bits, evidence)` runs once per partition index `i` (scheduled largest first by `PartitionRunner`). Each entry of `sources` is `(&KmerIndex, n_genomes)`, where `n_genomes` is the number of columns that source contributes (> 1 for an already-merged source). Code: `partition_merge/mod.rs` (orchestration, pass 1) and `partition_merge/pass2.rs` (per-group fill pipeline).
Each entry in `sources` is `(&KmerPartition, n_genomes)` where `n_genomes` is the column count that source contributes (> 1 when the source is itself a merged index). **First merge, Presence mode**: when `n_dst_genomes == 1`, `TypedLayer::<()>::init_presence_matrix` is called on every existing dst layer first. It materialises the implicit single-genome matrix (`presence/col_000000.pbiv`, all true) so that source columns can be appended to it.
**First merge, Presence mode**: when `n_dst_genomes == 1`, `Layer::<()>::init_presence_matrix` is called on every existing destination layer before any source column is appended. This creates `presence/col_000000.pbiv` set all-true (genome 0 is present in every slot). The dst layers are opened once per partition and shared by both passes.
**Pass 1 — classify kmers** **Pass 1 — new k-mers into a de Bruijn graph.** Only the sources' `unitigs.bin` *paths* are needed (`KmerLayer::new(..).unitigs_path()`); no source layer is opened. The files are streamed (at most `max_open = 2` open at once), and every k-mer absent from all dst layers is pushed into one `GraphDeBruijn`. All new k-mers of all sources go into a **single** new layer per merge, built by `materialize_layer` (unitigs + MPHF + evidence).
Iterate all kmers from all source partitions (via `UnitigFileReader` + canonical kmer iteration). For each kmer, probe the destination `LayeredMap<()>`: **New-layer matrix.** If a new layer exists, its matrix is created with `n_dst_genomes` absent columns (all zero/false) and closed immediately. This keeps the column-count invariant, and every source group then appends to it like to the existing layers.
- **Hit**: kmer already in the destination; record for Pass 2. **Pass 2 — by consecutive groups of `SRC_GROUP = 2` sources, in source order.** Per group:
- **Miss**: push kmer into a `GraphDeBruijn` accumulator.
**New layer construction** 1. open the group's source layers for this partition;
2. `MatrixBuilder::resume` every dst matrix (existing layers + new layer), check `n_cols == n_dst_genomes + col_base`, append the group's `group_n` columns;
3. run the pipeline: stream each source layer's k-mers, `hash_batch` + `fill_sub_matrix` on the source, locate each k-mer in the dst layers (`find_batch_grouped`) or else in the new layer's MPHF, and write the values column-major into the group's builders;
4. close the columns, then the matrices; drop the group's layers; `col_base += group_n`.
If the accumulator is non-empty, compute de Bruijn unitigs and call `Layer::<()>::build(&new_layer_dir, block_bits)`. All kmers absent from the destination — across **all** sources — accumulate into a **single** graph, producing one new layer per merge operation (not one per source). Only one group's source layers and destination columns are alive at a time, so the number of live mappings per partition does not depend on the number of sources. Columns are appended in source order, so every genome lands in the same absolute column as in a single-session merge.
**Pass 2 — fill column builders** **In-memory columns.** The group's columns are built in RAM (`MatrixBuilder::add_mem_col` → `MemBitVecBuilder` / `MemCompactIntVecBuilder`) and written once, sequentially, at `close` (`.building` file, then `rename`). The files are byte-identical to the former mmap-backed builders' output. There is no sparse file and no `MmapMut` on this path, so a block the FS cannot allocate is an `io::Error` instead of a SIGBUS. Files are not `fsync`ed: one sync per column file made merge ~8× slower. RAM per partition is at most `SRC_GROUP × src_n × N_part × (1/8 | 1)` bytes (presence | counts). History and figures: `merge_mmap_audit.md`; scaling limits: `merge_scaling.md`.
For each source and each of its layers, re-iterate unitigs and look up stored values via `SrcLayerData::lookup(kmer, src_n)`:
- `SrcLayerData::SetMembership` — no data directory exists; every kmer returns `vec![1; n_genomes]`
- `SrcLayerData::Presence` — reads `PersistentBitMatrix` from `presence/`
- `SrcLayerData::Count` — reads `PersistentCompactIntMatrix` from `counts/`
Hits are routed to `exist_builders[dst_layer][src_col]`; misses are routed to `new_src_builders[src_col]`.
**Column prepending for new layers**
Before source columns are written to the new layer, `n_dst_genomes` absent columns (all-zero / all-false) are prepended — one per genome already in the index — so the column count invariant holds immediately after layer creation.
**Close and update metadata**
Close all builders; update `presence/meta.json` or `counts/meta.json` with `{"n": N, "n_cols": n_dst_genomes + n_src_total}`; increment `PartitionMeta::n_layers` if a new layer was added.
### 4. Update index metadata ### 4. Update index metadata
@@ -129,7 +115,7 @@ impl Layer<PersistentBitMatrix> {
} }
``` ```
Each appends one column file to the matrix subdirectory (`counts/` or `presence/`). In `merge_partition`, columns are written directly via `PersistentBitVecBuilder` / `PersistentCompactIntVecBuilder` rather than through these helpers, but the invariant they enforce is the same. Each appends one column file to the matrix subdirectory (`counts/` or `presence/`). In `merge_partition`, columns are written through `MatrixBuilder::add_mem_col` (in-memory builders) rather than through these helpers, but the invariant they enforce is the same.
--- ---
@@ -139,7 +125,7 @@ After any merge, **every layer in every partition has exactly `n_genomes` column
Maintained by three mechanisms: Maintained by three mechanisms:
1. **Existing layers**: `n_src_total` columns appended (one per source genome). 1. **Existing layers**: `n_src_total` columns appended (one per source genome), group by group.
2. **New layers created during merge**: `n_dst_genomes` absent columns prepended before source columns. 2. **New layers created during merge**: `n_dst_genomes` absent columns prepended before source columns.
3. **First merge, Presence mode**: `init_presence_matrix` retroactively creates `presence/col_0` all-true for genome 0. 3. **First merge, Presence mode**: `init_presence_matrix` retroactively creates `presence/col_0` all-true for genome 0.
@@ -161,7 +147,7 @@ The invariant is a precondition of `LayeredStore` aggregation traits: `col_weigh
## On-disk impact ## On-disk impact
After merging `G` genomes (sources_0 contributes `G0`, subsequent sources the rest): After merging `G` genomes (sources_0 contributes `G0`, subsequent sources the rest), *before* `finalize` packs each matrix (sparse by default, `matrix.pbmx` with `--dense`):
``` ```
partitions/ partitions/
@@ -0,0 +1,96 @@
# Merge: mmap audit (2026-11-17)
Audit of the memory-mapping behaviour of `obikmerge::partition_merge::merge_partition`, triggered by two production failures on a 192-core NUMA server merging 186 freshly built mono-genome presence indexes (256 partitions, data on BeeGFS `/hoyt`). Status: **A + C implemented (2026-11-17)**, B deferred. See [Implementation](#implementation-2026-11-17). Implementation prompt: `prompts/merge_mmap_refactor.md` (repository root, outside the doc tree).
## Observed failures
| Run | Flags | Failure | When |
|---|---|---|---|
| 1 | none | `Bus error` (SIGBUS) | seconds after `merge_partitions` starts |
| 2 | `--cpu-max 32` | `memory allocation of 32768 bytes failed` → abort, many threads at once | after ~1 h, 128/256 partitions done |
Run 2 rules out "one partition too large for RAM": partitions are scheduled largest first (LFD, `merge.rs:289`), so at 50 % every remaining partition is smaller than one already merged. Small `malloc` failures on a 700 GB host are not RAM exhaustion either. Linux would OOM-kill (SIGKILL) instead. A NULL from `malloc` means `mmap`/`brk` returned `ENOMEM`: `vm.max_map_count` (default 65 530), `RLIMIT_AS`, or strict overcommit. Root access is not available on the compute server, so the fix must be in the code.
## Inventory: mappings held by one `merge_partition` call
Counts read from the code; totals assume 1 layer per source partition and 1 pre-existing dst layer (estimate, not measured).
| Resource | Where | Maps | Lifetime | Needed? |
|---|---|---|---|---|
| Source layers, fully opened | `partition_merge.rs:219-227` → `KmerLayer::open` → `MphfLayer::open` (`mphf_layer.rs:171`) | 4 per layer: MPHF (`MphfOnly`), `unitigs.bin` (direct access), `evidence.bin`, `unitigs.bin` again (`LayerEvidence::open`, `mphf_layer.rs:100`). ≈ **740** | whole partition | Only the MPHF. Pass 2 calls `hash_batch` (MPHF only, `mphf_layer.rs:258`) and `fill_sub_matrix` (matrix: `Implicit` for mono-genome, zero maps). Pass 1 only needs the `unitigs.bin` *path*. |
| Writable columns, created up front | `partition_merge.rs:327`, `:349` → `PersistentBitVecBuilder::new` / `PersistentCompactIntVecBuilder::new` | `set_len` (sparse) + `MmapMut`, `n_src_total × (L_dst + 1)` ≈ **370** | whole pass 2 | Only the columns of the ≤ 2 sources being streamed (`max_open = 2`). |
| dst layers, new-layer MPHF, sequential readers | various | ≈ 10 | variable | yes |
≈ **1 100 maps per partition**, > 95 % idle at any instant. × ~32 concurrent partitions (`--cpu-max 32`) ≈ 35 k, plus stacks and glibc arenas of several hundred pipeline threads. That is within reach of 65 530. Without a cap (192 cores) it is several times over. Consistent with run 2, **not measured**. Whether run 1's SIGBUS comes from the same exhaustion or from failure 4 below is **not established**.
## Weaknesses
1. **Source layers opened in full** although merge needs MPHF + matrix only. Evidence and both `unitigs.bin` maps are dead weight.
2. **All sources opened up front, for the whole partition.** This contradicts the point of partitioning, which is to work on a bounded sub-part of the index. The cost of one partition grows linearly with the number of sources, whatever the partition size.
3. **All destination columns created up front**, the same mismatch on the write side.
4. **Sparse files written through `MmapMut`.** A page whose block cannot be allocated (quota, full FS, BeeGFS) faults with SIGBUS. That cannot be caught and reports nothing.
5. **No bound on live mappings.** `obipipeline::throttle` only bounds open read streams.
6. *(minor)* `obicompactvec::mmap_file::open_mmap_file` validates magic and header length, not file length against `n`.
Side finding: `merge_parallelism.md` documents a `MemoryBudget` semaphore and `--budget-fraction`. `obisys::MemoryBudget` still exists but `merge.rs` no longer uses it (scheduling is `PartitionRunner`). That page is stale on this point.
## Plan
**A. Source-group streaming in pass 2** (`obikmerge`). Process sources in consecutive groups of `G` (G = 2, matching `max_open`). Per group: open its layers, resume each dst matrix (existing layers + new layer), append its `src_n` columns, fill them through the existing pipeline, close, drop. Pass 1 opens no source layer: it builds paths through `KmerLayer::new(..).unitigs_path()` (valid on `Empty`). Live maps per partition become independent of the source count. Column order is unchanged (sources visited in order, columns appended in order), so the on-disk result is identical. Fixes 2, 3, 5.
**C. In-memory column builders** (`obicompactvec`). Under A only `G × src_n × (L_dst+1)` columns are alive. Build them in RAM (bits: n/8 B; counts: n B + overflow map) and write once with `write_all` to the `.building` path, then publish. No sparse file and no `MmapMut` on the merge write path, so `ENOSPC`/quota becomes an `io::Error`. The on-disk format must be byte-identical to the current builders' output. Exposed through `MatrixBuilder`/`ColBuilder` so `obikselect`/`obikfilter` can adopt it later. Fixes 4.
**B. Slots-only layer opening** (`obikindex`, the authority crate). **Deferred to a separate work item** (decision 2026-11-17). Consumers whose kmers are members by construction need MPHF + matrix, never evidence. The same `hash_batch` + `fill_sub_matrix` pattern without `find` appears in three crates: `obikmerge` (`partition_merge.rs:489`), `obikfilter` (`partition_iter.rs:106`), `obikphylo` (`family_scan.rs:289`). `family_scan.rs:402` hashes `variants`; whether those are guaranteed members is unchecked. Fixes 1.
- **Rejected: lazy evidence inside `MphfLayer`.** `find`/`confirm` return `Option<usize>`, so an evidence open error at first use cannot be propagated (panic, or a signature change across the query path). It would also add an "initialised?" branch to the hottest path, and hide in runtime state a distinction that is structural: external query (`find`, needs evidence) vs in-layer iteration (`index`/`index_batch`, no evidence).
- **Rejected: a standalone parallel type** next to `KmerLayer`/`TypedLayer`. It would duplicate count/presence detection and matrix access, and two opening paths would drift.
- **Retained: composition.** A slots layer = `MphfOnly` + matrix, carrying `hash_batch`, `fill_sub_matrix`, `n_cols`. The full layer = slots layer + evidence, keeping `find`/`confirm`. `KmerLayer` exposes both openings (`open()` / e.g. `open_slots()`) over a single count/presence detection. Open point: `MphfLayer::n` currently comes from the evidence (`evidence.len()` / `fingerprint.n()`). A slots layer must take it from the matrix (`n_rows`; `layer_meta.json` for `Implicit`), so the two must be checked to always agree.
Why deferred: after A, B only takes merge from ~8 to ~2 maps per partition, which is negligible. It touches the `obikindex` query path and needs its own session with `query` performance checks.
Order: oracle first (end-to-end merge golden output), then A → C. A alone removes the scaling problem; C removes the SIGBUS class.
## Implementation (2026-11-17)
**A — source-group streaming** (`obikmerge/src/partition_merge/`, split into `mod.rs` + `pass2.rs`). As planned: pass 1 builds paths only; the new-layer matrix is created with its absent columns and closed; pass 2 iterates `sources.chunks(SRC_GROUP = 2)`, resuming each dst matrix, checking `n_cols == n_dst_genomes + col_base` (`InvalidData` otherwise), and appending the group's columns. Prerequisite: `MatrixBuilder::n_cols` (`obicompactvec`).
**C — in-memory columns** (`obicompactvec`). New `mem_builder.rs` (`MemBitVecBuilder`, `MemCompactIntVecBuilder`), `add_mem_col` on both matrix builders and on `MatrixBuilder`, and `ColBuilder::MemBit`/`MemInt`. Format writing is factored rather than duplicated: `format::write_pciv_header`/`write_pciv_tail` (also used by `finalize_pciv`), `mmap_file::write_std_header`/`write_built`. Byte identity with the mmap builders is unit-tested for n ∈ {0, 1, 63, 64, 65, 10 000}, with overwrites and > 2 048 overflow entries. Merge uses `add_mem_col` for source columns; `add_absent_col` is unchanged (never written through its mapping).
Deviations from the plan:
- **No `sync_all` in `close`.** With one forced sync per column file, `merge_partitions` went from 7.4 s to 63.6 s on the benchmark; without it, 7.0 s (measured, macOS). The mmap builders never synced either. The `rename` still prevents publishing a partial file. Caveat: on a write-back network FS an allocation error may only surface at `close(2)`, which `File`'s drop ignores.
- `add_mem_col` returns `ColBuilder`, not `io::Result<ColBuilder>`: it does no I/O.
- The mmap count builder (`PersistentCompactIntVecBuilder`) writes in place, without `.building`; the in-memory one goes through `.building` + `rename`. Same final bytes.
- `obikmerge`'s `matrix_builder_tests` were removed: they only exercised `obicompactvec::MatrixBuilder` and are covered by `obicompactvec/src/tests/{matrix_builder,mem_builder}.rs`. `tempfile` dev-dependency of `obikmerge` removed with them.
**Verification.** Two golden-output oracles, recorded on the unmodified code:
- `scripts/smoke_test_merge.sh`: synthetic, 4 genomes with shared fragments and counts ≥ 255, presence/count/`--dense`; frozen source indexes;
- `scripts/bench_merge_oracle.sh`: 20 real specimen presence indexes, sorted `dump --debug` hash, 85 681 700 k-mers.
Both pass after every step. The existing `global_index_presence` (October binary) gives the same hash.
`merge_partitions` wall time on the benchmark: 7.40 s (ref) → 7.91 s (A) → 7.07 s (A + C) → 8.05 s (after the module split, same code). Single runs; the spread is noise. RSS 243–256 MB throughout. The benchmark is too small to show either the mapping or the memory effect.
**Live mappings per partition (estimate, not measured).** Assumptions: 186 mono-genome sources, 1 layer each, 1 pre-existing dst layer.
| | Source layers | Dst columns | Other | Total |
|---|---|---|---|---|
| Before | 186 × 4 ≈ 744 | 186 × 2 ≈ 372 (`MmapMut`) | ≈ 10 | **≈ 1 100** |
| After A + C | 2 × 4 = 8 | 0 (in RAM) | ≈ 10 | **≈ 20**, independent of the source count |
B would take the 8 down to ~2.
**Follow-up — sparse pack (same day).** `pack_sparse_bit_matrix`/`pack_sparse_compact_int_matrix` used to open the columnar matrix directly, which maps every column file of a layer at once: `n_genomes` maps per partition, since layers are packed one after another. Both now call the dense pack first. It concatenates the column files one at a time without mapping them, so the sparse matrix is built from a single `matrix.pbmx`/`.pcmx` mapping. This deliberately reverses an earlier choice ("never materialise a packed copy just to read it back once"). It costs one extra write and read of the dense matrix: `pack` went from 1.11 s to 1.60 s on the benchmark (single run). As a side effect, a stale `matrix.pbmx` that predates newer columnar data is now re-packed instead of being used as is (`pack_bit_matrix` compares `n_cols`). Both oracles pass.
## Re-merging a merge output (resolved 2026-11-18)
Symptom: `I/O error: No such file or directory` in `merge_partitions` when a merge output was one of the sources. Cause (verified): `finalize_indexed` packs the output matrices (`sparse_meta.json` + dict/PFIV files, or `matrix.pbmx`), so `presence/`/`counts/` no longer holds `meta.json` + `col_*`. `choose_base` picks the largest source as destination, i.e. the merged one; bootstrap copies the packed layout and `MatrixBuilder::resume` → `MatrixMeta::load` finds no `meta.json`.
Fix: a packed index can only be a *source* (pass 2 reads it through `PersistentBitMatrix`/`PersistentIntMatrix::open`, which handle every layout). `choose_base` now filters candidates with `can_be_destination` (`obikmerge/src/merge.rs`): one layer is probed with `detect_storage`; `Columnar` qualifies, `Implicit` only for a single genome (materialized on the fly), `Packed`/`Sparse` never. With a single source there is nothing to append and the filter is skipped. When several sources are given and none qualifies, `run` returns an explicit error before any work. Consequence: the result's column order follows the chosen base first, so it can differ from the command-line order.
Not done: in-place unpack of a packed index to allow it as destination, or a destination-side append on the packed format — the sparse incremental strategy is tracked in `merge_scaling.md`.
Oracle: `scripts/smoke_test_merge.sh` scenario 2 (A B → AB, then AB C D) is compared against the flat merge by genome label (not against the golden dumps, whose layer placement differs), and merging only packed indexes is checked to be refused.
Residual risk, accepted: read-only maps still SIGBUS if a file is truncated by another process during a merge. That is inherent to `mmap`.
+41
View File
@@ -0,0 +1,41 @@
# Merge: scaling to thousands of genomes (open, 2026-11-17)
Open design question raised during the mmap refactor (`merge_mmap_audit.md`). Target: merges of ~6 000 genomes of a few Gb each. Nothing here is implemented or measured. Figures are estimates, labelled as such.
## What scales with what (after the mmap refactor)
| Cost | Depends on | Status |
|---|---|---|
| Live source layers + dst columns per partition (merge pass 2) | `SRC_GROUP × src_n`, not the source count | fixed by plan A |
| Pass-2 column RAM | `SRC_GROUP × src_n × N_part × (1/8 \| 1)` B | bounded, see below |
| Pass-1 de Bruijn graph | ≥ ~19 B per *new* k-mer (`HashMap<u64, AtomicU8>`, verified in `obidebruinj/src/debruijn/graph.rs`) | unchanged |
| Columnar intermediate files | `n_genomes × n_layers × n_partitions` (6 000 × 2 × 256 ≈ 3 M files, estimate) | **does not scale** |
| Sparse `pack` | used to map **all** columns of a layer at once (`ColumnarBitMatrix::open`); layers are packed one after another, so `n_genomes` maps per partition (186 × 192 concurrent ≈ 36 k, estimate) | **fixed 2026-11-17**: now goes columnar → dense `matrix.pbmx` (sequential concatenation, unmapped) → sparse, one mapping per partition |
`N_part`: k-mers of one dst partition over all layers. Pass-2 column RAM stays below pass-1 graph RAM for one-shot merges. It dominates only for **incremental** merges, where few genomes are added to a large index: the new layer is small, the existing layers are large. Example (estimate): `N_total = 1e12`, counts, 192 concurrent partitions → ~1.5 TB. A memory-budgeted concurrency cap is needed there whatever the column strategy. `obisys::MemoryBudget` exists but is no longer used by merge.
## Why a columnar intermediate at all
The sparse format (`bitmatrix/sparse.rs`) is row-major and deduplicated: per slot, either one genome (`singleton.pfiv`) or a `dict_id` (`multi.pfiv`) into a dictionary of distinct genome sets (`dict_offsets` + `dict_values.bin`). Its builder only offers `push_row(genomes)`, i.e. the **complete** genome set of a row, rows in **slot order**.
Merge discovers values **column by column** (source by source), in **random slot order** (MPHF order). The columnar files are a transposition buffer: they absorb random per-column writes, and `pack` transposes them row by row.
## Options to build the final format directly
1. **All sources at once.** That is exactly what plan A removed. No.
2. **Whole matrix in RAM.** `N_part × n_genomes / 8` B per partition: ~3 TB at 6 000 genomes and `N_part = 4e9` (estimate). No.
3. **External sort of `(slot, genome)` pairs.** Volume ∝ non-zeros ≈ Σ genome k-mers (~6 000 × 3e9). That is huge, while the final format is small precisely because genome sets repeat. No.
4. **Incremental sparse update per source group** (most promising). Per group:
- fill the group's columns in RAM (as now);
- stream the partition's existing sparse matrix **sequentially in slot order**;
- for each row: new set = old set ∪ group bits. Through the dictionary this is a transition `(old dict_id, group bit pattern) → new dict_id`, which can be memoised in a table;
- write the new sparse matrix sequentially, then publish.
No columnar files, no `pack`, no per-genome mapping, sequential I/O only. It also removes the re-merge bug: it starts from an existing sparse matrix instead of requiring a columnar one.
The cost is **one full pass over the partition's sparse matrix per group**. With `SRC_GROUP = 2` and 6 000 genomes that is 3 000 rewrites, far too many. Groups must therefore be large (64–256 sources), and group column RAM grows with them: `group × N_part / 8` B, e.g. 32 GB per partition for 64 sources at `N_part = 4e9` (estimate). One parameter trades passes against RAM, and it should be sized from a memory budget. The count matrix (`sparse_intmatrix.rs`) has its own format and needs the same analysis.
## Related findings
- The module header of `bitmatrix/sparse.rs` says "not used by any production code path yet". This is stale: `finalize_indexed` packs to sparse by default.
- The new layer of a merge is not byte-reproducible run to run: unitig segmentation varies, so the MPHF, evidence and matrix row order vary with it (measured 2026-11-17, `layer_1/*` differs in all 256 partitions; `layer_0` and metadata are identical). `obikmer index` shows the same behaviour. Merge oracles therefore compare sorted dumps, not file hashes. Where the non-determinism originates (parallel graph fill or unitig walk) is not established.
+1
View File
@@ -509,6 +509,7 @@ Provided finalisations:
| `hellinger_dist_matrix()` | `√partial_hellinger[i,j] / √2` | | `hellinger_dist_matrix()` | `√partial_hellinger[i,j] / √2` |
| `hellinger_euclidean_dist_matrix()` | `√partial_hellinger[i,j]` | | `hellinger_euclidean_dist_matrix()` | `√partial_hellinger[i,j]` |
| `threshold_mash_dist_matrix(k, t)` | Mash distance, derived from `threshold_jaccard_dist_matrix(t)` — no separate partial | | `threshold_mash_dist_matrix(k, t)` | Mash distance, derived from `threshold_jaccard_dist_matrix(t)` — no separate partial |
| `threshold_hamming_dist_matrix(t)` | `union[i,j] − inter[i,j]` (k-mers present in exactly one genome) — no separate partial |
### BitPartials ### BitPartials
+2 -2
View File
@@ -13,11 +13,11 @@
| `query` | Query an index with sequences and annotate matches | | `query` | Query an index with sequences and annotate matches |
| `dump` | Dump all indexed k-mers as CSV (kmer + per-genome counts or presence); supports the shared [kmer filtering](implementation/filtering.md) system; `--head N` limits output to the first N k-mers | | `dump` | Dump all indexed k-mers as CSV (kmer + per-genome counts or presence); supports the shared [kmer filtering](implementation/filtering.md) system; `--head N` limits output to the first N k-mers |
| `annotate` | Add or update genome metadata from a CSV file; or dump metadata as CSV | | `annotate` | Add or update genome metadata from a CSV file; or dump metadata as CSV |
| `phylo` | Compute pairwise evolutionary-distance proxies between genomes (`--metric jaccard\|mash\|hamming\|bray-curtis\|relfreq-bray-curtis\|euclidean\|relfreq-euclidean\|hellinger\|hellinger-euclidean`); optionally build NJ/UPGMA trees; `--presence-threshold N` sets the minimum count to consider a k-mer present when computing Jaccard/Mash on count indexes (default 1); optionally a central-position SNP/Sankoff calibration with exports for TNT/PhyG/IQ-TREE (see [evolutionary distances](theory/evolutionary_distances.md)) | | `phylo` | Genome-vs-genome distances, trees and SNP-based pseudo-alignments. `--distance` selects either a whole-genome distance computed from k-mer content (`jaccard`, `mash`, `hamming`, `bray-curtis`, `relfreq-bray-curtis`, `euclidean`, `relfreq-euclidean`, `hellinger`, `hellinger-euclidean`; `--presence-threshold N` converts counts to presence for Jaccard/Mash/Hamming on count indexes) or a `snp-*` distance (`snp-raw`, `snp-jc`, `snp-k2p`, `snp-k81`, `snp-f81`, `snp-t92`, `snp-tn93`, `snp-tv`; optional `--gamma-shape`) computed from central-position SNPs of k-mer families. `--nj`/`--upgma` build trees (Jaccard if no `--distance`); `--shared-kmers` writes the shared-k-mer count matrix on its own. SNP analyses read the sibling annex, built on demand (`--sibling-annex` forces it), with site selection/sampling options (`--n-sites`, `--quartet`, `--pair-balance`, `--entropy`, `--free-loss`, `--no-ambiguity`/`--only-diploid`, `--session`), a pseudo-alignment export (`--pseudo-alignment`, `--fasta`, `--vcf`) and Sankoff cost calibration with TNT/PhyG/IQ-TREE exports. See [evolutionary distances](theory/evolutionary_distances.md) |
| `unitig` | Build a global de Bruijn graph across all partitions and enumerate its unitigs as FASTA; supports the shared [kmer filtering](implementation/filtering.md) system | | `unitig` | Build a global de Bruijn graph across all partitions and enumerate its unitigs as FASTA; supports the shared [kmer filtering](implementation/filtering.md) system |
| `select` | Project and/or aggregate genome columns into a new or in-place index; the column-axis counterpart of `filter` (see [select](implementation/select.md)) | | `select` | Project and/or aggregate genome columns into a new or in-place index; the column-axis counterpart of `filter` (see [select](implementation/select.md)) |
| `estimate` | Estimate approximate-index parameters (z, evidence bits, FP rates) before indexing | | `estimate` | Estimate approximate-index parameters (z, evidence bits, FP rates) before indexing |
| `reindex` | Convert an index's evidence in-place: exact ↔ approx | | `convert` | Convert an index's evidence representation (exact / approximate / hybrid), in place |
| `utils` | Miscellaneous index utilities: `--new-label NEW=OLD` renames a genome label; `--upgrade-index` adds missing `layer_meta.json` to old indexes | | `utils` | Miscellaneous index utilities: `--new-label NEW=OLD` renames a genome label; `--upgrade-index` adds missing `layer_meta.json` to old indexes |
| `pack` | Pack per-column matrix files into single-file format to reduce query I/O | | `pack` | Pack per-column matrix files into single-file format to reduce query I/O |
+84
View File
@@ -2873,6 +2873,90 @@ user, no external consumers to break). Scoped to the distance matrix only:
`--shared-kmers` and `--raw-snp-counts` are counts, not distances, and keep `--shared-kmers` and `--raw-snp-counts` are counts, not distances, and keep
their existing CSV-only format. their existing CSV-only format.
## VCF export of the pseudo-alignment
`phylo --vcf` writes the drawn sites of `--pseudo-alignment` as a VCF
(`--fasta`/`--vcf` choose the formats, FASTA is the default, both can be
combined). A family has no genomic coordinate, so coordinates are virtual and
derived from where the scan finds the family: `CHROM = obikmer_<partition>_<layer>`,
`POS = family_idx + 1` (`scan_layer_families`' iteration-order index inside
that layer). Stable for a given index + annex, sorted by construction because
`on_family` is replayed in `family_idx` order. A global dense rank or a bit
packing of (layer, partition, idx) into one `POS` was rejected: it needs
per-layer minorant counts and can exceed VCF's 2^31-1 bound.
`ID` is `CanonicalKmer::family_id()`: centre forced to `A` on the
k-mer and on its revcomp, each canonicalised, minimum of the two (centre
printed as `N`). Index-independent, hence the key for merging VCFs from
different indexes. No `INFO/LOCUS` duplicate (dropped: same string as `ID`, costly
at 5M loci); `QUAL`/`FILTER`/`INFO` are all `.`. An alternative defined as the minimum of
`central_canonical_neighbors()` was dropped: it canonicalises 4 k-mers where
this one needs 2, with no other benefit. `family_id` idempotence and
family-wide equality are tested exhaustively for k = 3, 5, 7.
`SurvivingFamily` now carries the minorant k-mer (the scan callback passes it),
and `SnpAlignment` a `SiteLocus` per column. Alleles are the unambiguous cells
of the IUPAC matrix, in the FASTA's convention (central base of the member's
canonical form, not necessarily the `ID` k-mer's strand).
Absence handling. `masked_state` keeps absence (`m == 0`, `-`) as a real state
unless `--free-loss`, where it becomes missing (`?`). A first version mapped
`-` and `?` both to `.`: on a toy 4-genome index half of the lines then had
`ALT = .` (presence/absence-only sites, selected by `survives_masking`), i.e.
polymorphic sites rendered as monomorphic. Absence became an extra allele,
first written `*`; that was a misuse: in VCF 4.5 (section on the ALT column)
`*` is "allele missing due to overlapping deletion", and `<*>` is the gVCF
unspecified allele, `DEL`/`INS`/`DUP`/`INV`/`CNV` are structural-variant
types. Now the symbolic allele `<ABS>` (free-form ID, declared by a
`##ALT=<ID=ABS,...>` line, always written in the header, listed after the
bases); `?` and IUPAC ambiguity codes stay `.` (haploid GT cannot express a
genome carrying several forms; superseded below). A symbolic ALT formally
asks for a padding base before `POS`; moot here, positions being virtual.
Checked with `bcftools view`/`stats` and `tabix`: no warning, the sites with
`<ABS>` are classed "others" by `bcftools stats`, and
`-e 'ALT="<ABS>"'` drops exactly them. The writer needs no knowledge of
`--free-loss`: `-` simply never occurs with it. A reference-genome option
(`--vcf-ref`) was implemented then dropped: `REF` is always the per-site
majority base (alphabetical on ties).
Ploidy. VCF allows any ploidy in `GT` (number of alleles written, variable
per sample/site), but an obikmer cell only says *which* forms a genome
carries, not their multiplicity (`A/A/G` and `A/G/G` are both `R`), nor
whether several forms mean heterozygosity or paralogues. Compromise chosen:
`--no-ambiguity` => haploid (ambiguous cells are already `?`); otherwise
diploid, single base `X` -> `X/X`, two-base code -> `X/Y`, `-` -> `<ABS>/<ABS>`, and
three/four-base codes or `?` -> `./.`. `REF` is the majority allele counting
each copy once (so a het contributes 1+1, a homozygote 2). Counting copies
for a real dosage (polyploids) would need k-mer counts (`--with-counts`),
not used by the pseudo-alignment scan today. `--only-diploid` (selection option, `conflicts_with --no-ambiguity`) adds a
third `masked_state` rule: `count_ones() > 2` -> missing. It lives in
`SamplingSpec`/`masked_state`, i.e. the post-hoc masking and
`survives_masking`, so FASTA, Sankoff and `snp-*` see it too. Unit-tested
(`masked_state`); the toy index has no 3/4-member cells, so the end-to-end
run is identical to the default mode.
First version of this flag read the persisted annexes under the default
convention (a third variant was judged not worth it). Rejected: the draw
would then describe other genomes than the alignment, and the pair-balance
weights and entropy are computed from the carrier set. Replaced by
`AmbiguityPolicy { Keep, Forbid, OnlyDiploid }` (one type instead of
`free_loss`/`no_ambiguity`/`only_diploid` booleans in `masked_state`,
`is_carrier`, `SamplingSpec` and every annex key). Annex headers store
`AmbiguityPolicy::header_flags` (0 = Keep, 1 = Forbid, as the old bit, 2 =
OnlyDiploid), so annexes written before the change stay valid; each policy
has its own file (`*.diploid.*` for the new one). `--sibling-annex` no longer
precomputes everything (`AnnexRequest::everything` removed, a rebuild only
costs the base annex): it builds the derived annexes named by the other
options of the command (`AnnexSelection`: `--entropy`, `--quartet`,
`--pair-balance`, `--family-overlap`/`--min-shared-family`, under the one
policy given); the rest is built on demand as before. A rebuild still drops
every derived annex (their `family_idx` numbering is tied to the base one).
Checked on the toy index: `--sibling-annex` alone leaves only
`siblings.psib`; `--quartet --only-diploid` adds `quartet.diploid.pbiv` only.
Toy index: default mode 187
lines, no `ALT = .`; `--no-ambiguity` 93 lines (the sites whose polymorphism
was only ambiguity codes disappear, since those cells became `?`).
## References ## References
The Mash mutation-rate model this discussion contrasts with: The Mash mutation-rate model this discussion contrasts with:
+3 -1
View File
@@ -226,7 +226,9 @@ Design decision (user): the way genomes are counted depends on `--no-ambiguity`,
## Implemented: `--quartet` (2026-10-13) ## Implemented: `--quartet` (2026-10-13)
`--quartet` keeps only the families with at least two called states each carried by at least two genomes (`masking::is_quartet_informative` for the raw masks, `survives_quartet` for the exact check on masked states and `--exclude-genome`). Absence never counts as a state; with `--no-ambiguity`, multi-member genomes are left out of the count. The persisted annex is a plain `PersistentBitVec` (one bit per non-monomorphic family, compact ordinal) — no ad hoc format; the `--no-ambiguity` convention is carried by the file name (`quartet.pbiv`, `quartet.noamb.pbiv`) because that header has no room for a flag. It is a safe pre-filter: excluding genomes can only destroy a bipartition, never create one. `--quartet` keeps only the families with at least two called states each carried by at least two genomes (`masking::is_quartet_informative` for the raw masks, `survives_quartet` for the exact check on masked states and `--exclude-genome`). Genomes are read through `masked_state`, exactly as in the alignment: absence is a state of its own unless `--free-loss` (then it is missing and counts for nothing), and with `--no-ambiguity` multi-member genomes are left out of the count. The persisted annex is a plain `PersistentBitVec` (one bit per non-monomorphic family, compact ordinal) — no ad hoc format; its counting convention (`QuartetConvention`: ambiguity policy × `--free-loss`) is carried by the file name (`quartet.pbiv`, `quartet.noamb.pbiv`, `quartet.diploid.pbiv` for `--free-loss`; `quartet.abs.*.pbiv` for absence-as-a-state) because that header has no room for a flag.
*Correction (2026-11-19).* The first version never counted absence, whatever `--free-loss` said, which contradicted the alignment: without `--free-loss` a genome lacking the family has the state `∅`, so `A A ∅ ∅` separates two pairs yet was rejected. Fixed in `has_bipartition` (index 0 is now counted, and is only ever populated when absence is a state), in the annex (one file per `QuartetConvention`) and in `--sample-stats` (informative sites follow the same rule). The old file names keep the `--free-loss` semantics, which is exactly the old rule, so annexes already built with `--free-loss` stay valid; the sampling measured below used `--free-loss` and is unaffected. It is a safe pre-filter: excluding genomes can only destroy a bipartition, never create one.
Plumbing: `AnnexRequest` (which annexes, per convention) and `ensure_layer_family_annexes` (one shared scan for entropy, quartet and overlap) replace the entropy-only builder; the sampling parameters travel as one `SamplingSpec { free_loss, no_ambiguity, quartet, entropy_bias }` instead of a growing positional list. Plumbing: `AnnexRequest` (which annexes, per convention) and `ensure_layer_family_annexes` (one shared scan for entropy, quartet and overlap) replace the entropy-only builder; the sampling parameters travel as one `SamplingSpec { free_loss, no_ambiguity, quartet, entropy_bias }` instead of a growing positional list.
+107
View File
@@ -0,0 +1,107 @@
# VCF pseudo-alignment
[`phylo --vcf`](../usage/phylo.md) writes the sites of a SNP pseudo-alignment as a VCF file, `<prefix>_alignment.vcf`. This page describes exactly what that file contains, because the k-mer families it is built from are not ordinary genomic variants and a few conventions follow from that.
## Why a VCF, and what a site is
The pseudo-alignment (see [Central-position SNP model](../theory/phylogeny/snp_based/central_snp_model.md)) has one column per *family*: the up to four k-mers that share all their bases except the central one. A genome that carries one of these k-mers shows a particular base at the centre, and the columns of the alignment are those central bases, genome by genome. The FASTA output keeps this as a matrix of letters. The VCF output lists the same columns as sites with alleles and genotypes, which is the form that variant-oriented tools (`bcftools`, population-genetics software, genotype viewers) expect.
Nothing here comes from a reference genome. The file has no real coordinates, and the sites are not located on a chromosome: the positions below are labels that make the file valid and reproducible, not map positions.
## Producing the file
```
obikmer phylo --vcf --n-sites 5000000 -o run1 my_index
```
`--vcf` implies `--pseudo-alignment`, so `--n-sites N|all` is required, together with an already-built sibling annex (see [`phylo`](../usage/phylo.md)). `--fasta` and `--vcf` can be given together to get both files from the same draw; with neither, only the FASTA is written. Every option that shapes the draw (`--free-loss`, `--no-ambiguity`, `--only-diploid`, `--quartet`, `--exclude-genome`, ...) applies to the VCF exactly as to the FASTA, since both describe the same sites.
## Layout
An illustrative excerpt (columns are tab-separated in the file):
```
##fileformat=VCFv4.5
##source=obikmer
##contig=<ID=obikmer_0_0>
##contig=<ID=obikmer_0_1>
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype (diploid)">
#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT idx0 idx1 idx2 idx3
obikmer_0_0 440 CGATTNGTTAA A G . . . GT 0/0 1/1 0/0 0/0
obikmer_0_0 1511 CACGANTGTAG A G,T . . . GT 0/0 0/0 1/1 2/2
obikmer_0_1 9 TTACGNAGCTA C * . . . GT 0/0 0/0 1/1 0/0
```
There is one sample column per genome that takes part in the draw, named after the genome label, in the index's own order. A genome removed with `--exclude-genome` or `--min-shared-family` has no column at all. `QUAL`, `FILTER` and `INFO` are always `.`: obikmer applies no variant filter and attaches no annotation, and leaving them empty keeps a file of millions of sites compact.
## Coordinates: `CHROM` and `POS`
An index is organised in partitions, and each partition holds one or more layers of k-mers (see [Index construction and on-disk layout](index_layout.md)). A family always belongs to exactly one layer of one partition, and the families of a layer are visited in a fixed order. The VCF uses that structure as coordinates:
- `CHROM` is `obikmer_<partition>_<layer>`, one virtual contig per layer of each partition in which at least one site was drawn;
- `POS` is the rank of the family inside that layer, starting at 1.
For a given index and a given sibling annex, a family therefore always receives the same `CHROM` and `POS`, whatever the draw, the options or the number of sites requested. Two runs on the same index can be compared position by position. This stability is lost if the sibling annex is rebuilt, or if the index is rebuilt, merged or filtered, because the families are then enumerated again. Positions are never comparable between two different indexes; use `ID` for that. Even rebuilding an index from the very same genomes may enumerate the families in another order: on a small test index built twice from identical input, the family `CGATTNGTTAA` had `POS` 440 the first time and 371 the second, with the same `ID`.
The file is sorted by partition, then layer, then position, which is also the order of the `##contig` lines, so tools that require a sorted file accept it directly. Positions are sparse: only the drawn families are written.
## Identifier: `ID`
The `ID` column holds a label for the family that does not depend on the index. It is built from one of the family's k-mers by replacing its central base with `N` (for example `CGATTNGTTAA` for k = 11), after choosing, among the family's possible orientations, the one that sorts first. All the members of a family, whichever strand they were read on, give the same label.
This is what makes two files comparable even when they come from different indexes: the same family has the same `ID` in both, while its `CHROM` and `POS` will generally differ. It is also unique within a file, since two different families never share a label.
## Alleles: `REF` and `ALT`
The alleles are the bases carried at the centre of the family's k-mers, as in the FASTA. Two points are worth knowing.
First, the base is the central base of the member's *canonical* form (the smaller of the k-mer and its reverse complement). That is not necessarily the strand of the `ID` label, so the letters should not be matched to the `ID` sequence by position; they are labels for the allele, not a reading of that sequence.
Second, `REF` carries no meaning of reference: it is simply the most frequent allele among the genomes of the draw, each allele copy counted once, with the alphabetical order A < C < G < T deciding ties. `ALT` lists the remaining observed bases in the order A, C, G, T, separated by commas. There is no option to choose a reference genome.
### Absence: the `<ABS>` allele
A genome may carry none of the family's k-mers. Without `--free-loss` this absence is treated as a real state (the family was lost, or is absent from that genome), which the FASTA shows as `-`. In the VCF it becomes one more allele, the symbolic allele `<ABS>`, listed last in `ALT` and declared in the header by `##ALT=<ID=ABS,Description="Family absent from the genome">`. A genome without the family receives the genotype of `<ABS>`; for a site with `ALT` equal to `G,T,<ABS>` that is `3/3`. A site that differs between genomes only by presence versus absence has `ALT` equal to `<ABS>`.
The VCF specification gives the plain character `*` a meaning of its own (an allele missing because of an overlapping deletion described by another record), which is not the situation here, so it is deliberately not used. `<ABS>` is a free-form symbolic identifier, outside the reserved structural-variant names (`DEL`, `INS`, `DUP`, `INV`, `CNV`) and outside `<*>`, which belongs to reference blocks. Tools that only look at sequence alleles treat the site as having an alternative that is not a base.
With `--free-loss`, absence is declared missing data instead: `<ABS>` never appears, and the genomes concerned get a missing genotype.
## Genotypes
Every genome gets one `GT` value per site. The ploidy of these values is a convention chosen from the options, and the header says which (`Genotype (diploid)` or `Genotype (haploid)`).
| Cell in the alignment | Diploid (default, `--only-diploid`) | Haploid (`--no-ambiguity`) |
|---|---|---|
| one base `X` | `X/X`, for example `0/0` or `1/1` | `X`, for example `0` |
| two bases (IUPAC `R Y S W K M`) | the heterozygote, for example `0/1` | does not occur |
| three or four bases (`B D H V N`) | `./.` (not expressible) | does not occur |
| absent family (`<ABS>`) | `<ABS>/<ABS>`, for example `3/3` | the index of `<ABS>` |
| missing (`?`) | `./.` | `.` |
The two allele indices of a diploid genotype are always written in increasing order (`0/1`, never `1/0`), and the separator is `/` because no phasing information exists.
Why diploid by default? A genome can carry several forms of the same family, which the alignment records as an IUPAC ambiguity code. A diploid genotype can express up to two forms, and `--only-diploid` removes the genomes that would need more (three or four forms) before the sites are selected, so that no `./.` comes from them. `--no-ambiguity` goes further and treats every such genome as missing, which makes all genotypes haploid.
This is a convention, not a measurement. The file records which forms a genome carries, not how many copies of each: `A/A/G` and `A/G/G` both give the code `R` and the genotype `0/1`. Likewise, a genome with two forms may be heterozygous at one locus, or it may carry two similar regions of its genome (paralogues, repeats); the file cannot tell these apart. Do not read the genotypes as calls from sequencing reads.
## Which sites are written
Only the families that survive the draw and its masking appear. Within those:
- a site needs at least one genome with a base or (in diploid mode) a two-base code, because `REF` is a base; a site where every genome is missing, absent or has an ambiguity code too wide for the mode is left out of the VCF while remaining in the FASTA;
- the log line at the end of the run gives the number of sites written out of the number drawn, so the difference is visible.
A site is selected because its genomes show at least two distinct states, and absence counts as a state. A site that differs between genomes only by presence versus absence is therefore a legitimate site, written with `ALT` equal to `<ABS>`.
## Using the file with other tools
- The contigs are virtual and have no declared length, and the file refers to no reference. Tools that need one (annotation, liftover) cannot use it; tools that only read sites and genotypes (`bcftools view`, `query`, `stats`, population-genetics software, PLINK-style converters) can.
- If a tool rejects multiallelic sites or the `<ABS>` allele, `bcftools view -m2 -M2` keeps biallelic sites only, and `bcftools view -e 'ALT="<ABS>"'` removes the sites involving absence. The file as written was read by `bcftools` and indexed by `tabix` without complaint on a small test index, in both diploid and haploid modes. `bcftools stats` counts the sites with `<ABS>` among the "others" (symbolic alleles) rather than as SNPs.
- If a tool expects haploid data, run `phylo` with `--no-ambiguity` rather than converting the genotypes afterwards.
## See also
- [`phylo`](../usage/phylo.md): all the options that shape the draw.
- [Sampling design](../theory/phylogeny/snp_based/sampling_design.md): how the sites are drawn.
- [Central-position SNP model](../theory/phylogeny/snp_based/central_snp_model.md): what a family is.
@@ -21,7 +21,7 @@ A family where every genome carries the same single state has entropy 0 (uninfor
## Quartet restriction ## Quartet restriction
A family only constrains the shape of a tree if its carrier genomes show a bipartition: two distinct states, each carried by at least two genomes. With fewer carriers, or with a single variant against a uniform background, every topology explains the family equally well. In a skimming dataset many families have only two or three carriers, so a uniform draw spends a large share of its quota on them. `--quartet` removes them before the draw. The test counts called states only, so absence never forms one of the two pairs. A family only constrains the shape of a tree if its carrier genomes show a bipartition: two distinct states, each carried by at least two genomes. With fewer carriers, or with a single variant against a uniform background, every topology explains the family equally well. In a skimming dataset many families have only two or three carriers, so a uniform draw spends a large share of its quota on them. `--quartet` removes them before the draw. The test counts the states of the alignment: without `--free-loss`, absence is a state like any other and can form one of the two pairs; with `--free-loss` it is missing data and never does.
## Pair balance ## Pair balance
+8
View File
@@ -12,6 +12,14 @@ obikmer index -o OUTPUT [OPTIONS] [INPUTS...]
|---|---| |---|---|
| `INPUTS...` | Input sequence files or directories (FASTA/FASTQ/GenBank, gzip optional). If omitted, reads from stdin. | | `INPUTS...` | Input sequence files or directories (FASTA/FASTQ/GenBank, gzip optional). If omitted, reads from stdin. |
## Which files are read
A file named explicitly on the command line is always read, whatever its extension: the format (FASTA, FASTQ or GenBank) and the compression (gzip, bzip2, xz or zstd) are detected from the content. If the content is not a sequence file, the command reports an error rather than skipping the file.
When an input is a directory, the files inside it are selected by extension: `.fasta`, `.fa`, `.fna` (the extension of NCBI genome downloads), `.ffn`, `.fas`, `.fsa`, `.fastq`, `.fq`, `.gbff`, `.gbk` and `.gb`, each optionally followed by `.gz`.
A small input spread over many partitions leaves most partitions empty. This is valid: the index is built normally and empty partitions simply hold no k-mers.
## Options ## Options
| Option | Default | Description | | Option | Default | Description |
+17 -1
View File
@@ -20,10 +20,26 @@ obikmer merge -o OUTPUT SOURCE... [OPTIONS]
| `--force` | off | Overwrite an existing output directory | | `--force` | off | Overwrite an existing output directory |
| `--force-presence` | off | Store the merged index as presence/absence even if all sources have counts | | `--force-presence` | off | Store the merged index as presence/absence even if all sources have counts |
| `--rename-duplicates` | off | Disambiguate duplicate genome labels (`.1`, `.2`, …) instead of failing | | `--rename-duplicates` | off | Disambiguate duplicate genome labels (`.1`, `.2`, …) instead of failing |
| `--budget-fraction` | `0.5` | Fraction of available RAM reserved as the memory budget for parallel partition merging | | `--dense` | off | Store the merged index's presence/absence matrices in the dense packed format instead of the default sparse one (see [pack](pack.md)) |
## Behaviour ## Behaviour
The output mode is chosen automatically: if every source index stores counts, the merged index stores counts too; otherwise it is presence/absence. `--force-presence` forces presence/absence regardless of the sources. The output mode is chosen automatically: if every source index stores counts, the merged index stores counts too; otherwise it is presence/absence. `--force-presence` forces presence/absence regardless of the sources.
By default, merging two indexes that share a genome label fails with an error; `--rename-duplicates` instead appends a numeric suffix to keep both copies. By default, merging two indexes that share a genome label fails with an error; `--rename-duplicates` instead appends a numeric suffix to keep both copies.
The merged index is written in the sparse packed format by default, because it is usually several times smaller on disk; `--dense` keeps the dense format, which is faster when whole genome columns are read at once, for example for distance matrices. The [pack](pack.md) page explains the trade-off between the two formats.
## Resource usage
The merge works partition by partition, and processes several partitions in parallel. Within a partition, the source indexes are read two at a time rather than all at once. The memory and the number of files mapped into memory for one partition therefore depend on how many k-mers the partition holds, not on how many indexes are being merged: merging hundreds of genomes maps about as many files per partition as merging three. The partitions do grow as more diverse genomes are added, so the total memory still rises with the size of the merged collection.
Each genome column being filled is held in memory and written to disk in one go once it is complete. A full disk or an exhausted quota is therefore reported as an ordinary error rather than crashing the program. The largest memory cost comes from the k-mers that are new to the merged index, which are assembled in memory, partition by partition, before being written out.
## Merging an index that was already merged
An index produced by `merge` can be used as one of the sources of a later merge, which is how a collection is extended without starting again from the original genomes. There is one rule: among the sources, at least one must not have been through a merge. A merged index is stored in a compact packed format (see [pack](pack.md)) that is read very efficiently but cannot be extended in place. The merge therefore takes a freshly built index as its starting point and brings the packed index in as a source. A typical case is a merged index of many genomes plus one or a few newly built single-genome indexes, and this works.
If every source has already been merged or packed, the command stops before doing any work and says so. In that case, merge the original single-genome indexes together instead.
The genome columns of the result follow the order of the starting index first, then the other sources, so they may not come out in the order the sources were listed. The genome labels always identify each column unambiguously.
+15 -7
View File
@@ -48,13 +48,15 @@ See [Central-position SNP model](../theory/phylogeny/snp_based/central_snp_model
| Option | Description | | Option | Description |
|---|---| |---|---|
| `--sibling-annex` | Build (or rebuild) the sibling-count/minorant annex, then every annex derived from it (entropy, quartet, pair weight, family overlap, under both `--no-ambiguity` conventions). Every option in this section needs the sibling annex, and builds it by itself, under an exclusive lock on the index, if it is missing (as it is after `migrate-mphf`); running `--sibling-annex` first does the same work once, which is the safe way to start several `phylo` runs on the same index at the same time | | `--sibling-annex` | Build (or rebuild) the sibling-count/minorant annex. Rebuilding it makes the annexes derived from it obsolete, so they are dropped; then only the derived annexes that the other options of the same command read are rebuilt, each under the `--no-ambiguity` / `--only-diploid` convention given: `--entropy` the entropy annex, `--quartet` the quartet annex, `--pair-balance` the pair-weight and family-overlap annexes, `--family-overlap` or `--min-shared-family` the family-overlap annex. For example `--sibling-annex --quartet --only-diploid` prepares exactly the quartet annex of that convention, and `--sibling-annex` alone prepares none. An annex that is not prepared is built later, on demand, by the computation that reads it. Every option in this section needs the sibling annex, and builds it by itself, under an exclusive lock on the index, if it is missing (as it is after `migrate-mphf`); running `--sibling-annex` first does the same work once, which is the safe way to start several `phylo` runs on the same index at the same time |
| `--sibling-stats` | Write `<prefix>_siblings.csv`: the family-size distribution, per genome and globally | | `--sibling-stats` | Write `<prefix>_siblings.csv`: the family-size distribution, per genome and globally |
| `--sibling-hist` | Print the global family-size histogram (1-4 members) only | | `--sibling-hist` | Print the global family-size histogram (1-4 members) only |
| `--family-overlap` | Write `<prefix>_family_overlap.csv`: for every genome pair, how many variable families both genomes carry a call for | | `--family-overlap` | Write `<prefix>_family_overlap.csv`: for every genome pair, how many variable families both genomes carry a call for |
| `--shannon` | Write `<prefix>_entropy.csv`: per-family Shannon entropy, one row per family, full unsampled scan | | `--shannon` | Write `<prefix>_entropy.csv`: per-family Shannon entropy, one row per family, full unsampled scan |
| `--shannon-summary` | Write `<prefix>_shannon_summary.csv`: the same scan as `--shannon`, reduced to a distribution (see below) — a few hundred thousand lines at most, whatever the size of the index | | `--shannon-summary` | Write `<prefix>_shannon_summary.csv`: the same scan as `--shannon`, reduced to a distribution (see below) — a few hundred thousand lines at most, whatever the size of the index |
| `--pseudo-alignment` | Write `<prefix>_alignment.fasta`: a SNP-only pseudo-alignment. Requires `--n-sites N\|all` | | `--pseudo-alignment` | Write `<prefix>_alignment.fasta`: a SNP-only pseudo-alignment. Requires `--n-sites N\|all` |
| `--fasta` | Choose FASTA as the pseudo-alignment format (the default when neither `--fasta` nor `--vcf` is given). Implies `--pseudo-alignment` |
| `--vcf` | Also (or instead) write `<prefix>_alignment.vcf`, the same sites as a VCF. Implies `--pseudo-alignment`; combinable with `--fasta` |
### `--sibling-stats` ### `--sibling-stats`
@@ -75,12 +77,17 @@ See [Central-position SNP model](../theory/phylogeny/snp_based/central_snp_model
`<prefix>_alignment.fasta` — one record per non-excluded genome, one column per variable family (family size ≥ 2). Each site is IUPAC-coded from the genome's presence mask at that family: a single observed form → the plain base; several forms → the matching IUPAC ambiguity code; no form → `-`. `<prefix>_alignment.fasta` — one record per non-excluded genome, one column per variable family (family size ≥ 2). Each site is IUPAC-coded from the genome's presence mask at that family: a single observed form → the plain base; several forms → the matching IUPAC ambiguity code; no form → `-`.
#### VCF output (`--vcf`)
The same drawn sites can be written as `<prefix>_alignment.vcf`, with `--fasta` and `--vcf` freely combined (FASTA alone remains the default). The coordinates, identifiers, alleles and genotypes of this file follow conventions of their own, described on the [VCF pseudo-alignment](../formats/vcf.md) page. The genotype ploidy follows `--no-ambiguity` (haploid) and `--only-diploid` (diploid, with the genomes carrying three or four forms of a family removed beforehand).
| Option | Default | Description | | Option | Default | Description |
|---|---|---| |---|---|---|
| `--n-sites N\|all` | none (mandatory here) | Number of families to draw, or `all` to use every variable family that passes the other filters, with no sampling. A number is a target, not a guarantee. | | `--n-sites N\|all` | none (mandatory here) | Number of families to draw, or `all` to use every variable family that passes the other filters, with no sampling. A number is a target, not a guarantee. |
| `--free-loss` | off | Treat a genome carrying none of a family's observed members as missing data (`?`) instead of `-` | | `--free-loss` | off | Treat a genome carrying none of a family's observed members as missing data (`?`) instead of `-` |
| `--no-ambiguity` | off | Treat a genome carrying more than one member of a family as missing data (`?`) instead of an IUPAC ambiguity code | | `--no-ambiguity` | off | Treat a genome carrying more than one member of a family as missing data (`?`) instead of an IUPAC ambiguity code |
| `--quartet` | off | Keep only the families that can separate two pairs of genomes from each other: at least two distinct called states, each carried by at least two genomes. Which genomes count as carriers follows `--no-ambiguity` | | `--only-diploid` | off | Milder than `--no-ambiguity` (the two cannot be combined): keep the genomes carrying one or two members of a family (a base, or a two-base IUPAC code such as `R`) and treat those carrying three or four as missing data (`?`) |
| `--quartet` | off | Keep only the families that can separate two pairs of genomes from each other: at least two distinct states, each carried by at least two genomes. Absence counts as a state unless `--free-loss` is given. Which genomes count as carriers follows `--no-ambiguity` |
| `--pair-balance` | off | Favour the families covering the pairs of genomes that share the fewest families across the index, so that every pair is better represented in the sample. Which genomes count as carriers follows `--no-ambiguity` | | `--pair-balance` | off | Favour the families covering the pairs of genomes that share the fewest families across the index, so that every pair is better represented in the sample. Which genomes count as carriers follows `--no-ambiguity` |
| `--sample-stats` | off | Write three CSV files describing what the draw contains (see "Describing a draw" below) | | `--sample-stats` | off | Write three CSV files describing what the draw contains (see "Describing a draw" below) |
| `--sample-plan` | off | Write `<prefix>_sample_plan.csv`: how the draw was spread over the layers of the index | | `--sample-plan` | off | Write `<prefix>_sample_plan.csv`: how the draw was spread over the layers of the index |
@@ -129,9 +136,9 @@ With `--n-sites all` (or, for a `snp-*` `--distance` value, without `--n-sites`)
### `--quartet`: keeping only families that can resolve a tree ### `--quartet`: keeping only families that can resolve a tree
A family carried by fewer than four genomes, or whose genomes all share one state except a single variant, fits every tree topology equally well and so tells a tree nothing: to separate two pairs of genomes from each other (`AB|CD`) it takes at least two distinct states, each carried by at least two genomes. `--quartet` restricts the draw to the families that show this pattern. Absence is never a state here: a genome that does not carry the family does not count, with or without `--free-loss`. A genome carrying several members of the family counts as a carrier of its own combination of bases, unless `--no-ambiguity` is given, in which case it is left out, as in the alignment. A family carried by fewer than four genomes, or whose genomes all share one state except a single variant, fits every tree topology equally well and so tells a tree nothing: to separate two pairs of genomes from each other (`AB|CD`) it takes at least two distinct states, each carried by at least two genomes. `--quartet` restricts the draw to the families that show this pattern. The states are those of the alignment. Without `--free-loss`, the absence of a family is a state of its own, so two genomes that carry a base and two genomes that lack the family do form two pairs. With `--free-loss`, absence is missing data and never forms a pair. A genome carrying several members of the family counts as a carrier of its own combination of bases, unless `--no-ambiguity` is given, in which case it is left out, as in the alignment.
The restriction is applied before the draw, from a persisted annex (one bit per variable family, built the first time `--quartet` is used with a given `--no-ambiguity` convention, in one pass shared with the other annexes, and built for both conventions by `--sibling-annex`), then checked exactly on each family drawn, taking `--exclude-genome` into account. The quota given to `--n-sites` is therefore spent only on usable families. The restriction is applied before the draw, from a persisted annex (one bit per variable family, built the first time `--quartet` is used with a given `--no-ambiguity` / `--only-diploid` convention and with or without `--free-loss`, in one pass shared with the other annexes, or ahead of time by `--sibling-annex --quartet`), then checked exactly on each family drawn, taking `--exclude-genome` into account. The quota given to `--n-sites` is therefore spent only on usable families.
### Describing a draw: `--sample-stats`, `--sample-plan` ### Describing a draw: `--sample-stats`, `--sample-plan`
@@ -155,7 +162,7 @@ A call is any state other than the empty one: absence is never a call, with or w
Two genomes that are distantly related share few families, so a uniform draw covers such a pair with few sites, far fewer than a close pair: the pairs that matter for the deep branches of a tree are the worst represented. `--pair-balance` compensates. Each family is weighted by the mean, over the pairs of genomes carrying it, of 1 divided by the number of families that pair shares across the whole index (the matrix `--family-overlap` writes): a family carried by a rarely-sharing pair weighs more. It is the mean, not the sum, so a family carried by most genomes — in skimming data often a high-copy region — is not favoured merely for covering many pairs. Two genomes that are distantly related share few families, so a uniform draw covers such a pair with few sites, far fewer than a close pair: the pairs that matter for the deep branches of a tree are the worst represented. `--pair-balance` compensates. Each family is weighted by the mean, over the pairs of genomes carrying it, of 1 divided by the number of families that pair shares across the whole index (the matrix `--family-overlap` writes): a family carried by a rarely-sharing pair weighs more. It is the mean, not the sum, so a family carried by most genomes — in skimming data often a high-copy region — is not favoured merely for covering many pairs.
Each layer's share of the draw is proportional to the total weight of its families, and a family's chance of being drawn is proportional to its own weight. The weights are persisted (one value per variable family, built the first time `--pair-balance` is used with a given `--no-ambiguity` convention, after the family-overlap matrix it depends on, and by `--sibling-annex` for both conventions). They combine with `--entropy` and `--quartet`: the weight multiplies the entropy kernel, and `--quartet` restricts the families first. Each layer's share of the draw is proportional to the total weight of its families, and a family's chance of being drawn is proportional to its own weight. The weights are persisted (one value per variable family, built the first time `--pair-balance` is used with a given `--no-ambiguity` / `--only-diploid` convention, after the family-overlap matrix it depends on, or ahead of time by `--sibling-annex --pair-balance`). They combine with `--entropy` and `--quartet`: the weight multiplies the entropy kernel, and `--quartet` restricts the families first.
### `--entropy MU` / `--entropy-sd SIGMA`: biasing the sample toward informative families ### `--entropy MU` / `--entropy-sd SIGMA`: biasing the sample toward informative families
@@ -163,7 +170,7 @@ By default, `--n-sites N` draws families uniformly. With `--entropy`/`--entropy-
The first `phylo` run on a given index that uses `--entropy`/`--entropy-sd` pays a one-time extra cost (every candidate family's entropy is computed once and saved alongside the index); later runs, even with different `MU`/`SIGMA`, reuse that saved data. The first `phylo` run on a given index that uses `--entropy`/`--entropy-sd` pays a one-time extra cost (every candidate family's entropy is computed once and saved alongside the index); later runs, even with different `MU`/`SIGMA`, reuse that saved data.
The entropy that weights the draw follows `--no-ambiguity`, so that it describes the same genomes as the alignment. Without the flag, every genome carrying the family counts, and a genome carrying several members of it is a state of its own (`entropy15`). With the flag, only the genomes carrying exactly one member count, since the others become missing data (`entropy4`). The two entropies are saved separately: switching `--no-ambiguity` builds the other one once, without discarding the first, and `--sibling-annex` builds both. The entropy that weights the draw follows `--no-ambiguity` and `--only-diploid`, so that it describes the same genomes as the alignment. By default, every genome carrying the family counts, and a genome carrying several members of it is a state of its own (`entropy15`). With `--no-ambiguity`, only the genomes carrying exactly one member count, since the others become missing data (`entropy4`). With `--only-diploid`, the genomes carrying one or two members count and those carrying three or four become missing data. The three entropies are saved separately: switching convention builds the missing one once, without discarding the others, and `--sibling-annex --entropy` builds the one of the convention given.
## Sankoff calibration and phylogenetic exports ## Sankoff calibration and phylogenetic exports
@@ -222,7 +229,8 @@ CSV matrix layout (`_dist.csv`, `_shared.csv`, `_family_overlap.csv`): header `g
| `<prefix>_shannon_summary.csv` | `--shannon-summary` | CSV table | families per entropy bin, number of carriers and number of single-member carriers | | `<prefix>_shannon_summary.csv` | `--shannon-summary` | CSV table | families per entropy bin, number of carriers and number of single-member carriers |
| `<prefix>_sample_genomes.csv`, `<prefix>_sample_carriers.csv`, `<prefix>_sample_overlap.csv` | `--sample-stats` | CSV | what the draw contains: per genome, per number of carriers, per pair of genomes | | `<prefix>_sample_genomes.csv`, `<prefix>_sample_carriers.csv`, `<prefix>_sample_overlap.csv` | `--sample-stats` | CSV | what the draw contains: per genome, per number of carriers, per pair of genomes |
| `<prefix>_sample_plan.csv` | `--sample-plan` | CSV table | per layer: eligible families, weight, quota, drawn, kept | | `<prefix>_sample_plan.csv` | `--sample-plan` | CSV table | per layer: eligible families, weight, quota, drawn, kept |
| `<prefix>_alignment.fasta` | `--pseudo-alignment` | FASTA | SNP-only pseudo-alignment, IUPAC-coded | | `<prefix>_alignment.fasta` | `--pseudo-alignment` or `--fasta` | FASTA | SNP-only pseudo-alignment, IUPAC-coded |
| `<prefix>_alignment.vcf` | `--vcf` | VCF | Same sites as the FASTA, with stable virtual coordinates and a k-mer-based `ID` |
### Sankoff calibration and exports ### Sankoff calibration and exports
+2 -2
View File
@@ -19,7 +19,7 @@ INDEX_PATH="${RUN_DIR}/specimen_index_count/${SPECIMEN}"
STATS_DIR="${RUN_DIR}/stats/indexing_count" STATS_DIR="${RUN_DIR}/stats/indexing_count"
STATS_FILE="${STATS_DIR}/${SPECIMEN}.stats" STATS_FILE="${STATS_DIR}/${SPECIMEN}.stats"
mkdir -p "${STATS_DIR}" mkdir -p "${STATS_DIR}" "$(dirname "${INDEX_PATH}")"
r1="${READS_DIR}/reads_R1.fastq.gz" r1="${READS_DIR}/reads_R1.fastq.gz"
r2="${READS_DIR}/reads_R2.fastq.gz" r2="${READS_DIR}/reads_R2.fastq.gz"
@@ -41,7 +41,7 @@ trap 'rm -f "${STDERR_LOG}"' EXIT
--label "${SPECIMEN}" \ --label "${SPECIMEN}" \
--meta "species=${species}" \ --meta "species=${species}" \
"${r1}" "${r2}" \ "${r1}" "${r2}" \
2>"${STDERR_LOG}" 2>"${STDERR_LOG}" || { cat "${STDERR_LOG}" >&2; exit 1; }
cat "${STDERR_LOG}" >&2 cat "${STDERR_LOG}" >&2
+2 -2
View File
@@ -19,7 +19,7 @@ INDEX_PATH="${RUN_DIR}/specimen_index_presence/${SPECIMEN}"
STATS_DIR="${RUN_DIR}/stats/indexing_presence" STATS_DIR="${RUN_DIR}/stats/indexing_presence"
STATS_FILE="${STATS_DIR}/${SPECIMEN}.stats" STATS_FILE="${STATS_DIR}/${SPECIMEN}.stats"
mkdir -p "${STATS_DIR}" mkdir -p "${STATS_DIR}" "$(dirname "${INDEX_PATH}")"
r1="${READS_DIR}/reads_R1.fastq.gz" r1="${READS_DIR}/reads_R1.fastq.gz"
r2="${READS_DIR}/reads_R2.fastq.gz" r2="${READS_DIR}/reads_R2.fastq.gz"
@@ -40,7 +40,7 @@ trap 'rm -f "${STDERR_LOG}"' EXIT
--label "${SPECIMEN}" \ --label "${SPECIMEN}" \
--meta "species=${species}" \ --meta "species=${species}" \
"${r1}" "${r2}" \ "${r1}" "${r2}" \
2>"${STDERR_LOG}" 2>"${STDERR_LOG}" || { cat "${STDERR_LOG}" >&2; exit 1; }
cat "${STDERR_LOG}" >&2 cat "${STDERR_LOG}" >&2
+1
View File
@@ -64,6 +64,7 @@ nav:
- Predicates and taxonomy paths: usage/predicates.md - Predicates and taxonomy paths: usage/predicates.md
- Formats: - Formats:
- Index construction and on-disk layout: formats/index_layout.md - Index construction and on-disk layout: formats/index_layout.md
- VCF pseudo-alignment: formats/vcf.md
- Architecture notes: architecture.md - Architecture notes: architecture.md
watch: watch:
+2
View File
@@ -56,6 +56,8 @@ nav:
- PersistentBitVec: implementation/persistent_bit_vec.md - PersistentBitVec: implementation/persistent_bit_vec.md
- Merge command: implementation/merge.md - Merge command: implementation/merge.md
- Merge parallelism & memory: implementation/merge_parallelism.md - Merge parallelism & memory: implementation/merge_parallelism.md
- "Merge mmap audit (2026-11)": implementation/merge_mmap_audit.md
- "Merge scaling (open)": implementation/merge_scaling.md
- Kmer filtering: implementation/filtering.md - Kmer filtering: implementation/filtering.md
- Select command: implementation/select.md - Select command: implementation/select.md
- obitaxonomy crate: implementation/obitaxonomy.md - obitaxonomy crate: implementation/obitaxonomy.md
+150
View File
@@ -0,0 +1,150 @@
# Prompt — refonte mmap de `obikmerge` (plans A puis C)
> À coller tel quel au début d'une nouvelle session Claude Code ouverte à la racine du dépôt obikmer.
> Conçu pour être suivi par **Sonnet** : toutes les décisions d'architecture sont fixées ci-dessous. Le plan B de l'audit (ouverture « slots only » dans `obikindex`) est **hors périmètre** de cette session (voir étape 5).
---
## Contexte
Lis d'abord `DevDocMD/implementation/merge_mmap_audit.md`. C'est le diagnostic complet, et ce prompt en implémente le plan.
En résumé : `obikmerge::partition_merge::merge_partition` ouvre, pour **chaque** partition, **toutes** les couches de **toutes** les sources (4 mmaps chacune). Elle crée aussi **toutes** les colonnes destination d'avance, en fichiers creux mappés en écriture (`set_len` + `MmapMut`). Avec 186 sources, ça fait environ 1 100 mappings par partition, ce qui épuise `vm.max_map_count` en production (pas d'accès root pour le relever). L'écriture via `MmapMut` dans des fichiers creux transforme aussi tout refus d'allocation disque en SIGBUS.
Objectif : le nombre de mappings vivants par partition ne doit plus dépendre du nombre de sources, et l'écriture des colonnes du merge ne doit plus passer par `MmapMut`. **Le résultat sur disque doit rester strictement identique.**
## Règles (non négociables)
- Appelle `mcp__serena__initial_instructions` avant tout travail de code. Utilise Serena (`find_symbol`, `get_symbols_overview`, `find_referencing_symbols`, `replace_symbol_body`, `replace_content`) pour naviguer et éditer le Rust.
- **Aucune commande VCS mutante** (git/jj : commit, add, mv, rm, describe, push…). Lecture seule autorisée.
- **Aucune nouvelle dépendance**, ni changement de version.
- `obikindex` fait autorité : cette session n'y modifie et n'y ajoute rien.
- Pas de fonctions libres ad hoc dans `obikmerge` pour combler un manque d'API. Si une opération générique manque sur un type de `obicompactvec`, ajoute-la comme méthode sur ce type.
- Fichiers source sous ~500 lignes. `partition_merge.rs` en fait déjà 605 : il **doit** être découpé (étape 6).
- **Ne jamais `.collect()` un flux de k-mers ou de unitigs** pour satisfaire une contrainte de trait.
- Échanges en **français**, documentation en **anglais**.
- **Point d'arrêt obligatoire** à la fin de chaque étape. Résume ce qui a été fait et les résultats de tests (sortie réelle), puis **attends mon feu vert** avant l'étape suivante.
- Si un test ou l'oracle échoue et que tu ne trouves pas la cause en une tentative, **arrête-toi et expose le problème**, sans contournement.
## Étape 1 — Oracle de non-régression (avant toute modification de code)
### 1a. Oracle synthétique (rapide, couvre comptage et merge en deux temps)
1. Lis `scripts/smoke_test_index.sh` (modèle de style, options de `obikmer index`) et les `--help` de `obikmer index`, `merge` et `dump`.
2. Écris `scripts/smoke_test_merge.sh`, sur le même modèle :
- génère 4 FASTA aléatoires déterministes A, B, C, D, petits, avec **des fragments partagés** entre eux. Le merge doit à la fois retrouver des k-mers déjà présents dans la destination (chemin « couche existante ») et en créer de nouveaux (chemin « nouvelle couche ») ;
- indexe chacun deux fois : présence (défaut) et comptage (`--with-counts`). Plusieurs partitions (`-p 4`) ;
- scénario 1 : `merge A B C D` (présence), puis la même chose en comptage ;
- scénario 2 : `merge A B` → AB, puis `merge AB C D`. Ce scénario couvre une source multi-génome et une destination à plusieurs couches ;
- scénario 3 : comme le scénario 1, avec `--dense` ;
- pour chaque sortie : `obikmer dump` → `sort` → un fichier texte ;
- deux modes : `--record DIR` écrit les sorties triées dans `DIR`, `--check DIR` les régénère et fait un `diff` avec `DIR`. Affiche PASS/FAIL clairement, avec un code de sortie cohérent.
3. Lance `--record benchmark/run/merge_oracle/synthetic` **avec le code actuel, non modifié**. Vérifie que les fichiers ne sont pas vides et que les colonnes de génomes ont du sens.
4. Lance aussitôt `--check` sur le même code pour vérifier que le script est déterministe.
### 1b. Oracle sur données réelles : `benchmark/`
`benchmark/run/specimen_index_presence/` contient 20 index présence construits à partir de vrais génomes (bactéries + une levure, lectures simulées). Lis `benchmark/README.md` et `benchmark/merge_presence.sh` (le merge de référence y utilise `--force --force-presence`).
État constaté au 2026-11-17, à revérifier :
- `specimen_index_count/` est **vide** : pas de piste comptage sans relancer `gmake index_count` (GNU make ≥ 4.3 : `gmake` sur macOS). **Demande-moi** avant de le lancer, car ça prend du temps ;
- `reference_index/` (`.npz`) est **absent** : `verify_merge_*.py` est inutilisable, ne t'appuie pas dessus ;
- `global_index_presence/` a été produit par un binaire antérieur au code actuel (release du 2026-10-22) : **ce n'est pas la référence**.
Procédure :
1. Recompile le binaire release à partir du **code actuel non modifié** (`cargo build --release` dans `src/`). Ne touche pas aux artefacts existants de `benchmark/run/`.
2. Lance le merge des 20 index avec les mêmes options que `merge_presence.sh`, mais vers un répertoire **dédié**, par exemple `benchmark/run/merge_oracle/bench/merged`, jamais `benchmark/run/global_index_presence`. Garde le log stderr : le rapport de stages (wall et RSS de `merge_partitions`) sert de référence de performance.
3. `obikmer dump` du résultat → `sort` → compresse (`zstd`) + `sha256sum`, dans `benchmark/run/merge_oracle/bench/`. La sortie peut être volumineuse : compare d'abord les hash, et ne fais un `diff` que s'ils diffèrent.
4. Contrôle croisé informatif : fais la même chose (dump trié + hash) sur `benchmark/run/global_index_presence`. Si les hash diffèrent, **signale-le sans conclure** : l'écart peut venir d'évolutions légitimes du code depuis octobre.
5. Écris ces étapes dans un script `scripts/bench_merge_oracle.sh` avec deux modes : `--record DIR` et `--check DIR`. En mode `--check`, il affiche aussi le temps de `merge_partitions` avant/après.
Après chaque étape de code (3, 4, 6), les **deux** oracles doivent passer : `smoke_test_merge.sh --check` et `bench_merge_oracle.sh --check`. Si `merge_partitions` est plus de 20 % plus lent qu'avant sur le benchmark, signale-le au point d'arrêt.
**Point d'arrêt.**
## Étape 2 — `MatrixBuilder::n_cols` (petit prérequis, `obicompactvec`)
Ajoute `pub fn n_cols(&self) -> usize` à `obicompactvec::MatrixBuilder` (`src/obicompactvec/src/matrix_builder.rs`), en déléguant au builder Bit ou Int. Si `PersistentCompactIntMatrixBuilder` n'expose pas `n_cols()`, ajoute-le sur ce type. Ajoute un test unitaire. `cargo test -p obicompactvec`.
Pas de point d'arrêt séparé : enchaîne sur l'étape 3.
## Étape 3 — Plan A : passe 2 par groupes de sources (`obikmerge`)
Fichier : `src/obikmerge/src/partition_merge.rs`, fonction `merge_partition`.
### 3a. Passe 1 sans ouvrir les couches sources
Construis la liste des chemins `unitigs.bin` sans `open()`. Pour chaque source et chaque `l in 0..partition.n_layers()`, prends `KmerLayer::new(&partition, l).unitigs_path()` (ce chemin est valide sur la variante `Empty`). Garde le filtre `exists()`. Ne touche pas au reste de la passe 1 (`build_graph`, `materialize_layer`, ouverture de `new_mphf`).
### 3b. Nouvelle couche : matrice initialisée seule
Si `any_new`, crée la matrice de la nouvelle couche avec `MatrixBuilder::new(presence, n_new, data_dir)`, ajoute les `n_dst_genomes` colonnes absentes, puis **ferme**. Les colonnes sources seront ajoutées par `resume`, groupe par groupe.
### 3c. Boucle sur les groupes
- `const SRC_GROUP: usize = 2;` (aligné sur `max_open`), avec un commentaire qui explique pourquoi.
- Itère sur `sources.chunks(SRC_GROUP)`, **dans l'ordre**, en tenant un `col_base` = nombre de colonnes sources déjà ajoutées avant ce groupe.
- Pour chaque groupe :
1. ouvre les couches de ses sources pour cette partition (`partition.layer(l)?.open()` puis `Arc::new`), comme aujourd'hui mais **seulement pour ce groupe** ;
2. `group_n` = somme des `src_n` du groupe ;
3. pour chaque couche dst existante, puis pour la nouvelle couche si `any_new` : `MatrixBuilder::resume`, vérifie `n_cols() == n_dst_genomes + col_base`, puis `add_col()` × `group_n` ;
4. construis les `pass2_items` du groupe. Le `col_offset` est **local au groupe** (0..group_n) ;
5. lance le pipeline de passe 2 existant, inchangé dans sa logique (Flat → Transform → Sink), sur ces items et ces builders. Les vecteurs `exist_sink`/`new_sink` ne contiennent que les colonnes du groupe ;
6. vérifie l'erreur de passe 2, ferme toutes les `ColBuilder` puis tous les `MatrixBuilder` du groupe ;
7. **drop** les couches du groupe avant de passer au suivant ;
8. `col_base += group_n`.
- L'initialisation `init_presence_matrix` (premier merge en présence) reste **avant** la boucle, inchangée.
- `dst_layers` (ouverts une seule fois) et `new_mphf` restent partagés par tous les groupes.
Invariant à documenter dans un commentaire : les colonnes sont ajoutées dans l'ordre des sources, donc la colonne absolue `n_dst_genomes + col_base + offset_local + g` est la même qu'avant la refonte.
### 3d. Vérification
`cargo test -p obikmerge`, `cargo clippy -p obikmerge`, puis `scripts/smoke_test_merge.sh --check benchmark/run/merge_oracle/synthetic` et `scripts/bench_merge_oracle.sh --check benchmark/run/merge_oracle/bench` (recompile le release avant). Les deux **doivent** passer sans aucun diff.
**Point d'arrêt.**
## Étape 4 — Plan C : colonnes construites en mémoire (`obicompactvec`)
But : plus aucun `MmapMut` ni fichier creux pour les colonnes **remplies** par le merge.
1. Lis `PersistentBitVecBuilder` (`bitvec.rs` : `new`, `set`, `close`), `PersistentCompactIntVecBuilder` (`builder.rs` : `new`, `set`, `close`), `format.rs` (`finalize_pciv`, `HEADER_SIZE`) et `mmap_file.rs` (`building_path`, `publish_built`). Note **exactement** les octets écrits à la fermeture : en-tête, données, overflow, `.building` puis `rename`.
2. Ajoute deux types :
- `MemBitVecBuilder` (dans `bitvec.rs`, ou un nouveau `mem_builder.rs` si `bitvec.rs` dépasse 500 lignes) : `words: Vec<u64>`, `n`, `path` ; `new(n, path)`, `set(slot, bool)`, `close() -> io::Result<()>` ;
- `MemCompactIntVecBuilder` : `primary: Vec<u8>`, `overflow: HashMap<usize, u32>`, `n`, `path` ; `new`, `set(slot, u32)` (même sémantique sentinelle 255 que le builder persistant), `close()`.
`close()` écrit avec un `BufWriter` dans `building_path(path)`, produit le **même fichier, octet pour octet**, que le builder persistant, fait `flush` + `sync_all`, puis `publish_built(path)`. Réutilise les fonctions de format existantes plutôt que de dupliquer la mise en page de l'en-tête. S'il faut factoriser l'écriture de l'en-tête ou de l'overflow pour la partager, fais-le dans `format.rs` / `mmap_file.rs`.
3. Dans `matrix_builder.rs` : ajoute les variantes `ColBuilder::MemBit(MemBitVecBuilder)` et `ColBuilder::MemInt(MemCompactIntVecBuilder)` (`set_val` et `close` les gèrent), plus `MatrixBuilder::add_mem_col() -> io::Result<ColBuilder>`. Ce dernier incrémente `n_cols` exactement comme `add_col`. Si les builders de matrice sous-jacents n'exposent ni le chemin de la prochaine colonne ni l'incrément, ajoute-leur une méthode `add_mem_col` symétrique de `add_col`. Le nommage des colonnes reste privé à la crate.
4. Tests unitaires (`src/obicompactvec/src/tests/`) : pour n ∈ {0, 1, 63, 64, 65, 10 000} et des valeurs pseudo-aléatoires (incluant des comptes ≥ 255), comparer **octet par octet** le fichier produit par le builder mémoire et celui du builder persistant. Vérifier aussi qu'une matrice construite avec `add_mem_col` se relit via `PersistentBitMatrix::open` / `PersistentIntMatrix::open`.
5. Dans `merge_partition`, remplace `add_col()` par `add_mem_col()` pour les colonnes sources (couches existantes et nouvelle couche). `add_absent_col` reste tel quel : on n'écrit jamais dans son mapping, donc aucun risque de SIGBUS.
6. Ajoute un commentaire sur le coût mémoire : `SRC_GROUP × src_n × (L_dst + 1) × (n/8 ou n)` octets au plus par partition.
7. `cargo test -p obicompactvec -p obikmerge`, clippy, puis `smoke_test_merge.sh --check` et `bench_merge_oracle.sh --check` (release recompilé).
`obikselect` et `obikfilter` ne sont **pas** migrés dans cette session.
**Point d'arrêt.**
## Étape 5 — Plan B : hors périmètre
**Ne pas implémenter.** Le plan B (ouverture « slots only » des couches dans `obikindex`) a été tranché par le développeur : ce sera un chantier séparé (forme retenue : composition, voir la section « Plan B » de `merge_mmap_audit.md`). Après l'étape 3, son gain en mmaps pour le merge est négligeable (~8 → ~2 par partition). Il touche en revanche le chemin de requête d'`obikindex` et concerne trois crates.
Les couches sources de la passe 2 continuent donc d'être ouvertes par `partition.layer(l)?.open()`, mais seulement groupe par groupe (étape 3). Ne crée aucun type ni méthode « lookup only » dans `obikindex` ou `obikmerge`.
## Étape 6 — Découpage, doc, finition
1. Découpe `partition_merge.rs` en module `partition_merge/` (par exemple `mod.rs` pour l'orchestration, `pass1.rs`, `pass2.rs`, `tests.rs`), chaque fichier sous 500 lignes, sans changement de comportement.
2. Supprime tout code devenu mort : `find_referencing_symbols` sur ce que la refonte a remplacé. Signale tout ce qui est devenu injustifié plutôt que de le garder par inertie.
3. `cargo test --workspace`, `cargo clippy --workspace`, `cargo check --workspace --features numa` si la feature existe, `scripts/smoke_test_index.sh`, `scripts/smoke_test_merge.sh --check` et `scripts/bench_merge_oracle.sh --check` (release recompilé ; reporte le temps de `merge_partitions` avant/après).
4. Documentation :
- `DevDocMD/implementation/merge_mmap_audit.md` : passe le statut à « implemented », en indiquant ce qui a été fait et les écarts éventuels au plan ;
- `DevDocMD/implementation/merge.md`, section « Algorithm » : décris le traitement par groupes de sources et les colonnes en mémoire ;
- vérifie si `UserDocMD/` décrit le comportement mémoire du merge, et le cas échéant mets-le à jour, en prose, sans code Rust ;
- mets à jour la `nav:` des `mkdocs*.yml` si un fichier a été ajouté, puis lance `qmd embed`.
5. Rapport final : liste des fichiers modifiés, résultats de tests, et estimation du nombre de mmaps par partition avant/après (calculée, marquée comme estimée). Termine par la commande de mesure à lancer en production, qui ne demande pas root :
```bash
PID=$(pgrep -f 'obikmer merge'); while sleep 60; do
echo "$(date +%T) maps=$(wc -l < /proc/$PID/maps) threads=$(ls /proc/$PID/task | wc -l)"; done
```
Ne commite rien : c'est moi qui le ferai.
+61
View File
@@ -0,0 +1,61 @@
# Prompt — où optimiser l'étape `pack` du merge (diagnostic d'abord)
> À coller tel quel au début d'une nouvelle session Claude Code ouverte à la racine du dépôt obikmer.
> Cette session est un **diagnostic** : elle produit une analyse chiffrée et des pistes classées, pas du code. Toute modification du code attend mon feu vert explicite, piste par piste.
---
## Contexte
`obikmer merge` se termine par une étape `finalize`, dont l'essentiel est `pack` : les matrices de présence/comptage, écrites en colonnes pendant le merge (`MatrixBuilder`), sont transposées en un format compact par partition (sparse par défaut). Le point d'entrée est `KmerIndex::finalize_indexed` (`src/obikindex/src/index/builder.rs`, étape « pack » vers la ligne 109), qui appelle `KmerIndex::pack_matrices(sparse)` (`src/obikindex/src/index/kmer_index.rs`). Les primitives sont dans `src/obicompactvec/src/` : `pack_bit_matrix` (`bitmatrix/packed.rs`), `pack_compact_int_matrix` (`intmatrix.rs`), `pack_sparse_compact_int_matrix` (`sparse_intmatrix.rs`), et le côté sparse des bits dans `bitmatrix/`.
Lis d'abord `DevDocMD/implementation/merge_scaling.md` (état du pack sparse : passage par une matrice dense séquentielle puis sparse, un seul mapping par partition) et `DevDocMD/implementation/merge_mmap_audit.md`.
## Mesure de référence (production, 2026-11-18)
Merge de 186 sources (`Cedrus`), 256 partitions, binaire glibc compilé localement, **192 CPU non contraints**, 17,5 milliards de k-mers nouveaux au total. Mode **présence**, compaction par défaut, donc **sparse** (pas de `--dense`) :
| Étape | Durée | Efficacité CPU | RSS max |
|---|---|---|---|
| `merge_partitions` | 748 s | 141×/192 (73 %) | 354 Go |
| `pack` | 231 s | 65×/192 (34 %) | **503,7 Go** |
La RAM de la machine est de 754 Go. Le RSS max est celui du process : il inclut les pages de fichiers mappés, pas seulement la mémoire anonyme. **Je n'ai pas encore séparé les deux.**
Progression de `pack` (partitions terminées par tranche de 10 %, début à 04:08:09) :
`+53 s` (10 %) · `+35 s` (20 %) · `+13 s` (30 %) · `+48 s` (40 %) · `+24 s` (50 %) · `+9 s` · `+6 s` · `+6 s` · `+6 s` · `+11 s` (100 %), puis `+20 s` avant la fin de l'étape. Les partitions les plus grosses passent en premier (tri décroissant dans `merge.rs`), mais je ne sais pas si `pack` les traite dans le même ordre ni si l'ordre est le bon.
## Ce que je veux savoir
1. **Où part le temps** de `pack` ? Lecture des colonnes, transposition, écriture, `sync`/`fsync`, attente d'un verrou, déséquilibre entre partitions ? Efficacité de 34 % sur 192 workers : qu'est-ce qui les bloque ?
2. **Où part la mémoire** (504 Go) ? Combien est de la mémoire anonyme, combien des pages de fichiers mappés ? Est-elle proportionnelle au nombre de workers ? Quelle est la mémoire vivante par partition en cours de `pack` ?
3. **Quel est le parallélisme réel** : combien de partitions `pack` traite-t-il en même temps, et qu'est-ce qui le fixe (`PartitionRunner`, `--cpu-max`, NUMA) ?
4. Y a-t-il du **travail inutile** (relectures, copies intermédiaires, colonnes vides ou toutes-absentes ré-écrites, étape dense qui pourrait être évitée) ?
## Méthode (impérative)
- **Lire et mesurer avant de proposer.** Pas de récit plausible sans preuve : tout ce qui n'est pas vérifié par lecture du code ou par mesure est marqué « hypothèse ».
- Pour mesurer sans la production : `scripts/docker_debug/` fournit un conteneur Linux (build glibc ou musl) qui lance un merge de N sources synthétiques ou dérivées de génomes réels (`run.sh`, variables `GEN`, `REAL`, `GENOME_LEN`, `TARGET`, `TOOL=heaptrack` pour glibc) avec un moniteur CSV de mappings et de RSS. Les génomes de 2 Mb donnent des partitions assez grosses pour que `pack` ait quelque chose à faire. `heaptrack` fonctionne sur le build glibc (pas sur le build musl statique). `gdb` ne fonctionne pas sous Rosetta.
- Pour séparer mémoire anonyme et fichiers mappés, utilise `/proc/PID/status` (`RssAnon`, `RssFile`, `RssShmem`), **jamais `smaps` en entier** sur un process à beaucoup de mappings (il verrouille `mmap_lock` et ralentit le process).
- Les estimations de coût (octets, secondes) sont marquées **estimées** ; les chiffres observés sont marqués **mesurés**.
## Règles du projet (non négociables)
- Avant tout travail de code : `mcp__serena__initial_instructions`, puis Serena (`find_symbol`, `get_symbols_overview`, `find_referencing_symbols`) pour naviguer. `rust-analyzer-mcp` et `qmd` quand ils servent. Consulte `context7` avant d'utiliser une API de bibliothèque externe.
- **Aucune commande VCS mutante** (git/jj : commit, add, mv, rm, describe, push…). Lecture seule autorisée.
- **Aucune nouvelle dépendance, aucun changement de version** sans mon accord explicite. Si un problème de bibliothèque se pose, expose-le et propose des alternatives.
- **Une question appelle une réponse, pas une action.** Ne modifie aucun fichier de `src/` sans demande explicite. Cette session est en lecture seule sur le code.
- Le code existant est une hypothèse, pas une vérité : signale tout ce qui est devenu injustifié plutôt que de le garder par inertie.
- Si tu proposes d'ajouter une opération à une structure partagée, mets-la comme méthode générique sur cette structure (crate `obicompactvec`), pas comme fonction libre dans le consommateur.
- Ne jamais matérialiser en mémoire une structure volumineuse qui peut être streamée.
- Échanges en **français**, documentation en **anglais**.
## Livrable
Un rapport (dans la conversation, pas dans un fichier, sauf si je le demande) avec :
1. Le chemin de code de `pack`, de `finalize_indexed` jusqu'aux écritures disque, résumé en quelques lignes par étape.
2. Pour chaque question ci-dessus : ce qui est **mesuré**, ce qui est **hypothèse**, et comment vérifier l'hypothèse.
3. Une liste de **pistes d'optimisation classées** par gain attendu/coût/risque. Pour chaque piste : ce qu'elle changerait, ce qu'elle risque, et comment on verrait qu'elle marche. Aucune implémentation.
4. Les vérifications de non-régression à prévoir : le résultat sur disque doit rester **identique octet pour octet** (voir les oracles `scripts/smoke_test_merge.sh` et `scripts/bench_merge_oracle.sh`).
Termine par la question : « Quelle piste veux-tu que j'implémente ? » et attends ma réponse.
+156
View File
@@ -0,0 +1,156 @@
#!/usr/bin/env bash
# bench_merge_oracle.sh — golden-output + timing oracle for `obikmer merge`
# on the real-data benchmark (benchmark/run/specimen_index_presence, 20
# presence indexes built from real genomes).
#
# Companion of smoke_test_merge.sh for refactors of obikmerge (see
# DevDocMD/implementation/merge_mmap_audit.md). Same merge options as
# benchmark/merge_presence.sh (--force --force-presence), but the output goes
# to a dedicated directory, never to benchmark/run/global_index_presence.
#
# What it does:
# 1. builds the release `obikmer`
# 2. merges every specimen presence index (sorted, as merge_presence.sh)
# and keeps the stderr log (stage report: wall/RSS of merge_partitions)
# 3. `obikmer dump --debug` of the result → header first, rows sorted
# (LC_ALL=C) → sha256 of that sorted text, plus a zstd copy
# 4. --record DIR: writes DIR/merged (the merged index), DIR/merge.stderr.log, DIR/dump.sha256,
# DIR/dump.sorted.csv.zst, and (informative cross-check)
# DIR/global_index_presence.sha256 = same dump hash computed on the
# existing benchmark/run/global_index_presence (older binary — a
# difference there is reported, not judged)
# --check DIR: merges into a scratch directory, compares the hash with
# DIR/dump.sha256 (full diff against the zstd copy only on mismatch),
# and prints merge_partitions wall/RSS and pack wall before → after
#
# Usage:
# scripts/bench_merge_oracle.sh (--record DIR | --check DIR) [--keep]
set -euo pipefail
KEEP=0
MODE=""
DIR=""
while [ $# -gt 0 ]; do
case "$1" in
--record) MODE=record; DIR="$2"; shift 2 ;;
--check) MODE=check; DIR="$2"; shift 2 ;;
--keep) KEEP=1; shift ;;
*) echo "unknown argument: $1" >&2; exit 2 ;;
esac
done
[ -n "$MODE" ] || { echo "usage: $0 (--record DIR | --check DIR) [--keep]" >&2; exit 2; }
REPO_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
case "$DIR" in /*) ;; *) DIR="$PWD/$DIR" ;; esac
RUN_DIR="$REPO_ROOT/benchmark/run"
IDX_DIR="$RUN_DIR/specimen_index_presence"
GLOBAL_OLD="$RUN_DIR/global_index_presence"
BIN="$REPO_ROOT/src/target/release/obikmer"
mkdir -p "$DIR" "$RUN_DIR/merge_oracle"
# Scratch space on the same file system as the benchmark data (merge's
# bootstrap hard-links from its base source; sort needs room for temp files).
WORK="$(mktemp -d "$RUN_DIR/merge_oracle/.work.XXXXXX")"
cleanup() {
if [ "$KEEP" -eq 1 ]; then echo "kept: $WORK"; else rm -rf "$WORK"; fi
}
trap cleanup EXIT
fail() {
echo "FAIL: $1" >&2
exit 1
}
# ── 1. build ────────────────────────────────────────────────────────────────
if ! (cd "$REPO_ROOT/src" && cargo build --release -q -p obikmer --bin obikmer) > "$WORK/build.log" 2>&1; then
cat "$WORK/build.log" >&2
fail "cargo build --release failed"
fi
sorted_dump() { # sorted_dump INDEX OUT_TXT — header first, rows sorted
local raw="$WORK/dump.raw.csv"
"$BIN" dump --debug "$1" > "$raw" 2> "$WORK/dump.log" \
|| { cat "$WORK/dump.log" >&2; fail "obikmer dump $1 exited non-zero"; }
{ head -n 1 "$raw"
tail -n +2 "$raw" | LC_ALL=C sort -S 25% --parallel=8 -T "$WORK"
} > "$2"
rm -f "$raw"
[ "$(head -n 2 "$2" | wc -l)" -eq 2 ] || fail "empty dump of $1"
}
hash_of() { sha256sum "$1" | cut -d' ' -f1; }
stage_field() { # stage_field LOG FIELD [STAGE] — FIELD of the 'done stage=STAGE' line (default merge_partitions)
perl -pe 's/\e\[[0-9;]*m//g' "$1" \
| grep "done stage=${3:-merge_partitions} " \
| sed -n "s/.*$2=\([^ ]*\( [KMG]*B\)\{0,1\}\).*/\1/p" | tail -n 1
}
# ── 2. merge ────────────────────────────────────────────────────────────────
mapfile -t SOURCES < <(find "$IDX_DIR" -mindepth 1 -maxdepth 1 -type d | sort)
[ "${#SOURCES[@]}" -gt 0 ] || fail "no indexes found in $IDX_DIR"
if [ "$MODE" = record ]; then
MERGED="$DIR/merged"
LOG="$DIR/merge.stderr.log"
else
MERGED="$WORK/merged"
LOG="$WORK/merge.stderr.log"
fi
echo "merging ${#SOURCES[@]} presence indexes → $MERGED"
if ! "$BIN" merge --output "$MERGED" --force --force-presence "${SOURCES[@]}" \
> /dev/null 2> "$LOG"; then
tail -n 40 "$LOG" >&2
fail "obikmer merge exited non-zero"
fi
WALL="$(stage_field "$LOG" wall_secs)"
RSS="$(stage_field "$LOG" rss)"
PACK_WALL="$(stage_field "$LOG" wall_secs pack)"
echo "merge_partitions: wall=${WALL}s rss=${RSS} pack: wall=${PACK_WALL}s"
# ── 3. sorted dump + hash ───────────────────────────────────────────────────
DUMP="$WORK/dump.sorted.csv"
sorted_dump "$MERGED" "$DUMP"
H="$(hash_of "$DUMP")"
echo "dump: $(($(wc -l < "$DUMP") - 1)) k-mers, sha256 $H"
# ── 4. record or check ──────────────────────────────────────────────────────
if [ "$MODE" = record ]; then
echo "$H" > "$DIR/dump.sha256"
zstd -q -T0 -f "$DUMP" -o "$DIR/dump.sorted.csv.zst"
if [ -d "$GLOBAL_OLD" ]; then
OLD_DUMP="$WORK/old.sorted.csv"
sorted_dump "$GLOBAL_OLD" "$OLD_DUMP"
HO="$(hash_of "$OLD_DUMP")"
echo "$HO" > "$DIR/global_index_presence.sha256"
if [ "$HO" = "$H" ]; then
echo "cross-check: global_index_presence (older binary) dump is identical"
else
echo "cross-check: global_index_presence (older binary) dump DIFFERS" \
"($(($(wc -l < "$OLD_DUMP") - 1)) k-mers, sha256 $HO) — informative only"
fi
fi
echo "RECORDED → $DIR"
exit 0
fi
[ -f "$DIR/dump.sha256" ] || fail "no reference in $DIR (run --record first)"
REF_H="$(cat "$DIR/dump.sha256")"
REF_WALL="$(stage_field "$DIR/merge.stderr.log" wall_secs)"
REF_RSS="$(stage_field "$DIR/merge.stderr.log" rss)"
REF_PACK="$(stage_field "$DIR/merge.stderr.log" wall_secs pack)"
echo "merge_partitions wall: ${REF_WALL}s → ${WALL}s rss: ${REF_RSS} → ${RSS}"
echo "pack wall: ${REF_PACK}s → ${PACK_WALL}s"
python3 -c "import sys; r,n=float(sys.argv[1]),float(sys.argv[2]); print(f'merge_partitions wall ratio: {n/r:.2f}' + (' (> 1.20: SLOWER THAN 20 %)' if n > 1.2*r else ''))" \
"$REF_WALL" "$WALL" || true
if [ "$H" = "$REF_H" ]; then
echo "PASS: dump identical to reference ($H)"
exit 0
fi
echo "FAIL: dump hash differs (ref $REF_H, now $H)" >&2
zstd -q -d -c "$DIR/dump.sorted.csv.zst" | { diff - "$DUMP" || true; } | head -n 40 >&2
exit 1
+16
View File
@@ -0,0 +1,16 @@
# Linux debug environment for the merge mapping/memory-growth hunt.
# Same toolchain as .gitea/workflows/release.yml for the musl target
# (cargo-zigbuild + zig cc), plus a glibc toolchain and diagnostic tools.
FROM rust:1-bookworm
RUN apt-get update && apt-get install -y --no-install-recommends \
build-essential cmake pkg-config autoconf automake libtool \
heaptrack valgrind gdb strace procps python3 python3-pip \
&& rm -rf /var/lib/apt/lists/*
RUN pip install ziglang --quiet --break-system-packages \
&& cargo install cargo-zigbuild \
&& rustup target add x86_64-unknown-linux-musl
RUN ZIG=$(python3 -c "import ziglang, os; print(os.path.join(os.path.dirname(ziglang.__file__), 'zig'))") \
&& printf '#!/bin/sh\nexec "%s" cc -target x86_64-linux-musl "$@"\n' "$ZIG" > /usr/local/bin/x86_64-linux-musl-gcc \
&& printf '#!/bin/sh\nexec "%s" c++ -target x86_64-linux-musl "$@"\n' "$ZIG" > /usr/local/bin/x86_64-linux-musl-g++ \
&& chmod +x /usr/local/bin/x86_64-linux-musl-gcc /usr/local/bin/x86_64-linux-musl-g++
WORKDIR /work
+6
View File
@@ -0,0 +1,6 @@
#!/usr/bin/env bash
# Builds the debug image. PLATFORM defaults to linux/amd64 (Rosetta on Apple Silicon).
set -euo pipefail
cd "$(dirname "${BASH_SOURCE[0]}")"
PLATFORM="${PLATFORM:-linux/amd64}"
docker build --platform "$PLATFORM" -t "obikmer-debug:${PLATFORM//\//-}" .
+82
View File
@@ -0,0 +1,82 @@
#!/usr/bin/env bash
# Index compatibility test between two obikmer binaries that differ only by
# their ptr_hash version (A = baseline, B = candidate).
#
# Usage: compat_ptrhash.sh VOL_A VOL_B OUT
# VOL_A / VOL_B docker volumes holding a musl release build (CARGO_TARGET_DIR)
# OUT host directory receiving report.txt
#
# What it checks (everything compared by md5 of dump/query output):
# T0 on-disk bytes of an index built by A vs by B (informational)
# T1 A-built single-genome indexes read by A and by B
# T2 merges: A merging A-built sources (reference), B merging A-built
# sources, B merging B-built sources; each result dumped by A and by B
# T3 queries (hits and misses) on indexes read by A and by B
# T4 incremental merge into an existing A-built merged index, by A and B
set -u
VOL_A="$1"; VOL_B="$2"; OUT="$3"
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
REPO="$(cd "$HERE/../.." && pwd)"
mkdir -p "$OUT"
docker run --rm --platform linux/amd64 \
-v "$VOL_A:/va:ro" -v "$VOL_B:/vb:ro" \
-v "$REPO/benchmark/run/genomes:/genomes_ro:ro" -v "$OUT:/out" \
obikmer-debug:linux-amd64 bash -c '
set -u
R=x86_64-unknown-linux-musl/release/obikmer
mkdir -p /tmp/bin && cp /va/$R /tmp/bin/A && cp /vb/$R /tmp/bin/B
A=/tmp/bin/A; B=/tmp/bin/B
W=/tmp/w; rm -rf $W; mkdir -p $W/ia $W/ib
exec > >(tee /out/report.txt) 2>&1
echo "A sha: $(sha256sum $A | cut -c1-16) B sha: $(sha256sum $B | cut -c1-16)"
mapfile -t G < <(ls -S /genomes_ro/*.fna.gz | sed -n 8,12p)
echo "genomes: ${#G[@]}"; printf " %s\n" "${G[@]##*/}"
mkdir -p $W/fa; n=0; for gz in "${G[@]}"; do n=$((n+1)); g=$W/fa/L$n.fasta; zcat "$gz" > "$g"
$A index -k 21 -m 9 --theta 0 -p 256 --label L$n -o $W/ia/L$n.idx "$g" 2>$W/ia/L$n.err >/dev/null || { echo "A index failed L$n"; tail -2 $W/ia/L$n.err; }
$B index -k 21 -m 9 --theta 0 -p 256 --label L$n -o $W/ib/L$n.idx "$g" 2>$W/ib/L$n.err >/dev/null || { echo "B index failed L$n"; tail -2 $W/ib/L$n.err; }
done
echo; echo "== T0 on-disk bytes, index built by A vs by B"
nd=$(diff -rq $W/ia $W/ib | wc -l); nt=$(find $W/ia -type f | wc -l)
echo "files differing: $nd / $nt"
diff -rq $W/ia $W/ib | sed "s#$W/##g" | awk "{print \$2}" | sed -E "s#/partitions/.*/#/…/#" | sort | uniq -c | sort -rn | head -8
h() { md5sum | cut -c1-12; }
echo; echo "== T1 A-built single-genome indexes: dump by A vs by B"
for i in 1 2 3 4 5; do
a=$($A dump $W/ia/L$i.idx 2>/dev/null | h); b=$($B dump $W/ia/L$i.idx 2>/dev/null | h)
echo "L$i A:$a B:$b $([ "$a" = "$b" ] && echo OK || echo DIFF)"
done
cp_src() { rm -rf $2; mkdir -p $2; for i in $1; do cp -r $3/L$i.idx $2/; done; }
echo; echo "== T2 merges of 4 genomes (presence)"
cp_src "1 2 3 4" $W/s_aa $W/ia; cp_src "1 2 3 4" $W/s_ba $W/ia; cp_src "1 2 3 4" $W/s_bb $W/ib
$A merge --output $W/M_AA --force --force-presence $W/s_aa/L*.idx 2>$W/m_aa.err; echo "M_AA rc=$?"
$B merge --output $W/M_BA --force --force-presence $W/s_ba/L*.idx 2>$W/m_ba.err; echo "M_BA rc=$? (B merging A-built sources)"
$B merge --output $W/M_BB --force --force-presence $W/s_bb/L*.idx 2>$W/m_bb.err; echo "M_BB rc=$? (B merging B-built sources)"
for X in M_AA M_BA M_BB; do for T in A B; do
eval "BIN=\$$T"; $BIN dump $W/$X 2>/dev/null > $W/d_${X}_$T.csv
echo "dump $X by $T: $(wc -l < $W/d_${X}_$T.csv) lines raw:$(h < $W/d_${X}_$T.csv) sorted:$(LC_ALL=C sort -S 2G $W/d_${X}_$T.csv | h)"
done; done
echo; echo "== T3 queries"
head -c 60000 $W/fa/L1.fasta > $W/q.fa
python3 - <<PY >> $W/q.fa
import random
random.seed(1)
print(">rand")
print("".join(random.choice("ACGT") for _ in range(5000)))
PY
for X in ia/L1.idx M_AA M_BA M_BB; do for T in A B; do
eval "BIN=\$$T"; echo "query $X by $T: $($BIN query $W/$X $W/q.fa 2>/dev/null | h)"
done; done
echo; echo "== T4 incremental merge (existing A-built merged index + a 5th genome)"
for T in A B; do
eval "BIN=\$$T"; rm -rf $W/inc_$T $W/dst_$T; cp -r $W/M_AA $W/dst_$T; cp -r $W/ia/L5.idx $W/l5_$T.idx
$BIN merge --output $W/inc_$T --force --force-presence $W/dst_$T $W/l5_$T.idx 2>$W/inc_$T.err; echo "inc by $T rc=$?"
$BIN dump $W/inc_$T 2>/dev/null | h
done
echo; echo done
'
+39
View File
@@ -0,0 +1,39 @@
#!/usr/bin/env bash
# Follow-up to compat_ptrhash.sh: determinism control (A built twice) and the
# incremental merge with its stderr shown. Usage: compat_ptrhash2.sh VOL_A VOL_B OUT
set -u
VOL_A="$1"; VOL_B="$2"; OUT="$3"
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"; REPO="$(cd "$HERE/../.." && pwd)"
mkdir -p "$OUT"
docker run --rm --platform linux/amd64 -v "$VOL_A:/va:ro" -v "$VOL_B:/vb:ro" \
-v "$REPO/benchmark/run/genomes:/genomes_ro:ro" -v "$OUT:/out" obikmer-debug:linux-amd64 bash -c '
set -u
R=x86_64-unknown-linux-musl/release/obikmer
mkdir -p /tmp/bin && cp /va/$R /tmp/bin/A && cp /vb/$R /tmp/bin/B; A=/tmp/bin/A; B=/tmp/bin/B
W=/tmp/w; rm -rf $W; mkdir -p $W/i1 $W/i2 $W/ib $W/fa
exec > >(tee /out/report2.txt) 2>&1
mapfile -t G < <(ls -S /genomes_ro/*.fna.gz | sed -n 8,12p)
n=0; for gz in "${G[@]}"; do n=$((n+1)); zcat "$gz" > $W/fa/L$n.fasta; done
h() { md5sum | cut -c1-12; }
echo "== C1 determinism control: same binary A, same input, built twice"
$A index -k 21 -m 9 --theta 0 -p 256 --label L1 -o $W/i1/L1.idx $W/fa/L1.fasta >/dev/null 2>&1
$A index -k 21 -m 9 --theta 0 -p 256 --label L1 -o $W/i2/L1.idx $W/fa/L1.fasta >/dev/null 2>&1
$B index -k 21 -m 9 --theta 0 -p 256 --label L1 -o $W/ib/L1.idx $W/fa/L1.fasta >/dev/null 2>&1
echo "A vs A: $(diff -rq $W/i1 $W/i2 | wc -l) / $(find $W/i1 -type f | wc -l) files differ"
echo "A vs B: $(diff -rq $W/i1 $W/ib | wc -l) / $(find $W/i1 -type f | wc -l) files differ"
for f in mphf.bin unitigs.bin evidence.bin unitigs.bin.idx; do
p=$(find $W/i1 -name $f | head -1); rel=${p#$W/i1/}
echo " $f AvsA: $(find $W/i1 -name $f | while read q; do cmp -s $q $W/i2/${q#$W/i1/} || echo x; done | wc -l)/256 differ AvsB: $(find $W/i1 -name $f | while read q; do cmp -s $q $W/ib/${q#$W/i1/} || echo x; done | wc -l)/256 differ"
done
echo; echo "== C2 incremental merge, stderr shown"
for i in 2 3 4 5; do $A index -k 21 -m 9 --theta 0 -p 256 --label L$i -o $W/i1/L$i.idx $W/fa/L$i.fasta >/dev/null 2>&1; done
mkdir -p $W/s; for i in 1 2 3; do cp -r $W/i1/L$i.idx $W/s/; done
$A merge --output $W/M3 --force --force-presence $W/s/L*.idx 2>/dev/null; echo "M3 rc=$?"
for T in A B; do
eval "BIN=\$$T"; rm -rf $W/inc_$T $W/dst_$T $W/l4_$T.idx; cp -r $W/M3 $W/dst_$T; cp -r $W/i1/L4.idx $W/l4_$T.idx
$BIN merge --output $W/inc_$T --force --force-presence $W/dst_$T $W/l4_$T.idx 2>$W/inc_$T.err; rc=$?
echo "inc by $T rc=$rc"; sed "s/\x1b\[[0-9;]*m//g" $W/inc_$T.err | tail -3
[ $rc = 0 ] && echo " dump sorted: $($BIN dump $W/inc_$T 2>/dev/null | LC_ALL=C sort -S 2G | h)"
done
echo done
'
+33
View File
@@ -0,0 +1,33 @@
#!/usr/bin/env python3
"""gen_fasta.py N OUTDIR [GENOME_LEN] — N deterministic synthetic genomes.
Each genome = unique random sequence + a random subset of a shared fragment
pool (some fragments lightly mutated), so a merge both hits k-mers already
in the destination and creates new ones, like related genomes would.
"""
import random, sys
n, out = int(sys.argv[1]), sys.argv[2]
glen = int(sys.argv[3]) if len(sys.argv) > 3 else 150_000
seed = int(sys.argv[4]) if len(sys.argv) > 4 else 20261118
rng = random.Random(seed)
def rnd(m):
return "".join(rng.choices("ACGT", k=m))
def mutate(s, rate=0.01):
s = list(s)
for i in range(len(s)):
if rng.random() < rate:
s[i] = rng.choice("ACGT")
return "".join(s)
pool = [rnd(2000) for _ in range(300)] # 600 kb shared pool
for g in range(n):
parts = [rnd(glen)]
for f in rng.sample(range(len(pool)), 60):
parts.append(mutate(pool[f]) if rng.random() < 0.3 else pool[f])
rng.shuffle(parts)
with open(f"{out}/G{g:03d}.fasta", "w") as fh:
for j, p in enumerate(parts):
fh.write(f">G{g:03d}_{j}\n{p}\n")
+39
View File
@@ -0,0 +1,39 @@
#!/usr/bin/env python3
"""gen_fasta_real.py N OUTDIR GENOMES_DIR [WINDOW] [SEED] [MUT]
N derived genomes built from *real* genomes (GENOMES_DIR/*.fna.gz): each one is
a random window of WINDOW bp of a randomly chosen real genome, point-mutated at
rate MUT (default 3 %, which alters ~47 % of 21-mers), so a merge both finds
k-mers already present in the destination and creates new ones, with realistic
sequence statistics (no dedicated degenerate structure).
"""
import glob, gzip, random, sys
n, out, gdir = int(sys.argv[1]), sys.argv[2], sys.argv[3]
window = int(sys.argv[4]) if len(sys.argv) > 4 else 300_000
seed = int(sys.argv[5]) if len(sys.argv) > 5 else 20261118
mut = float(sys.argv[6]) if len(sys.argv) > 6 else 0.03
rng = random.Random(seed)
def read_fna(path):
seq = []
with gzip.open(path, "rt") as fh:
for line in fh:
if not line.startswith(">"):
seq.append(line.strip().upper())
return "".join(seq)
genomes = [read_fna(p) for p in sorted(glob.glob(f"{gdir}/*.fna.gz"))]
genomes = [g for g in genomes if len(g) > window]
for i in range(n):
g = rng.choice(genomes)
start = rng.randrange(0, len(g) - window)
s = list(g[start:start + window])
for j in range(len(s)):
if s[j] in "ACGT" and rng.random() < mut:
s[j] = rng.choice("ACGT")
seq = "".join(c if c in "ACGT" else "A" for c in s)
with open(f"{out}/R{i:03d}.fasta", "w") as fh:
fh.write(f">R{i:03d}\n")
for k in range(0, len(seq), 80):
fh.write(seq[k:k + 80] + "\n")
+11
View File
@@ -0,0 +1,11 @@
#!/usr/bin/env bash
# monitor.sh PID OUT [INTERVAL_S] — samples mappings / VM / commit figures until PID exits.
PID=$1; OUT=$2; DT=${3:-5}
printf 'time,maps,vmsize_kb,vmdata_kb,vmrss_kb,threads,committed_as_kb,commit_limit_kb\n' >"$OUT"
while kill -0 "$PID" 2>/dev/null; do
maps=$(wc -l <"/proc/$PID/maps" 2>/dev/null) || break
st=$(grep -E '^(VmSize|VmData|VmRSS|Threads):' "/proc/$PID/status" | awk '{printf "%s,", $2}')
mi=$(grep -E '^(Committed_AS|CommitLimit):' /proc/meminfo | awk '{printf "%s,", $2}')
echo "$(date +%T),$maps,${st}${mi%,}" >>"$OUT"
sleep "$DT"
done
+91
View File
@@ -0,0 +1,91 @@
#!/usr/bin/env bash
# run.sh — builds obikmer (release, debug symbols) and merges the benchmark presence
# indexes in the container while sampling mappings/RSS/commit in parallel.
#
# Env:
# PLATFORM linux/amd64 (default, Rosetta) | linux/arm64
# GEN if set: build GEN synthetic indexes (gen_fasta.py) instead of using
# the benchmark ones; GENOME_LEN (default 150000), K (21), M (9), PART (256)
# REAL if set (with GEN): derive genomes from benchmark/run/genomes (gen_fasta_real.py)
# SRC_DIR alternative source tree (default: <repo>/src); VARIANT: volume suffix;
# LOCKED=0 to allow Cargo.lock changes (needed when a dep is added)
# NSRC number of source indexes to merge (default: all 20)
# TARGET gnu (default, glibc) | musl (static, same chain as the release CI)
# TOOL none (default) | heaptrack (heaptrack output goes to OUT/)
# MAXMAP if set, vm.max_map_count is lowered to this inside the VM for the
# run and restored afterwards (needs --privileged; affects the whole
# OrbStack VM while it runs)
# MALLOC_ARENA_MAX, MALLOC_MMAP_THRESHOLD_, MALLOC_TRIM_THRESHOLD_ passed through
# OUT host output directory (default: ./out/<timestamp>)
set -euo pipefail
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
REPO="$(cd "$HERE/../.." && pwd)"
PLATFORM="${PLATFORM:-linux/amd64}"
NSRC="${NSRC:-20}"
TOOL="${TOOL:-none}"
TARGET="${TARGET:-gnu}"
OUT="${OUT:-$HERE/out/$(date +%Y%m%d-%H%M%S)}"
mkdir -p "$OUT"
IMG="obikmer-debug:${PLATFORM//\//-}"
VOL="obikmer-debug-target-${PLATFORM//\//-}-${TARGET}${VARIANT:+-$VARIANT}"
docker run --rm --privileged --platform "$PLATFORM" \
-v "${SRC_DIR:-$REPO/src}:/work/src:$([ -n "${SRC_DIR:-}" ] && echo rw || echo ro)" \
-v "$REPO/benchmark/run/specimen_index_presence:/idx_ro:ro" \
-v "$VOL:/target" \
-v "$OUT:/out" \
-v "$HERE/monitor.sh:/monitor.sh:ro" \
-v "$HERE/gen_fasta.py:/gen_fasta.py:ro" \
-v "$HERE/gen_fasta_real.py:/gen_fasta_real.py:ro" \
-v "$REPO/benchmark/run/genomes:/genomes_ro:ro" \
-e CARGO_TARGET_DIR=/target -e CARGO_PROFILE_RELEASE_DEBUG=1 \
-e GEN="${GEN:-}" -e LOCKED_FLAG="$([ "${LOCKED:-1}" = 0 ] || echo --locked)" -e SEED="${SEED:-20261118}" -e REAL="${REAL:-}" -e GENOME_LEN="${GENOME_LEN:-150000}" -e K="${K:-21}" -e M="${M:-9}" -e PART="${PART:-256}" -e RUST_LOG="${RUST_LOG:-}" -e NSRC="$NSRC" -e TARGET="$TARGET" -e TOOL="$TOOL" -e MAXMAP="${MAXMAP:-}" \
-e MALLOC_ARENA_MAX="${MALLOC_ARENA_MAX:-}" \
-e MALLOC_MMAP_THRESHOLD_="${MALLOC_MMAP_THRESHOLD_:-}" \
-e MALLOC_TRIM_THRESHOLD_="${MALLOC_TRIM_THRESHOLD_:-}" \
"$IMG" bash -c '
set -euo pipefail
for v in MALLOC_ARENA_MAX MALLOC_MMAP_THRESHOLD_ MALLOC_TRIM_THRESHOLD_ RUST_LOG; do
[ -n "${!v}" ] || unset $v
done
if [ "$TARGET" = musl ]; then
PKG_CONFIG_ALLOW_CROSS=1 cargo zigbuild --release $LOCKED_FLAG --target x86_64-unknown-linux-musl \
--manifest-path /work/src/Cargo.toml --bin obikmer 2>&1 | tail -3
BIN=/target/x86_64-unknown-linux-musl/release/obikmer
else
cargo build --release $LOCKED_FLAG --manifest-path /work/src/Cargo.toml --bin obikmer 2>&1 | tail -3
BIN=/target/release/obikmer
fi
file "$BIN" 2>/dev/null | cut -c1-150 || true
# merge writes into its sources (lock files…): work on a copy, never on benchmark/
rm -rf /tmp/idx && mkdir -p /tmp/idx
if [ -n "$GEN" ]; then
mkdir -p /tmp/fa
if [ -n "$REAL" ]; then python3 /gen_fasta_real.py "$GEN" /tmp/fa /genomes_ro "$GENOME_LEN" "$SEED"
else python3 /gen_fasta.py "$GEN" /tmp/fa "$GENOME_LEN" "$SEED"; fi
NSRC=$GEN
echo "indexing $GEN synthetic genomes (k=$K m=$M p=$PART)…"
ls /tmp/fa/*.fasta | xargs -P 8 -I{} sh -c \
"b=\$(basename {} .fasta); $BIN index -k $K -m $M --theta 0 -p $PART --label \$b -o /tmp/idx/\$b.idx {} >/dev/null 2>&1 || echo index failed: \$b"
else
cp -r /idx_ro/. /tmp/idx/
fi
mapfile -t SRC < <(find /tmp/idx -mindepth 1 -maxdepth 1 -type d | sort | head -n "$NSRC")
echo "sources: ${#SRC[@]} target: $TARGET tool: $TOOL max_map_count: $(cat /proc/sys/vm/max_map_count)"
if [ -n "$MAXMAP" ]; then
OLD=$(cat /proc/sys/vm/max_map_count); sysctl -w vm.max_map_count="$MAXMAP" >/dev/null
trap "sysctl -w vm.max_map_count=$OLD >/dev/null" EXIT
fi
rm -rf /tmp/merged
CMD=("$BIN" merge --output /tmp/merged --force --force-presence "${SRC[@]}")
if [ "$TOOL" = heaptrack ]; then CMD=(heaptrack -o /out/heaptrack "${CMD[@]}"); fi
"${CMD[@]}" 2>/out/merge.stderr &
PID=$!
/monitor.sh "$PID" /out/monitor.csv 1 &
MON=$!
set +e; wait "$PID"; RC=$?; set -e
wait "$MON" || true
echo "merge exit code: $RC" | tee /out/exit_code.txt
tail -5 /out/merge.stderr
'
echo "outputs: $OUT"
+102
View File
@@ -0,0 +1,102 @@
#!/usr/bin/env bash
# monitor_pack.sh — 1 Hz sampler for I/O- and memory-bound stages (pack, merge_partitions).
#
# Usage:
# scripts/monitor_pack.sh --pid PID OUT_DIR [OUTPUT_INDEX_DIR] # attach to a running process
# scripts/monitor_pack.sh --cmd OUT_DIR [OUTPUT_INDEX_DIR] -- CMD... # start CMD and sample it
# env: DT=1 (seconds between samples)
#
# Writes OUT_DIR/samples.csv (one row per sample) and OUT_DIR/env.txt (context
# needed to read it). Linux only. Never reads smaps (it takes mmap_lock and
# slows a process with many mappings); only /proc/PID/{status,io,task/*/stat}
# and /proc/meminfo.
#
# Columns of samples.csv
# t epoch seconds
# rss_anon_kb, rss_file_kb, rss_shmem_kb from /proc/PID/status
# threads thread count
# n_R n_D n_S threads running / in uninterruptible sleep (I/O wait) / sleeping
# (futex, channel, sleep: NOT disk). n_D is the direct count of
# workers blocked on I/O.
# maps number of mappings (wc -l /proc/PID/maps)
# read_bytes write_bytes /proc/PID/io (cumulative, bytes submitted to the
# block layer; buffered writes counted at page dirtying)
# rchar wchar /proc/PID/io (cumulative, bytes through read()/write() syscalls)
# dirty_kb writeback_kb cached_kb mem_avail_kb /proc/meminfo
#
# Pair it with the process's own log (RUST_LOG=info, plus obisys=debug for the
# "activated N worker(s)" lines and obikindex=debug for per-partition timings)
# to align stages on wall-clock time: the log timestamps are UTC ISO-8601, `t`
# is epoch seconds.
set -euo pipefail
MODE=${1:-}; shift || true
case "$MODE" in
--pid) PID=${1:?PID}; shift ;;
--cmd) ;;
*) sed -n 2,12p "$0"; exit 2 ;;
esac
OUT=${1:?OUT_DIR}; shift
IDX=""
if [ "${1:-}" != "" ] && [ "${1:-}" != "--" ]; then IDX=$1; shift; fi
mkdir -p "$OUT"
if [ "$MODE" = "--cmd" ]; then
[ "${1:-}" = "--" ] || { echo "--cmd needs: -- CMD..." >&2; exit 2; }
shift
"$@" 2>"$OUT/cmd.stderr" &
PID=$!
fi
{
echo "date_utc: $(date -u +%FT%TZ)"
echo "pid: $PID"
echo "nproc: $(nproc)"
echo "kernel: $(uname -r)"
grep -E 'MemTotal|SwapTotal' /proc/meminfo
for k in vm.dirty_ratio vm.dirty_background_ratio vm.dirty_bytes vm.dirty_background_bytes \
vm.dirty_expire_centisecs vm.dirty_writeback_centisecs vm.max_map_count; do
echo "$k: $(sysctl -n $k 2>/dev/null || echo '?')"
done
echo "ulimit -n: $(ulimit -n)"
echo "cmdline: $(tr '\0' ' ' </proc/$PID/cmdline 2>/dev/null)"
if [ -n "$IDX" ]; then
echo "output_dir: $IDX"
echo "fs_type: $(stat -f -c %T "$IDX" 2>/dev/null || echo '?')"
echo "mount: $(df -P "$IDX" 2>/dev/null | tail -1)"
command -v beegfs-ctl >/dev/null && echo "beegfs: $(beegfs-ctl --getentryinfo "$IDX" 2>&1 | head -8 | tr '\n' ' ')"
fi
env | grep -E '^(MALLOC_|RUST_LOG|RAYON_)' || true
} >"$OUT/env.txt"
DT=${DT:-1}
printf 't,rss_anon_kb,rss_file_kb,rss_shmem_kb,threads,n_R,n_D,n_S,maps,read_bytes,write_bytes,rchar,wchar,dirty_kb,writeback_kb,cached_kb,mem_avail_kb\n' >"$OUT/samples.csv"
while [ -d "/proc/$PID" ]; do
t=$(date +%s)
st=$(awk '/^RssAnon:/{a=$2} /^RssFile:/{f=$2} /^RssShmem:/{s=$2} /^Threads:/{th=$2} END{printf "%s,%s,%s,%s",a,f,s,th}' "/proc/$PID/status" 2>/dev/null) || break
# thread states: field 3 of /proc/PID/task/TID/stat, after the ")" closing comm
states=$(cat /proc/$PID/task/*/stat 2>/dev/null | sed -E 's/^.*\) (.).*/\1/' | sort | uniq -c \
| awk '{c[$2]=$1} END{printf "%d,%d,%d", c["R"]+0, c["D"]+0, c["S"]+0}')
maps=$(wc -l <"/proc/$PID/maps" 2>/dev/null || echo 0)
io=$(awk '/^rchar:/{rc=$2} /^wchar:/{wc=$2} /^read_bytes:/{r=$2} /^write_bytes:/{w=$2} END{printf "%s,%s,%s,%s",r,w,rc,wc}' "/proc/$PID/io" 2>/dev/null || echo ",,,")
mi=$(awk '/^Dirty:/{d=$2} /^Writeback:/{w=$2} /^Cached:/{c=$2} /^MemAvailable:/{m=$2} END{printf "%s,%s,%s,%s",d,w,c,m}' /proc/meminfo)
echo "$t,$st,$states,$maps,$io,$mi" >>"$OUT/samples.csv"
sleep "$DT"
done
if [ "$MODE" = "--cmd" ]; then wait "$PID" || echo "command exit code: $?" >>"$OUT/env.txt"; fi
# ── summary ──────────────────────────────────────────────────────────────────
awk -F, 'NR==1{next}
{ n++; if(NR==2){t0=$1; r0=$10; w0=$11}
if($2>ma)ma=$2; if($3>mf)mf=$3; if($4>ms)ms=$4; if($5>mt)mt=$5
if($6>mR)mR=$6; if($7>mD)mD=$7; if($8>mS)mS=$8; sR+=$6; sD+=$7; sS+=$8
if($14>md)md=$14; if($15>mw)mw=$15; if($9>mm)mm=$9
t1=$1; r1=$10; w1=$11 }
END{ if(!n)exit
printf "samples: %d over %ds\n", n, t1-t0
printf "peak RssAnon %.1f GB | RssFile %.1f GB | RssShmem %.1f GB | threads %d | maps %d\n", ma/1e6, mf/1e6, ms/1e6, mt, mm
printf "threads by state (mean / max): R %.1f / %d | D(I/O wait) %.1f / %d | S %.1f / %d\n", sR/n, mR, sD/n, mD, sS/n, mS
printf "peak Dirty %.1f GB | Writeback %.1f GB\n", md/1e6, mw/1e6
printf "block I/O over the run: read %.1f GB, write %.1f GB\n", (r1-r0)/1e9, (w1-w0)/1e9 }' "$OUT/samples.csv" | tee "$OUT/summary.txt"
+23 -1
View File
@@ -125,4 +125,26 @@ if [ "$FOUND" -ne 1 ]; then
fail "no candidate k-mer round-tripped (tried ${#CANDIDATES[@]}) — likely a real regression, not a low-complexity fixture" fail "no candidate k-mer round-tripped (tried ${#CANDIDATES[@]}) — likely a real regression, not a low-complexity fixture"
fi fi
echo "PASS: index+query round-trip OK — $N_KMERS kmers indexed, query k-mer '$QUERY_KMER' found" # ── 5. regressions: (a) many partitions against a tiny input leaves most of
# them empty (no layer 0 on disk) — must still build; (b) an NCBI-style
# `.fna.gz` name must be read, not silently skipped (it used to be
# taken for a directory and ignored, yielding an empty index).
count_kmers() { # $1 = output index dir, rest = obikmer index args
local out="$1"; shift
local log="$WORK/regress.log"
( cd "$REPO_ROOT/src" && cargo run -q -p obikmer --bin obikmer -- \
index -k "$K" -m "$M" --theta 0 -o "$out" "$@" ) > "$log" 2>&1 \
|| { cat "$log" >&2; fail "obikmer index $* exited non-zero"; }
grep -o '[0-9]* total kmers indexed' "$log" | grep -o '^[0-9]*' || echo 0
}
N_EMPTY="$(count_kmers "$WORK/empty_parts.idx" -p 256 "$WORK/test.fasta")"
[ "$N_EMPTY" = "$N_KMERS" ] \
|| fail "-p 256 (mostly empty partitions) indexed $N_EMPTY kmers, expected $N_KMERS"
gzip -c "$WORK/test.fasta" > "$WORK/test.fna.gz"
N_GZ="$(count_kmers "$WORK/gz.idx" -p "$PARTITIONS" "$WORK/test.fna.gz")"
[ "$N_GZ" = "$N_KMERS" ] \
|| fail ".fna.gz input indexed $N_GZ kmers, expected $N_KMERS (input skipped?)"
echo "PASS: index+query round-trip OK — $N_KMERS kmers indexed, query k-mer '$QUERY_KMER' found; empty-partition and .fna.gz regressions OK"
+217
View File
@@ -0,0 +1,217 @@
#!/usr/bin/env bash
# smoke_test_merge.sh — end-to-end golden-output oracle for `obikmer merge`
#
# Non-regression oracle for refactors of obikmerge (see
# DevDocMD/implementation/merge_mmap_audit.md). Like smoke_test_index.sh it
# drives the real CLI, never the library APIs.
#
# What it does:
# 1. builds `obikmer` (debug)
# 2. generates (on --record) 4 small deterministic FASTA files A, B, C, D sharing
# fragments, so that merge both hits k-mers already present in the
# destination ("existing layer" path) and creates new ones ("new layer"
# path). D carries a short motif repeated > 255 times so the count track
# exercises the compact-int overflow (counts >= 255).
# 3. on --record, indexes each genome twice (presence, --with-counts) and
# freezes the indexes in DIR/sources; --check reuses them (see below),
# with -p PARTITIONS partitions
# 4. runs the merge scenarios:
# s1_presence / s1_count merge A B C D
# s3_dense merge A B C D --dense (presence)
# Scenario 2 (merge A B -> AB, then merge AB C D: a packed merge output
# re-merged as a *source*) is checked relatively, not against the
# golden dumps: its table (k-mer x genome label) must equal s1_presence's,
# whatever the column order and layer placement. A packed index cannot
# be a merge *destination* (its matrices are no longer column-by-column,
# see `can_be_destination` in obikmerge); `choose_base` therefore picks
# an unpacked source, and merging only packed indexes is refused with a
# clear error, which this script also checks.
# 5. dumps every result with `obikmer dump --debug` (partition + layer
# columns included, so a k-mer moving between layers is caught), keeps
# the CSV header first and sorts the rows (LC_ALL=C)
# 6. --record DIR writes the sorted dumps to DIR;
# --check DIR regenerates them and diffs against DIR
#
# Usage:
# scripts/smoke_test_merge.sh (--record DIR | --check DIR) [-p PARTITIONS] [--keep]
set -euo pipefail
K=21
M=9
PARTITIONS=4
KEEP=0
MODE=""
DIR=""
while [ $# -gt 0 ]; do
case "$1" in
--record) MODE=record; DIR="$2"; shift 2 ;;
--check) MODE=check; DIR="$2"; shift 2 ;;
-p) PARTITIONS="$2"; shift 2 ;;
--keep) KEEP=1; shift ;;
*) echo "unknown argument: $1" >&2; exit 2 ;;
esac
done
[ -n "$MODE" ] || { echo "usage: $0 (--record DIR | --check DIR) [-p N] [--keep]" >&2; exit 2; }
REPO_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
case "$DIR" in /*) ;; *) DIR="$PWD/$DIR" ;; esac
WORK="$(mktemp -d -t obikmer_merge_smoke.XXXXXX)"
cleanup() {
if [ "$KEEP" -eq 1 ]; then echo "kept: $WORK"; else rm -rf "$WORK"; fi
}
trap cleanup EXIT
fail() {
echo "FAIL: $1" >&2
exit 1
}
# ── 1. build ────────────────────────────────────────────────────────────────
if ! (cd "$REPO_ROOT/src" && cargo build -q -p obikmer --bin obikmer) > "$WORK/build.log" 2>&1; then
cat "$WORK/build.log" >&2
fail "cargo build failed"
fi
BIN="$REPO_ROOT/src/target/debug/obikmer"
run() { # run LOGNAME CMD... — runs obikmer, dumps its log on failure
local log="$WORK/$1.log"; shift
if ! "$BIN" "$@" > "$log" 2>&1; then
cat "$log" >&2
fail "obikmer $1 exited non-zero"
fi
}
# ── 2. deterministic genomes with shared fragments ──────────────────────────
python3 - "$WORK" <<'EOF'
import random, sys
out = sys.argv[1]
rng = random.Random(20261117)
def rnd(n): return "".join(rng.choice("ACGT") for _ in range(n))
frag = {i: rnd(600) for i in range(5)} # shared fragments F0..F4
motif = rnd(40) # tandem motif (counts >= 255)
genomes = {
"A": [rnd(1500), frag[0], rnd(300), frag[1]],
"B": [frag[1], rnd(1200), frag[2], motif * 3],
"C": [frag[0], rnd(800), frag[2], rnd(400), frag[3]],
"D": [frag[3], rnd(1000), motif * 300, frag[4], rnd(200), frag[0][:250]],
}
for name, parts in genomes.items():
with open(f"{out}/{name}.fasta", "w") as f:
for j, p in enumerate(parts):
f.write(f">{name}_{j}\n{p}\n")
EOF
# ── 3. source indexes: built on --record, frozen in DIR/sources, reused on
# --check. `obikmer index` is not byte-stable run to run (unitig
# segmentation, hence unitigs.bin size, varies — measured 2026-11-17),
# and the merge base is the source with the largest unitigs.bin
# (`choose_base`), so re-indexing on every check would make the merge
# base — and the column order — flip between A and C. Freezing the
# sources makes this an oracle of `merge` alone. Copied (not linked)
# into WORK, since merge's bootstrap hard-links from its base source.
if [ "$MODE" = record ]; then
for G in A B C D; do
run "index_${G}_pres" index -k "$K" -m "$M" --theta 0 -p "$PARTITIONS" \
--label "$G" -o "$WORK/$G.pres.idx" "$WORK/$G.fasta"
run "index_${G}_cnt" index -k "$K" -m "$M" --theta 0 -p "$PARTITIONS" \
--with-counts --label "$G" -o "$WORK/$G.cnt.idx" "$WORK/$G.fasta"
done
rm -rf "$DIR/sources"
mkdir -p "$DIR/sources"
cp -R "$WORK"/*.idx "$DIR/sources/"
else
[ -d "$DIR/sources" ] || fail "no frozen source indexes in $DIR/sources (run --record first)"
cp -R "$DIR/sources"/*.idx "$WORK/"
fi
# ── 4. merge scenarios ──────────────────────────────────────────────────────
OUT="$WORK/out"
mkdir -p "$OUT"
dump() { # dump NAME INDEX — header first, rows sorted
local raw="$WORK/$1.dump.csv"
if ! "$BIN" dump --debug "$2" > "$raw" 2> "$WORK/dump_$1.log"; then
cat "$WORK/dump_$1.log" >&2
fail "obikmer dump $1 exited non-zero"
fi
{ head -n 1 "$raw"; tail -n +2 "$raw" | LC_ALL=C sort; } > "$OUT/$1.csv"
}
for T in pres cnt; do
run "s1_$T" merge -o "$WORK/s1_$T.idx" \
"$WORK/A.$T.idx" "$WORK/B.$T.idx" "$WORK/C.$T.idx" "$WORK/D.$T.idx"
dump "s1_$T" "$WORK/s1_$T.idx"
done
run s3_dense merge --dense -o "$WORK/s3_dense.idx" \
"$WORK/A.pres.idx" "$WORK/B.pres.idx" "$WORK/C.pres.idx" "$WORK/D.pres.idx"
dump s3_dense "$WORK/s3_dense.idx"
# scenario 2: re-merge a packed merge output as a source (checked against s1)
run s2_ab merge -o "$WORK/s2_ab.idx" "$WORK/A.pres.idx" "$WORK/B.pres.idx"
run s2 merge -o "$WORK/s2.idx" "$WORK/s2_ab.idx" "$WORK/C.pres.idx" "$WORK/D.pres.idx"
for N in s1_pres s2; do
"$BIN" dump "$WORK/$N.idx" > "$WORK/plain_$N.csv" 2> "$WORK/dump_plain_$N.log" \
|| { cat "$WORK/dump_plain_$N.log" >&2; fail "obikmer dump $N exited non-zero"; }
done
cat > "$WORK/compare_tables.py" <<'PY'
import csv, sys
def table(path):
rows = list(csv.reader(open(path)))
head = rows[0]
return {r[0]: {head[i]: r[i] for i in range(1, len(head))} for r in rows[1:]}
a, b = table(sys.argv[1]), table(sys.argv[2])
if a != b:
print(f"k-mers: {len(a)} vs {len(b)}; only in flat: {len(set(a) - set(b))}; "
f"only in re-merge: {len(set(b) - set(a))}", file=sys.stderr)
sys.exit(1)
PY
python3 "$WORK/compare_tables.py" "$WORK/plain_s1_pres.csv" "$WORK/plain_s2.csv" \
|| fail "scenario 2 (re-merge of a packed output) differs from the flat merge"
echo "ok s2 (re-merge of a packed output) equals the flat merge"
if "$BIN" merge -o "$WORK/s2_bad.idx" "$WORK/s2_ab.idx" "$WORK/s1_pres.idx" > "$WORK/s2_bad.log" 2>&1; then
fail "merging only packed indexes should be refused"
fi
grep -q "none of the sources can be the destination" "$WORK/s2_bad.log" \
|| { cat "$WORK/s2_bad.log" >&2; fail "refusal of an all-packed merge lacks the expected message"; }
echo "ok merge of only packed indexes is refused with a clear message"
# sanity: no dump may be empty (header + at least one row)
for f in "$OUT"/*.csv; do
[ "$(wc -l < "$f")" -gt 1 ] || fail "empty dump: $(basename "$f")"
done
# ── 5. record or check ──────────────────────────────────────────────────────
if [ "$MODE" = record ]; then
mkdir -p "$DIR"
rm -f "$DIR"/*.csv
cp "$OUT"/*.csv "$DIR"/
echo "RECORDED: $(ls "$OUT" | wc -l | tr -d ' ') dumps → $DIR"
wc -l "$DIR"/*.csv
exit 0
fi
[ -d "$DIR" ] || fail "reference directory not found: $DIR"
STATUS=0
for ref in "$DIR"/*.csv; do
name="$(basename "$ref")"
if [ ! -f "$OUT/$name" ]; then
echo "FAIL $name (not regenerated)"; STATUS=1
elif diff -q "$ref" "$OUT/$name" > /dev/null; then
echo "ok $name"
else
echo "FAIL $name"; { diff "$ref" "$OUT/$name" || true; } | head -20; STATUS=1
fi
done
for new in "$OUT"/*.csv; do
[ -f "$DIR/$(basename "$new")" ] || { echo "FAIL $(basename "$new") (no reference)"; STATUS=1; }
done
if [ "$STATUS" -eq 0 ]; then
echo "PASS: all merge dumps identical to $DIR"
else
echo "FAIL: merge dumps differ from $DIR" >&2
fi
exit "$STATUS"
+53
View File
@@ -0,0 +1,53 @@
#!/usr/bin/env bash
# watch_merge.sh PID OUT.csv — samples a running process and kills it past a threshold.
#
# Every INTERVAL seconds appends one CSV row: number of memory mappings
# (lines of /proc/PID/maps), VmSize, VmData, VmRSS, threads, and the
# system-wide Committed_AS / CommitLimit. If the mapping count exceeds
# MAPS_MAX, VmRSS exceeds RSS_MAX_GB, or VmSize exceeds VMSIZE_MAX_GB (stay
# below `ulimit -v`: past it, allocations fail instead of the guard firing), the process receives SIGTERM (then
# SIGKILL after 30 s) and the reason is written to OUT.csv.reason — a guard
# so a runaway merge ends with a clean message instead of a failed
# allocation (or the OOM killer on a shared node).
#
# Reading /proc/PID/maps is itself slow when the process has hundreds of
# thousands of mappings (it held the kernel's mmap lock in earlier
# measurements): keep INTERVAL >= 5. It never reads /proc/PID/smaps.
#
# Usage: scripts/watch_merge.sh PID OUT.csv
# env: INTERVAL=5 MAPS_MAX=500000 RSS_MAX_GB=400 VMSIZE_MAX_GB=440
#
# Typical use (Linux only):
# nohup scripts/watch_merge.sh "$(pgrep -f 'obikmer merge' | head -1)" /tmp/watch.csv &
set -u
PID=${1:?usage: watch_merge.sh PID OUT.csv}
OUT=${2:?usage: watch_merge.sh PID OUT.csv}
INTERVAL=${INTERVAL:-5}
MAPS_MAX=${MAPS_MAX:-500000}
RSS_MAX_KB=$(( ${RSS_MAX_GB:-400} * 1024 * 1024 ))
VMSIZE_MAX_KB=$(( ${VMSIZE_MAX_GB:-440} * 1024 * 1024 ))
[ -r "/proc/$PID/maps" ] || { echo "cannot read /proc/$PID/maps" >&2; exit 1; }
printf 'time,maps,vmsize_kb,vmdata_kb,vmrss_kb,threads,committed_as_kb,commit_limit_kb\n' >"$OUT"
while kill -0 "$PID" 2>/dev/null; do
maps=$(wc -l <"/proc/$PID/maps" 2>/dev/null) || break
st=$(grep -E '^(VmSize|VmData|VmRSS|Threads):' "/proc/$PID/status" 2>/dev/null | awk '{printf "%s,", $2}')
rss=$(grep '^VmRSS:' "/proc/$PID/status" 2>/dev/null | awk '{print $2}')
vsz=$(grep '^VmSize:' "/proc/$PID/status" 2>/dev/null | awk '{print $2}')
mi=$(grep -E '^(Committed_AS|CommitLimit):' /proc/meminfo | awk '{printf "%s,", $2}')
echo "$(date +%T),$maps,${st}${mi%,}" >>"$OUT"
reason=""
[ "$maps" -gt "$MAPS_MAX" ] && reason="mappings $maps > MAPS_MAX $MAPS_MAX"
[ -n "${rss:-}" ] && [ "$rss" -gt "$RSS_MAX_KB" ] && reason="VmRSS ${rss} kB > RSS_MAX_GB"
[ -n "${vsz:-}" ] && [ "$vsz" -gt "$VMSIZE_MAX_KB" ] && reason="VmSize ${vsz} kB > VMSIZE_MAX_GB"
if [ -n "$reason" ]; then
echo "$(date +%T) GUARD: $reason — sending SIGTERM to $PID" | tee "$OUT.reason" >&2
kill -TERM "$PID" 2>/dev/null
for _ in $(seq 30); do kill -0 "$PID" 2>/dev/null || exit 0; sleep 1; done
kill -KILL "$PID" 2>/dev/null
exit 0
fi
sleep "$INTERVAL"
done
+10 -20
View File
@@ -422,7 +422,7 @@ version = "3.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "faf9468729b8cbcea668e36183cb69d317348c2e08e994829fb56ebfdfbaac34" checksum = "faf9468729b8cbcea668e36183cb69d317348c2e08e994829fb56ebfdfbaac34"
dependencies = [ dependencies = [
"windows-sys 0.61.2", "windows-sys 0.52.0",
] ]
[[package]] [[package]]
@@ -758,7 +758,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb" checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb"
dependencies = [ dependencies = [
"libc", "libc",
"windows-sys 0.61.2", "windows-sys 0.52.0",
] ]
[[package]] [[package]]
@@ -1040,7 +1040,7 @@ checksum = "3640c1c38b8e4e43584d8df18be5fc6b0aa314ce6ebf51b53313d4306cca8e46"
dependencies = [ dependencies = [
"hermit-abi", "hermit-abi",
"libc", "libc",
"windows-sys 0.61.2", "windows-sys 0.52.0",
] ]
[[package]] [[package]]
@@ -1408,10 +1408,8 @@ version = "0.1.0"
dependencies = [ dependencies = [
"crossbeam-channel", "crossbeam-channel",
"hashbrown 0.17.1", "hashbrown 0.17.1",
"obifastwrite",
"obikseq", "obikseq",
"rayon", "rayon",
"tracing",
] ]
[[package]] [[package]]
@@ -1492,14 +1490,10 @@ dependencies = [
"ndarray", "ndarray",
"niffler", "niffler",
"obicompactvec", "obicompactvec",
"obidebruinj",
"obikentropy",
"obikseq", "obikseq",
"obipipeline",
"obiread", "obiread",
"obiskio", "obiskio",
"obisys", "obisys",
"obitaxonomy",
"ptr_hash", "ptr_hash",
"rayon", "rayon",
"serde", "serde",
@@ -1537,7 +1531,7 @@ dependencies = [
[[package]] [[package]]
name = "obikmer" name = "obikmer"
version = "1.4.1" version = "1.4.5"
dependencies = [ dependencies = [
"clap", "clap",
"csv", "csv",
@@ -1556,7 +1550,6 @@ dependencies = [
"obikrope", "obikrope",
"obikselect", "obikselect",
"obikseq", "obikseq",
"obiksession",
"obikstats", "obikstats",
"obipipeline", "obipipeline",
"obiread", "obiread",
@@ -1565,6 +1558,7 @@ dependencies = [
"rayon", "rayon",
"serde", "serde",
"serde_yaml", "serde_yaml",
"tikv-jemallocator",
"tracing", "tracing",
"tracing-subscriber", "tracing-subscriber",
] ]
@@ -1582,7 +1576,6 @@ dependencies = [
"obiskio", "obiskio",
"obisys", "obisys",
"rayon", "rayon",
"tempfile",
"tracing", "tracing",
] ]
@@ -1601,7 +1594,6 @@ dependencies = [
"obikseq", "obikseq",
"obipipeline", "obipipeline",
"obiread", "obiread",
"obiskbuilder",
"obiskio", "obiskio",
"obisys", "obisys",
"petgraph", "petgraph",
@@ -1647,7 +1639,6 @@ dependencies = [
name = "obikrope" name = "obikrope"
version = "0.1.0" version = "0.1.0"
dependencies = [ dependencies = [
"bytes",
"criterion2", "criterion2",
] ]
@@ -1725,7 +1716,6 @@ name = "obiskbuilder"
version = "0.1.0" version = "0.1.0"
dependencies = [ dependencies = [
"criterion2", "criterion2",
"lazy_static",
"obikentropy", "obikentropy",
"obikrope", "obikrope",
"obikseq", "obikseq",
@@ -1985,9 +1975,9 @@ dependencies = [
[[package]] [[package]]
name = "ptr_hash" name = "ptr_hash"
version = "2.1.2" version = "2.1.5"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b22cf1be774892680e056d3387db0dcaa4fd46964b270671ef661e127186b969" checksum = "0d2381547e2e36c72a455c1cc418d5c60ecd6e94f4819d7560456108e99a4efa"
dependencies = [ dependencies = [
"bitvec", "bitvec",
"cacheline-ef", "cacheline-ef",
@@ -2279,7 +2269,7 @@ dependencies = [
"errno", "errno",
"libc", "libc",
"linux-raw-sys", "linux-raw-sys",
"windows-sys 0.61.2", "windows-sys 0.52.0",
] ]
[[package]] [[package]]
@@ -2611,7 +2601,7 @@ dependencies = [
"getrandom 0.4.3", "getrandom 0.4.3",
"once_cell", "once_cell",
"rustix", "rustix",
"windows-sys 0.61.2", "windows-sys 0.52.0",
] ]
[[package]] [[package]]
@@ -3016,7 +3006,7 @@ version = "0.1.11"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c2a7b1c03c876122aa43f3020e6c3c3ee5c05081c9a00739faf7503aeba10d22" checksum = "c2a7b1c03c876122aa43f3020e6c3c3ee5c05081c9a00739faf7503aeba10d22"
dependencies = [ dependencies = [
"windows-sys 0.61.2", "windows-sys 0.52.0",
] ]
[[package]] [[package]]
@@ -3,6 +3,7 @@ use std::io;
use std::path::{Path, PathBuf}; use std::path::{Path, PathBuf};
use crate::bitvec::PersistentBitVecBuilder; use crate::bitvec::PersistentBitVecBuilder;
use crate::mem_builder::MemBitVecBuilder;
use crate::meta::MatrixMeta; use crate::meta::MatrixMeta;
use crate::tempbitvec::TempBitVec; use crate::tempbitvec::TempBitVec;
use crate::tempintvec::TempCompactIntVec; use crate::tempintvec::TempCompactIntVec;
@@ -43,6 +44,14 @@ impl PersistentBitMatrixBuilder {
PersistentBitVecBuilder::new(self.n, &path) PersistentBitVecBuilder::new(self.n, &path)
} }
/// Same as [`add_col`](Self::add_col) — next column, `n_cols` + 1 — but
/// the column is built in RAM and written once at `close`.
pub fn add_mem_col(&mut self) -> MemBitVecBuilder {
let path = col_path(&self.dir, self.n_cols);
self.n_cols += 1;
MemBitVecBuilder::new(self.n, &path)
}
pub fn add_col_ones(&mut self) -> io::Result<PersistentBitVecBuilder> { pub fn add_col_ones(&mut self) -> io::Result<PersistentBitVecBuilder> {
let path = col_path(&self.dir, self.n_cols); let path = col_path(&self.dir, self.n_cols);
self.n_cols += 1; self.n_cols += 1;
+1
View File
@@ -18,6 +18,7 @@ mod packed;
mod pairwise; mod pairwise;
mod persistent; mod persistent;
mod sparse; mod sparse;
mod sparse_columnar;
pub use builder::PersistentBitMatrixBuilder; pub use builder::PersistentBitMatrixBuilder;
pub use group_ops::batch_presence_counts; pub use group_ops::batch_presence_counts;
+1 -1
View File
@@ -105,7 +105,7 @@ impl PackedBitMatrix {
/// Reads just the `n_cols` field from an existing packed matrix's header, /// Reads just the `n_cols` field from an existing packed matrix's header,
/// without mapping the file. Used by `pack_bit_matrix` to tell a genuinely /// without mapping the file. Used by `pack_bit_matrix` to tell a genuinely
/// complete pack from a stale one that predates a later column-widening. /// complete pack from a stale one that predates a later column-widening.
fn packed_bit_matrix_n_cols(path: &Path) -> io::Result<usize> { pub(super) fn packed_bit_matrix_n_cols(path: &Path) -> io::Result<usize> {
let mut f = File::open(path)?; let mut f = File::open(path)?;
let mut header = [0u8; PBMX_HEADER]; let mut header = [0u8; PBMX_HEADER];
f.read_exact(&mut header)?; f.read_exact(&mut header)?;
+47 -38
View File
@@ -31,14 +31,12 @@ use ndarray::{Array1, Array2};
use crate::traits::{BitPartials, ColumnWeights}; use crate::traits::{BitPartials, ColumnWeights};
use crate::eliasfano::{EliasFano, EliasFanoBuilder}; use crate::eliasfano::{EliasFano, EliasFanoBuilder};
use crate::fixedintvec::{PersistentFixedIntVec, PersistentFixedIntVecBuilder, bit_width_for_range}; use crate::fixedintvec::{PersistentFixedIntVec, PersistentFixedIntVecBuilder, bit_width_for_range};
use crate::meta::field; use crate::meta::{MatrixMeta, field};
use crate::rankselect::{PersistentRankSelectBitVec, PersistentRankSelectBitVecBuilder}; use crate::rankselect::{PersistentRankSelectBitVec, PersistentRankSelectBitVecBuilder};
use super::PersistentBitMatrix; use super::PersistentBitMatrix;
use super::col_path; use super::col_path;
use super::columnar::ColumnarBitMatrix; use super::packed::{PackedBitMatrix, pack_bit_matrix, packed_bit_matrix_n_cols};
use super::packed::PackedBitMatrix;
use crate::views::nonzero_triples;
fn is_multi_path(dir: &Path) -> PathBuf { dir.join("is_multi.prsb") } fn is_multi_path(dir: &Path) -> PathBuf { dir.join("is_multi.prsb") }
fn singleton_path(dir: &Path) -> PathBuf { dir.join("singleton.pfiv") } fn singleton_path(dir: &Path) -> PathBuf { dir.join("singleton.pfiv") }
@@ -501,12 +499,12 @@ impl PersistentSparseBitMatrixBuilder {
/// existing index. /// existing index.
/// ///
/// Processes rows in batches of [`BUILD_FROM_DENSE_BATCH`] instead of /// Processes rows in batches of [`BUILD_FROM_DENSE_BATCH`] instead of
/// one row at a time: `nonzero_triples(..., sorted_slots=true)` walks /// one row at a time: [`BitSliceView::for_each_set_in`] walks each
/// each column *once per batch*, sequentially — matching how /// column *once per batch*, word by word, visiting only the set bits —
/// `PersistentBitMatrix::nonzero_iter` already batches slots for /// rather than `fill_row`'s row-major loop, which re-reads every column
/// queries — rather than `fill_row`'s row-major loop, which re-reads /// with one `get()` per cell and jumps between all `n_cols` columns on
/// every column with one `get()` per cell and jumps between all /// every single row. Columns are visited in ascending order, so each
/// `n_cols` column files on every single row. /// row's genome list comes out sorted, as `push_row` requires.
pub fn build_from_dense(dense: &PersistentBitMatrix, dir: &Path) -> io::Result<Self> { pub fn build_from_dense(dense: &PersistentBitMatrix, dir: &Path) -> io::Result<Self> {
let n = dense.n(); let n = dense.n();
let n_cols = dense.n_cols(); let n_cols = dense.n_cols();
@@ -515,9 +513,10 @@ impl PersistentSparseBitMatrixBuilder {
let mut start = 0; let mut start = 0;
while start < n { while start < n {
let end = (start + BUILD_FROM_DENSE_BATCH).min(n); let end = (start + BUILD_FROM_DENSE_BATCH).min(n);
let slots: Vec<usize> = (start..end).collect(); for c in 0..n_cols {
for (row, c, _) in nonzero_triples(&slots, n_cols, |c| dense.col_view(c), true) { dense.col_view(c).for_each_set_in(start..end, |slot| {
row_buffers[row].push(c as u32); row_buffers[slot - start].push(c as u32);
});
} }
for buf in row_buffers.iter_mut().take(end - start) { for buf in row_buffers.iter_mut().take(end - start) {
builder.push_row(buf); builder.push_row(buf);
@@ -615,37 +614,47 @@ impl BitPartials for PersistentSparseBitMatrix {
m m
} }
} }
/// sparse on-disk format in place, mirroring [`super::pack_bit_matrix`]'s
/// convention (old-format files removed only after the new format is /// Converts a layer's presence matrix (columnar `col_*.pbiv` or dense
/// fully written, so a crash mid-conversion leaves the previous, still /// `matrix.pbmx`) to the sparse on-disk format in place. Old-format files
/// valid format rather than a half-written one). Idempotent — does /// are removed only after the new format is fully written, so a crash
/// mid-conversion leaves a previous, still valid format. Idempotent: does
/// nothing if `is_multi.prsb` already exists. /// nothing if `is_multi.prsb` already exists.
/// ///
/// The dense→sparse transpose (`build_from_dense`) only needs `fill_row`, /// Columnar data is transposed straight into the sparse format, a window of
/// which costs the same `O(n_cols)` per row on `Columnar` as on `Packed` /// rows at a time, one column file open at a time
/// (see `ColumnarBitMatrix::fill_row`) — so this reads straight from /// ([`PersistentSparseBitMatrixBuilder::build_from_columnar`]): no dense
/// whichever dense form is already on disk. In particular, when this runs /// intermediate is written, and a worker holds no mapping of the columns
/// right after a build (the common case: `presence/` is still `Columnar`, /// whatever their number. A dense `matrix.pbmx` is only read when it is the
/// `matrix.pbmx` doesn't exist yet), it never materialises a full packed /// current form of the layer: no columnar data pending, or a `matrix.pbmx`
/// copy of the presence matrix just to read it back once and delete it. /// that already has as many columns as the columnar metadata says.
pub fn pack_sparse_bit_matrix(dir: &Path) -> io::Result<()> { pub fn pack_sparse_bit_matrix(dir: &Path) -> io::Result<()> {
if is_multi_path(dir).exists() { if is_multi_path(dir).exists() {
return Ok(()); return Ok(());
} }
let packed_path = dir.join("matrix.pbmx"); let packed_path = dir.join("matrix.pbmx");
if packed_path.exists() {
let dense = PersistentBitMatrix::Packed(PackedBitMatrix::open(&packed_path)?); if let Ok(meta) = MatrixMeta::load(dir) {
PersistentSparseBitMatrixBuilder::build_from_dense(&dense, dir)?.close()?; if packed_bit_matrix_n_cols(&packed_path).ok() != Some(meta.n_cols) {
drop(dense); // Columnar data is the freshest form (any `matrix.pbmx` here is
fs::remove_file(&packed_path)?; // stale, e.g. copied from a merge base before the layer was
} else { // widened): transpose it directly.
let dense = PersistentBitMatrix::Columnar(ColumnarBitMatrix::open(dir)?); PersistentSparseBitMatrixBuilder::build_from_columnar(dir, &meta)?.close()?;
let n_cols = dense.n_cols(); for c in 0..meta.n_cols {
PersistentSparseBitMatrixBuilder::build_from_dense(&dense, dir)?.close()?; fs::remove_file(col_path(dir, c))?;
drop(dense); }
for c in 0..n_cols { let _ = fs::remove_file(col_path(dir, c)); } fs::remove_file(dir.join("meta.json"))?;
let _ = fs::remove_file(dir.join("meta.json")); let _ = fs::remove_file(&packed_path);
return Ok(());
}
} }
Ok(())
// Dense packed form is the current one. `pack_bit_matrix` is a no-op when
// there is no columnar data left, and otherwise only drops the
// now-redundant column files.
pack_bit_matrix(dir)?;
let dense = PersistentBitMatrix::Packed(PackedBitMatrix::open(&packed_path)?);
PersistentSparseBitMatrixBuilder::build_from_dense(&dense, dir)?.close()?;
drop(dense);
fs::remove_file(&packed_path)
} }
@@ -0,0 +1,95 @@
//! Building the sparse presence matrix straight from the columnar
//! `col_*.pbiv` files, with no dense intermediate and no mapping.
//!
//! Each column is opened, a window of it is read with `pread`, and it is
//! closed again: a worker holds at most one column file at a time,
//! whatever the number of genomes. See [`crate::transpose`].
use std::fs::File;
use std::io;
use std::os::unix::fs::FileExt;
use std::path::Path;
use crate::bitvec::MAGIC;
use crate::meta::MatrixMeta;
use crate::transpose::{DEFAULT_MAX_HITS, DEFAULT_WINDOW_ROWS, Hit, transpose_windows};
use crate::views::BitSliceView;
use super::PersistentSparseBitMatrixBuilder;
use super::col_path;
/// `PBIV` header: magic(4) + pad(4) + n(8).
const PBIV_HEADER: usize = 16;
/// Reads the bits of `path` for rows `start..end` (`start` a multiple of 64)
/// into `words`, after checking the header against `n`.
fn read_bit_window(
path: &Path,
n: usize,
start: usize,
end: usize,
bytes: &mut Vec<u8>,
words: &mut Vec<u64>,
) -> io::Result<()> {
debug_assert_eq!(start % 64, 0);
let file = File::open(path)?;
let mut header = [0u8; PBIV_HEADER];
file.read_exact_at(&mut header, 0)?;
if header[..4] != MAGIC || u64::from_le_bytes(header[8..16].try_into().unwrap()) as usize != n {
return Err(io::Error::new(
io::ErrorKind::InvalidData,
format!("{}: not a PBIV column of {n} rows", path.display()),
));
}
let nw = (end - start).div_ceil(64);
bytes.clear();
bytes.resize(nw * 8, 0);
file.read_exact_at(bytes, (PBIV_HEADER + start / 8) as u64)?;
words.clear();
words.extend(bytes.chunks_exact(8).map(|b| u64::from_le_bytes(b.try_into().unwrap())));
Ok(())
}
impl PersistentSparseBitMatrixBuilder {
/// Builds the sparse matrix of the columnar layer in `dir` (`meta.json`
/// plus `col_*.pbiv`), writing the sparse files into `dir`. Same rows,
/// same genome order and so same bytes as going through a dense
/// `matrix.pbmx` (`build_from_dense`), without ever writing it.
pub fn build_from_columnar(dir: &Path, meta: &MatrixMeta) -> io::Result<Self> {
Self::build_from_columnar_with(dir, meta, DEFAULT_WINDOW_ROWS, DEFAULT_MAX_HITS)
}
/// [`build_from_columnar`](Self::build_from_columnar) with explicit window
/// size (rows) and non-zero-cell cap; see [`crate::transpose`].
pub fn build_from_columnar_with(
dir: &Path,
meta: &MatrixMeta,
window_rows: usize,
max_hits: usize,
) -> io::Result<Self> {
let mut builder = Self::new(meta.n, meta.n_cols, dir)?;
let mut bytes: Vec<u8> = Vec::new();
let mut words: Vec<u64> = Vec::new();
transpose_windows(
meta.n,
meta.n_cols,
window_rows,
max_hits,
(),
|c, start, end, hits: &mut Vec<Hit<()>>| {
read_bit_window(&col_path(dir, c), meta.n, start, end, &mut bytes, &mut words)?;
BitSliceView::new(&words, end - start).for_each_set_in(0..end - start, |r| {
hits.push((r as u32, c as u32, ()));
});
Ok(())
},
|_, batch| {
for i in 0..batch.len() {
builder.push_row(batch.cols(i));
}
Ok(())
},
)?;
Ok(builder)
}
}
+1 -1
View File
@@ -8,7 +8,7 @@ use crate::mmap_file::{building_path, create_mmap_file, open_mmap_file, publish_
use crate::reader::PersistentCompactIntVec; use crate::reader::PersistentCompactIntVec;
use crate::views::{BitSliceIter, BitSliceView, IntSliceView}; use crate::views::{BitSliceIter, BitSliceView, IntSliceView};
const MAGIC: [u8; 4] = *b"PBIV"; pub(crate) const MAGIC: [u8; 4] = *b"PBIV";
// Header: magic(4) + _pad(4) + n(8) = 16 bytes. // Header: magic(4) + _pad(4) + n(8) = 16 bytes.
// Data starts at offset 16, u64-aligned (mmap base is page-aligned, 16 % 8 == 0). // Data starts at offset 16, u64-aligned (mmap base is page-aligned, 16 % 8 == 0).
+48 -38
View File
@@ -1,5 +1,5 @@
use std::fs::OpenOptions; use std::fs::OpenOptions;
use std::io::{self, BufWriter, Seek, SeekFrom, Write as _}; use std::io::{self, BufWriter, Seek, SeekFrom, Write};
use std::path::Path; use std::path::Path;
pub const MAGIC: [u8; 4] = *b"PCIV"; pub const MAGIC: [u8; 4] = *b"PCIV";
@@ -54,53 +54,63 @@ pub fn parse_index_entry(data: &[u8], base: usize, i: usize) -> (usize, usize) {
// Sparse index target: ≤ 32 KB in L1 cache (16 B per entry → 2048 entries). // Sparse index target: ≤ 32 KB in L1 cache (16 B per entry → 2048 entries).
pub const L1_INDEX_ENTRIES: usize = 2048; pub const L1_INDEX_ENTRIES: usize = 2048;
/// Sparse-index `step` and entry count for `n_overflow` overflow entries:
/// no index (`step == 0`) while the overflow section fits the L1 target.
fn pciv_index_layout(n_overflow: usize) -> (usize, usize) {
if n_overflow <= L1_INDEX_ENTRIES {
(0, 0)
} else {
let step = n_overflow.div_ceil(L1_INDEX_ENTRIES);
(step, n_overflow.div_ceil(step))
}
}
/// Write the 40-byte PCIV header for a vector of `n` slots with
/// `n_overflow` overflow entries (index layout derived from it).
pub(crate) fn write_pciv_header<W: Write>(w: &mut W, n: usize, n_overflow: usize) -> io::Result<()> {
let (step, n_index) = pciv_index_layout(n_overflow);
w.write_all(&MAGIC)?;
w.write_all(&[0u8; 4])?;
w.write_all(&(n as u64).to_le_bytes())?;
w.write_all(&(n_overflow as u64).to_le_bytes())?;
w.write_all(&(n_index as u64).to_le_bytes())?;
w.write_all(&(step as u64).to_le_bytes())
}
/// Write the sections that follow the primary bytes:
/// overflow n_overflow × 12 B (slot: u64, value: u32), `entries` sorted by slot
/// index n_index × 16 B (slot: u64, pos: u64) sparse index
pub(crate) fn write_pciv_tail<W: Write>(w: &mut W, entries: &[(usize, u32)]) -> io::Result<()> {
for &(slot, value) in entries {
w.write_all(&(slot as u64).to_le_bytes())?;
w.write_all(&value.to_le_bytes())?;
}
let (step, _) = pciv_index_layout(entries.len());
if step > 0 {
for (block, chunk) in entries.chunks(step).enumerate() {
w.write_all(&(chunk[0].0 as u64).to_le_bytes())?;
w.write_all(&((block * step) as u64).to_le_bytes())?;
}
}
Ok(())
}
/// Finalise a PCIV file whose placeholder header and primary section are already on disk. /// Finalise a PCIV file whose placeholder header and primary section are already on disk.
/// ///
/// Truncates the file to `HEADER_SIZE + n`, then appends: /// Truncates the file to `HEADER_SIZE + n`, appends the overflow and index
/// overflow n_overflow × 12 B (slot: u64, value: u32) sorted by slot /// sections ([`write_pciv_tail`]) and overwrites the placeholder header at
/// index n_index × 16 B (slot: u64, pos: u64) sparse index /// offset 0 ([`write_pciv_header`]).
/// and overwrites the placeholder header at offset 0.
pub fn finalize_pciv(path: &Path, n: usize, entries: &[(usize, u32)]) -> io::Result<()> { pub fn finalize_pciv(path: &Path, n: usize, entries: &[(usize, u32)]) -> io::Result<()> {
let n_overflow = entries.len();
let file = OpenOptions::new().read(true).write(true).open(path)?; let file = OpenOptions::new().read(true).write(true).open(path)?;
file.set_len((HEADER_SIZE + n) as u64)?; file.set_len((HEADER_SIZE + n) as u64)?;
let mut w = BufWriter::new(file); let mut w = BufWriter::new(file);
w.seek(SeekFrom::End(0))?; w.seek(SeekFrom::End(0))?;
write_pciv_tail(&mut w, entries)?;
for &(slot, value) in entries {
w.write_all(&(slot as u64).to_le_bytes())?;
w.write_all(&value.to_le_bytes())?;
}
let step: usize = if n_overflow <= L1_INDEX_ENTRIES {
0
} else {
n_overflow.div_ceil(L1_INDEX_ENTRIES)
};
let n_index: usize = if step > 0 {
let count = n_overflow.div_ceil(step);
for (block, chunk) in entries.chunks(step).enumerate() {
let slot = chunk[0].0 as u64;
let pos = (block * step) as u64;
w.write_all(&slot.to_le_bytes())?;
w.write_all(&pos.to_le_bytes())?;
}
count
} else {
0
};
w.flush()?; w.flush()?;
let mut file = w.into_inner().map_err(|e| e.into_error())?; let mut file = w.into_inner().map_err(|e| e.into_error())?;
file.seek(SeekFrom::Start(0))?; file.seek(SeekFrom::Start(0))?;
file.write_all(&MAGIC)?; write_pciv_header(&mut file, n, entries.len())?;
file.write_all(&[0u8; 4])?;
file.write_all(&(n as u64).to_le_bytes())?;
file.write_all(&(n_overflow as u64).to_le_bytes())?;
file.write_all(&(n_index as u64).to_le_bytes())?;
file.write_all(&(step as u64).to_le_bytes())?;
file.flush() file.flush()
} }
+10 -1
View File
@@ -11,6 +11,7 @@ use crate::bitmatrix::{
use crate::builder::PersistentCompactIntVecBuilder; use crate::builder::PersistentCompactIntVecBuilder;
use crate::colgroup::{ColGroup, MatrixGroupOps, chunked_presence_count}; use crate::colgroup::{ColGroup, MatrixGroupOps, chunked_presence_count};
use crate::format::{HEADER_SIZE, OVERFLOW_ENTRY_SIZE}; use crate::format::{HEADER_SIZE, OVERFLOW_ENTRY_SIZE};
use crate::mem_builder::MemCompactIntVecBuilder;
use crate::meta::MatrixMeta; use crate::meta::MatrixMeta;
use crate::reader::PersistentCompactIntVec; use crate::reader::PersistentCompactIntVec;
use crate::sparse_intmatrix::{PersistentSparseCompactIntMatrix, is_present as sparse_is_present}; use crate::sparse_intmatrix::{PersistentSparseCompactIntMatrix, is_present as sparse_is_present};
@@ -297,7 +298,7 @@ impl PackedCompactIntMatrix {
/// without mapping the file. Used by `pack_compact_int_matrix` to tell a /// without mapping the file. Used by `pack_compact_int_matrix` to tell a
/// genuinely complete pack from a stale one that predates a later /// genuinely complete pack from a stale one that predates a later
/// column-widening. /// column-widening.
fn packed_int_matrix_n_cols(path: &Path) -> io::Result<usize> { pub(crate) fn packed_int_matrix_n_cols(path: &Path) -> io::Result<usize> {
let mut f = File::open(path)?; let mut f = File::open(path)?;
let mut header = [0u8; PCMX_HEADER]; let mut header = [0u8; PCMX_HEADER];
f.read_exact(&mut header)?; f.read_exact(&mut header)?;
@@ -717,6 +718,14 @@ impl PersistentCompactIntMatrixBuilder {
PersistentCompactIntVecBuilder::new(self.n, &path) PersistentCompactIntVecBuilder::new(self.n, &path)
} }
/// Same as [`add_col`](Self::add_col) — next column, `n_cols` + 1 — but
/// the column is built in RAM and written once at `close`.
pub fn add_mem_col(&mut self) -> MemCompactIntVecBuilder {
let path = col_path(&self.dir, self.n_cols);
self.n_cols += 1;
MemCompactIntVecBuilder::new(self.n, &path)
}
pub fn add_col_from(&mut self, src: &TempCompactIntVec) -> io::Result<()> { pub fn add_col_from(&mut self, src: &TempCompactIntVec) -> io::Result<()> {
src.make_persistent(&col_path(&self.dir, self.n_cols))?; src.make_persistent(&col_path(&self.dir, self.n_cols))?;
self.n_cols += 1; self.n_cols += 1;
+5
View File
@@ -9,16 +9,20 @@ mod int_group_ops;
mod intmatrix; mod intmatrix;
mod layer_meta; mod layer_meta;
mod matrix_builder; mod matrix_builder;
mod mem_builder;
mod meta; mod meta;
mod mmap_file; mod mmap_file;
mod rankselect; mod rankselect;
mod reader; mod reader;
mod sparse_intmatrix; mod sparse_intmatrix;
mod sparse_intmatrix_columnar;
mod storage_kind; mod storage_kind;
mod tempbitvec; mod tempbitvec;
mod tempintvec; mod tempintvec;
mod transpose;
pub mod traits; pub mod traits;
mod views; mod views;
mod views_scan;
pub use bitmatrix::{ pub use bitmatrix::{
PersistentBitMatrix, PersistentBitMatrixBuilder, PersistentSparseBitMatrix, PersistentBitMatrix, PersistentBitMatrixBuilder, PersistentSparseBitMatrix,
@@ -35,6 +39,7 @@ pub use intmatrix::{
}; };
pub use layer_meta::LayerMeta; pub use layer_meta::LayerMeta;
pub use mem_builder::{MemBitVecBuilder, MemCompactIntVecBuilder};
pub use mmap_file::{building_path, publish_built}; pub use mmap_file::{building_path, publish_built};
pub use matrix_builder::{ColBuilder, MatrixBuilder}; pub use matrix_builder::{ColBuilder, MatrixBuilder};
pub use rankselect::{PersistentRankSelectBitVec, PersistentRankSelectBitVecBuilder}; pub use rankselect::{PersistentRankSelectBitVec, PersistentRankSelectBitVecBuilder};
+31 -1
View File
@@ -10,13 +10,19 @@ use std::path::Path;
use crate::{PersistentBitMatrixBuilder as BitMB, PersistentBitVecBuilder}; use crate::{PersistentBitMatrixBuilder as BitMB, PersistentBitVecBuilder};
use crate::{PersistentCompactIntMatrixBuilder as IntMB, PersistentCompactIntVecBuilder}; use crate::{PersistentCompactIntMatrixBuilder as IntMB, PersistentCompactIntVecBuilder};
use crate::{TempBitVec, TempCompactIntVec}; use crate::{MemBitVecBuilder, MemCompactIntVecBuilder, TempBitVec, TempCompactIntVec};
// ── ColBuilder ──────────────────────────────────────────────────────────────── // ── ColBuilder ────────────────────────────────────────────────────────────────
/// One column being built. `Bit`/`Int` write through a mmap of a sparse
/// file as values are set; `MemBit`/`MemInt` keep the column in RAM and
/// write it once at `close` (same file, byte for byte — see
/// [`MemBitVecBuilder`]).
pub enum ColBuilder { pub enum ColBuilder {
Bit(PersistentBitVecBuilder), Bit(PersistentBitVecBuilder),
Int(PersistentCompactIntVecBuilder), Int(PersistentCompactIntVecBuilder),
MemBit(MemBitVecBuilder),
MemInt(MemCompactIntVecBuilder),
} }
impl ColBuilder { impl ColBuilder {
@@ -24,6 +30,8 @@ impl ColBuilder {
match self { match self {
ColBuilder::Bit(b) => b.set(slot, value > 0), ColBuilder::Bit(b) => b.set(slot, value > 0),
ColBuilder::Int(b) => b.set(slot, value), ColBuilder::Int(b) => b.set(slot, value),
ColBuilder::MemBit(b) => b.set(slot, value > 0),
ColBuilder::MemInt(b) => b.set(slot, value),
} }
} }
@@ -31,6 +39,8 @@ impl ColBuilder {
match self { match self {
ColBuilder::Bit(b) => b.close(), ColBuilder::Bit(b) => b.close(),
ColBuilder::Int(b) => b.close(), ColBuilder::Int(b) => b.close(),
ColBuilder::MemBit(b) => b.close(),
ColBuilder::MemInt(b) => b.close(),
} }
} }
} }
@@ -64,6 +74,15 @@ impl MatrixBuilder {
}) })
} }
/// Number of columns created so far (absent + data columns, including
/// those inherited from a resumed directory).
pub fn n_cols(&self) -> usize {
match self {
MatrixBuilder::Bit(b) => b.n_cols(),
MatrixBuilder::Int(b) => b.n_cols(),
}
}
/// Add a column with no data written (all-zero/false) — for columns /// Add a column with no data written (all-zero/false) — for columns
/// absent from a given source (e.g. destination-only genomes not /// absent from a given source (e.g. destination-only genomes not
/// contributed by the source being merged in). /// contributed by the source being merged in).
@@ -81,6 +100,17 @@ impl MatrixBuilder {
}) })
} }
/// Same as [`add_col`](Self::add_col) — next column, `n_cols` + 1 — but
/// the column is built in RAM and written once, sequentially, at
/// `close`: no sparse file, no writable mapping. Costs `n / 8` bytes
/// (presence) or `n` bytes + overflow (counts) of RAM until closed.
pub fn add_mem_col(&mut self) -> ColBuilder {
match self {
MatrixBuilder::Bit(b) => ColBuilder::MemBit(b.add_mem_col()),
MatrixBuilder::Int(b) => ColBuilder::MemInt(b.add_mem_col()),
}
}
/// Add a column already computed as a bit vector. /// Add a column already computed as a bit vector.
pub fn add_col_from_bit(&mut self, src: &TempBitVec) -> io::Result<()> { pub fn add_col_from_bit(&mut self, src: &TempBitVec) -> io::Result<()> {
match self { match self {
+122
View File
@@ -0,0 +1,122 @@
//! In-memory column builders: the content of a column is built in RAM and
//! written once, sequentially, when the builder is closed.
//!
//! They produce exactly the same file, byte for byte, as
//! [`PersistentBitVecBuilder`](crate::PersistentBitVecBuilder) /
//! [`PersistentCompactIntVecBuilder`](crate::PersistentCompactIntVecBuilder),
//! but never create a sparse file nor a writable mapping: a block the file
//! system cannot allocate (full disk, quota, network FS) is reported as an
//! `io::Error` by `close`, instead of a SIGBUS on the page fault that first
//! touches it. The price is that the whole column lives in RAM until
//! `close`: `n / 8` bytes for bits, `n` bytes (+ overflow map) for counts.
use std::collections::HashMap;
use std::io::{self, Write};
use std::path::{Path, PathBuf};
use crate::bitvec::{n_words, MAGIC as PBIV_MAGIC};
use crate::format::{write_pciv_header, write_pciv_tail};
use crate::mmap_file::{write_built, write_std_header};
// ── Bits (PBIV) ───────────────────────────────────────────────────────────────
/// In-memory counterpart of `PersistentBitVecBuilder` (`set` + `close` only).
pub struct MemBitVecBuilder {
words: Vec<u64>,
n: usize,
path: PathBuf,
}
impl MemBitVecBuilder {
/// All-false vector of `n` bits, to be written at `path` by `close`.
pub fn new(n: usize, path: &Path) -> Self {
Self { words: vec![0u64; n_words(n)], n, path: path.to_path_buf() }
}
#[inline]
pub fn len(&self) -> usize {
self.n
}
#[inline]
pub fn is_empty(&self) -> bool {
self.n == 0
}
#[inline]
pub fn set(&mut self, slot: usize, value: bool) {
debug_assert!(slot < self.n);
let bit = 1u64 << (slot & 63);
if value {
self.words[slot >> 6] |= bit;
} else {
self.words[slot >> 6] &= !bit;
}
}
/// Writes the PBIV file — header, then the words in native byte order,
/// exactly as the mmap-backed builder lays them out — and publishes it.
pub fn close(self) -> io::Result<()> {
let Self { words, n, path } = self;
write_built(&path, |w| {
write_std_header(w, PBIV_MAGIC, [0u8; 4], n)?;
for word in &words {
w.write_all(&word.to_ne_bytes())?;
}
Ok(())
})
}
}
// ── Compact ints (PCIV) ───────────────────────────────────────────────────────
/// In-memory counterpart of `PersistentCompactIntVecBuilder` (`set` +
/// `close` only). Same encoding: one byte per slot, values ≥ 255 stored as
/// the sentinel 255 plus an overflow entry.
pub struct MemCompactIntVecBuilder {
primary: Vec<u8>,
overflow: HashMap<usize, u32>,
n: usize,
path: PathBuf,
}
impl MemCompactIntVecBuilder {
/// All-zero vector of `n` slots, to be written at `path` by `close`.
pub fn new(n: usize, path: &Path) -> Self {
Self { primary: vec![0u8; n], overflow: HashMap::new(), n, path: path.to_path_buf() }
}
#[inline]
pub fn len(&self) -> usize {
self.n
}
#[inline]
pub fn is_empty(&self) -> bool {
self.n == 0
}
#[inline]
pub fn set(&mut self, slot: usize, value: u32) {
if value < 255 {
self.primary[slot] = value as u8;
self.overflow.remove(&slot);
} else {
self.primary[slot] = 255;
self.overflow.insert(slot, value);
}
}
/// Writes the PCIV file — header, primary bytes, overflow entries sorted
/// by slot, sparse index — and publishes it.
pub fn close(self) -> io::Result<()> {
let Self { primary, overflow, n, path } = self;
let mut entries: Vec<(usize, u32)> = overflow.into_iter().collect();
entries.sort_unstable_by_key(|&(slot, _)| slot);
write_built(&path, |w| {
write_pciv_header(w, n, entries.len())?;
w.write_all(&primary)?;
write_pciv_tail(w, &entries)
})
}
}
+33 -4
View File
@@ -11,7 +11,7 @@
//! between formats and aren't forced into a shared shape. //! between formats and aren't forced into a shared shape.
use std::fs::{self, File, OpenOptions}; use std::fs::{self, File, OpenOptions};
use std::io::{self, Seek, SeekFrom, Write as _}; use std::io::{self, BufWriter, Seek, SeekFrom, Write};
use std::path::{Path, PathBuf}; use std::path::{Path, PathBuf};
use memmap2::{Mmap, MmapMut}; use memmap2::{Mmap, MmapMut};
@@ -37,6 +37,37 @@ pub fn publish_built(path: &Path) -> io::Result<()> {
fs::rename(building_path(path), path) fs::rename(building_path(path), path)
} }
/// Writes the crate-wide standard header: `magic` + `pad` + `n as u64` (LE).
pub(crate) fn write_std_header<W: Write>(w: &mut W, magic: [u8; 4], pad: [u8; 4], n: usize) -> io::Result<()> {
w.write_all(&magic)?;
w.write_all(&pad)?;
w.write_all(&(n as u64).to_le_bytes())
}
/// Writes a whole file sequentially through `write` into
/// [`building_path`]`(path)`, flushes it, then publishes it at `path`. No
/// sparse file and no writable mapping: an allocation failure on the file
/// system (full disk, quota) surfaces as an `io::Error` from a `write`,
/// never as a SIGBUS on a page fault.
///
/// Deliberately no `sync_all`: one forced sync per file made the merge
/// ~8× slower (thousands of column files; measured 2026-11-17 on macOS),
/// and the mmap builders it replaces never synced either. Durability is the
/// OS's, like the rest of an index; the `rename` still guarantees that no
/// partially written file is ever published. Caveat: on a write-back
/// network FS an allocation error may only be reported at `close`, which
/// `File`'s drop ignores.
pub(crate) fn write_built(
path: &Path,
write: impl FnOnce(&mut BufWriter<File>) -> io::Result<()>,
) -> io::Result<()> {
let mut w = BufWriter::new(File::create(building_path(path))?);
write(&mut w)?;
w.flush()?;
drop(w.into_inner().map_err(|e| e.into_error())?);
publish_built(path)
}
/// Creates (or truncates) [`building_path`]`(path)` — the caller publishes it /// Creates (or truncates) [`building_path`]`(path)` — the caller publishes it
/// with [`publish_built`] once complete — writes the standard header — `magic` + /// with [`publish_built`] once complete — writes the standard header — `magic` +
/// `pad` (4 bytes right after `magic`; usually all-zero, but /// `pad` (4 bytes right after `magic`; usually all-zero, but
@@ -60,9 +91,7 @@ pub(crate) fn create_mmap_file(
.create(true) .create(true)
.truncate(true) .truncate(true)
.open(building_path(path))?; .open(building_path(path))?;
file.write_all(&magic)?; write_std_header(&mut file, magic, pad, n)?;
file.write_all(&pad)?;
file.write_all(&(n as u64).to_le_bytes())?;
if extra_header_zeros > 0 { if extra_header_zeros > 0 {
file.write_all(&vec![0u8; extra_header_zeros])?; file.write_all(&vec![0u8; extra_header_zeros])?;
} }
+26 -25
View File
@@ -33,6 +33,7 @@ use ndarray::{Array1, Array2};
use crate::bitmatrix::RowRank; use crate::bitmatrix::RowRank;
use crate::builder::PersistentCompactIntVecBuilder; use crate::builder::PersistentCompactIntVecBuilder;
use crate::eliasfano::{EliasFano, EliasFanoBuilder}; use crate::eliasfano::{EliasFano, EliasFanoBuilder};
use crate::meta::MatrixMeta;
use crate::reader::PersistentCompactIntVec; use crate::reader::PersistentCompactIntVec;
use crate::traits::{BitPartials, ColumnWeights, CountPartials}; use crate::traits::{BitPartials, ColumnWeights, CountPartials};
use crate::{PersistentIntMatrix, PersistentSparseBitMatrix, PersistentSparseBitMatrixBuilder}; use crate::{PersistentIntMatrix, PersistentSparseBitMatrix, PersistentSparseBitMatrixBuilder};
@@ -617,37 +618,37 @@ impl PersistentSparseCompactIntMatrixBuilder {
// ── pack_sparse_compact_int_matrix ────────────────────────────────────────── // ── pack_sparse_compact_int_matrix ──────────────────────────────────────────
/// Converts whichever dense count-matrix format is on disk at `dir` /// Converts a layer's count matrix (columnar `col_*.pciv` or dense
/// (`Packed`, i.e. `matrix.pcmx`, or `Columnar`, i.e. `col_*.pciv` + /// `matrix.pcmx`) to the sparse on-disk format in place, exactly like
/// `meta.json`) into the sparse on-disk format in place — generic over the /// [`crate::pack_sparse_bit_matrix`]: columnar data is transposed straight
/// dense source the same way [`super::bitmatrix::pack_sparse_bit_matrix`] /// into the sparse format, window by window, one column file open at a time
/// is: detects the format, transposes straight from it, and only removes /// (no dense intermediate, no mapping); a dense `matrix.pcmx` is read only
/// the old dense files once the new format is fully written (a crash /// when it is the current form. Old-format files are removed only once the
/// mid-conversion leaves the previous, still-valid format rather than a /// new format is fully written. Idempotent: a no-op if the sparse format is
/// half-written one). Idempotent — a no-op if `singleton_values.pciv` /// already present.
/// already exists.
pub fn pack_sparse_compact_int_matrix(dir: &Path) -> io::Result<()> { pub fn pack_sparse_compact_int_matrix(dir: &Path) -> io::Result<()> {
use crate::intmatrix::{ColumnarCompactIntMatrix, PackedCompactIntMatrix, col_path}; use crate::intmatrix::{PackedCompactIntMatrix, pack_compact_int_matrix, packed_int_matrix_n_cols};
if is_present(dir) { if is_present(dir) {
return Ok(()); return Ok(());
} }
let packed_path = dir.join("matrix.pcmx"); let packed_path = dir.join("matrix.pcmx");
if packed_path.exists() {
let dense = PersistentIntMatrix::Packed(PackedCompactIntMatrix::open(&packed_path)?); if let Ok(meta) = MatrixMeta::load(dir) {
PersistentSparseCompactIntMatrixBuilder::build_from_dense(&dense, dir)?.close()?; if packed_int_matrix_n_cols(&packed_path).ok() != Some(meta.n_cols) {
drop(dense); PersistentSparseCompactIntMatrixBuilder::build_from_columnar(dir, &meta)?.close()?;
fs::remove_file(&packed_path)?; for c in 0..meta.n_cols {
} else { fs::remove_file(crate::intmatrix::col_path(dir, c))?;
let dense = PersistentIntMatrix::Columnar(ColumnarCompactIntMatrix::open(dir)?); }
let n_cols = dense.n_cols(); fs::remove_file(dir.join("meta.json"))?;
PersistentSparseCompactIntMatrixBuilder::build_from_dense(&dense, dir)?.close()?; let _ = fs::remove_file(&packed_path);
drop(dense); return Ok(());
for c in 0..n_cols {
let _ = fs::remove_file(col_path(dir, c));
} }
let _ = fs::remove_file(dir.join("meta.json"));
} }
Ok(())
pack_compact_int_matrix(dir)?;
let dense = PersistentIntMatrix::Packed(PackedCompactIntMatrix::open(&packed_path)?);
PersistentSparseCompactIntMatrixBuilder::build_from_dense(&dense, dir)?.close()?;
drop(dense);
fs::remove_file(&packed_path)
} }
@@ -0,0 +1,148 @@
//! Building the sparse count matrix straight from the columnar
//! `col_*.pciv` files, with no dense intermediate and no mapping. Counterpart
//! of `bitmatrix::sparse_columnar`; see [`crate::transpose`].
//!
//! A `PCIV` column is `header(40) | primary(n bytes) | overflow(n_overflow
//! × 12 B, sorted by slot) | index`. A primary byte `255` is a sentinel: the
//! true value (≥ 255) is the overflow entry of that slot. A window therefore
//! reads its primary bytes, then looks up the overflow entries of its
//! sentinels (binary search for the first one, one contiguous read for the
//! rest).
use std::fs::File;
use std::io;
use std::os::unix::fs::FileExt;
use std::path::Path;
use crate::format::{HEADER_SIZE, MAGIC, OVERFLOW_ENTRY_SIZE, parse_overflow_entry};
use crate::intmatrix::col_path;
use crate::meta::MatrixMeta;
use crate::sparse_intmatrix::PersistentSparseCompactIntMatrixBuilder;
use crate::transpose::{DEFAULT_MAX_HITS, DEFAULT_WINDOW_ROWS, Hit, transpose_windows};
fn invalid(path: &Path, what: &str) -> io::Error {
io::Error::new(io::ErrorKind::InvalidData, format!("{}: {what}", path.display()))
}
/// Appends the non-zero cells of column `c` (file `path`) for rows
/// `start..end` to `hits`, as `(row - start, c, value)`.
fn read_int_window(
path: &Path,
n: usize,
c: usize,
start: usize,
end: usize,
primary: &mut Vec<u8>,
hits: &mut Vec<Hit<u32>>,
) -> io::Result<()> {
let file = File::open(path)?;
let mut header = [0u8; HEADER_SIZE];
file.read_exact_at(&mut header, 0)?;
if header[..4] != MAGIC || u64::from_le_bytes(header[8..16].try_into().unwrap()) as usize != n {
return Err(invalid(path, &format!("not a PCIV column of {n} rows")));
}
let n_overflow = u64::from_le_bytes(header[16..24].try_into().unwrap()) as usize;
primary.clear();
primary.resize(end - start, 0);
file.read_exact_at(primary, (HEADER_SIZE + start) as u64)?;
// Slots (relative to `start`) whose primary byte is the 255 sentinel,
// pushed to `hits` with a placeholder and patched below.
let first_hit = hits.len();
let mut n_sentinels = 0usize;
let mut chunks = primary.chunks_exact(8);
let mut base = 0usize;
for chunk in &mut chunks {
if u64::from_le_bytes(chunk.try_into().unwrap()) != 0 {
for (j, &b) in chunk.iter().enumerate() {
if b != 0 {
n_sentinels += (b == 255) as usize;
hits.push(((base + j) as u32, c as u32, b as u32));
}
}
}
base += 8;
}
for (j, &b) in chunks.remainder().iter().enumerate() {
if b != 0 {
n_sentinels += (b == 255) as usize;
hits.push(((base + j) as u32, c as u32, b as u32));
}
}
if n_sentinels == 0 {
return Ok(());
}
// Overflow entries of this window: first entry with slot >= start, found
// by binary search over the on-disk section, then one contiguous read.
let ov_base = (HEADER_SIZE + n) as u64;
let slot_at = |i: usize| -> io::Result<usize> {
let mut b = [0u8; 8];
file.read_exact_at(&mut b, ov_base + (i * OVERFLOW_ENTRY_SIZE) as u64)?;
Ok(u64::from_le_bytes(b) as usize)
};
let (mut lo, mut hi) = (0usize, n_overflow);
while lo < hi {
let mid = lo + (hi - lo) / 2;
if slot_at(mid)? < start {
lo = mid + 1;
} else {
hi = mid;
}
}
if lo + n_sentinels > n_overflow {
return Err(invalid(path, "fewer overflow entries than sentinel bytes"));
}
let mut raw = vec![0u8; n_sentinels * OVERFLOW_ENTRY_SIZE];
file.read_exact_at(&mut raw, ov_base + (lo * OVERFLOW_ENTRY_SIZE) as u64)?;
let mut k = 0usize;
for h in &mut hits[first_hit..] {
if h.2 == 255 {
let (slot, value) = parse_overflow_entry(&raw, 0, k);
if slot != start + h.0 as usize {
return Err(invalid(path, "overflow entry does not match its sentinel slot"));
}
h.2 = value;
k += 1;
}
}
Ok(())
}
impl PersistentSparseCompactIntMatrixBuilder {
/// Builds the sparse count matrix of the columnar layer in `dir`
/// (`meta.json` plus `col_*.pciv`), writing the sparse files into `dir`.
/// Same rows, same cells and so same bytes as going through a dense
/// `matrix.pcmx` (`build_from_dense`), without ever writing it.
pub fn build_from_columnar(dir: &Path, meta: &MatrixMeta) -> io::Result<Self> {
Self::build_from_columnar_with(dir, meta, DEFAULT_WINDOW_ROWS, DEFAULT_MAX_HITS)
}
/// [`build_from_columnar`](Self::build_from_columnar) with explicit window
/// size (rows) and non-zero-cell cap; see [`crate::transpose`].
pub fn build_from_columnar_with(
dir: &Path,
meta: &MatrixMeta,
window_rows: usize,
max_hits: usize,
) -> io::Result<Self> {
let mut builder = Self::new(meta.n, meta.n_cols, dir)?;
let mut primary: Vec<u8> = Vec::new();
transpose_windows(
meta.n,
meta.n_cols,
window_rows,
max_hits,
0u32,
|c, start, end, hits| read_int_window(&col_path(dir, c), meta.n, c, start, end, &mut primary, hits),
|_, batch| {
for i in 0..batch.len() {
builder.push_row(batch.cols(i), batch.vals(i));
}
Ok(())
},
)?;
Ok(builder)
}
}
+14
View File
@@ -249,6 +249,20 @@ fn partial_threshold_jaccard_consistent() {
} }
} }
#[test]
fn threshold_hamming_counts_kmers_present_in_one_genome_only() {
// At threshold 2 the genomes (columns) hold the k-mers (rows):
// c0 = {r0, r2}, c1 = {r1}, c2 = {r1}.
let (_d, m) = make_matrix(&[&[3, 0, 2], &[1, 2, 0], &[0, 3, 1]]);
let h = m.threshold_hamming_dist_matrix(2);
let expected = [[0, 3, 3], [3, 0, 0], [3, 0, 0]];
for i in 0..3 {
for j in 0..3 {
assert_eq!(h[[i, j]], expected[i][j], "[{i},{j}]");
}
}
}
#[test] #[test]
fn single_col_distance_matrix_is_zero() { fn single_col_distance_matrix_is_zero() {
let (_d, m) = make_matrix(&[&[1, 2, 3]]); let (_d, m) = make_matrix(&[&[1, 2, 3]]);
@@ -0,0 +1,27 @@
use tempfile::tempdir;
use crate::MatrixBuilder;
#[test]
fn n_cols_counts_absent_data_and_resumed_columns() {
for presence in [true, false] {
let dir = tempdir().unwrap();
let data = dir.path().join("m");
let mut mb = MatrixBuilder::new(presence, 5, &data).unwrap();
assert_eq!(mb.n_cols(), 0);
mb.add_absent_col().unwrap();
assert_eq!(mb.n_cols(), 1);
let mut c = mb.add_col().unwrap();
c.set_val(2, 7);
c.close().unwrap();
assert_eq!(mb.n_cols(), 2);
mb.close().unwrap();
let mut mb = MatrixBuilder::resume(presence, &data).unwrap();
assert_eq!(mb.n_cols(), 2, "presence={presence}");
mb.add_absent_col().unwrap();
assert_eq!(mb.n_cols(), 3);
mb.close().unwrap();
}
}
+117
View File
@@ -0,0 +1,117 @@
use std::fs;
use tempfile::tempdir;
use crate::{
MatrixBuilder, MemBitVecBuilder, MemCompactIntVecBuilder, PersistentBitMatrix,
PersistentBitVecBuilder, PersistentCompactIntVecBuilder, PersistentIntMatrix,
};
const SIZES: [usize; 6] = [0, 1, 63, 64, 65, 10_000];
/// Deterministic xorshift64* — no RNG dependency needed for test data.
struct Rng(u64);
impl Rng {
fn next(&mut self) -> u64 {
self.0 ^= self.0 >> 12;
self.0 ^= self.0 << 25;
self.0 ^= self.0 >> 27;
self.0.wrapping_mul(0x2545_F491_4F6C_DD1D)
}
}
/// `(slot, value)` writes in random order, with overwrites; for counts,
/// about a third of the values are ≥ 255 (overflow), enough at n = 10 000
/// to exceed `L1_INDEX_ENTRIES` and exercise the sparse index section.
fn writes(n: usize, seed: u64, max_small: u64) -> Vec<(usize, u32)> {
let mut rng = Rng(seed);
(0..2 * n)
.map(|_| {
let slot = (rng.next() % n as u64) as usize;
let r = rng.next();
let value = match r % 3 {
0 => 255 + (r >> 8) as u32 % 100_000, // overflow
1 => 0, // also clears a previous overflow
_ => (r >> 8) as u32 % max_small as u32,
};
(slot, value)
})
.collect()
}
#[test]
fn mem_bitvec_is_byte_identical_to_persistent() {
for n in SIZES {
let dir = tempdir().unwrap();
let (pp, mp) = (dir.path().join("p.pbiv"), dir.path().join("m.pbiv"));
let mut p = PersistentBitVecBuilder::new(n, &pp).unwrap();
let mut m = MemBitVecBuilder::new(n, &mp);
for (slot, v) in writes(n, 0x9E37_79B9 + n as u64, 2) {
p.set(slot, v > 0);
m.set(slot, v > 0);
}
p.close().unwrap();
m.close().unwrap();
assert_eq!(fs::read(&pp).unwrap(), fs::read(&mp).unwrap(), "n = {n}");
}
}
#[test]
fn mem_compact_int_is_byte_identical_to_persistent() {
for n in SIZES {
let dir = tempdir().unwrap();
let (pp, mp) = (dir.path().join("p.pciv"), dir.path().join("m.pciv"));
let mut p = PersistentCompactIntVecBuilder::new(n, &pp).unwrap();
let mut m = MemCompactIntVecBuilder::new(n, &mp);
for (slot, v) in writes(n, 0xC2B2_AE35 + n as u64, 255) {
p.set(slot, v);
m.set(slot, v);
}
p.close().unwrap();
m.close().unwrap();
assert_eq!(fs::read(&pp).unwrap(), fs::read(&mp).unwrap(), "n = {n}");
}
}
/// `add_mem_col` columns, interleaved with absent columns and continued
/// across a `resume`, read back through the regular matrix readers.
#[test]
fn add_mem_col_matrix_reads_back() {
for presence in [true, false] {
let n = 100;
let dir = tempdir().unwrap();
let data = dir.path().join(if presence { "presence" } else { "counts" });
let value = |c: usize, slot: usize| -> u32 {
if (slot + c).is_multiple_of(3) { (slot * 7 + c) as u32 % 400 + 1 } else { 0 }
};
let mut mb = MatrixBuilder::new(presence, n, &data).unwrap();
mb.add_absent_col().unwrap();
let mut c1 = mb.add_mem_col();
(0..n).for_each(|s| c1.set_val(s, value(1, s)));
c1.close().unwrap();
mb.close().unwrap();
let mut mb = MatrixBuilder::resume(presence, &data).unwrap();
assert_eq!(mb.n_cols(), 2);
let mut c2 = mb.add_mem_col();
(0..n).for_each(|s| c2.set_val(s, value(2, s)));
c2.close().unwrap();
assert_eq!(mb.n_cols(), 3);
mb.close().unwrap();
for s in 0..n {
let expect = [0, value(1, s), value(2, s)];
if presence {
let m = PersistentBitMatrix::open(dir.path()).unwrap();
assert_eq!(m.n_cols(), 3);
let want: Vec<bool> = expect.iter().map(|&v| v > 0).collect();
assert_eq!(&*m.row(s), want.as_slice(), "slot {s}");
} else {
let m = PersistentIntMatrix::open(dir.path()).unwrap();
assert_eq!(m.n_cols(), 3);
assert_eq!(&*m.row(s), &expect, "slot {s}");
}
}
}
}
+4
View File
@@ -5,9 +5,13 @@ mod eliasfano;
mod fixedintvec; mod fixedintvec;
mod int_group_ops; mod int_group_ops;
mod intmatrix; mod intmatrix;
mod matrix_builder;
mod mem_builder;
mod rankselect; mod rankselect;
mod sparse; mod sparse;
mod sparse_intmatrix; mod sparse_intmatrix;
mod transpose_columnar;
mod views_scan;
use tempfile::tempdir; use tempfile::tempdir;
@@ -0,0 +1,156 @@
//! The windowed columnar → sparse transposition must produce the very bytes
//! the dense route (`pack` to `matrix.pbmx`/`.pcmx`, then `build_from_dense`)
//! produces, whatever the window size and non-zero-cell cap — including the
//! retry-with-a-smaller-window path.
use std::fs;
use std::path::Path;
use tempfile::tempdir;
use crate::meta::MatrixMeta;
use crate::{
PersistentBitMatrix, PersistentBitMatrixBuilder, PersistentCompactIntMatrixBuilder,
PersistentIntMatrix, PersistentSparseBitMatrixBuilder, PersistentSparseCompactIntMatrixBuilder,
pack_bit_matrix, pack_compact_int_matrix,
};
const N: usize = 5_000 + 37; // not a multiple of 64
const N_COLS: usize = 7;
fn xorshift(s: &mut u64) -> u64 {
*s ^= *s << 13;
*s ^= *s >> 7;
*s ^= *s << 17;
*s
}
/// Column `c`: all zero, very sparse, sparse, dense, ... so that rows have
/// from 0 to `N_COLS` cells and whole words are zero in some columns.
fn cell(c: usize, s: &mut u64) -> Option<u32> {
let r = xorshift(s);
let keep = match c % 4 {
0 => false,
1 => r % 97 == 0,
2 => r % 7 == 0,
_ => r % 2 == 0,
};
keep.then(|| match (r >> 20) % 11 {
0 => 255,
1 => 256,
2 => 70_000,
v => (v as u32) + 1,
})
}
fn build_bits(root: &Path) {
let mut b = PersistentBitMatrixBuilder::new(N, &root.join("presence")).unwrap();
let mut s = 0x9E37_79B9_7F4A_7C15u64;
for c in 0..N_COLS {
let mut cb = b.add_col().unwrap();
for slot in 0..N {
cb.set(slot, cell(c, &mut s).is_some());
}
cb.close().unwrap();
}
b.close().unwrap();
}
fn build_ints(root: &Path) {
let mut b = PersistentCompactIntMatrixBuilder::new(N, &root.join("counts")).unwrap();
let mut s = 0x9E37_79B9_7F4A_7C15u64;
for c in 0..N_COLS {
let mut cb = b.add_col().unwrap();
for slot in 0..N {
if let Some(v) = cell(c, &mut s) {
cb.set(slot, v);
}
}
cb.close().unwrap();
}
b.close().unwrap();
}
/// Sparse-format files of `dir` (columnar leftovers and dense files excluded),
/// as `(name, bytes)` sorted by name.
fn sparse_files(dir: &Path) -> Vec<(String, Vec<u8>)> {
let mut v: Vec<(String, Vec<u8>)> = fs::read_dir(dir)
.unwrap()
.map(|e| e.unwrap().path())
.filter(|p| {
let n = p.file_name().unwrap().to_string_lossy().into_owned();
!(n.starts_with("col_") || n == "meta.json" || n.starts_with("matrix."))
})
.map(|p| (p.file_name().unwrap().to_string_lossy().into_owned(), fs::read(&p).unwrap()))
.collect();
v.sort();
v
}
const WINDOWS: &[(usize, usize)] = &[
(64, usize::MAX / 4), // smallest window, never too dense
(192, 50), // forces the retry-with-half-window path
(4096, 10_000), // a few windows
(crate::transpose::DEFAULT_WINDOW_ROWS, crate::transpose::DEFAULT_MAX_HITS),
];
#[test]
fn bit_columnar_transpose_is_byte_identical_to_the_dense_route() {
let ref_root = tempdir().unwrap();
build_bits(ref_root.path());
let ref_dir = ref_root.path().join("presence");
pack_bit_matrix(&ref_dir).unwrap();
let dense = PersistentBitMatrix::open(ref_root.path()).unwrap();
PersistentSparseBitMatrixBuilder::build_from_dense(&dense, &ref_dir).unwrap().close().unwrap();
drop(dense);
let expected = sparse_files(&ref_dir);
assert!(!expected.is_empty());
for &(window, max_hits) in WINDOWS {
let root = tempdir().unwrap();
build_bits(root.path());
let dir = root.path().join("presence");
let meta = MatrixMeta::load(&dir).unwrap();
PersistentSparseBitMatrixBuilder::build_from_columnar_with(&dir, &meta, window, max_hits)
.unwrap()
.close()
.unwrap();
assert_eq!(sparse_files(&dir), expected, "window={window} max_hits={max_hits}");
}
}
#[test]
fn int_columnar_transpose_is_byte_identical_to_the_dense_route() {
let ref_root = tempdir().unwrap();
build_ints(ref_root.path());
let ref_dir = ref_root.path().join("counts");
pack_compact_int_matrix(&ref_dir).unwrap();
let dense = PersistentIntMatrix::open(ref_root.path()).unwrap();
PersistentSparseCompactIntMatrixBuilder::build_from_dense(&dense, &ref_dir).unwrap().close().unwrap();
drop(dense);
let expected = sparse_files(&ref_dir);
assert!(!expected.is_empty());
for &(window, max_hits) in WINDOWS {
let root = tempdir().unwrap();
build_ints(root.path());
let dir = root.path().join("counts");
let meta = MatrixMeta::load(&dir).unwrap();
PersistentSparseCompactIntMatrixBuilder::build_from_columnar_with(&dir, &meta, window, max_hits)
.unwrap()
.close()
.unwrap();
assert_eq!(sparse_files(&dir), expected, "window={window} max_hits={max_hits}");
}
}
#[test]
fn columnar_transpose_rejects_a_column_of_another_length() {
let root = tempdir().unwrap();
build_bits(root.path());
let dir = root.path().join("presence");
let mut meta = MatrixMeta::load(&dir).unwrap();
meta.n += 64; // metadata claims more rows than the column files hold
let err = PersistentSparseBitMatrixBuilder::build_from_columnar(&dir, &meta).err().expect("must fail");
assert_eq!(err.kind(), std::io::ErrorKind::InvalidData);
}
+71
View File
@@ -0,0 +1,71 @@
use crate::views::BitSliceView;
/// Deterministic pseudo-random words (xorshift), thinned so that many words
/// are zero: the case the word-wise scan has to skip correctly.
fn words(n_words: usize, seed: u64, thin: u32) -> Vec<u64> {
let mut s = seed | 1;
(0..n_words)
.map(|_| {
let mut w = u64::MAX;
for _ in 0..thin {
s ^= s << 13;
s ^= s >> 7;
s ^= s << 17;
w &= s;
}
w
})
.collect()
}
fn naive(v: &BitSliceView<'_>, r: std::ops::Range<usize>) -> Vec<usize> {
r.filter(|&s| v.get(s)).collect()
}
fn scanned(v: &BitSliceView<'_>, r: std::ops::Range<usize>) -> Vec<usize> {
let mut out = Vec::new();
v.for_each_set_in(r, |s| out.push(s));
out
}
#[test]
fn for_each_set_in_matches_get_loop_on_all_boundaries() {
// Lengths around word boundaries; padding bits of the last word are zero.
for n in [0usize, 1, 63, 64, 65, 127, 128, 129, 200, 4096, 4097] {
for (seed, thin) in [(1u64, 0u32), (7, 1), (11, 3), (13, 6)] {
let mut w = words(n.div_ceil(64), seed, thin);
if n % 64 != 0 {
if let Some(last) = w.last_mut() {
*last &= (1u64 << (n % 64)) - 1;
}
}
let v = BitSliceView::new(&w, n);
let cuts = [0, 1, 63, 64, 65, 100, 127, 128, 129, n / 2, n.saturating_sub(1), n];
for &a in &cuts {
for &b in &cuts {
if a <= n && b <= n {
assert_eq!(scanned(&v, a..b), naive(&v, a..b), "n={n} seed={seed} thin={thin} range={a}..{b}");
}
}
}
}
}
}
#[test]
fn for_each_set_in_empty_and_reversed_ranges_do_nothing() {
let w = [u64::MAX; 2];
let v = BitSliceView::new(&w, 128);
assert!(scanned(&v, 5..5).is_empty());
#[allow(clippy::reversed_empty_ranges)]
let rev = 9..3;
assert!(scanned(&v, rev).is_empty());
}
#[test]
#[should_panic]
fn for_each_set_in_rejects_range_past_the_view() {
let w = [u64::MAX; 2];
let v = BitSliceView::new(&w, 100);
scanned(&v, 0..101);
}
+10
View File
@@ -129,6 +129,16 @@ pub trait CountPartials: ColumnWeights {
jaccard_to_mash(&self.threshold_jaccard_dist_matrix(threshold), k) jaccard_to_mash(&self.threshold_jaccard_dist_matrix(threshold), k)
} }
/// Hamming distance on presence/absence: the number of k-mers present
/// (count ≥ `threshold`) in exactly one of the two genomes, i.e. the union
/// minus the intersection that [`partial_threshold_jaccard`] already gives.
///
/// [`partial_threshold_jaccard`]: Self::partial_threshold_jaccard
fn threshold_hamming_dist_matrix(&self, threshold: u32) -> Array2<u64> {
let (inter, union) = self.partial_threshold_jaccard(threshold);
union - inter
}
fn relfreq_bray_dist_matrix(&self) -> Array2<f64> { fn relfreq_bray_dist_matrix(&self) -> Array2<f64> {
let global = self.col_weights(); let global = self.col_weights();
let mut m = self.partial_relfreq_bray(&global).mapv(|v| 1.0 - v); let mut m = self.partial_relfreq_bray(&global).mapv(|v| 1.0 - v);
+136
View File
@@ -0,0 +1,136 @@
//! Windowed column → row transposition, shared by the presence and count
//! packers.
//!
//! A columnar matrix stores one file per genome; the sparse formats are
//! built row by row, each row needing the cells of *every* column. Holding
//! every column mapped (or open) at once costs one mapping (or descriptor)
//! per column and per worker. This module never holds more than one column:
//! it cuts the rows into windows and, for each window, visits the columns
//! one after the other through a caller-supplied reader, collecting the
//! non-zero cells; the cells are then regrouped by row and handed to the
//! caller. No intermediate dense matrix is written.
//!
//! Memory is bounded by `max_hits`, not by the partition size: when a window
//! turns out denser than expected it is abandoned and retried with half as
//! many rows.
use std::io;
/// Default rows per window. A multiple of 64, so that a bit column's window
/// starts on a word boundary. 2 M rows = 256 KiB per bit column and read.
pub const DEFAULT_WINDOW_ROWS: usize = 1 << 21;
/// Default cap on the non-zero cells held for one window.
pub const DEFAULT_MAX_HITS: usize = 4 << 20;
/// Smallest window the retry logic shrinks to.
const MIN_WINDOW_ROWS: usize = 64;
/// One non-zero cell of a window: row relative to the window start, column,
/// value. `V = ()` for presence matrices.
pub(crate) type Hit<V> = (u32, u32, V);
/// The rows of one window, regrouped by row, columns ascending within a row.
pub(crate) struct RowBatch<V> {
offsets: Vec<u32>,
cols: Vec<u32>,
vals: Vec<V>,
}
impl<V: Copy> RowBatch<V> {
fn new() -> Self {
Self { offsets: Vec::new(), cols: Vec::new(), vals: Vec::new() }
}
/// Number of rows in the window.
pub(crate) fn len(&self) -> usize {
self.offsets.len().saturating_sub(1)
}
/// Columns of row `i`, ascending.
pub(crate) fn cols(&self, i: usize) -> &[u32] {
&self.cols[self.offsets[i] as usize..self.offsets[i + 1] as usize]
}
/// Values of row `i`, in the order of [`cols`](Self::cols).
pub(crate) fn vals(&self, i: usize) -> &[V] {
&self.vals[self.offsets[i] as usize..self.offsets[i + 1] as usize]
}
/// Stable counting sort of `hits` by row. `hits` was produced column by
/// column in ascending column order, so stability leaves each row's
/// columns ascending, as `push_row` requires, without a per-row sort.
fn fill(&mut self, rows: usize, hits: &[Hit<V>], zero: V) {
self.offsets.clear();
self.offsets.resize(rows + 1, 0);
for &(r, _, _) in hits {
self.offsets[r as usize + 1] += 1;
}
for i in 0..rows {
self.offsets[i + 1] += self.offsets[i];
}
self.cols.clear();
self.cols.resize(hits.len(), 0);
self.vals.clear();
self.vals.resize(hits.len(), zero);
// `next[r]` = write position of the next cell of row `r`; reuse the
// tail of `offsets` would alias, so keep a small separate cursor.
let mut next: Vec<u32> = self.offsets[..rows].to_vec();
for &(r, c, v) in hits {
let p = next[r as usize] as usize;
self.cols[p] = c;
self.vals[p] = v;
next[r as usize] += 1;
}
}
}
/// Transposes an `n`-row, `n_cols`-column matrix window by window.
///
/// `read_col(c, start, end, hits)` must append to `hits` every non-zero cell
/// of column `c` in rows `start..end`, as `(row - start, c, value)`.
/// `on_batch(start, batch)` receives each completed window, in ascending row
/// order; `batch` row `i` is matrix row `start + i`.
///
/// `window_rows` is rounded down to a multiple of 64 (minimum 64), and `start`
/// is always a multiple of 64.
pub(crate) fn transpose_windows<V: Copy>(
n: usize,
n_cols: usize,
window_rows: usize,
max_hits: usize,
zero: V,
mut read_col: impl FnMut(usize, usize, usize, &mut Vec<Hit<V>>) -> io::Result<()>,
mut on_batch: impl FnMut(usize, &RowBatch<V>) -> io::Result<()>,
) -> io::Result<()> {
let window_max = (window_rows / 64 * 64).clamp(MIN_WINDOW_ROWS, u32::MAX as usize / 2);
// `Hit` rows are u32 and offsets count hits in u32.
let max_hits = max_hits.min(u32::MAX as usize / 2);
let mut w = window_max;
let mut hits: Vec<Hit<V>> = Vec::new();
let mut batch = RowBatch::new();
let mut start = 0usize;
while start < n {
let end = (start + w).min(n);
hits.clear();
let mut too_dense = false;
for c in 0..n_cols {
read_col(c, start, end, &mut hits)?;
if hits.len() > max_hits && w > MIN_WINDOW_ROWS {
too_dense = true;
break;
}
}
if too_dense {
w = (w / 2 / 64 * 64).max(MIN_WINDOW_ROWS);
continue;
}
batch.fill(end - start, &hits, zero);
on_batch(start, &batch)?;
start = end;
if hits.len() < max_hits / 4 && w < window_max {
w = (w * 2).min(window_max);
}
}
Ok(())
}
+48
View File
@@ -0,0 +1,48 @@
//! Word-wise scans over a [`BitSliceView`]. Kept apart from `views.rs`, which
//! is already past the size budget; these are inherent methods of the same type.
use std::ops::Range;
use crate::views::BitSliceView;
impl BitSliceView<'_> {
/// Calls `f(slot)` for every set bit with `slot` in `range`, in ascending
/// order.
///
/// Walks the underlying `u64` words and jumps from set bit to set bit
/// with `trailing_zeros`, so the cost is proportional to the number of
/// words plus the number of set bits — not to the number of slots, as a
/// `get(slot)` loop would be. Words that are entirely zero cost one load
/// and one test, which is what makes this the right traversal for the
/// sparse presence columns `build_from_dense` reads.
///
/// Panics if `range.end > self.len()`.
#[inline]
pub fn for_each_set_in(&self, range: Range<usize>, mut f: impl FnMut(usize)) {
assert!(range.end <= self.n, "range end {} past view length {}", range.end, self.n);
if range.start >= range.end {
return;
}
let first = range.start >> 6;
let last = (range.end - 1) >> 6;
for w in first..=last {
let mut word = self.words[w];
if word == 0 {
continue;
}
if w == first {
word &= !0u64 << (range.start & 63);
}
if w == last {
let tail = range.end & 63;
if tail != 0 {
word &= (1u64 << tail) - 1;
}
}
while word != 0 {
f((w << 6) | word.trailing_zeros() as usize);
word &= word - 1;
}
}
}
}
-2
View File
@@ -5,11 +5,9 @@ edition = "2021"
[dependencies] [dependencies]
obikseq = { path = "../obikseq" } obikseq = { path = "../obikseq" }
obifastwrite = { path = "../obifastwrite" }
hashbrown = { version = "0.17", features = ["rayon"] } hashbrown = { version = "0.17", features = ["rayon"] }
rayon = "1" rayon = "1"
crossbeam-channel = "0.5" crossbeam-channel = "0.5"
tracing = "0.1"
[features] [features]
test-utils = [] test-utils = []
+1 -1
View File
@@ -81,7 +81,7 @@ impl IndexDump for Arc<KmerIndex> {
if remaining.load(Ordering::Relaxed) == 0 { return Ok(vec![]); } if remaining.load(Ordering::Relaxed) == 0 { return Ok(vec![]); }
let mut buf = Vec::<u8>::new(); let mut buf = Vec::<u8>::new();
let try_write = |buf: &mut Vec<u8>, row: &[u32], prefix: &str| -> bool { let try_write = |buf: &mut Vec<u8>, row: &[u32], prefix: &str| -> bool {
match remaining.fetch_update(Ordering::SeqCst, Ordering::SeqCst, |cur| { match remaining.try_update(Ordering::SeqCst, Ordering::SeqCst, |cur| {
if cur > 0 { Some(cur - 1) } else { None } if cur > 0 { Some(cur - 1) } else { None }
}) { }) {
Err(_) => false, Err(_) => false,
-4
View File
@@ -5,13 +5,9 @@ edition = "2024"
[dependencies] [dependencies]
obikseq = { path = "../obikseq" } obikseq = { path = "../obikseq" }
obitaxonomy = { path = "../obitaxonomy" }
obiskio = { path = "../obiskio" } obiskio = { path = "../obiskio" }
obisys = { path = "../obisys", default-features = false } obisys = { path = "../obisys", default-features = false }
obicompactvec = { path = "../obicompactvec" } obicompactvec = { path = "../obicompactvec" }
obidebruinj = { path = "../obidebruinj" }
obipipeline = { path = "../obipipeline" }
obikentropy = { path = "../obikentropy" }
cacheline-ef = "2.0" cacheline-ef = "2.0"
epserde = "0.13" epserde = "0.13"
ptr_hash = { version = "2.1", features = ["epserde"] } ptr_hash = { version = "2.1", features = ["epserde"] }
+77 -3
View File
@@ -280,7 +280,7 @@ impl KmerIndex {
/// on-disk formats (`PersistentSparseBitMatrix` for presence, /// on-disk formats (`PersistentSparseBitMatrix` for presence,
/// `PersistentSparseCompactIntMatrix` for counts) — smaller and faster /// `PersistentSparseCompactIntMatrix` for counts) — smaller and faster
/// for single-row access on real, sparse data; column-oriented access /// for single-row access on real, sparse data; column-oriented access
/// (`--metric` distance matrices) is much slower on the sparse format. /// (`phylo --distance` matrices) is much slower on the sparse format.
/// ///
/// Public — and not moved out of this crate, despite covering the same /// Public — and not moved out of this crate, despite covering the same
/// "maintenance operation on an already-built index" territory as /// "maintenance operation on an already-built index" territory as
@@ -296,7 +296,14 @@ impl KmerIndex {
}; };
let n = self.n_partitions(); let n = self.n_partitions();
let order: Vec<usize> = (0..n).collect(); // Largest partitions first (LFD), so the long jobs do not end up as
// stragglers at the tail. Sizes are `stat()` only; the sort is stable,
// ties keep partition-number order. Partitions are independent, so
// the order has no effect on what is written.
let sizes: Vec<u64> = (0..n).map(|i| self.pending_matrix_bytes(i)).collect();
let mut order: Vec<usize> = (0..n).collect();
order.sort_by_key(|&i| std::cmp::Reverse(sizes[i]));
let mut timings: Vec<(usize, u64, std::time::Duration)> = Vec::with_capacity(n);
let pb = progress_bar("pack", n as u64, "partitions"); let pb = progress_bar("pack", n as u64, "partitions");
obisys::PartitionRunner::new().run( obisys::PartitionRunner::new().run(
&order, &order,
@@ -327,14 +334,50 @@ impl KmerIndex {
} }
Ok(()) Ok(())
}, },
|_, _, _| { |i, _, elapsed| {
tracing::debug!(
partition = i,
matrix_bytes = sizes[i],
secs = elapsed.as_secs_f64(),
"pack: partition done"
);
timings.push((i, sizes[i], elapsed));
pb.inc(1); pb.inc(1);
}, },
)?; )?;
pb.finish_and_clear(); pb.finish_and_clear();
report_pack_timings(&timings);
Ok(()) Ok(())
} }
/// Bytes of matrix data (`presence/` and `counts/` of every layer) still
/// present in partition `i`, from `stat()` only. Before packing this is
/// the columnar data, i.e. a proxy for the work `pack_matrices` has to do
/// on that partition. Missing partition or directories count as 0.
pub fn pending_matrix_bytes(&self, i: usize) -> u64 {
let Ok(n_layers) = self.n_layers(i) else {
return 0;
};
let mut total = 0u64;
for l in 0..n_layers {
let Ok(layer_dir) = self.layer_dir(i, l) else {
continue;
};
for sub in ["presence", "counts"] {
let Ok(entries) = fs::read_dir(layer_dir.join(sub)) else {
continue;
};
total += entries
.flatten()
.filter_map(|e| e.metadata().ok())
.filter(|m| m.is_file())
.map(|m| m.len())
.sum::<u64>();
}
}
total
}
// /// Write a `layer_meta.json` in any layer directory that is missing one. // /// Write a `layer_meta.json` in any layer directory that is missing one.
// /// // ///
// /// Old indexes were built before this file was required. The number of // /// Old indexes were built before this file was required. The number of
@@ -394,3 +437,34 @@ impl KmerIndex {
// Ok(()) // Ok(())
// } // }
} }
/// One `info!` summary of `pack_matrices` per-partition timings: the sum of
/// partition times (divide by the stage wall time to get the mean number of
/// concurrently busy workers), the median and maximum, and the slowest
/// partitions with their pre-pack matrix size. `timings` holds
/// `(partition, matrix_bytes, elapsed)`.
fn report_pack_timings(timings: &[(usize, u64, std::time::Duration)]) {
if timings.is_empty() {
return;
}
let mut by_time: Vec<&(usize, u64, std::time::Duration)> = timings.iter().collect();
by_time.sort_by_key(|t| std::cmp::Reverse(t.2));
let sum: f64 = timings.iter().map(|t| t.2.as_secs_f64()).sum();
let bytes: u64 = timings.iter().map(|t| t.1).sum();
let median = by_time[by_time.len() / 2].2.as_secs_f64();
tracing::info!(
"pack: {} partition(s), sum of partition times {:.0}s, median {:.1}s, max {:.1}s, {:.1} GB of matrix data",
timings.len(),
sum,
median,
by_time[0].2.as_secs_f64(),
bytes as f64 / 1e9,
);
for (i, b, d) in by_time.iter().take(5).map(|t| (t.0, t.1, t.2)) {
tracing::info!(
"pack: slowest partition {i}: {:.1}s, {:.2} GB of matrix data",
d.as_secs_f64(),
b as f64 / 1e9,
);
}
}
@@ -135,6 +135,12 @@ impl PrivateBuilder for KmerIndex {
mode: &IndexMode, mode: &IndexMode,
block_bits: u8, block_bits: u8,
) -> Result<usize, SKError> { ) -> Result<usize, SKError> {
// A partition that received no k-mer has no layer 0 on disk at all
// (small inputs against many partitions): nothing to build, not an
// error.
if self.partition(i).map_err(|e| io::Error::other(e.to_string()))?.n_layers() == 0 {
return Ok(0);
}
let layer0 = self let layer0 = self
.layer0(i) .layer0(i)
.map_err(|e| io::Error::other(e.to_string()))?; .map_err(|e| io::Error::other(e.to_string()))?;
@@ -208,6 +214,10 @@ impl PrivateBuilder for KmerIndex {
} }
fn remove_build_artifacts(&self, i: usize) { fn remove_build_artifacts(&self, i: usize) {
// No layer 0 = empty partition: no build artifacts to remove.
if self.partition(i).map_or(false, |p| p.n_layers() == 0) {
return;
}
let layer0 = match self.layer0(i) { let layer0 = match self.layer0(i) {
Ok(layer) => layer, Ok(layer) => layer,
Err(e) => { Err(e) => {
+8 -2
View File
@@ -1,6 +1,6 @@
[package] [package]
name = "obikmer" name = "obikmer"
version = "1.4.1" version = "1.4.5"
edition = "2024" edition = "2024"
[[bin]] [[bin]]
@@ -24,7 +24,6 @@ obikstats = { path = "../obikstats" }
obikquery = { path = "../obikquery" } obikquery = { path = "../obikquery" }
obikidxcache = { path = "../obikidxcache" } obikidxcache = { path = "../obikidxcache" }
obikphylo = { path = "../obikphylo" } obikphylo = { path = "../obikphylo" }
obiksession = { path = "../obiksession" }
obikrope = { path = "../obikrope" } obikrope = { path = "../obikrope" }
obifastwrite = { path = "../obifastwrite" } obifastwrite = { path = "../obifastwrite" }
obiskbuilder = { path = "../obiskbuilder" } obiskbuilder = { path = "../obiskbuilder" }
@@ -37,6 +36,13 @@ serde_yaml = "0.9"
tracing = "0.1.44" tracing = "0.1.44"
tracing-subscriber = { version = "0.3", features = ["fmt", "env-filter"] } tracing-subscriber = { version = "0.3", features = ["fmt", "env-filter"] }
# musl's built-in allocator fragments badly under the merge's multi-threaded
# allocation pattern (map count and RSS drift, see DevDocMD/implementation/
# merge_mmap_audit.md). Static musl builds use jemalloc instead; glibc and
# macOS builds keep the system allocator.
[target.'cfg(target_env = "musl")'.dependencies]
tikv-jemallocator = "0.5.4"
[features] [features]
default = ["numa"] default = ["numa"]
numa = ["obisys/numa"] numa = ["obisys/numa"]
+1 -1
View File
@@ -19,7 +19,7 @@ pub struct PackArgs {
/// Pack presence and count matrices into the dense on-disk format instead /// Pack presence and count matrices into the dense on-disk format instead
/// of the default sparse, deduplicated one. Dense is faster for /// of the default sparse, deduplicated one. Dense is faster for
/// column-oriented access (`--metric` distance matrices); sparse is /// column-oriented access (`phylo --distance` matrices); sparse is
/// smaller and faster for single-row access on real, sparse data. /// smaller and faster for single-row access on real, sparse data.
#[arg(long, default_value_t = false)] #[arg(long, default_value_t = false)]
pub dense: bool, pub dense: bool,
@@ -0,0 +1,51 @@
//! `--pseudo-alignment` / `--fasta` / `--vcf` file output.
use std::io::{BufWriter, Write};
use obikphylo::siblings::SnpAlignment;
use tracing::info;
use super::args::PhyloArgs;
fn create(path: &str) -> BufWriter<std::fs::File> {
BufWriter::new(std::fs::File::create(path).unwrap_or_else(|e| {
eprintln!("error creating {path}: {e}");
std::process::exit(1);
}))
}
/// Writes the drawn alignment as FASTA (default, or `--fasta`) and/or VCF
/// (`--vcf`) next to `-o <prefix>`.
pub(super) fn write_alignment_outputs(
alignment: &SnpAlignment,
labels: &[String],
args: &PhyloArgs,
) {
let prefix = args.output.as_ref().map(|p| p.display().to_string());
let path_for = |suffix: &str| match &prefix {
Some(p) => format!("{p}_alignment.{suffix}"),
None => format!("alignment.{suffix}"),
};
let n_sites = alignment.sequences.first().map_or(0, Vec::len);
let n_genomes = alignment.genome_indices.len();
if args.align.fasta || !args.align.vcf {
let path = path_for("fasta");
let mut f = create(&path);
for (&g, seq) in alignment.genome_indices.iter().zip(alignment.sequences.iter()) {
obifastwrite::write_plain_record(seq, &labels[g], &mut f).unwrap();
}
info!("pseudo-alignment ({n_sites} site(s), {n_genomes} genome(s)) → {path}");
}
if args.align.vcf {
let path = path_for("vcf");
let mut f = create(&path);
let written = alignment.write_vcf(labels, !args.sample.no_ambiguity, &mut f).unwrap_or_else(|e| {
eprintln!("error writing {path}: {e}");
std::process::exit(1);
});
f.flush().unwrap();
info!("pseudo-alignment VCF ({written}/{n_sites} site(s), {n_genomes} genome(s)) → {path}");
}
}
-506
View File
@@ -1,506 +0,0 @@
use std::path::PathBuf;
use clap::Args;
use obikphylo::{DistanceMetric, GammaShape, SnpDistanceKind};
/// `--gamma-shape` CLI value: a fixed `alpha`, or `auto`/`estimate` to fit
/// it from the data (see [`GammaShape::Auto`]). Kept as its own tiny type
/// (rather than parsing straight to `GammaShape`) so the CLI layer stays
/// free of `GammaShape::Disabled`, which the `Option<GammaShapeArg>` around
/// this already expresses on its own (`None` = flag not given at all).
#[derive(Debug, Clone, Copy)]
pub enum GammaShapeArg {
Fixed(f64),
Auto,
}
impl From<GammaShapeArg> for GammaShape {
fn from(arg: GammaShapeArg) -> Self {
match arg {
GammaShapeArg::Fixed(a) => GammaShape::Fixed(a),
GammaShapeArg::Auto => GammaShape::Auto,
}
}
}
fn parse_gamma_shape(s: &str) -> Result<GammaShapeArg, String> {
if s.eq_ignore_ascii_case("auto") || s.eq_ignore_ascii_case("estimate") {
return Ok(GammaShapeArg::Auto);
}
s.parse::<f64>()
.map(GammaShapeArg::Fixed)
.map_err(|_| format!("invalid --gamma-shape value {s:?}: expected a positive number or \"auto\"/\"estimate\""))
}
/// `--n-sites` CLI value: a number of sites to draw, or `all` for every site
/// — the draw then keeps every variable family that passes the other filters,
/// with no sampling involved. Its own tiny type (like [`GammaShapeArg`]) so
/// the two cases stay distinct up to the call that needs the target, instead
/// of an in-band sentinel (a negative number, `0`, ...) that a typo could hit.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum NSitesArg {
All,
N(usize),
}
impl NSitesArg {
/// The target to hand to the sampler: `None` = exhaustive.
pub fn target(self) -> Option<usize> {
match self {
NSitesArg::All => None,
NSitesArg::N(n) => Some(n),
}
}
}
fn parse_n_sites(s: &str) -> Result<NSitesArg, String> {
if s.eq_ignore_ascii_case("all") {
return Ok(NSitesArg::All);
}
s.parse::<usize>()
.map(NSitesArg::N)
.map_err(|_| format!("invalid --n-sites value {s:?}: expected a non-negative integer or \"all\""))
}
/// `--distance` value — either one of `obikphylo::DistanceMetric`'s
/// whole-index metrics (routed to `IndexCache::distance`) or one of
/// `obikphylo::SnpDistanceKind`'s `snp-*` corrections (routed to
/// `obikphylo::siblings::distance_matrix`, over a tally built by
/// `sampled_scan`/`PairwiseReducer`, the sibling-annex pipeline) — two
/// genuinely different code paths behind one CLI vocabulary, see
/// `DevDocMD/theory/evolutionary_distances.md`, "`--distance` unification".
#[derive(clap::ValueEnum, Clone, Copy, Debug)]
pub enum DistanceArg {
Jaccard,
Mash,
Hamming,
BrayCurtis,
#[value(name = "relfreq-bray-curtis")]
RelfreqBrayCurtis,
Euclidean,
#[value(name = "relfreq-euclidean")]
RelfreqEuclidean,
Hellinger,
#[value(name = "hellinger-euclidean")]
HellingerEuclidean,
#[value(name = "snp-raw")]
SnpRaw,
#[value(name = "snp-jc")]
SnpJc,
#[value(name = "snp-k2p")]
SnpK2p,
#[value(name = "snp-k81")]
SnpK81,
#[value(name = "snp-f81")]
SnpF81,
#[value(name = "snp-t92")]
SnpT92,
#[value(name = "snp-tn93")]
SnpTn93,
#[value(name = "snp-tv")]
SnpTv,
}
/// The distance matrix this invocation must compute, if any: the one asked for
/// with `--distance`, else `jaccard` when `--nj`, `--upgma` or
/// `--shared-kmers` needs a matrix to work from, else none — a matrix is never
/// computed (nor written) just because nothing said otherwise.
pub fn distance_to_compute(
distance: Option<DistanceArg>,
nj: bool,
upgma: bool,
shared_kmers: bool,
) -> Option<DistanceArg> {
distance.or_else(|| (nj || upgma || shared_kmers).then_some(DistanceArg::Jaccard))
}
#[cfg(test)]
mod distance_tests {
use super::*;
#[test]
fn nothing_asked_nothing_computed() {
assert!(distance_to_compute(None, false, false, false).is_none());
}
#[test]
fn tree_and_shared_kmer_outputs_imply_jaccard() {
for (nj, upgma, shared) in [(true, false, false), (false, true, false), (false, false, true)] {
assert!(matches!(
distance_to_compute(None, nj, upgma, shared),
Some(DistanceArg::Jaccard)
));
}
}
#[test]
fn an_explicit_distance_wins() {
assert!(matches!(
distance_to_compute(Some(DistanceArg::Mash), true, true, true),
Some(DistanceArg::Mash)
));
assert!(matches!(
distance_to_compute(Some(DistanceArg::SnpJc), false, false, false),
Some(DistanceArg::SnpJc)
));
}
}
impl DistanceArg {
/// `Some` for the whole-index metrics, `None` for `snp-*` values.
pub fn as_classic(self) -> Option<DistanceMetric> {
Some(match self {
DistanceArg::Jaccard => DistanceMetric::Jaccard,
DistanceArg::Mash => DistanceMetric::Mash,
DistanceArg::Hamming => DistanceMetric::Hamming,
DistanceArg::BrayCurtis => DistanceMetric::BrayCurtis,
DistanceArg::RelfreqBrayCurtis => DistanceMetric::RelfreqBrayCurtis,
DistanceArg::Euclidean => DistanceMetric::Euclidean,
DistanceArg::RelfreqEuclidean => DistanceMetric::RelfreqEuclidean,
DistanceArg::Hellinger => DistanceMetric::Hellinger,
DistanceArg::HellingerEuclidean => DistanceMetric::HellingerEuclidean,
_ => return None,
})
}
/// `Some` for the `snp-*` values, `None` for the whole-index metrics.
pub fn as_snp(self) -> Option<SnpDistanceKind> {
Some(match self {
DistanceArg::SnpRaw => SnpDistanceKind::Raw,
DistanceArg::SnpJc => SnpDistanceKind::Jc,
DistanceArg::SnpK2p => SnpDistanceKind::K2p,
DistanceArg::SnpK81 => SnpDistanceKind::K81,
DistanceArg::SnpF81 => SnpDistanceKind::F81,
DistanceArg::SnpT92 => SnpDistanceKind::T92,
DistanceArg::SnpTn93 => SnpDistanceKind::Tn93,
DistanceArg::SnpTv => SnpDistanceKind::Tv,
_ => return None,
})
}
}
/// Genome-vs-genome distance computation: the whole-index `--distance` path
/// (classic metrics + `snp-*` corrections/NJ/UPGMA), annex construction
/// (`--sibling-annex`), annex diagnostics (`--sibling-stats`,
/// `--sibling-hist`), entropy reporting (`--shannon`), SNP pseudo-alignment
/// sampling (`--pseudo-alignment`, `--n-sites`, `--free-loss`,
/// `--no-ambiguity`, `--entropy`/`--entropy-sd`), Sankoff cost-matrix
/// calibration (`--sankoff`, `--sankoff-ratio-ceiling`) and its TNT/PhyG/
/// IQ-TREE exports (`--tnt`, `--phyg`, `--iqtree`/`--iqtree-min-freq`,
/// `--sankoff-cost-scale`), and Family Overlap (`--family-overlap`,
/// `--min-shared-family`).
#[derive(Args)]
pub struct PhyloArgs {
/// Index directory
pub index: PathBuf,
/// Exclude a genome (by its exact label) — from `--pseudo-alignment`'s
/// sampling (a family whose only polymorphism lived in an excluded
/// genome is discarded during sampling, not filtered afterward — see
/// `obikphylo::siblings::sampled_scan`'s own docs) and from the
/// distance matrix / shared-kmer matrix CSV output (row and column
/// both dropped; the underlying computation itself is unaffected).
/// Repeatable.
#[arg(long = "exclude-genome", value_name = "LABEL")]
pub exclude_genome: Vec<String>,
/// Auto-exclude any genome whose mean shared-variable-family count
/// against every other genome (`FamilyOverlap::mean_row` — the same
/// per-row statistic `--family-overlap`'s own matrix shows) falls below
/// this threshold — same exclusion machinery as `--exclude-genome`,
/// applied on top of it rather than instead of it. Applies to the
/// `snp-*` `--distance`/`--pseudo-alignment`/`--sankoff` computations
/// below (all sibling-annex-based); does *not* affect the whole-index
/// `--distance` metrics (jaccard, hamming, bray-curtis, ...) or their
/// matrix/NJ/UPGMA output — a genome with too little SNP-family
/// coverage to trust is a different concern from one whose plain k-mer
/// profile is simply divergent. Requires the Family Overlap annex
/// (built on demand if missing, same as every other annex here — see
/// `obikphylo::siblings::extensions::SiblingExt::family_overlap`'s own
/// docs).
#[arg(long, value_name = "N")]
pub min_shared_family: Option<f64>,
/// Build (or rebuild) the sibling-count/minorant annex — independent of
/// the distance metric below, meant to be run routinely, ahead of any
/// SNP-family distance computation that will later consume it.
#[arg(long)]
pub sibling_annex: bool,
/// Tally the sibling-count distribution (CSV) of an already-built annex
/// (run with `--sibling-annex` first, in this invocation or an earlier
/// one). A separate, occasional diagnostic pass — not run every time the
/// annex itself is (re)built.
#[arg(long)]
pub sibling_stats: bool,
/// Print just the global family-size histogram (1-4 members) of an
/// already-built annex — the `global` row `--sibling-stats` also
/// writes, but without the per-genome breakdown, so it skips
/// `--sibling-stats`'s cross-partition resolution entirely (annex bits
/// only).
#[arg(long)]
pub sibling_hist: bool,
/// Write the Family Overlap matrix (CSV) — number of shared *variable*
/// families per genome pair, index-wide (`obikphylo::siblings::FamilyOverlap`).
/// Built on demand if missing (same as every other annex here); no
/// `--sibling-annex` prerequisite beyond that. Every genome is written,
/// unfiltered by `--exclude-genome`/`--min-shared-family` — a raw
/// coverage diagnostic, not a computation those exclusions are meant to
/// protect.
#[arg(long)]
pub family_overlap: bool,
/// Write a per-family Shannon entropy report (CSV) — requires an
/// already-built sibling annex (`--sibling-annex` first, in this
/// invocation or an earlier one). Always a full, unsampled scan of
/// every family (`--n-sites`/`--entropy`/`--entropy-sd` below only
/// apply to `--pseudo-alignment`, not this).
#[arg(long)]
pub shannon: bool,
/// Write the same entropy scan as `--shannon` as a distribution
/// (`<prefix>_shannon_summary.csv`): the number of families per entropy
/// kind, number of genomes carrying the family, number of those without
/// ambiguity and 0.05-bit entropy bin — a few hundred thousand lines at
/// most, whatever the size of the index, instead of one line per family.
/// Requires an already-built sibling annex, like `--shannon`.
#[arg(long)]
pub shannon_summary: bool,
/// Write a SNP-only pseudo-alignment (FASTA) — requires
/// `--n-sites <N|all>` and an already-built sibling annex
/// (`--sibling-annex` first, in this invocation or an earlier one).
#[arg(long)]
pub pseudo_alignment: bool,
/// Number of variable sites to draw index-wide, or `all` for every site
/// that passes the other filters (no sampling then). Mandatory for
/// `--pseudo-alignment`/`--sankoff`; optional for a `snp-*`
/// `--distance` value, where omitting it also means every site. A
/// number is a target, not a guarantee (proportional per-layer
/// sampling; see `obikphylo::siblings::sampled_scan`'s own docs).
#[arg(long = "n-sites", value_name = "N|all", value_parser = parse_n_sites)]
pub n_sites: Option<NSitesArg>,
/// In `--pseudo-alignment`, treat a genome carrying none of a family's
/// observed members (`∅`) as missing data (`?`) rather than a real
/// character state.
#[arg(long)]
pub free_loss: bool,
/// In `--pseudo-alignment`, treat a genome carrying more than one
/// member of a family (ambiguous) as missing data (`?`) rather than an
/// IUPAC ambiguity code.
#[arg(long)]
pub no_ambiguity: bool,
/// In `--pseudo-alignment`/`--sankoff`/`snp-*` `--distance`, keep only
/// the families that can separate two pairs of genomes from each other
/// (a bipartition): at least two distinct called states, each carried
/// by at least two genomes. A singleton variant, or a family carried by
/// fewer than four genomes, fits every topology and tells a tree
/// nothing. Which genomes count as carriers follows `--no-ambiguity`.
#[arg(long)]
pub quartet: bool,
/// In `--pseudo-alignment`/`--sankoff`/`snp-*` `--distance`, favour the
/// families that cover the pairs of genomes sharing the fewest families
/// across the index: each family is weighted by the mean, over the pairs
/// of genomes carrying it, of 1 / (number of families the pair shares).
/// Evens out how well each pair of genomes is represented in the sample.
/// Which genomes count as carriers follows `--no-ambiguity`.
#[arg(long)]
pub pair_balance: bool,
/// Write what the draw contains, from the very sites it kept
/// (requires `--pseudo-alignment`, `--sankoff` or a `snp-*`
/// `--distance`): `<prefix>_sample_genomes.csv` (per genome, the sites it
/// has a call in and the informative sites it takes part in),
/// `<prefix>_sample_carriers.csv` (sites and informative sites per number
/// of carriers) and `<prefix>_sample_overlap.csv` (the sample's own
/// genome × genome overlap matrix, to compare with `--family-overlap`).
#[arg(long)]
pub sample_stats: bool,
/// Write how the draw was spread over the layers of the index
/// (requires `--pseudo-alignment`, `--sankoff` or a `snp-*`
/// `--distance`): `<prefix>_sample_plan.csv`, one line per layer with the
/// families eligible, their total weight, the layer's quota, the families
/// drawn and the ones kept after masking.
#[arg(long)]
pub sample_plan: bool,
/// Entropy-biased sampling target (Gaussian kernel mean) for
/// `--pseudo-alignment` — activates biasing as soon as this or
/// `--entropy-sd` is given; the other defaults to 1.0/0.5.
#[arg(long)]
pub entropy: Option<f64>,
/// Entropy-biased sampling kernel width (Gaussian standard deviation)
/// for `--pseudo-alignment` — see `--entropy`.
#[arg(long)]
pub entropy_sd: Option<f64>,
/// Persist the `snp-*` `--distance` sample (the pairwise substitution
/// tally) in `DIR` across separate `obikmer phylo` invocations: a later
/// run against the same index with the exact same
/// `--n-sites`/`--free-loss`/`--no-ambiguity`/`--quartet`/`--exclude-genome`/
/// `--min-shared-family`/`--entropy`/`--entropy-sd` restores it instead
/// of resampling. `DIR` is created if it doesn't exist. If it already
/// holds a sample built under *different* selection parameters, this
/// run is rejected with an error listing what changed, unless
/// `--session-force` is also given (which discards the old sample and
/// starts fresh under this run's parameters). Only covers a `snp-*`
/// `--distance` today — `--sankoff`/`--tnt`/`--phyg`/`--iqtree` don't
/// read or write a `--session` yet.
#[arg(long, value_name = "DIR")]
pub session: Option<PathBuf>,
/// With `--session DIR`: overwrite its saved selection parameters (and
/// discard its cached sample) instead of erroring out when this run's
/// parameters don't match. No effect without `--session`.
#[arg(long, requires = "session")]
pub session_force: bool,
/// Calibrate a 16-state Sankoff cost matrix (and its matching
/// pseudo-alignment) from an already-built sibling annex — requires
/// `--n-sites <N|all>`, and shares `--free-loss`/`--no-ambiguity`/`--quartet`/
/// `--entropy`/`--entropy-sd` with `--pseudo-alignment` (one draw, same
/// selection feeds both the alignment and every calibration tally).
#[arg(long)]
pub sankoff: bool,
/// Exclude genome pairs whose raw SNP ratio exceeds this value from the
/// base-pair (composition) calibration `--sankoff` pools — a pair this
/// close to substitution saturation carries no information about the
/// true substitution spectrum. Does *not* gate the cardinality
/// calibration (see `obikphylo::siblings::CardinalityTally`'s own
/// docs for why).
#[arg(long, default_value = "0.5")]
pub sankoff_ratio_ceiling: f64,
/// Also write <prefix>_sankoff.tnt, a ready-to-run TNT script (`proc
/// <file>;`) for the same matrix/alignment `--sankoff` computes —
/// recoded to TNT's default xread alphabet (0-9A-F only; TNT rejects
/// the wider IUPAC set `--sankoff`'s own output uses unless `nstates
/// dna` is set, which imposes TNT's own incompatible DNA encoding
/// instead) with integer-scaled costs (TNT's smatrix/cost commands
/// reject decimals). Implies `--sankoff`.
#[arg(long)]
pub tnt: bool,
/// Also write <prefix>_sankoff.tcm and <prefix>_sankoff.pg, a
/// custom-alphabet cost matrix and a ready-to-run PhyG script (`read`/
/// `search`/`report`) for the same matrix/alignment `--sankoff`
/// computes. Reuses `--sankoff`'s own `_sankoff.fasta` directly — PhyG's
/// `tcm:` alphabet is read from the matrix file itself, so the IUPAC+`0`
/// alphabet needs no recoding here, unlike `--tnt`. Implies `--sankoff`.
#[arg(long)]
pub phyg: bool,
/// Also write <prefix>_iqtree.model and <prefix>_iqtree.fasta, a
/// custom-model file and a matching recoded alignment for genuine
/// maximum-likelihood inference with IQ-TREE (`iqtree3 -s ...
/// --seqtype MORPH -m ...+ASC`) — real branch lengths, unlike
/// `--tnt`/`--phyg`'s parsimony step counts. The model is the
/// reversible `Q(i,j) = R(i,j)·π_j` construction: `R` (exchangeability,
/// symmetric) recovered from the same calibrated cost matrix
/// `--sankoff` computes, `π` the real empirical state frequencies
/// counted from the alignment. Only the states that actually occur in
/// this alignment are kept, compactly renumbered (IQ-TREE infers its
/// state count from the alignment itself, and a gap in the numbering
/// would silently misalign the model file). Implies `--sankoff`.
#[arg(long)]
pub iqtree: bool,
/// Under `--iqtree --free-loss`, also recode to `?` (the same
/// missing-data treatment as `-`) any state whose empirical frequency
/// in the alignment falls below this threshold — not just genuinely
/// absent calls. States encoding 3 or 4 simultaneously-observed central
/// bases (IUPAC `V`/`H`/`K`.../`N` for 3, `N` for 4) are rare by
/// construction and often land in exactly this low-frequency range —
/// more likely assembly/detection noise than a genuine, widely-preserved
/// multi-way polymorphism, the same "sampling failure, not true signal"
/// reasoning `--free-loss` already applies to absence. No effect
/// without `--free-loss` (there is no missing-data symbol to recode to
/// otherwise). `<prefix>_iqtree_states.csv` reports the frequency
/// actually used to decide.
#[arg(long, default_value = "0.001")]
pub iqtree_min_freq: f64,
/// Scale factor applied before rounding real-valued costs to the
/// integers both `--tnt`'s smatrix/cost commands and `--phyg`'s `tcm:`
/// matrix require. Keep this small: the total tree score is this scale
/// times the sum of per-character costs across every character, and
/// there are hints in TNT's own manual that at least some of its
/// internal accumulators are 32-bit — a large scale risks a silent
/// integer overflow (undetectable, not just a crash) far more costly
/// than the resolution a bigger factor would buy. Shared between `--tnt`
/// and `--phyg` rather than split into two flags: both scale the same
/// calibrated matrix for the same reason (integer-only cost commands).
#[arg(long, default_value = "100")]
pub sankoff_cost_scale: f64,
/// Distance matrix to compute and write — none is computed unless this
/// is given, or `--nj`/`--upgma`/`--shared-kmers` is (they need a matrix
/// and use `jaccard` then, see [`distance_to_compute`]). Either a
/// whole-index metric (`jaccard`,
/// `mash`, `hamming`, `bray-curtis`, ...) or a `snp-*` correction over
/// the central-position SNP substitution spectrum (`snp-raw`, `snp-jc`,
/// `snp-k2p`, `snp-k81`, `snp-f81`, `snp-t92`, `snp-tn93`, `snp-tv`) —
/// the latter route to a different computation entirely
/// (`obikphylo::siblings::sampled_scan` + `distance_matrix`, requires
/// `--sibling-annex` first; see
/// `DevDocMD/theory/evolutionary_distances.md`, "`--distance`
/// unification" for the full catalog and why LogDet/Tajima-Nei/F84/
/// HKY85 aren't offered yet).
#[arg(long, value_enum)]
pub distance: Option<DistanceArg>,
/// Rate-heterogeneity correction (Jin-Nei gamma shape parameter `α`)
/// for `snp-*` `--distance` values that support it
/// (`obikphylo::SnpDistanceKind::supports_gamma`: every one except
/// `snp-raw`/`snp-tv`, which have nothing to correct/are deliberately
/// uncorrected). Has no effect on the whole-index metrics. Rejected at
/// runtime if given alongside an unsupported `--distance` value.
/// Either a positive number (fixed `alpha`) or `auto`/`estimate` to fit
/// `alpha` from the data itself — method-of-moments over the
/// over-dispersion of substitution rate across partitions.
#[arg(long, value_name = "ALPHA|auto", value_parser = parse_gamma_shape)]
pub gamma_shape: Option<GammaShapeArg>,
/// Minimum count to consider a kmer present when computing Jaccard on count indexes
#[arg(long, default_value = "1")]
pub presence_threshold: u32,
/// Write the primary distance matrix as plain CSV instead of the
/// default relaxed-PHYLIP format (`n` on the first line, then one
/// `label<TAB>value...` row per genome — no 10-character label
/// truncation, unlike strict PHYLIP, not yet offered here). PHYLIP is
/// the default because it's what external NJ tools (PHYLIP `neighbor`,
/// FastME, T-REX, SplitsTree) actually read; CSV stays available for
/// scripting/inspection. Only affects the primary distance matrix —
/// `--shared-kmers` keeps its own CSV-only format regardless of this
/// flag.
#[arg(long)]
pub csv: bool,
/// Also output the shared-kmer count matrix (CSV)
#[arg(long)]
pub shared_kmers: bool,
/// Compute and write a Neighbor-Joining tree (Newick)
#[arg(long)]
pub nj: bool,
/// Compute and write a UPGMA tree (Newick)
#[arg(long)]
pub upgma: bool,
/// Output prefix: <prefix>_dist.csv, <prefix>_shared.csv, <prefix>_nj.nwk,
/// <prefix>_upgma.nwk. If omitted, the distance matrix is written to stdout.
#[arg(short, long)]
pub output: Option<PathBuf>,
}
+136
View File
@@ -0,0 +1,136 @@
use std::path::PathBuf;
use clap::Args;
use super::super::values::{DistanceArg, GammaShapeArg, parse_gamma_shape};
/// Distance estimation and tree reconstruction.
#[derive(Args)]
#[command(next_help_heading = "Distance estimation and tree reconstruction")]
pub struct DistanceArgs {
/// Distance used to compare genomes
///
/// Computes a genome-by-genome distance matrix. Two families of
/// distances are available.
///
/// Whole-genome distances, based on the k-mer content of each genome.
/// The first three compare which k-mers are present or absent (on an
/// index that stores counts, a k-mer is present when its count reaches
/// --presence-threshold):
/// jaccard Jaccard distance
/// mash Mash distance, a divergence estimate derived
/// from the Jaccard index
/// hamming number of k-mers present in one genome and
/// absent from the other
/// The others compare k-mer counts, and need an index that stores them:
/// bray-curtis Bray-Curtis dissimilarity on k-mer counts
/// relfreq-bray-curtis Bray-Curtis dissimilarity on counts divided
/// by each genome's total count
/// euclidean Euclidean distance on k-mer counts
/// relfreq-euclidean Euclidean distance on relative frequencies
/// hellinger Hellinger distance on k-mer counts
/// hellinger-euclidean Euclidean distance on square roots of
/// relative frequencies (unnormalised Hellinger)
///
/// SNP-based distances, calculated from the nucleotide differences at
/// the central position of related k-mers:
/// snp-raw proportion of differing sites, no correction
/// snp-jc Jukes-Cantor correction
/// snp-k2p Kimura two-parameter correction
/// snp-k81 Kimura three-parameter correction
/// snp-f81 Felsenstein 1981 correction
/// snp-t92 Tamura 1992 correction
/// snp-tn93 Tamura-Nei 1993 correction
/// snp-tv proportion of transversions only, no correction
/// They use every eligible variable site, or a sample drawn with
/// --n-sites.
#[arg(
long,
value_enum,
verbatim_doc_comment,
hide_possible_values = true,
value_name = "DISTANCE"
)]
pub distance: Option<DistanceArg>,
/// Minimum k-mer count to count as present
///
/// Converts counts into presence/absence for the Jaccard, Mash and
/// Hamming distances and for the shared k-mer counts, when the index
/// stores counts: a k-mer is present in a genome only when its count reaches
/// this threshold. It has no effect on indexes without counts, nor on
/// the distances that use the counts themselves.
#[arg(long, default_value = "1")]
pub presence_threshold: u32,
/// Account for different evolution rates among sites
///
/// Applies a gamma rate-heterogeneity correction to SNP-based
/// distances. A positive number is the gamma shape parameter α. `auto`
/// (or `estimate`) estimates α from the data, using the variation of
/// the substitution rate across the partitions of the index.
///
/// Available for every snp-* distance except snp-raw and snp-tv, which
/// have no evolutionary correction to refine. It has no effect on
/// whole-genome distances, and combining it with a distance that does
/// not support it is an error.
#[arg(long, value_name = "ALPHA|auto", value_parser = parse_gamma_shape)]
pub gamma_shape: Option<GammaShapeArg>,
/// Save and reuse the sample of a SNP-based distance
///
/// Saves, in this directory, the sample's tally of pairwise differences
/// between genomes, so that a later run on the same index can reuse it
/// instead of drawing again. The directory is created if needed.
///
/// The saved sample is reused only if these options are identical:
/// --n-sites, --free-loss, --no-ambiguity, --quartet, --exclude-genome,
/// --min-shared-family, --entropy and --entropy-sd. Otherwise the run
/// stops and lists the differences, unless --session-force is given.
///
/// For now only SNP-based distances use sessions; Sankoff analyses and
/// the TNT, PhyG and IQ-TREE exports neither read nor write them.
#[arg(long, value_name = "DIR")]
pub session: Option<PathBuf>,
/// Replace a saved sample made with other parameters
///
/// With --session, discards the saved sample when its parameters differ
/// from this run's, and draws a new one. Without this option, a
/// difference is an error.
#[arg(long, requires = "session")]
pub session_force: bool,
/// Write the distance matrix as CSV instead of PHYLIP
///
/// By default the main matrix is written in relaxed PHYLIP format: the
/// number of genomes on the first line, then one row per genome made of
/// its label (never truncated) and its distances to the others. CSV is
/// easier to load in scripts or spreadsheets. Only the main matrix is
/// affected: the --shared-kmers matrix is always CSV.
#[arg(long)]
pub csv: bool,
/// Also write the number of k-mers shared by each pair of genomes
///
/// Writes a genome-by-genome CSV matrix of shared k-mer counts, to
/// `PREFIX_shared.csv`.
#[arg(long)]
pub shared_kmers: bool,
/// Build a Neighbor-Joining tree
///
/// Builds the tree from the distance matrix and writes it in Newick
/// format to `PREFIX_nj.nwk`. If --distance is not given, the Jaccard
/// distance is used.
#[arg(long)]
pub nj: bool,
/// Build a UPGMA tree
///
/// Builds a tree with the Unweighted Pair Group Method with Arithmetic
/// Mean and writes it in Newick format to `PREFIX_upgma.nwk`. If
/// --distance is not given, the Jaccard distance is used.
#[arg(long)]
pub upgma: bool,
}
+133
View File
@@ -0,0 +1,133 @@
//! Command-line arguments of `obikmer phylo`.
//!
//! The options are split by theme into flattened groups, each carrying its own
//! help heading and its own help texts (short for `-h`, long for `--help`).
//! Those texts are written for biologists: no internal names, no design
//! rationale — that belongs in the documentation.
mod distance;
mod outputs;
mod sampling;
mod select;
use std::path::PathBuf;
use clap::Args;
use obikphylo::siblings::{AmbiguityPolicy, AnnexSelection};
pub use distance::DistanceArgs;
pub use outputs::{AlignArgs, CostsArgs};
pub use sampling::SamplingArgs;
pub use select::{AnnexArgs, SelectArgs};
/// Text shown at the end of `obikmer phylo --help`.
pub const WORKFLOW_HELP: &str = "\
Typical workflows
Compare genome-wide k-mer content and infer a Neighbor-Joining tree:
obikmer phylo INDEX --distance jaccard --nj -o out
Compute SNP-based distances and infer a Neighbor-Joining tree:
obikmer phylo INDEX --distance snp-k2p --n-sites 200000 --nj -o out
Generate a SNP pseudo-alignment:
obikmer phylo INDEX --pseudo-alignment --n-sites 200000 -o out
Prepare a maximum-likelihood analysis for IQ-TREE:
obikmer phylo INDEX --iqtree --n-sites 200000 -o out
Option dependencies
--fasta, --vcf imply --pseudo-alignment
--tnt, --phyg, --iqtree imply --sankoff
--nj, --upgma use the Jaccard distance when --distance
is not given
--pseudo-alignment, --sankoff require --n-sites
--session-force requires --session";
/// Genome comparison and phylogenetic analysis
///
/// Compares the genomes stored in an index and performs phylogenetic
/// analyses. Two complementary approaches are available:
///
/// - Whole-genome distances compare the k-mer content of the genomes,
/// without any alignment.
/// - SNP-based distances and analyses compare nucleotide variants at the
/// central position of related k-mers.
///
/// Distance matrices can be used to build Neighbor-Joining or UPGMA trees.
/// SNP-based analyses can also produce pseudo-alignments and calibrated
/// substitution costs, for phylogenetic inference with TNT, PhyG or
/// IQ-TREE.
///
/// SNP families
///
/// SNP-based analyses use groups of k-mers that differ only at their
/// central nucleotide. These groups are called families. A variable family
/// is one character of the analysis, and the variants a genome carries
/// give its state for that character.
#[derive(Args)]
#[command(verbatim_doc_comment)]
pub struct PhyloArgs {
/// Index directory
pub index: PathBuf,
/// Prefix of the output files
///
/// Output files are named after this prefix: for example
/// `PREFIX_dist.phy` (or `PREFIX_dist.csv` with --csv), `PREFIX_nj.nwk`
/// and `PREFIX_upgma.nwk`, plus the files described with the other
/// options.
///
/// Without this option, the main distance matrix is written to standard
/// output, and the other files are written in the current directory
/// under their name without prefix (for example `nj.nwk`).
#[arg(short, long)]
pub output: Option<PathBuf>,
#[command(flatten)]
pub dist: DistanceArgs,
#[command(flatten)]
pub select: SelectArgs,
#[command(flatten)]
pub sample: SamplingArgs,
#[command(flatten)]
pub annex: AnnexArgs,
#[command(flatten)]
pub align: AlignArgs,
#[command(flatten)]
pub costs: CostsArgs,
}
impl PhyloArgs {
/// How genomes carrying several members of a family are treated.
pub fn ambiguity(&self) -> AmbiguityPolicy {
AmbiguityPolicy::from_flags(self.sample.no_ambiguity, self.sample.only_diploid)
}
/// The derived annexes the options of this invocation read, all under
/// [`ambiguity`](Self::ambiguity).
pub fn annex_selection(&self) -> AnnexSelection {
AnnexSelection {
policy: self.ambiguity(),
free_loss: self.sample.free_loss,
entropy: self.sample.entropy.is_some(),
quartet: self.sample.quartet,
overlap: self.annex.family_overlap
|| self.select.min_shared_family.is_some()
|| self.sample.pair_balance,
pair_weight: self.sample.pair_balance,
}
}
/// Whether a pseudo-alignment must be drawn and written (`--pseudo-alignment`,
/// or any of its output-format flags).
pub fn alignment_output(&self) -> bool {
self.align.pseudo_alignment || self.align.fasta || self.align.vcf
}
}
+132
View File
@@ -0,0 +1,132 @@
use clap::Args;
/// SNP pseudo-alignment.
///
/// A pseudo-alignment has one column per variable family and one state per
/// genome. Unlike a conventional sequence alignment, its columns come
/// directly from groups of related k-mers of the index.
#[derive(Args)]
#[command(next_help_heading = "SNP pseudo-alignment")]
pub struct AlignArgs {
/// Write a SNP pseudo-alignment
///
/// Writes an alignment with one column per selected variable family;
/// the states are the central nucleotides of the k-mers of the family.
/// The format is FASTA unless --vcf is given.
///
/// The number of sites is set with --n-sites (required). The options of
/// the "SNP character selection and sampling" group control which
/// families are included and how ambiguous states are written.
#[arg(long)]
pub pseudo_alignment: bool,
/// Write the pseudo-alignment in FASTA format
///
/// Writes `PREFIX_alignment.fasta`. This is the default format. Implies
/// --pseudo-alignment; can be combined with --vcf to get both formats.
#[arg(long)]
pub fasta: bool,
/// Write the pseudo-alignment in VCF format
///
/// Writes `PREFIX_alignment.vcf`. Each scanned index layer is a contig
/// named `obikmer_<partition>_<layer>`; `POS` is the position of the
/// family within the layer and `ID` is the masked canonical k-mer of the
/// family. Implies --pseudo-alignment; can be combined with --fasta to
/// get both formats.
#[arg(long)]
pub vcf: bool,
}
/// Sankoff cost calibration and exports for phylogenetic software.
///
/// Sankoff parsimony lets different changes between states have different
/// costs. These costs are estimated here from the observed variation, and
/// exported for external phylogenetic software.
#[derive(Args)]
#[command(next_help_heading = "Sankoff cost calibration and phylogenetic software exports")]
pub struct CostsArgs {
/// Estimate a Sankoff cost matrix and its alignment
///
/// Estimates, from the selected variable families, the cost of changing
/// from each of the 16 possible states to each other state, for weighted
/// parsimony. Writes `PREFIX_sankoff_matrix.csv` (costs),
/// `PREFIX_sankoff_params.yaml` (parameters) and
/// `PREFIX_sankoff.fasta` (alignment).
///
/// Needs --n-sites. The selection and sampling options are the same as
/// for pseudo-alignments, and a single sample of families serves both
/// the alignment and the estimation of the costs.
#[arg(long)]
pub sankoff: bool,
/// Leave very divergent genome pairs out of the calibration
///
/// Genome pairs whose raw SNP ratio exceeds this value are not used to
/// estimate the costs between nucleotides: at such divergence, repeated
/// substitutions at the same site blur the original substitution
/// pattern. The costs related to the number of variants are not
/// affected.
#[arg(long, default_value = "0.5")]
pub sankoff_ratio_ceiling: f64,
/// Export the Sankoff analysis for TNT
///
/// Writes `PREFIX_sankoff.tnt`, a script ready to run in TNT with
/// `proc FILE;`. States are recoded with TNT's default alphabet (0-9 and
/// A-F) and costs are turned into integers with --sankoff-cost-scale,
/// as TNT requires. Implies --sankoff.
#[arg(long)]
pub tnt: bool,
/// Export the Sankoff analysis for PhyG
///
/// Writes:
/// - `PREFIX_sankoff.tcm`: the cost matrix;
/// - `PREFIX_sankoff.pg`: the PhyG script.
/// The alignment is the `PREFIX_sankoff.fasta` written by --sankoff,
/// used without recoding. Costs are turned into integers with
/// --sankoff-cost-scale. Implies --sankoff.
#[arg(long, verbatim_doc_comment)]
pub phyg: bool,
/// Export a model and alignment for IQ-TREE maximum likelihood
///
/// Writes:
/// - `PREFIX_iqtree.model`: a substitution model derived from the
/// calibrated Sankoff costs;
/// - `PREFIX_iqtree.fasta`: the alignment with recoded states.
/// The model combines the calibrated costs with the frequencies of the
/// states observed in the alignment. Only the states that actually
/// occur are kept, renumbered consecutively.
///
/// Run IQ-TREE with:
/// iqtree3 -s PREFIX_iqtree.fasta --seqtype MORPH -m ...+ASC
/// giving the model file to -m. Unlike the parsimony analyses of TNT
/// and PhyG, maximum likelihood estimates branch lengths.
///
/// Implies --sankoff.
#[arg(long, verbatim_doc_comment)]
pub iqtree: bool,
/// Treat rare states as missing in the IQ-TREE export
///
/// With --iqtree and --free-loss, the states whose frequency in the
/// alignment is below this value are recoded as missing data (`?`).
/// States combining three or four nucleotides are typically that rare,
/// and are more likely noise than a real polymorphism.
///
/// No effect without --free-loss. The frequency actually applied is
/// reported in `PREFIX_iqtree_states.csv`.
#[arg(long, default_value = "0.001")]
pub iqtree_min_freq: f64,
/// Factor turning costs into integers for TNT and PhyG
///
/// TNT and PhyG only accept integer costs: the calibrated costs are
/// multiplied by this factor, then rounded. Keep it small. The tree
/// score is multiplied by the same factor, and a very large value can
/// overflow integers in TNT without any error.
#[arg(long, default_value = "100")]
pub sankoff_cost_scale: f64,
}
+127
View File
@@ -0,0 +1,127 @@
use clap::Args;
use super::super::values::{NSitesArg, parse_n_sites};
/// SNP character selection and sampling.
///
/// SNP-based analyses can use every eligible variable family, or a subset
/// drawn from the index. These options control which characters are
/// selected, which genomes and states are kept, and how the sample is spread.
#[derive(Args)]
#[command(next_help_heading = "SNP character selection and sampling")]
pub struct SamplingArgs {
/// Number of variable sites to select
///
/// Target number of variable families to draw from the index. Either a
/// positive integer, or `all` to keep every family that passes the other
/// filters, with no sampling.
///
/// Required for pseudo-alignments and Sankoff analyses. Optional for
/// SNP-based distances: when omitted, every eligible site is used.
///
/// Sites are drawn in proportion across the layers of the index, so the
/// number actually selected can differ slightly from the target.
#[arg(long = "n-sites", value_name = "N|all", value_parser = parse_n_sites)]
pub n_sites: Option<NSitesArg>,
/// Treat genomes lacking a family as missing data (?)
///
/// When a genome carries none of the variants found in a family, its
/// state is coded as missing data (`?`) instead of a real state.
///
/// Applies to every SNP-based analysis: distances, pseudo-alignments and
/// the Sankoff calibration with its exports.
#[arg(long)]
pub free_loss: bool,
/// Treat genomes with several variants as missing data (?)
///
/// A genome carrying more than one nucleotide variant of a family is
/// coded as missing data (`?`) instead of an IUPAC ambiguity code.
///
/// This also decides which genomes count as carriers for --quartet and
/// --pair-balance.
#[arg(long)]
pub no_ambiguity: bool,
/// Keep genomes with one or two variants per family
///
/// A genome with one variant is written as a single nucleotide, a genome
/// with two variants as the matching two-base IUPAC code. Genomes with
/// three or four variants are coded as missing data (`?`).
///
/// A milder alternative to --no-ambiguity; the two cannot be combined.
#[arg(long, conflicts_with = "no_ambiguity")]
pub only_diploid: bool,
/// Keep families that can tell tree topologies apart
///
/// A family is kept only when at least two different states are each
/// found in at least two genomes. A variant seen in a single genome, or
/// a family carried by fewer than four genomes, fits every arrangement
/// of the genomes and teaches nothing about the tree.
///
/// The states are those of the alignment. Without --free-loss, absence
/// is a state of its own, so two genomes with a base and two genomes
/// lacking the family form two pairs. With --free-loss, absence is
/// missing data and never forms a pair. Which genomes count as carriers
/// also follows --no-ambiguity. Applies to pseudo-alignments, Sankoff
/// analyses and SNP-based distances.
#[arg(long)]
pub quartet: bool,
/// Favour families covering poorly represented genome pairs
///
/// Each family gets a weight: the mean, over the pairs of genomes
/// carrying it, of 1 / (number of families the pair shares). Families
/// covering pairs that share little information are therefore drawn
/// more often, which balances how well each pair is represented in the
/// sample.
///
/// Which genomes count as carriers follows --no-ambiguity. Applies to
/// pseudo-alignments, Sankoff analyses and SNP-based distances.
#[arg(long)]
pub pair_balance: bool,
/// Favour families of a given Shannon entropy
///
/// Families are drawn with a weight following a bell-shaped curve
/// centred on this entropy. Giving this option, or --entropy-sd, turns
/// the weighting on; the other one then takes its default (centre 1.0
/// bit, standard deviation 0.5 bit).
#[arg(long)]
pub entropy: Option<f64>,
/// Width of the entropy weighting
///
/// Standard deviation of the bell-shaped curve used by --entropy.
/// Giving this option also turns the weighting on.
#[arg(long)]
pub entropy_sd: Option<f64>,
/// Write statistics about the selected sample
///
/// Writes three CSV files describing the sites actually drawn:
/// - `PREFIX_sample_genomes.csv`: for each genome, the number of sites
/// where it has a called state and of informative sites;
/// - `PREFIX_sample_carriers.csv`: the number of sites and of
/// informative sites by number of carrier genomes;
/// - `PREFIX_sample_overlap.csv`: the genome-by-genome matrix of
/// families shared within the sample.
/// The overlap matrix can be compared with the one written by
/// --family-overlap, to check how well the sample represents the index.
///
/// Needs --pseudo-alignment, --sankoff or a SNP-based distance.
#[arg(long, verbatim_doc_comment)]
pub sample_stats: bool,
/// Write how the sites were spread over the index layers
///
/// Writes `PREFIX_sample_plan.csv`, one line per layer: number of
/// eligible families, their total weight, the layer's quota, the number
/// of families drawn and the number kept after masking.
///
/// Needs --pseudo-alignment, --sankoff or a SNP-based distance.
#[arg(long)]
pub sample_plan: bool,
}
+113
View File
@@ -0,0 +1,113 @@
use clap::Args;
/// Genome selection and filtering.
#[derive(Args)]
#[command(next_help_heading = "Genome selection and filtering")]
pub struct SelectArgs {
/// Exclude a genome from the analysis (repeatable)
///
/// Removes the genome identified by its exact label. Repeat the option
/// to exclude several genomes.
///
/// For SNP-based analyses, excluded genomes are removed before the
/// variable sites are selected: a family that was variable only because
/// of an excluded genome is no longer considered.
///
/// Excluded genomes are also removed from the distance and shared-k-mer
/// matrices; the distances between the remaining genomes are not
/// otherwise modified.
#[arg(long = "exclude-genome", value_name = "LABEL")]
pub exclude_genome: Vec<String>,
/// Exclude genomes sharing too few variable families
///
/// Removes the genomes whose average number of variable families shared
/// with the other genomes is below N. The average is the per-genome mean
/// of the matrix written by --family-overlap. This comes on top of
/// --exclude-genome.
///
/// The filter removes genomes that carry too little information for
/// SNP-based comparisons. It applies to SNP-based distances,
/// pseudo-alignments and Sankoff analyses, not to whole-genome distances
/// such as Jaccard: a genome with little SNP information is not
/// necessarily a genome whose overall k-mer content is divergent.
///
/// The family-overlap matrix is computed automatically if needed.
#[arg(long, value_name = "N")]
pub min_shared_family: Option<f64>,
}
/// SNP family preparation and diagnostics.
#[derive(Args)]
#[command(next_help_heading = "SNP family preparation and diagnostics")]
pub struct AnnexArgs {
/// Prepare in advance the tables used by SNP analyses
///
/// SNP-based distances, pseudo-alignments, Sankoff analyses and entropy
/// calculations rely on precomputed tables describing the families of
/// k-mers that differ only at their central nucleotide. They are
/// prepared automatically when first needed, so this option is never
/// required. It lets you prepare them beforehand, for example before a
/// series of analyses on the same index, or force them to be rebuilt.
///
/// Rebuilding the main table invalidates the tables derived from it.
/// Among those, only the ones requested in the same command are rebuilt
/// immediately (--entropy, --quartet, --pair-balance, --family-overlap
/// and --min-shared-family), following the treatment of ambiguous
/// states given by --no-ambiguity or --only-diploid. The others are
/// rebuilt when next needed.
#[arg(long)]
pub sibling_annex: bool,
/// Write the distribution of family sizes, by genome
///
/// Writes `PREFIX_siblings.csv`. For each genome, it gives the number of
/// families of each size (1 to 4 members) in which the genome carries at
/// least one member. A final `global` row gives the number of families
/// of each size over the whole index. Mostly a diagnostic.
#[arg(long)]
pub sibling_stats: bool,
/// Write the global distribution of family sizes
///
/// Writes `PREFIX_sibling_hist.csv`, the number of families having 1 to
/// 4 members over the whole index. These are the figures of the `global`
/// row of --sibling-stats, without the per-genome breakdown, and they
/// are faster to obtain.
#[arg(long)]
pub sibling_hist: bool,
/// Write the number of variable families shared by each genome pair
///
/// Writes `PREFIX_family_overlap.csv`, a genome-by-genome matrix giving,
/// for each pair of genomes, the number of variable families they share
/// over the whole index. It tells how much SNP information is available
/// to compare each pair, and helps spot genomes that are poorly covered.
///
/// Every genome is included, whatever --exclude-genome and
/// --min-shared-family say.
#[arg(long)]
pub family_overlap: bool,
/// Write the Shannon entropy of every family
///
/// Writes `PREFIX_entropy.csv`, one line per family. All families are
/// scanned, with no sampling: --n-sites, --entropy and --entropy-sd do
/// not apply.
#[arg(long)]
pub shannon: bool,
/// Summarise the distribution of family entropy
///
/// Same complete scan as --shannon, but writes a compact summary,
/// `PREFIX_shannon_summary.csv`, instead of one line per family. It
/// counts the families according to:
/// - entropy category;
/// - number of genomes carrying the family;
/// - number of those carriers without ambiguity;
/// - entropy bins of 0.05 bits.
/// It has a few hundred thousand lines at most, whatever the size of
/// the index.
#[arg(long, verbatim_doc_comment)]
pub shannon_summary: bool,
}
+10 -3
View File
@@ -100,7 +100,14 @@ fn drop_ascertainment_noninformative(alignment: &SnpAlignment) -> SnpAlignment {
.collect() .collect()
}) })
.collect(); .collect();
SnpAlignment { sequences, genome_indices: alignment.genome_indices.clone() } let sites = alignment
.sites
.iter()
.zip(keep.iter())
.filter(|&(_, &k)| k)
.map(|(&site, _)| site)
.collect();
SnpAlignment { sequences, genome_indices: alignment.genome_indices.clone(), sites }
} }
/// Recode every occurrence of a byte in `symbols` to `-` — the same /// Recode every occurrence of a byte in `symbols` to `-` — the same
@@ -119,7 +126,7 @@ fn recode_symbols_as_absent(alignment: &SnpAlignment, symbols: &[u8]) -> SnpAlig
.collect() .collect()
}) })
.collect(); .collect();
SnpAlignment { sequences, genome_indices: alignment.genome_indices.clone() } SnpAlignment { sequences, genome_indices: alignment.genome_indices.clone(), sites: alignment.sites.clone() }
} }
fn compact_alphabet(alignment: &SnpAlignment, free_loss: bool) -> CompactAlphabet { fn compact_alphabet(alignment: &SnpAlignment, free_loss: bool) -> CompactAlphabet {
@@ -384,7 +391,7 @@ mod tests {
fn alignment(sequences: Vec<Vec<u8>>) -> SnpAlignment { fn alignment(sequences: Vec<Vec<u8>>) -> SnpAlignment {
let genome_indices = (0..sequences.len()).collect(); let genome_indices = (0..sequences.len()).collect();
SnpAlignment { sequences, genome_indices } SnpAlignment { sequences, genome_indices, sites: Vec::new() }
} }
#[test] #[test]
+75 -88
View File
@@ -1,4 +1,7 @@
mod alignment_out;
mod args; mod args;
mod shared;
mod values;
mod iqtree; mod iqtree;
mod phyg; mod phyg;
mod phylip; mod phylip;
@@ -19,7 +22,9 @@ use obikphylo::{Metrics, neighbor_joining, upgma};
use obisys::{Reporter, Stage}; use obisys::{Reporter, Stage};
use tracing::info; use tracing::info;
use args::{DistanceArg, NSitesArg, distance_to_compute}; use alignment_out::write_alignment_outputs;
use shared::write_shared_kmers;
use values::{DistanceArg, NSitesArg, distance_to_compute};
use iqtree::write_iqtree; use iqtree::write_iqtree;
use phyg::write_sankoff_phyg; use phyg::write_sankoff_phyg;
use stats::{write_sample_plan, write_sample_stats}; use stats::{write_sample_plan, write_sample_stats};
@@ -27,7 +32,7 @@ use phylip::write_phylip_relaxed;
use sankoff::{write_sankoff_alignment_fasta, write_sankoff_matrix_csv, write_sankoff_params}; use sankoff::{write_sankoff_alignment_fasta, write_sankoff_matrix_csv, write_sankoff_params};
use tnt::write_sankoff_tnt; use tnt::write_sankoff_tnt;
pub use args::PhyloArgs; pub use args::{PhyloArgs, WORKFLOW_HELP};
pub fn run(args: PhyloArgs) { pub fn run(args: PhyloArgs) {
let idx = Arc::new(KmerIndex::open(&args.index).unwrap_or_else(|e| { let idx = Arc::new(KmerIndex::open(&args.index).unwrap_or_else(|e| {
@@ -54,7 +59,7 @@ pub fn run(args: PhyloArgs) {
// `cache.distance(...)` computation is unaffected either way. // `cache.distance(...)` computation is unaffected either way.
let exclude_mask: Vec<bool> = { let exclude_mask: Vec<bool> = {
let mut mask = vec![false; n]; let mut mask = vec![false; n];
for label in &args.exclude_genome { for label in &args.select.exclude_genome {
match labels.iter().position(|l| l == label) { match labels.iter().position(|l| l == label) {
Some(i) => mask[i] = true, Some(i) => mask[i] = true,
None => { None => {
@@ -77,13 +82,13 @@ pub fn run(args: PhyloArgs) {
// ── Sibling-count/minorant annex (independent of the distance metric) ── // ── Sibling-count/minorant annex (independent of the distance metric) ──
// Meant to be (re)built routinely, ahead of any SNP-family distance // Meant to be (re)built routinely, ahead of any SNP-family distance
// computation that will later consume it. // computation that will later consume it.
if args.sibling_annex { if args.annex.sibling_annex {
// Writes into the index directory; `build_sibling_annex` holds the // Writes into the index directory; `build_sibling_annex` holds the
// index's exclusive lock for the duration (`KmerIndex::with_lock`), // index's exclusive lock for the duration (`KmerIndex::with_lock`),
// so a concurrent run on the same index can't corrupt these writes. // so a concurrent run on the same index can't corrupt these writes.
info!("building sibling-count/minorant annex"); info!("building sibling-count/minorant annex");
let t = Stage::start("sibling_annex"); let t = Stage::start("sibling_annex");
cache.build_sibling_annex().unwrap_or_else(|e| { cache.build_sibling_annex(&args.annex_selection()).unwrap_or_else(|e| {
eprintln!("error building sibling annex: {e}"); eprintln!("error building sibling annex: {e}");
std::process::exit(1); std::process::exit(1);
}); });
@@ -91,7 +96,7 @@ pub fn run(args: PhyloArgs) {
} }
// ── Sibling-count distribution (`--sibling-stats`) ────────────────────────── // ── Sibling-count distribution (`--sibling-stats`) ──────────────────────────
if args.sibling_stats { if args.annex.sibling_stats {
let t = Stage::start("sibling_stats"); let t = Stage::start("sibling_stats");
let stats = cache.sibling_annex_stats().unwrap_or_else(|e| { let stats = cache.sibling_annex_stats().unwrap_or_else(|e| {
eprintln!("error computing sibling-annex stats: {e}"); eprintln!("error computing sibling-annex stats: {e}");
@@ -124,7 +129,7 @@ pub fn run(args: PhyloArgs) {
} }
// ── Family-size histogram (`--sibling-hist`) ──────────────────────────────── // ── Family-size histogram (`--sibling-hist`) ────────────────────────────────
if args.sibling_hist { if args.annex.sibling_hist {
let t = Stage::start("sibling_hist"); let t = Stage::start("sibling_hist");
let counts = cache.sibling_family_size_histogram().unwrap_or_else(|e| { let counts = cache.sibling_family_size_histogram().unwrap_or_else(|e| {
eprintln!("error computing sibling family-size histogram: {e}"); eprintln!("error computing sibling family-size histogram: {e}");
@@ -151,9 +156,9 @@ pub fn run(args: PhyloArgs) {
} }
// ── Family Overlap matrix (`--family-overlap`) ────────────────────────────── // ── Family Overlap matrix (`--family-overlap`) ──────────────────────────────
if args.family_overlap { if args.annex.family_overlap {
let t = Stage::start("family_overlap"); let t = Stage::start("family_overlap");
let overlap = cache.family_overlap(args.no_ambiguity).unwrap_or_else(|e| { let overlap = cache.family_overlap(args.ambiguity()).unwrap_or_else(|e| {
eprintln!("error computing family overlap: {e}"); eprintln!("error computing family overlap: {e}");
std::process::exit(1); std::process::exit(1);
}); });
@@ -178,7 +183,7 @@ pub fn run(args: PhyloArgs) {
} }
// ── Shannon entropy report (`--shannon`) ──────────────────────────────────── // ── Shannon entropy report (`--shannon`) ────────────────────────────────────
if args.shannon { if args.annex.shannon {
let path = args.output.as_ref() let path = args.output.as_ref()
.map(|p| format!("{}_entropy.csv", p.display())) .map(|p| format!("{}_entropy.csv", p.display()))
.unwrap_or_else(|| "entropy.csv".into()); .unwrap_or_else(|| "entropy.csv".into());
@@ -193,7 +198,7 @@ pub fn run(args: PhyloArgs) {
} }
// ── Shannon entropy distribution (`--shannon-summary`) ────────────────────── // ── Shannon entropy distribution (`--shannon-summary`) ──────────────────────
if args.shannon_summary { if args.annex.shannon_summary {
let path = args.output.as_ref() let path = args.output.as_ref()
.map(|p| format!("{}_shannon_summary.csv", p.display())) .map(|p| format!("{}_shannon_summary.csv", p.display()))
.unwrap_or_else(|| "shannon_summary.csv".into()); .unwrap_or_else(|| "shannon_summary.csv".into());
@@ -213,9 +218,9 @@ pub fn run(args: PhyloArgs) {
// `--pseudo-alignment`/`--sankoff`/`snp-*` `--distance` only, never the // `--pseudo-alignment`/`--sankoff`/`snp-*` `--distance` only, never the
// whole-index metrics' `kept`/`--shared-kmers` output (see // whole-index metrics' `kept`/`--shared-kmers` output (see
// `args::PhyloArgs::min_shared_family`'s own docs on why). // `args::PhyloArgs::min_shared_family`'s own docs on why).
let snp_exclude_mask: Vec<bool> = match args.min_shared_family { let snp_exclude_mask: Vec<bool> = match args.select.min_shared_family {
Some(threshold) => { Some(threshold) => {
let overlap = cache.family_overlap(args.no_ambiguity).unwrap_or_else(|e| { let overlap = cache.family_overlap(args.ambiguity()).unwrap_or_else(|e| {
eprintln!("error computing family overlap: {e}"); eprintln!("error computing family overlap: {e}");
std::process::exit(1); std::process::exit(1);
}); });
@@ -241,19 +246,19 @@ pub fn run(args: PhyloArgs) {
// Shared by `--pseudo-alignment` and `--sankoff` — same activation rule: // Shared by `--pseudo-alignment` and `--sankoff` — same activation rule:
// either flag given activates entropy-biased sampling, the other // either flag given activates entropy-biased sampling, the other
// defaults to 1.0/0.5. // defaults to 1.0/0.5.
let entropy_bias = if args.entropy.is_some() || args.entropy_sd.is_some() { let entropy_bias = if args.sample.entropy.is_some() || args.sample.entropy_sd.is_some() {
Some(EntropyBias { Some(EntropyBias {
mu: args.entropy.unwrap_or(1.0), mu: args.sample.entropy.unwrap_or(1.0),
sigma: args.entropy_sd.unwrap_or(0.5), sigma: args.sample.entropy_sd.unwrap_or(0.5),
}) })
} else { } else {
None None
}; };
let sampling = SamplingSpec { let sampling = SamplingSpec {
free_loss: args.free_loss, free_loss: args.sample.free_loss,
no_ambiguity: args.no_ambiguity, ambiguity: args.ambiguity(),
quartet: args.quartet, quartet: args.sample.quartet,
pair_balance: args.pair_balance, pair_balance: args.sample.pair_balance,
entropy_bias, entropy_bias,
}; };
@@ -264,7 +269,7 @@ pub fn run(args: PhyloArgs) {
// `algorithms::scan`'s own module docs) — persistence needs to be // `algorithms::scan`'s own module docs) — persistence needs to be
// rebuilt on top of the new single `sampled_scan` trunk instead of // rebuilt on top of the new single `sampled_scan` trunk instead of
// silently becoming a no-op. // silently becoming a no-op.
if args.session.is_some() { if args.dist.session.is_some() {
eprintln!( eprintln!(
"error: --session is temporarily unavailable (being rebuilt on top of the unified \ "error: --session is temporarily unavailable (being rebuilt on top of the unified \
sampling trunk) — rerun without --session/--session-force" sampling trunk) — rerun without --session/--session-force"
@@ -282,23 +287,23 @@ pub fn run(args: PhyloArgs) {
// instead of each running its own scan (which used to silently draw a // instead of each running its own scan (which used to silently draw a
// different sample per flag — see `obikphylo::siblings::LayerReducer`'s // different sample per flag — see `obikphylo::siblings::LayerReducer`'s
// own module docs). // own module docs).
let sankoff_requested = args.sankoff || args.tnt || args.phyg || args.iqtree; let sankoff_requested = args.costs.sankoff || args.costs.tnt || args.costs.phyg || args.costs.iqtree;
let distance = distance_to_compute(args.distance, args.nj, args.upgma, args.shared_kmers); let distance = distance_to_compute(args.dist.distance, args.dist.nj, args.dist.upgma);
let snp_kind = distance.and_then(DistanceArg::as_snp); let snp_kind = distance.and_then(DistanceArg::as_snp);
let need_alignment = args.pseudo_alignment || sankoff_requested; let need_alignment = args.alignment_output() || sankoff_requested;
let need_pairwise = sankoff_requested || snp_kind.is_some(); let need_pairwise = sankoff_requested || snp_kind.is_some();
// `None` = every site (`--n-sites all`, or `--n-sites` omitted where it is optional) // `None` = every site (`--n-sites all`, or `--n-sites` omitted where it is optional)
let target = args.n_sites.and_then(NSitesArg::target); let target = args.sample.n_sites.and_then(NSitesArg::target);
if args.pseudo_alignment && args.n_sites.is_none() { if args.alignment_output() && args.sample.n_sites.is_none() {
eprintln!("error: --pseudo-alignment requires --n-sites <N|all>"); eprintln!("error: --pseudo-alignment/--fasta/--vcf require --n-sites <N|all>");
std::process::exit(1); std::process::exit(1);
} }
if sankoff_requested && args.n_sites.is_none() { if sankoff_requested && args.sample.n_sites.is_none() {
eprintln!("error: --sankoff requires --n-sites <N|all>"); eprintln!("error: --sankoff requires --n-sites <N|all>");
std::process::exit(1); std::process::exit(1);
} }
if (args.sample_stats || args.sample_plan) && !(need_alignment || need_pairwise) { if (args.sample.sample_stats || args.sample.sample_plan) && !(need_alignment || need_pairwise) {
eprintln!( eprintln!(
"error: --sample-stats/--sample-plan describe a draw: they need --pseudo-alignment, \ "error: --sample-stats/--sample-plan describe a draw: they need --pseudo-alignment, \
--sankoff or a snp-* --distance" --sankoff or a snp-* --distance"
@@ -310,7 +315,7 @@ pub fn run(args: PhyloArgs) {
need_alignment.then(|| obikphylo::siblings::AlignmentReducer::new(&cache, &snp_exclude_mask)); need_alignment.then(|| obikphylo::siblings::AlignmentReducer::new(&cache, &snp_exclude_mask));
let mut pairwise_reducer = need_pairwise.then(|| obikphylo::siblings::PairwiseReducer::new(n)); let mut pairwise_reducer = need_pairwise.then(|| obikphylo::siblings::PairwiseReducer::new(n));
let mut stats_reducer = let mut stats_reducer =
args.sample_stats.then(|| obikphylo::siblings::SampleStatsReducer::new(n, &snp_exclude_mask)); args.sample.sample_stats.then(|| obikphylo::siblings::SampleStatsReducer::new(n, &snp_exclude_mask));
let mut sample_plan = None; let mut sample_plan = None;
if need_alignment || need_pairwise { if need_alignment || need_pairwise {
@@ -347,7 +352,7 @@ pub fn run(args: PhyloArgs) {
if let Some(reducer) = stats_reducer { if let Some(reducer) = stats_reducer {
write_sample_stats(&reducer.finish(), &labels, &snp_exclude_mask, &args.output); write_sample_stats(&reducer.finish(), &labels, &snp_exclude_mask, &args.output);
} }
if args.sample_plan { if args.sample.sample_plan {
write_sample_plan(&sample_plan.expect("scanned above: need_alignment || need_pairwise"), &args.output); write_sample_plan(&sample_plan.expect("scanned above: need_alignment || need_pairwise"), &args.output);
} }
@@ -357,21 +362,9 @@ pub fn run(args: PhyloArgs) {
let alignment = alignment_reducer.map(obikphylo::siblings::AlignmentReducer::finish); let alignment = alignment_reducer.map(obikphylo::siblings::AlignmentReducer::finish);
// ── SNP pseudo-alignment (`--pseudo-alignment`) ───────────────────────────── // ── SNP pseudo-alignment (`--pseudo-alignment`) ─────────────────────────────
if args.pseudo_alignment { if args.alignment_output() {
let alignment = alignment.as_ref().expect("scanned above: need_alignment"); let alignment = alignment.as_ref().expect("scanned above: need_alignment");
write_alignment_outputs(alignment, &labels, &args);
let path = args.output.as_ref()
.map(|p| format!("{}_alignment.fasta", p.display()))
.unwrap_or_else(|| "alignment.fasta".into());
let mut f = BufWriter::new(std::fs::File::create(&path).unwrap_or_else(|e| {
eprintln!("error creating {path}: {e}");
std::process::exit(1);
}));
let n_sites = alignment.sequences.first().map_or(0, Vec::len);
for (&g, seq) in alignment.genome_indices.iter().zip(alignment.sequences.iter()) {
obifastwrite::write_plain_record(seq, &labels[g], &mut f).unwrap();
}
info!("pseudo-alignment ({n_sites} site(s), {} genome(s)) → {path}", alignment.genome_indices.len());
} }
// ── Sankoff cost-matrix calibration (`--sankoff`, `--tnt`, `--phyg`, `--iqtree`) ── // ── Sankoff cost-matrix calibration (`--sankoff`, `--tnt`, `--phyg`, `--iqtree`) ──
@@ -387,13 +380,13 @@ pub fn run(args: PhyloArgs) {
let b = obikphylo::siblings::SankoffBundle::finish( let b = obikphylo::siblings::SankoffBundle::finish(
alignment, alignment,
pairwise, pairwise,
args.sankoff_ratio_ceiling, args.costs.sankoff_ratio_ceiling,
&snp_exclude_mask, &snp_exclude_mask,
); );
let p_card = cardinality_transition_probs(&b.cardinality_tally); let p_card = cardinality_transition_probs(&b.cardinality_tally);
let p_comp = composition_transition_probs(&b.base_pair_tally); let p_comp = composition_transition_probs(&b.base_pair_tally);
let matrix = pairwise_cost_matrix(&p_card, &p_comp, args.free_loss); let matrix = pairwise_cost_matrix(&p_card, &p_comp, args.sample.free_loss);
write_sankoff_matrix_csv(&matrix, &args.output); write_sankoff_matrix_csv(&matrix, &args.output);
write_sankoff_params( write_sankoff_params(
@@ -401,43 +394,54 @@ pub fn run(args: PhyloArgs) {
&p_card, &p_card,
&b.base_pair_tally, &b.base_pair_tally,
&p_comp, &p_comp,
args.sankoff_ratio_ceiling, args.costs.sankoff_ratio_ceiling,
&args.output, &args.output,
); );
write_sankoff_alignment_fasta(&b.alignment, &labels, &args.output, args.free_loss); write_sankoff_alignment_fasta(&b.alignment, &labels, &args.output, args.sample.free_loss);
if args.tnt { if args.costs.tnt {
write_sankoff_tnt( write_sankoff_tnt(
&matrix, &matrix,
&b.alignment, &b.alignment,
&labels, &labels,
&args.output, &args.output,
args.sankoff_cost_scale, args.costs.sankoff_cost_scale,
args.free_loss, args.sample.free_loss,
); );
} }
if args.phyg { if args.costs.phyg {
write_sankoff_phyg(&matrix, &args.output, args.sankoff_cost_scale); write_sankoff_phyg(&matrix, &args.output, args.costs.sankoff_cost_scale);
} }
if args.iqtree { if args.costs.iqtree {
write_iqtree( write_iqtree(
&matrix, &matrix,
&b.alignment, &b.alignment,
&labels, &labels,
&args.output, &args.output,
args.free_loss, args.sample.free_loss,
args.iqtree_min_freq, args.costs.iqtree_min_freq,
); );
} }
bundle = Some(b); bundle = Some(b);
} }
// ── Distance computation: classic whole-index metric vs. `snp-*` ─────────── // ── Distance computation: classic whole-index metric vs. `snp-*` ───────────
// Nothing below (matrix, shared-kmer matrix, NJ/UPGMA trees) is wanted // Nothing below (distance matrix, NJ/UPGMA trees) is wanted unless a
// unless a distance was asked for (`--distance`) or implied // distance was asked for (`--distance`) or implied (`--nj`/`--upgma`): stop
// (`--nj`/`--upgma`/`--shared-kmers`): stop here rather than compute and // here rather than compute and write a matrix nobody requested.
// write a matrix nobody requested. // `--shared-kmers` alone only asks for its own matrix, not for a distance.
let Some(distance) = distance else { let Some(distance) = distance else {
if args.dist.shared_kmers {
info!("computing shared k-mer counts for {n} genome(s)");
let t = Stage::start("shared_kmers");
let shared = cache.shared_kmers(args.dist.presence_threshold).unwrap_or_else(|e| {
eprintln!("error computing shared k-mer counts: {e}");
std::process::exit(1);
});
rep.push(t.stop());
let kept: Vec<usize> = (0..n).filter(|&i| !exclude_mask[i]).collect();
write_shared_kmers(&shared, &labels, &kept, &args.output);
}
rep.print(); rep.print();
return; return;
}; };
@@ -446,10 +450,10 @@ pub fn run(args: PhyloArgs) {
let (matrix, shared_kmers) = match distance.as_classic() { let (matrix, shared_kmers) = match distance.as_classic() {
Some(metric) => { Some(metric) => {
info!("computing {metric:?} distances for {n} genome(s)"); info!("computing {metric:?} distances for {n} genome(s)");
let need_shared = args.shared_kmers || args.nj || args.upgma; let need_shared = args.dist.shared_kmers;
let t = Stage::start("distance"); let t = Stage::start("distance");
let result = cache let result = cache
.distance(metric, need_shared, args.presence_threshold) .distance(metric, need_shared, args.dist.presence_threshold)
.unwrap_or_else(|e| { .unwrap_or_else(|e| {
eprintln!("error computing distances: {e}"); eprintln!("error computing distances: {e}");
std::process::exit(1); std::process::exit(1);
@@ -458,12 +462,12 @@ pub fn run(args: PhyloArgs) {
(result.matrix, result.shared_kmers) (result.matrix, result.shared_kmers)
} }
None => { None => {
if args.shared_kmers { if args.dist.shared_kmers {
eprintln!("error: --shared-kmers has no meaning for a snp-* --distance value"); eprintln!("error: --shared-kmers has no meaning for a snp-* --distance value");
std::process::exit(1); std::process::exit(1);
} }
let kind = snp_kind.expect("DistanceArg is always classic or snp"); let kind = snp_kind.expect("DistanceArg is always classic or snp");
let gamma_shape = args.gamma_shape.map_or(GammaShape::Disabled, Into::into); let gamma_shape = args.dist.gamma_shape.map_or(GammaShape::Disabled, Into::into);
let t = Stage::start("snp_distance"); let t = Stage::start("snp_distance");
let matrix = match &bundle { let matrix = match &bundle {
Some(b) => { Some(b) => {
@@ -495,7 +499,7 @@ pub fn run(args: PhyloArgs) {
// ── Distance matrix → relaxed PHYLIP (default) or CSV (`--csv`) ──────────── // ── Distance matrix → relaxed PHYLIP (default) or CSV (`--csv`) ────────────
let write_dist = |w: &mut dyn Write| { let write_dist = |w: &mut dyn Write| {
if args.csv { if args.dist.csv {
write!(w, "genome").unwrap(); write!(w, "genome").unwrap();
for &j in &kept { write!(w, ",{}", labels[j]).unwrap(); } for &j in &kept { write!(w, ",{}", labels[j]).unwrap(); }
writeln!(w).unwrap(); writeln!(w).unwrap();
@@ -513,7 +517,7 @@ pub fn run(args: PhyloArgs) {
match &args.output { match &args.output {
Some(prefix) => { Some(prefix) => {
let suffix = if args.csv { "_dist.csv" } else { "_dist.phy" }; let suffix = if args.dist.csv { "_dist.csv" } else { "_dist.phy" };
let path = format!("{}{suffix}", prefix.display()); let path = format!("{}{suffix}", prefix.display());
let mut f = BufWriter::new(std::fs::File::create(&path).unwrap_or_else(|e| { let mut f = BufWriter::new(std::fs::File::create(&path).unwrap_or_else(|e| {
eprintln!("error creating {path}: {e}"); eprintln!("error creating {path}: {e}");
@@ -530,29 +534,12 @@ pub fn run(args: PhyloArgs) {
} }
// ── Shared-kmer matrix → CSV ────────────────────────────────────────────── // ── Shared-kmer matrix → CSV ──────────────────────────────────────────────
if args.shared_kmers { if let Some(shared) = shared_kmers.as_ref().filter(|_| args.dist.shared_kmers) {
if let Some(shared) = &shared_kmers { write_shared_kmers(shared, &labels, &kept, &args.output);
let path = args.output.as_ref()
.map(|p| format!("{}_shared.csv", p.display()))
.unwrap_or_else(|| "shared.csv".into());
let mut f = BufWriter::new(std::fs::File::create(&path).unwrap_or_else(|e| {
eprintln!("error creating {path}: {e}");
std::process::exit(1);
}));
write!(f, "genome").unwrap();
for &j in &kept { write!(f, ",{}", labels[j]).unwrap(); }
writeln!(f).unwrap();
for &i in &kept {
write!(f, "{}", labels[i]).unwrap();
for &j in &kept { write!(f, ",{}", shared[[i, j]]).unwrap(); }
writeln!(f).unwrap();
}
info!("shared-kmer matrix → {path}");
}
} }
// ── NJ tree ──────────────────────────────────────────────────────────────── // ── NJ tree ────────────────────────────────────────────────────────────────
if args.nj { if args.dist.nj {
let tree = neighbor_joining(&matrix, &labels).unwrap_or_else(|e| { let tree = neighbor_joining(&matrix, &labels).unwrap_or_else(|e| {
eprintln!("error computing NJ tree: {e}"); eprintln!("error computing NJ tree: {e}");
std::process::exit(1); std::process::exit(1);
@@ -569,7 +556,7 @@ pub fn run(args: PhyloArgs) {
} }
// ── UPGMA tree ─────────────────────────────────────────────────────────────── // ── UPGMA tree ───────────────────────────────────────────────────────────────
if args.upgma { if args.dist.upgma {
let newick = upgma(&matrix, &labels).to_newick(); let newick = upgma(&matrix, &labels).to_newick();
let path = args.output.as_ref() let path = args.output.as_ref()
.map(|p| format!("{}_upgma.nwk", p.display())) .map(|p| format!("{}_upgma.nwk", p.display()))
+39
View File
@@ -0,0 +1,39 @@
//! `--shared-kmers`: the genome-by-genome matrix of shared k-mer counts.
use std::fs::File;
use std::io::{BufWriter, Write};
use std::path::PathBuf;
use ndarray::Array2;
use tracing::info;
/// Write `shared` (rows and columns of the genomes in `kept` only) as CSV to
/// `PREFIX_shared.csv`, or `shared.csv` without a prefix.
pub(super) fn write_shared_kmers(
shared: &Array2<u64>,
labels: &[String],
kept: &[usize],
output: &Option<PathBuf>,
) {
let path = output
.as_ref()
.map(|p| format!("{}_shared.csv", p.display()))
.unwrap_or_else(|| "shared.csv".into());
let mut f = BufWriter::new(File::create(&path).unwrap_or_else(|e| {
eprintln!("error creating {path}: {e}");
std::process::exit(1);
}));
write!(f, "genome").unwrap();
for &j in kept {
write!(f, ",{}", labels[j]).unwrap();
}
writeln!(f).unwrap();
for &i in kept {
write!(f, "{}", labels[i]).unwrap();
for &j in kept {
write!(f, ",{}", shared[[i, j]]).unwrap();
}
writeln!(f).unwrap();
}
info!("shared-kmer matrix → {path}");
}
+170
View File
@@ -0,0 +1,170 @@
use obikphylo::{DistanceMetric, GammaShape, SnpDistanceKind};
/// `--gamma-shape` CLI value: a fixed `alpha`, or `auto`/`estimate` to fit
/// it from the data (see [`GammaShape::Auto`]). Kept as its own tiny type
/// (rather than parsing straight to `GammaShape`) so the CLI layer stays
/// free of `GammaShape::Disabled`, which the `Option<GammaShapeArg>` around
/// this already expresses on its own (`None` = flag not given at all).
#[derive(Debug, Clone, Copy)]
pub enum GammaShapeArg {
Fixed(f64),
Auto,
}
impl From<GammaShapeArg> for GammaShape {
fn from(arg: GammaShapeArg) -> Self {
match arg {
GammaShapeArg::Fixed(a) => GammaShape::Fixed(a),
GammaShapeArg::Auto => GammaShape::Auto,
}
}
}
pub fn parse_gamma_shape(s: &str) -> Result<GammaShapeArg, String> {
if s.eq_ignore_ascii_case("auto") || s.eq_ignore_ascii_case("estimate") {
return Ok(GammaShapeArg::Auto);
}
s.parse::<f64>()
.map(GammaShapeArg::Fixed)
.map_err(|_| format!("invalid --gamma-shape value {s:?}: expected a positive number or \"auto\"/\"estimate\""))
}
/// `--n-sites` CLI value: a number of sites to draw, or `all` for every site
/// — the draw then keeps every variable family that passes the other filters,
/// with no sampling involved. Its own tiny type (like [`GammaShapeArg`]) so
/// the two cases stay distinct up to the call that needs the target, instead
/// of an in-band sentinel (a negative number, `0`, ...) that a typo could hit.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum NSitesArg {
All,
N(usize),
}
impl NSitesArg {
/// The target to hand to the sampler: `None` = exhaustive.
pub fn target(self) -> Option<usize> {
match self {
NSitesArg::All => None,
NSitesArg::N(n) => Some(n),
}
}
}
pub fn parse_n_sites(s: &str) -> Result<NSitesArg, String> {
if s.eq_ignore_ascii_case("all") {
return Ok(NSitesArg::All);
}
s.parse::<usize>()
.map(NSitesArg::N)
.map_err(|_| format!("invalid --n-sites value {s:?}: expected a non-negative integer or \"all\""))
}
/// `--distance` value — either one of `obikphylo::DistanceMetric`'s
/// whole-index metrics (routed to `IndexCache::distance`) or one of
/// `obikphylo::SnpDistanceKind`'s `snp-*` corrections (routed to
/// `obikphylo::siblings::distance_matrix`, over a tally built by
/// `sampled_scan`/`PairwiseReducer`, the sibling-annex pipeline) — two
/// genuinely different code paths behind one CLI vocabulary, see
/// `DevDocMD/theory/evolutionary_distances.md`, "`--distance` unification".
#[derive(clap::ValueEnum, Clone, Copy, Debug)]
pub enum DistanceArg {
Jaccard,
Mash,
Hamming,
BrayCurtis,
#[value(name = "relfreq-bray-curtis")]
RelfreqBrayCurtis,
Euclidean,
#[value(name = "relfreq-euclidean")]
RelfreqEuclidean,
Hellinger,
#[value(name = "hellinger-euclidean")]
HellingerEuclidean,
#[value(name = "snp-raw")]
SnpRaw,
#[value(name = "snp-jc")]
SnpJc,
#[value(name = "snp-k2p")]
SnpK2p,
#[value(name = "snp-k81")]
SnpK81,
#[value(name = "snp-f81")]
SnpF81,
#[value(name = "snp-t92")]
SnpT92,
#[value(name = "snp-tn93")]
SnpTn93,
#[value(name = "snp-tv")]
SnpTv,
}
/// The distance matrix this invocation must compute, if any: the one asked for
/// with `--distance`, else `jaccard` when `--nj` or `--upgma` needs a matrix to
/// work from, else none — a matrix is never computed (nor written) just
/// because nothing said otherwise. `--shared-kmers` asks for its own matrix,
/// not for a distance.
pub fn distance_to_compute(distance: Option<DistanceArg>, nj: bool, upgma: bool) -> Option<DistanceArg> {
distance.or_else(|| (nj || upgma).then_some(DistanceArg::Jaccard))
}
#[cfg(test)]
mod distance_tests {
use super::*;
#[test]
fn nothing_asked_nothing_computed() {
assert!(distance_to_compute(None, false, false).is_none());
}
#[test]
fn tree_outputs_imply_jaccard() {
for (nj, upgma) in [(true, false), (false, true)] {
assert!(matches!(distance_to_compute(None, nj, upgma), Some(DistanceArg::Jaccard)));
}
}
#[test]
fn an_explicit_distance_wins() {
assert!(matches!(
distance_to_compute(Some(DistanceArg::Mash), true, true),
Some(DistanceArg::Mash)
));
assert!(matches!(
distance_to_compute(Some(DistanceArg::SnpJc), false, false),
Some(DistanceArg::SnpJc)
));
}
}
impl DistanceArg {
/// `Some` for the whole-index metrics, `None` for `snp-*` values.
pub fn as_classic(self) -> Option<DistanceMetric> {
Some(match self {
DistanceArg::Jaccard => DistanceMetric::Jaccard,
DistanceArg::Mash => DistanceMetric::Mash,
DistanceArg::Hamming => DistanceMetric::Hamming,
DistanceArg::BrayCurtis => DistanceMetric::BrayCurtis,
DistanceArg::RelfreqBrayCurtis => DistanceMetric::RelfreqBrayCurtis,
DistanceArg::Euclidean => DistanceMetric::Euclidean,
DistanceArg::RelfreqEuclidean => DistanceMetric::RelfreqEuclidean,
DistanceArg::Hellinger => DistanceMetric::Hellinger,
DistanceArg::HellingerEuclidean => DistanceMetric::HellingerEuclidean,
_ => return None,
})
}
/// `Some` for the `snp-*` values, `None` for the whole-index metrics.
pub fn as_snp(self) -> Option<SnpDistanceKind> {
Some(match self {
DistanceArg::SnpRaw => SnpDistanceKind::Raw,
DistanceArg::SnpJc => SnpDistanceKind::Jc,
DistanceArg::SnpK2p => SnpDistanceKind::K2p,
DistanceArg::SnpK81 => SnpDistanceKind::K81,
DistanceArg::SnpF81 => SnpDistanceKind::F81,
DistanceArg::SnpT92 => SnpDistanceKind::T92,
DistanceArg::SnpTn93 => SnpDistanceKind::Tn93,
DistanceArg::SnpTv => SnpDistanceKind::Tv,
_ => return None,
})
}
}
+11 -6
View File
@@ -1,3 +1,8 @@
// Static musl builds: replace musl's allocator (see obikmer/Cargo.toml).
#[cfg(target_env = "musl")]
#[global_allocator]
static GLOBAL: tikv_jemallocator::Jemalloc = tikv_jemallocator::Jemalloc;
mod cli; mod cli;
mod cmd; mod cmd;
@@ -8,10 +13,11 @@ use tracing_subscriber::{EnvFilter, fmt};
#[derive(Parser)] #[derive(Parser)]
#[command(name = "obikmer2", about = "DNA k-mer tools", version)] #[command(name = "obikmer2", about = "DNA k-mer tools", version)]
struct Cli { struct Cli {
/// Hard ceiling on the number of CPU cores the process may use — bounds /// Maximum number of CPU cores to use
/// both the command's worker pool and every internal rayon pool. ///
/// Can only lower the budget, never raise it above the cores actually /// Caps every thread pool of the process. It can only lower the number of
/// available to the process. Defaults to that available count. /// cores, never raise it above those available. By default all available
/// cores may be used.
#[arg(long, global = true, value_name = "N")] #[arg(long, global = true, value_name = "N")]
cpu_max: Option<usize>, cpu_max: Option<usize>,
@@ -47,8 +53,7 @@ enum Commands {
Utils(cmd::utils::UtilsArgs), Utils(cmd::utils::UtilsArgs),
/// Convert an index's evidence representation (exact/approximate/hybrid), in place /// Convert an index's evidence representation (exact/approximate/hybrid), in place
Convert(cmd::convert::ConvertArgs), Convert(cmd::convert::ConvertArgs),
/// Genome-vs-genome distance matrix (+ optional NJ/UPGMA tree, sibling-annex-based #[command(after_long_help = cmd::phylo::WORKFLOW_HELP)]
/// SNP corrections, Sankoff/TNT/PhyG/IQ-TREE exports)
Phylo(cmd::phylo::PhyloArgs), Phylo(cmd::phylo::PhyloArgs),
} }
-2
View File
@@ -15,5 +15,3 @@ obisys = { path = "../obisys", default-features = false }
rayon = "1" rayon = "1"
tracing = "0.1.44" tracing = "0.1.44"
[dev-dependencies]
tempfile = "3"
+55 -3
View File
@@ -3,6 +3,7 @@ use std::fs;
use std::io; use std::io;
use std::path::{Path, PathBuf}; use std::path::{Path, PathBuf};
use obicompactvec::{PersistentBitMatrix, PersistentIntMatrix, StorageKind};
use obikalgorithm::Algorithm; use obikalgorithm::Algorithm;
use obikindex::layer::IndexMode; use obikindex::layer::IndexMode;
use obikindex::partition::PARTITIONS_SUBDIR; use obikindex::partition::PARTITIONS_SUBDIR;
@@ -176,7 +177,17 @@ impl Algorithm for Merge<'_> {
); );
} }
let base_idx = choose_base(sources, &src_genomes, self.mode); let Some(base_idx) = choose_base(sources, &src_genomes, self.mode) else {
return Err(
"merge: none of the sources can be the destination. Every candidate has its \
matrices packed (finalized index): a packed index can be merged in as a \
source, but at least one source must be unpacked — typically a freshly \
built mono-genome index. To extend a packed index, merge it with new \
mono-genome indexes; to combine several packed indexes, merge their \
original sources instead."
.into(),
);
};
let needs_approx = sources.iter().enumerate().any(|(i, src)| { let needs_approx = sources.iter().enumerate().any(|(i, src)| {
!is_trivial(&src_genomes[i], self.mode) !is_trivial(&src_genomes[i], self.mode)
&& matches!( && matches!(
@@ -490,7 +501,15 @@ fn index_unitig_size(src: &KmerIndex) -> u64 {
(0..n).map(|i| partition_unitig_bytes(src, i)).sum() (0..n).map(|i| partition_unitig_bytes(src, i)).sum()
} }
fn choose_base(sources: &[&KmerIndex], src_genomes: &[Vec<GenomeInfo>], mode: MergeMode) -> usize { /// The source whose partition data is copied as the merge destination: the
/// largest one (least data to merge in). `None` when several sources are
/// given and none of them can be resumed as a destination (see
/// [`can_be_destination`]).
fn choose_base(
sources: &[&KmerIndex],
src_genomes: &[Vec<GenomeInfo>],
mode: MergeMode,
) -> Option<usize> {
let needs_approx = sources.iter().enumerate().any(|(i, src)| { let needs_approx = sources.iter().enumerate().any(|(i, src)| {
!is_trivial(&src_genomes[i], mode) !is_trivial(&src_genomes[i], mode)
&& matches!( && matches!(
@@ -509,9 +528,42 @@ fn choose_base(sources: &[&KmerIndex], src_genomes: &[Vec<GenomeInfo>], mode: Me
IndexMode::Approx { .. } | IndexMode::Hybrid { .. } IndexMode::Approx { .. } | IndexMode::Hybrid { .. }
) )
}) })
// With a single source there is nothing to append: the bootstrap
// copy is the whole merge, whatever its matrix format.
.filter(|(i, src)| sources.len() == 1 || can_be_destination(src, &src_genomes[*i], mode))
.max_by_key(|(_, src)| index_unitig_size(src)) .max_by_key(|(_, src)| index_unitig_size(src))
.map(|(i, _)| i) .map(|(i, _)| i)
.unwrap() }
/// Whether `index` can be the destination of a merge: pass 2 appends the
/// other sources' columns to its matrices (`MatrixBuilder::resume`), which
/// needs them in the column-by-column format. A finalized index has its
/// matrices packed (sparse by default, or dense with `--dense`) into
/// single-file formats that cannot be extended — such an index can only
/// come in as a *source*. An implicit matrix (mono-genome presence) is
/// materialized on the fly, so it qualifies only for a single genome.
///
/// One layer is probed: `pack` rewrites every partition the same way. An
/// index with no k-mer at all has nothing to resume and qualifies.
fn can_be_destination(index: &KmerIndex, genomes: &[GenomeInfo], mode: MergeMode) -> bool {
let layer_dir = (0..index.n_partitions()).find_map(|i| {
let partition = index.partition(i).ok()?;
(partition.n_layers() > 0)
.then(|| partition.layer_dir(0).ok())
.flatten()
});
let Some(layer_dir) = layer_dir else {
return true;
};
let kind = match mode {
MergeMode::Presence => PersistentBitMatrix::detect_storage(&layer_dir),
MergeMode::Count => PersistentIntMatrix::detect_storage(&layer_dir),
};
match kind {
Ok(StorageKind::Columnar) => true,
Ok(StorageKind::Implicit) => genomes.len() == 1,
_ => false,
}
} }
/// Copies `src` (the base source's whole `partitions/` tree) into `dst` at /// Copies `src` (the base source's whole `partitions/` tree) into `dst` at
-605
View File
@@ -1,605 +0,0 @@
//! Merging a source partition's new layer into a destination partition's
//! de Bruijn graph union (pass 1) then column fill (pass 2). Free function,
//! not a `KmerIndex` method — `KmerIndex` stays the stateless
//! `Index { Partition { Layer } }` data model; this is one of the
//! algorithms that reads/writes through it, living here in `obikmerge`
//! instead.
use std::fs;
use std::path::PathBuf;
use std::sync::{Arc, Mutex};
use obipipeline::{
Pipeline, PipelineError, PipelineSender, SharedFlatFn, Stage, ThrottleGuard, WorkerPool,
make_sink, make_source, make_transform, throttle,
};
use tracing::debug;
use obicompactvec::{ColBuilder, MatrixBuilder};
use obikindex::layer::{IndexMode, KmerLayer, TypedLayer};
use obikindex::{KmerIndex, OKIError, OKIResult};
use obikindexer::{build_graph, materialize_layer};
use obikseq::CanonicalKmer;
use obiskio::UnitigFileReader;
use crate::MergeMode;
// `MatrixBuilder::new`/`resume` take a `presence: bool` — `MergeMode` maps
// to it directly (`Presence` → `true`, `Count` → `false`).
fn presence(mode: MergeMode) -> bool {
mode == MergeMode::Presence
}
#[cfg(test)]
mod matrix_builder_tests {
use tempfile::tempdir;
use obicompactvec::{PersistentBitMatrix, PersistentIntMatrix};
use super::{ColBuilder, MatrixBuilder};
/// Mirrors `merge_partition`'s "new layer" setup: absent (dst-genome)
/// columns get no data, then source columns are filled — all through one
/// continuous `MatrixBuilder` session, closed once at the end.
#[test]
fn new_layer_absent_then_source_columns_presence() {
let dir = tempdir().unwrap();
let data_dir = dir.path().join("presence");
let mut mb = MatrixBuilder::new(true, 3, &data_dir).unwrap();
// Two absent (dst-genome) columns.
mb.add_absent_col().unwrap();
mb.add_absent_col().unwrap();
// One source column, filled like pass 2 would.
let mut col = mb.add_col().unwrap();
match &mut col {
ColBuilder::Bit(b) => {
b.set(0, true);
b.set(1, false);
b.set(2, true);
}
ColBuilder::Int(_) => unreachable!(),
}
col.close().unwrap();
mb.close().unwrap();
let m = PersistentBitMatrix::open(dir.path()).unwrap();
assert_eq!(m.n_cols(), 3);
assert_eq!(&*m.row(0), &[false, false, true]);
assert_eq!(&*m.row(1), &[false, false, false]);
assert_eq!(&*m.row(2), &[false, false, true]);
}
#[test]
fn new_layer_absent_then_source_columns_count() {
let dir = tempdir().unwrap();
let data_dir = dir.path().join("counts");
let mut mb = MatrixBuilder::new(false, 2, &data_dir).unwrap();
mb.add_absent_col().unwrap();
let mut col = mb.add_col().unwrap();
match &mut col {
ColBuilder::Int(b) => {
b.set(0, 7);
b.set(1, 42);
}
ColBuilder::Bit(_) => unreachable!(),
}
col.close().unwrap();
mb.close().unwrap();
let m = PersistentIntMatrix::open(dir.path()).unwrap();
assert_eq!(m.n_cols(), 2);
assert_eq!(&*m.row(0), &[0u32, 7]);
assert_eq!(&*m.row(1), &[0u32, 42]);
}
/// Mirrors `merge_partition`'s "existing dst layer" setup: an
/// already-closed matrix (from a previous merge) gets more columns
/// appended via `resume`, without callers ever seeing `col_path`/
/// `MatrixMeta` — `n` itself is read back from the matrix directory.
#[test]
fn resume_appends_source_columns_to_existing_layer() {
let dir = tempdir().unwrap();
let data_dir = dir.path().join("presence");
// Previous merge: one dst-genome column already on disk.
let mut mb0 = MatrixBuilder::new(true, 3, &data_dir).unwrap();
mb0.add_absent_col().unwrap();
mb0.close().unwrap();
// This merge: resume and append two more source columns.
let mut mb = MatrixBuilder::resume(true, &data_dir).unwrap();
for vals in [[true, false, true], [false, true, false]] {
let mut col = mb.add_col().unwrap();
match &mut col {
ColBuilder::Bit(b) => {
for (slot, v) in vals.into_iter().enumerate() {
b.set(slot, v);
}
}
ColBuilder::Int(_) => unreachable!(),
}
col.close().unwrap();
}
mb.close().unwrap();
let m = PersistentBitMatrix::open(dir.path()).unwrap();
assert_eq!(m.n_cols(), 3);
assert_eq!(&*m.row(0), &[false, true, false]);
assert_eq!(&*m.row(1), &[false, false, true]);
assert_eq!(&*m.row(2), &[false, true, false]);
}
}
// ── dst cross-layer lookup ──────────────────────────────────────────────────
/// Batch membership lookup across `dst_layers` — first layer that carries
/// each kmer wins. Groups hits by layer index (as `(kmer batch index,
/// slot)` pairs) rather than returning them in query order, so pass 2 can
/// follow with one column-major write per `(layer, column)` instead of
/// interleaving across layers kmer by kmer. The kmers with no hit in any
/// layer come back as a separate index list, for the new-layer MPHF batch
/// lookup on a miss.
fn find_batch_grouped(
layers: &[KmerLayer],
kmers: &[CanonicalKmer],
) -> (Vec<Vec<(usize, usize)>>, Vec<usize>) {
let mut by_layer: Vec<Vec<(usize, usize)>> = vec![Vec::new(); layers.len()];
let mut not_found: Vec<usize> = Vec::new();
for (idx, &kmer) in kmers.iter().enumerate() {
match layers
.iter()
.enumerate()
.find_map(|(li, l)| l.find_slot(kmer).map(|slot| (li, slot)))
{
Some((li, slot)) => by_layer[li].push((idx, slot)),
None => not_found.push(idx),
}
}
(by_layer, not_found)
}
// ── merge_partition ────────────────────────────────────────────────────────
/// Merge `sources` into destination partition `i`.
///
/// Each entry in `sources` is `(index, n_genomes)` where `n_genomes` is the
/// number of genome columns that source contributes. A merged index
/// contributes more than one. The total new columns added to the
/// destination is `sum(n_genomes)`.
///
/// `n_dst_genomes` is the number of genome columns already in the
/// destination matrices (copied from the bootstrap source before this
/// call).
pub(crate) fn merge_partition(
dst: &KmerIndex,
i: usize,
sources: &[(&KmerIndex, usize)],
mode: MergeMode,
n_dst_genomes: usize,
block_bits: u8,
evidence: &IndexMode,
) -> OKIResult<usize> {
let dst_partition = dst.partition(i)?;
let n_dst_layers = dst_partition.n_layers();
let n_src_total: usize = sources.iter().map(|(_, n)| *n).sum();
// First merge in presence mode: materialize implicit single-genome
// presence matrices on existing layers (all slots true — every kmer in
// those layers belongs to genome_0) so a resumed builder can append
// source columns onto them below.
if n_dst_genomes == 1 && mode == MergeMode::Presence {
for l in 0..n_dst_layers {
let layer = dst_partition.layer(l)?.open()?;
TypedLayer::<()>::init_presence_matrix(layer.dir(), layer.n())?;
}
}
// dst's `0..n_dst_layers` layers and every source's own layers, for
// this one partition — all opened the same direct way, not via
// `IndexCache`. Every layer read here (dst's pre-existing ones, and
// every source's) is just as stable for the duration of this call: dst
// only ever gets a brand new layer *appended* below, never one of these
// read from, so there's no "dst isn't stable" distinction between the
// two loops — `IndexCache` still doesn't fit either one, for a single,
// uniform reason: `obikindexer::build_graph`/`obipipeline::WorkerPool`
// (pass 1/2 below) are plain-OS-thread pipelines requiring `Send + Sync
// + 'static`, and `IndexCache<'_>` always borrows `&KmerIndex` — no
// amount of logical stability changes that a borrowed cache can't cross
// that boundary, only owned `KmerLayer`s can (each owns its own mmap'd
// files, no reference back to the index it came from).
let dst_layers: Vec<KmerLayer> = (0..n_dst_layers)
.map(|l| dst_partition.layer(l)?.open())
.collect::<OKIResult<_>>()?;
let dst_layers = Arc::new(dst_layers);
let src_layers: Vec<Vec<Arc<KmerLayer>>> = sources
.iter()
.map(|(src, _)| -> OKIResult<Vec<Arc<KmerLayer>>> {
let partition = src.partition(i)?;
(0..partition.n_layers())
.map(|l| partition.layer(l)?.open().map(Arc::new))
.collect()
})
.collect::<OKIResult<_>>()?;
// ── Pass 1: pipeline — parallel file read + dst filter + graph fill ────
//
// Source : list of unitigs.bin paths (one per source × layer)
// Flat : open file, emit Vec<CanonicalKmer> batches (BeeGFS parallel I/O)
// Transform: filter via dst_layers — Arc<[KmerLayer]>: Sync
// Sink : push new kmers into GraphDeBruijn (single thread, no locks needed)
let mut unitig_paths: Vec<PathBuf> = Vec::new();
for layers in &src_layers {
for layer in layers {
let p = layer.unitigs_path();
if p.exists() {
unitig_paths.push(p);
}
}
}
let n_src_layers = unitig_paths.len();
debug!("partition {i}: de Bruijn graph build start — {n_src_layers} source layer(s)");
const BATCH: usize = 4096;
let n_workers = rayon::current_num_threads().min(16).max(4);
// At most 2 files open simultaneously: keeps n_workers-2 workers free
// for the Transform stage. Each open file monopolises one worker for the
// full duration of its read, so this must stay well below n_workers.
let max_open = 2;
let dst_filter = Arc::clone(&dst_layers);
let g = build_graph(
unitig_paths.into_iter(),
move |path: PathBuf, emit: &mut dyn FnMut(Vec<CanonicalKmer>)| -> OKIResult<()> {
let reader = UnitigFileReader::open_sequential(&path)?;
let mut batch: Vec<CanonicalKmer> = Vec::with_capacity(BATCH);
for (kmer, _, _) in reader.iter_indexed_canonical_kmers() {
batch.push(kmer);
if batch.len() == BATCH {
emit(std::mem::replace(&mut batch, Vec::with_capacity(BATCH)));
}
}
if !batch.is_empty() {
emit(batch);
}
Ok(())
},
move |kmer| dst_filter.iter().all(|l| l.find_slot(kmer).is_none()),
n_workers,
max_open,
)?;
let any_new = g.len() > 0;
debug!(
"partition {i}: de Bruijn graph done — {} new kmers",
g.len()
);
// Build new layer from de Bruijn graph if there are new kmers.
let new_layer_idx = n_dst_layers;
let new_layer_dir = KmerLayer::new(&dst_partition, new_layer_idx)
.dir()
.to_path_buf();
let n_new = if any_new {
debug!("partition {i}: unitig traversal start — {} nodes", g.len());
let n_nodes = materialize_layer(g, &new_layer_dir, block_bits, evidence)?;
debug!("partition {i}: MPHF build done");
n_nodes
} else {
drop(g);
0
};
let t_open = std::time::Instant::now();
let new_mphf: Option<Arc<obikindex::layer::MphfOnly>> = if any_new {
Some(Arc::new(obikindex::layer::MphfOnly::open(&new_layer_dir)?))
} else {
None
};
debug!(
"partition {i}: MPHF open in {:.3}s",
t_open.elapsed().as_secs_f64()
);
// ── Prepare matrix directories for the new layer ──────────────────────
// Absent columns (dst genomes) get an all-zero/false column. Source-genome
// columns are created as mutable builders for pass 2. `new_mb` is kept
// open (not `close`d) until pass 2 has filled every source column, so its
// `n_cols` bookkeeping stays in sync with the columns actually created.
let (new_src_builders, new_mb): (Vec<ColBuilder>, Option<MatrixBuilder>) = if any_new {
let data_dir = match mode {
MergeMode::Presence => new_layer_dir.join("presence"),
MergeMode::Count => new_layer_dir.join("counts"),
};
fs::create_dir_all(&data_dir)?;
let mut mb = MatrixBuilder::new(presence(mode), n_new, &data_dir).map_err(OKIError::Io)?;
for _ in 0..n_dst_genomes {
mb.add_absent_col().map_err(OKIError::Io)?;
}
let cols = (0..n_src_total)
.map(|_| mb.add_col().map_err(OKIError::Io))
.collect::<OKIResult<Vec<_>>>()?;
(cols, Some(mb))
} else {
(vec![], None)
};
let t_builders = std::time::Instant::now();
// Builders for existing layers: n_src_total per layer, resumed from
// each layer's own matrix directory (already holding n_dst_genomes
// columns from a previous merge). Columns land at
// n_dst_genomes .. n_dst_genomes + n_src_total - 1.
let mut exist_mbs: Vec<MatrixBuilder> = Vec::with_capacity(n_dst_layers);
let mut exist_builders: Vec<Vec<ColBuilder>> = Vec::with_capacity(n_dst_layers);
for l in 0..n_dst_layers {
let layer_dir = dst_layers[l].dir().to_path_buf();
let data_dir = match mode {
MergeMode::Presence => layer_dir.join("presence"),
MergeMode::Count => layer_dir.join("counts"),
};
let mut mb = MatrixBuilder::resume(presence(mode), &data_dir).map_err(OKIError::Io)?;
let cols = (0..n_src_total)
.map(|_| mb.add_col().map_err(OKIError::Io))
.collect::<OKIResult<Vec<_>>>()?;
exist_mbs.push(mb);
exist_builders.push(cols);
}
debug!(
"partition {i}: builders ready in {:.3}s",
t_builders.elapsed().as_secs_f64()
);
// ── Pass 2: fill builders (pipeline) ─────────────────────────────────
let t_pass2 = std::time::Instant::now();
// Collect source items before the pipeline so errors propagate via ?
// before any worker thread is spawned.
let mut pass2_items: Vec<(usize, usize, PathBuf, Arc<KmerLayer>)> = Vec::new();
{
let mut col_offset = 0usize;
for ((_, src_n), layers) in sources.iter().zip(src_layers.iter()) {
for layer in layers {
let unitigs_path = layer.unitigs_path();
if unitigs_path.exists() {
pass2_items.push((col_offset, *src_n, unitigs_path, Arc::clone(layer)));
}
}
col_offset += src_n;
}
}
enum Pass2Data {
SrcLayer((usize, usize, PathBuf, Arc<KmerLayer>, ThrottleGuard)),
RawBatch((usize, usize, Arc<KmerLayer>, Vec<CanonicalKmer>)),
/// One entry per `(dst layer, absolute column)` touched by this
/// batch, each carrying every `(slot, value)` pair for that column
/// together — column-major, so the sink below locks that column's
/// `ColBuilder` once per batch and writes the whole run, instead of
/// once per `(kmer, genome)` pair.
WriteBatch(Vec<(Option<usize>, usize, Vec<(usize, u32)>)>),
}
let exist_locked: Vec<Vec<Arc<Mutex<ColBuilder>>>> = exist_builders
.into_iter()
.map(|layer| layer.into_iter().map(|b| Arc::new(Mutex::new(b))).collect())
.collect();
let new_locked: Vec<Arc<Mutex<ColBuilder>>> = new_src_builders
.into_iter()
.map(|b| Arc::new(Mutex::new(b)))
.collect();
let exist_sink: Vec<Vec<Arc<Mutex<ColBuilder>>>> = exist_locked
.iter()
.map(|layer| layer.iter().map(Arc::clone).collect())
.collect();
let new_sink: Vec<Arc<Mutex<ColBuilder>>> = new_locked.iter().map(Arc::clone).collect();
let dst_layers_t2 = Arc::clone(&dst_layers);
let new_mphf_t2 = new_mphf.clone();
let pass2_err: Arc<Mutex<Option<String>>> = Arc::new(Mutex::new(None));
let err_cap2 = Arc::clone(&pass2_err);
let capacity = 2;
let throttled_pass2 = throttle(pass2_items.into_iter(), max_open);
let pipeline2 = Pipeline::new(
make_source!(
Pass2Data,
throttled_pass2.map(|t| {
let (col_offset, src_n, unitigs_path, src_layer) = t.item;
(col_offset, src_n, unitigs_path, src_layer, t.guard)
}),
SrcLayer
),
vec![
Stage::Flat(Arc::new(
move |data: Pass2Data,
push: &PipelineSender<Result<Pass2Data, PipelineError>>,
delta: &PipelineSender<isize>| {
if let Pass2Data::SrcLayer((
col_offset,
src_n,
unitigs_path,
src_layer,
_guard,
)) = data
{
// _guard dropped at end of block, releasing the slot.
// `src_layer` (MPHF + matrix) was already opened up
// front — only the unitigs.bin stream itself is
// opened (and throttled) here.
let reader = match UnitigFileReader::open_sequential(&unitigs_path) {
Ok(r) => r,
Err(e) => {
*err_cap2.lock().unwrap() = Some(e.to_string());
delta.send(-1).ok();
return;
}
};
const BATCH: usize = 4096;
let mut batch: Vec<CanonicalKmer> = Vec::with_capacity(BATCH);
let mut count: isize = 0;
for (kmer, _, _) in reader.iter_indexed_canonical_kmers() {
batch.push(kmer);
if batch.len() == BATCH {
let b = std::mem::replace(&mut batch, Vec::with_capacity(BATCH));
push.send(Ok(Pass2Data::RawBatch((
col_offset,
src_n,
Arc::clone(&src_layer),
b,
))))
.ok();
count += 1;
}
}
if !batch.is_empty() {
push.send(Ok(Pass2Data::RawBatch((
col_offset, src_n, src_layer, batch,
))))
.ok();
count += 1;
}
delta.send(count - 1).ok();
}
},
) as SharedFlatFn<Pass2Data>),
make_transform!(
Pass2Data,
{
move |(col_offset, src_n, src_layer, kmers): (
usize,
usize,
Arc<KmerLayer>,
Vec<CanonicalKmer>,
)|
-> Vec<(Option<usize>, usize, Vec<(usize, u32)>)> {
// Column-first, one pass for the whole batch: a single
// MPHF hash sweep (`hash_batch`) instead of one hash
// per kmer, then one read pass per source genome
// column (`fill_sub_matrix`) instead of one row read
// per kmer — matches the underlying matrix's own
// column-major layout.
let slots = src_layer.hash_batch(&kmers);
let mut cols: Vec<Vec<u32>> = (0..src_n)
.map(|_| Vec::with_capacity(kmers.len()))
.collect();
src_layer.fill_sub_matrix(&slots, &mut cols);
// Membership against dst, grouped by dst layer
// instead of interleaved kmer by kmer — same
// grouping `PartitionCache::find_presence_batch`
// does in `obikphylo`, via the shared primitive.
let (by_dst_layer, not_found) = find_batch_grouped(&dst_layers_t2, &kmers);
// Genome-major within each layer group (not kmer-major)
// so every `(layer, col)` pair's values land together —
// one entry per column, not one per `(kmer, genome)`.
let mut ops: Vec<(Option<usize>, usize, Vec<(usize, u32)>)> =
Vec::with_capacity(src_n * (by_dst_layer.len() + 1));
for (dst_layer, hits) in by_dst_layer.into_iter().enumerate() {
if hits.is_empty() {
continue;
}
for g in 0..src_n {
let col_vals: Vec<(usize, u32)> = hits
.iter()
.map(|&(idx, slot)| (slot, cols[g][idx]))
.collect();
ops.push((Some(dst_layer), col_offset + g, col_vals));
}
}
if let (Some(mphf), false) = (&new_mphf_t2, not_found.is_empty()) {
let not_found_kmers: Vec<CanonicalKmer> =
not_found.iter().map(|&idx| kmers[idx]).collect();
let new_slots = mphf.index_batch(&not_found_kmers);
for g in 0..src_n {
let col_vals: Vec<(usize, u32)> = not_found
.iter()
.zip(new_slots.iter())
.map(|(&idx, &slot)| (slot, cols[g][idx]))
.collect();
ops.push((None, col_offset + g, col_vals));
}
}
ops
}
},
RawBatch,
WriteBatch
),
],
make_sink!(
Pass2Data,
{
move |ops: Vec<(Option<usize>, usize, Vec<(usize, u32)>)>| {
for (layer_opt, col, col_vals) in ops {
let builder = match layer_opt {
Some(l) => &exist_sink[l][col],
None => &new_sink[col],
};
let mut b = builder.lock().unwrap();
for (slot, val) in col_vals {
b.set_val(slot, val);
}
}
}
},
WriteBatch
),
);
WorkerPool::new(pipeline2, n_workers, capacity).run();
debug!(
"partition {i}: pass2 pipeline done in {:.3}s",
t_pass2.elapsed().as_secs_f64()
);
if let Some(msg) = Arc::try_unwrap(pass2_err)
.unwrap_or_else(|_| panic!("pass2: pass2_err not uniquely owned"))
.into_inner()
.unwrap_or_else(|e| e.into_inner())
{
return Err(OKIError::InvalidData {
context: "merge pass2",
detail: msg,
});
}
let t_close = std::time::Instant::now();
// ── Close builders ──────────────────────────────────────────────────
for (mb, builders) in exist_mbs.into_iter().zip(exist_locked.into_iter()) {
for b in builders {
Arc::try_unwrap(b)
.unwrap_or_else(|_| panic!("pass2: exist_builder not uniquely owned"))
.into_inner()
.unwrap_or_else(|e| e.into_inner())
.close()?;
}
mb.close().map_err(OKIError::Io)?;
}
for b in new_locked {
Arc::try_unwrap(b)
.unwrap_or_else(|_| panic!("pass2: new_builder not uniquely owned"))
.into_inner()
.unwrap_or_else(|e| e.into_inner())
.close()?;
}
if let Some(mb) = new_mb {
mb.close().map_err(OKIError::Io)?;
}
debug!(
"partition {i}: builders closed in {:.3}s",
t_close.elapsed().as_secs_f64()
);
Ok(n_new)
}
+304
View File
@@ -0,0 +1,304 @@
//! Merging a source partition's new layer into a destination partition's
//! de Bruijn graph union (pass 1) then column fill (pass 2). Free function,
//! not a `KmerIndex` method — `KmerIndex` stays the stateless
//! `Index { Partition { Layer } }` data model; this is one of the
//! algorithms that reads/writes through it, living here in `obikmerge`
//! instead.
//!
//! Split: this file orchestrates one partition (pass 1, new-layer setup,
//! loop over source groups); [`pass2`] holds the per-group column-fill
//! pipeline.
mod pass2;
use std::fs;
use std::path::{Path, PathBuf};
use std::sync::Arc;
use tracing::debug;
use obicompactvec::{ColBuilder, MatrixBuilder};
use obikindex::layer::{IndexMode, KmerLayer, TypedLayer};
use obikindex::{KmerIndex, OKIError, OKIResult};
use obikindexer::{build_graph, materialize_layer};
use obikseq::CanonicalKmer;
use obiskio::UnitigFileReader;
use crate::MergeMode;
use pass2::run_pass2;
// `MatrixBuilder::new`/`resume` take a `presence: bool` — `MergeMode` maps
// to it directly (`Presence` → `true`, `Count` → `false`).
fn presence(mode: MergeMode) -> bool {
mode == MergeMode::Presence
}
// ── merge_partition ────────────────────────────────────────────────────────
/// Merge `sources` into destination partition `i`.
///
/// Each entry in `sources` is `(index, n_genomes)` where `n_genomes` is the
/// number of genome columns that source contributes. A merged index
/// contributes more than one. The total new columns added to the
/// destination is `sum(n_genomes)`.
///
/// `n_dst_genomes` is the number of genome columns already in the
/// destination matrices (copied from the bootstrap source before this
/// call).
pub(crate) fn merge_partition(
dst: &KmerIndex,
i: usize,
sources: &[(&KmerIndex, usize)],
mode: MergeMode,
n_dst_genomes: usize,
block_bits: u8,
evidence: &IndexMode,
) -> OKIResult<usize> {
let dst_partition = dst.partition(i)?;
let n_dst_layers = dst_partition.n_layers();
// First merge in presence mode: materialize implicit single-genome
// presence matrices on existing layers (all slots true — every kmer in
// those layers belongs to genome_0) so a resumed builder can append
// source columns onto them below.
if n_dst_genomes == 1 && mode == MergeMode::Presence {
for l in 0..n_dst_layers {
let layer = dst_partition.layer(l)?.open()?;
TypedLayer::<()>::init_presence_matrix(layer.dir(), layer.n())?;
}
}
// dst's `0..n_dst_layers` layers, opened once for the whole partition
// and shared by pass 1 (filter) and every pass-2 group (membership).
// Opened directly, not via `IndexCache`: `obikindexer::build_graph`/
// `obipipeline::WorkerPool` are plain-OS-thread pipelines requiring
// `Send + Sync + 'static`, and `IndexCache<'_>` always borrows
// `&KmerIndex` — only owned `KmerLayer`s can cross that boundary.
let dst_layers: Vec<KmerLayer> = (0..n_dst_layers)
.map(|l| dst_partition.layer(l)?.open())
.collect::<OKIResult<_>>()?;
let dst_layers = Arc::new(dst_layers);
// ── Pass 1: pipeline — parallel file read + dst filter + graph fill ────
//
// Source : list of unitigs.bin paths (one per source × layer)
// Flat : open file, emit Vec<CanonicalKmer> batches (BeeGFS parallel I/O)
// Transform: filter via dst_layers — Arc<[KmerLayer]>: Sync
// Sink : push new kmers into GraphDeBruijn (single thread, no locks needed)
//
// Only the unitigs.bin *paths* are needed here: no source layer is
// opened (`KmerLayer::new` is the unopened `Empty` variant, whose
// `unitigs_path` is valid), so pass 1 holds no source mapping.
let mut unitig_paths: Vec<PathBuf> = Vec::new();
for (src, _) in sources {
let partition = src.partition(i)?;
for l in 0..partition.n_layers() {
let p = KmerLayer::new(&partition, l).unitigs_path();
if p.exists() {
unitig_paths.push(p);
}
}
}
let n_src_layers = unitig_paths.len();
debug!("partition {i}: de Bruijn graph build start — {n_src_layers} source layer(s)");
const BATCH: usize = 4096;
let n_workers = rayon::current_num_threads().min(16).max(4);
// At most 2 files open simultaneously: keeps n_workers-2 workers free
// for the Transform stage. Each open file monopolises one worker for the
// full duration of its read, so this must stay well below n_workers.
let max_open = 2;
let dst_filter = Arc::clone(&dst_layers);
let g = build_graph(
unitig_paths.into_iter(),
move |path: PathBuf, emit: &mut dyn FnMut(Vec<CanonicalKmer>)| -> OKIResult<()> {
let reader = UnitigFileReader::open_sequential(&path)?;
let mut batch: Vec<CanonicalKmer> = Vec::with_capacity(BATCH);
for (kmer, _, _) in reader.iter_indexed_canonical_kmers() {
batch.push(kmer);
if batch.len() == BATCH {
emit(std::mem::replace(&mut batch, Vec::with_capacity(BATCH)));
}
}
if !batch.is_empty() {
emit(batch);
}
Ok(())
},
move |kmer| dst_filter.iter().all(|l| l.find_slot(kmer).is_none()),
n_workers,
max_open,
)?;
let any_new = g.len() > 0;
debug!(
"partition {i}: de Bruijn graph done — {} new kmers",
g.len()
);
// Build new layer from de Bruijn graph if there are new kmers.
let new_layer_idx = n_dst_layers;
let new_layer_dir = KmerLayer::new(&dst_partition, new_layer_idx)
.dir()
.to_path_buf();
let n_new = if any_new {
debug!("partition {i}: unitig traversal start — {} nodes", g.len());
let n_nodes = materialize_layer(g, &new_layer_dir, block_bits, evidence)?;
debug!("partition {i}: MPHF build done");
n_nodes
} else {
drop(g);
0
};
let t_open = std::time::Instant::now();
let new_mphf: Option<Arc<obikindex::layer::MphfOnly>> = if any_new {
Some(Arc::new(obikindex::layer::MphfOnly::open(&new_layer_dir)?))
} else {
None
};
debug!(
"partition {i}: MPHF open in {:.3}s",
t_open.elapsed().as_secs_f64()
);
let matrix_dir = |layer_dir: &Path| match mode {
MergeMode::Presence => layer_dir.join("presence"),
MergeMode::Count => layer_dir.join("counts"),
};
// ── New layer matrix: absent (dst-genome) columns only ────────────────
// Created and closed here; every source group below resumes it and
// appends its own columns, exactly like the existing dst layers.
let new_data_dir: Option<PathBuf> = if any_new {
let data_dir = matrix_dir(&new_layer_dir);
fs::create_dir_all(&data_dir)?;
let mut mb = MatrixBuilder::new(presence(mode), n_new, &data_dir).map_err(OKIError::Io)?;
for _ in 0..n_dst_genomes {
mb.add_absent_col().map_err(OKIError::Io)?;
}
mb.close().map_err(OKIError::Io)?;
Some(data_dir)
} else {
None
};
let exist_data_dirs: Vec<PathBuf> = dst_layers.iter().map(|l| matrix_dir(l.dir())).collect();
// ── Pass 2, by consecutive groups of sources ──────────────────────────
//
// Only one group's source layers and destination columns are alive at a
// time, so the number of live mappings per partition no longer grows
// with the number of sources. SRC_GROUP matches `max_open`: pass 2
// never streams more than `max_open` unitigs.bin files at once, so a
// larger group would only keep more layers/columns open without adding
// read parallelism.
//
// Invariant: groups are visited in source order and each appends its
// `group_n` columns in source order, so the absolute column of genome
// `g` of a source is `n_dst_genomes + col_base + offset_local + g` —
// the same column as when every source was processed in one session.
//
// Memory: the group's columns are built in RAM (`add_mem_col`), at most
// `SRC_GROUP × src_n × (L_dst + 1)` live columns per partition, each of
// `n_layer / 8` bytes (presence) or `n_layer` bytes + overflow (counts):
// about `SRC_GROUP × src_n × N_part × (1/8 | 1)` bytes, `N_part` being
// the partition's k-mer count over all dst layers. For mono-genome
// sources this stays below pass 1's de Bruijn graph (≥ ~19 B per new
// k-mer) unless the new layer is tiny next to the existing ones
// (incremental merge into a large index).
const SRC_GROUP: usize = 2;
let t_pass2 = std::time::Instant::now();
let mut col_base = 0usize;
for group in sources.chunks(SRC_GROUP) {
let group_n: usize = group.iter().map(|(_, n)| *n).sum();
// This group's source layers, for this partition only. Opened
// before any builder so errors propagate via `?` before any worker
// thread is spawned. `col_offset` is local to the group.
let mut items: Vec<(usize, usize, PathBuf, Arc<KmerLayer>)> = Vec::new();
let mut col_offset = 0usize;
for (src, src_n) in group {
let partition = src.partition(i)?;
for l in 0..partition.n_layers() {
let layer = Arc::new(partition.layer(l)?.open()?);
let unitigs_path = layer.unitigs_path();
if unitigs_path.exists() {
items.push((col_offset, *src_n, unitigs_path, layer));
}
}
col_offset += src_n;
}
let expected_cols = n_dst_genomes + col_base;
let open_group = |data_dir: &Path| -> OKIResult<(MatrixBuilder, Vec<ColBuilder>)> {
let mut mb = MatrixBuilder::resume(presence(mode), data_dir).map_err(OKIError::Io)?;
if mb.n_cols() != expected_cols {
return Err(OKIError::InvalidData {
context: "merge pass2",
detail: format!(
"{}: {} columns, expected {expected_cols}",
data_dir.display(),
mb.n_cols()
),
});
}
// In-memory columns: no sparse file / `MmapMut` on the write
// path, so a block the FS cannot allocate is an `io::Error` at
// `close`, not a SIGBUS (see `MatrixBuilder::add_mem_col`).
let cols = (0..group_n).map(|_| mb.add_mem_col()).collect();
Ok((mb, cols))
};
let (exist_mbs, exist_cols): (Vec<MatrixBuilder>, Vec<Vec<ColBuilder>>) = exist_data_dirs
.iter()
.map(|d| open_group(d))
.collect::<OKIResult<Vec<_>>>()?
.into_iter()
.unzip();
let (new_mb, new_cols) = match &new_data_dir {
Some(d) => {
let (mb, cols) = open_group(d)?;
(Some(mb), cols)
}
None => (None, vec![]),
};
// `items` (and with it this group's source layers) is consumed by
// the pipeline and dropped when it returns.
let (exist_cols, new_cols) = run_pass2(
items,
&dst_layers,
&new_mphf,
exist_cols,
new_cols,
n_workers,
max_open,
)?;
for (mb, cols) in exist_mbs.into_iter().zip(exist_cols) {
for c in cols {
c.close()?;
}
mb.close().map_err(OKIError::Io)?;
}
for c in new_cols {
c.close()?;
}
if let Some(mb) = new_mb {
mb.close().map_err(OKIError::Io)?;
}
col_base += group_n;
}
debug!(
"partition {i}: pass2 done in {:.3}s",
t_pass2.elapsed().as_secs_f64()
);
Ok(n_new)
}
+271
View File
@@ -0,0 +1,271 @@
//! Pass 2 of a partition merge, for one group of sources: stream each
//! source layer's k-mers, read their values from the source matrix, locate
//! them in dst (existing layers, else the new layer's MPHF) and write them
//! into the group's destination columns.
use std::path::PathBuf;
use std::sync::{Arc, Mutex};
use obipipeline::{
Pipeline, PipelineError, PipelineSender, SharedFlatFn, Stage, ThrottleGuard, WorkerPool,
make_sink, make_source, make_transform, throttle,
};
use obicompactvec::ColBuilder;
use obikindex::layer::KmerLayer;
use obikindex::{OKIError, OKIResult};
use obikseq::CanonicalKmer;
use obiskio::UnitigFileReader;
// ── dst cross-layer lookup ──────────────────────────────────────────────────
/// Batch membership lookup across `dst_layers` — first layer that carries
/// each kmer wins. Groups hits by layer index (as `(kmer batch index,
/// slot)` pairs) rather than returning them in query order, so pass 2 can
/// follow with one column-major write per `(layer, column)` instead of
/// interleaving across layers kmer by kmer. The kmers with no hit in any
/// layer come back as a separate index list, for the new-layer MPHF batch
/// lookup on a miss.
fn find_batch_grouped(
layers: &[KmerLayer],
kmers: &[CanonicalKmer],
) -> (Vec<Vec<(usize, usize)>>, Vec<usize>) {
let mut by_layer: Vec<Vec<(usize, usize)>> = vec![Vec::new(); layers.len()];
let mut not_found: Vec<usize> = Vec::new();
for (idx, &kmer) in kmers.iter().enumerate() {
match layers
.iter()
.enumerate()
.find_map(|(li, l)| l.find_slot(kmer).map(|slot| (li, slot)))
{
Some((li, slot)) => by_layer[li].push((idx, slot)),
None => not_found.push(idx),
}
}
(by_layer, not_found)
}
// ── per-group pipeline ─────────────────────────────────────────────────────
/// Pass 2 for one group of sources: stream each source layer's unitigs,
/// look its k-mers up in the source matrix and in dst (existing layers,
/// then the new layer's MPHF), and write the values into the group's
/// column builders. `items` carry group-local column offsets; the builders
/// are handed back (still open) for the caller to close.
#[allow(clippy::type_complexity)]
pub(super) fn run_pass2(
items: Vec<(usize, usize, PathBuf, Arc<KmerLayer>)>,
dst_layers: &Arc<Vec<KmerLayer>>,
new_mphf: &Option<Arc<obikindex::layer::MphfOnly>>,
exist_builders: Vec<Vec<ColBuilder>>,
new_builders: Vec<ColBuilder>,
n_workers: usize,
max_open: usize,
) -> OKIResult<(Vec<Vec<ColBuilder>>, Vec<ColBuilder>)> {
enum Pass2Data {
SrcLayer((usize, usize, PathBuf, Arc<KmerLayer>, ThrottleGuard)),
RawBatch((usize, usize, Arc<KmerLayer>, Vec<CanonicalKmer>)),
/// One entry per `(dst layer, column)` touched by this batch, each
/// carrying every `(slot, value)` pair for that column together —
/// column-major, so the sink below locks that column's `ColBuilder`
/// once per batch and writes the whole run, instead of once per
/// `(kmer, genome)` pair.
WriteBatch(Vec<(Option<usize>, usize, Vec<(usize, u32)>)>),
}
let exist_locked: Vec<Vec<Arc<Mutex<ColBuilder>>>> = exist_builders
.into_iter()
.map(|layer| layer.into_iter().map(|b| Arc::new(Mutex::new(b))).collect())
.collect();
let new_locked: Vec<Arc<Mutex<ColBuilder>>> = new_builders
.into_iter()
.map(|b| Arc::new(Mutex::new(b)))
.collect();
let exist_sink: Vec<Vec<Arc<Mutex<ColBuilder>>>> = exist_locked
.iter()
.map(|layer| layer.iter().map(Arc::clone).collect())
.collect();
let new_sink: Vec<Arc<Mutex<ColBuilder>>> = new_locked.iter().map(Arc::clone).collect();
let dst_layers_t2 = Arc::clone(dst_layers);
let new_mphf_t2 = new_mphf.clone();
let pass2_err: Arc<Mutex<Option<String>>> = Arc::new(Mutex::new(None));
let err_cap2 = Arc::clone(&pass2_err);
let capacity = 2;
let throttled_pass2 = throttle(items.into_iter(), max_open);
let pipeline2 = Pipeline::new(
make_source!(
Pass2Data,
throttled_pass2.map(|t| {
let (col_offset, src_n, unitigs_path, src_layer) = t.item;
(col_offset, src_n, unitigs_path, src_layer, t.guard)
}),
SrcLayer
),
vec![
Stage::Flat(Arc::new(
move |data: Pass2Data,
push: &PipelineSender<Result<Pass2Data, PipelineError>>,
delta: &PipelineSender<isize>| {
if let Pass2Data::SrcLayer((
col_offset,
src_n,
unitigs_path,
src_layer,
_guard,
)) = data
{
// _guard dropped at end of block, releasing the slot.
// `src_layer` (MPHF + matrix) was already opened by
// the caller — only the unitigs.bin stream itself is
// opened (and throttled) here.
let reader = match UnitigFileReader::open_sequential(&unitigs_path) {
Ok(r) => r,
Err(e) => {
*err_cap2.lock().unwrap() = Some(e.to_string());
delta.send(-1).ok();
return;
}
};
const BATCH: usize = 4096;
let mut batch: Vec<CanonicalKmer> = Vec::with_capacity(BATCH);
let mut count: isize = 0;
for (kmer, _, _) in reader.iter_indexed_canonical_kmers() {
batch.push(kmer);
if batch.len() == BATCH {
let b = std::mem::replace(&mut batch, Vec::with_capacity(BATCH));
push.send(Ok(Pass2Data::RawBatch((
col_offset,
src_n,
Arc::clone(&src_layer),
b,
))))
.ok();
count += 1;
}
}
if !batch.is_empty() {
push.send(Ok(Pass2Data::RawBatch((
col_offset, src_n, src_layer, batch,
))))
.ok();
count += 1;
}
delta.send(count - 1).ok();
}
},
) as SharedFlatFn<Pass2Data>),
make_transform!(
Pass2Data,
{
move |(col_offset, src_n, src_layer, kmers): (
usize,
usize,
Arc<KmerLayer>,
Vec<CanonicalKmer>,
)|
-> Vec<(Option<usize>, usize, Vec<(usize, u32)>)> {
// Column-first, one pass for the whole batch: a single
// MPHF hash sweep (`hash_batch`) instead of one hash
// per kmer, then one read pass per source genome
// column (`fill_sub_matrix`) instead of one row read
// per kmer — matches the underlying matrix's own
// column-major layout.
let slots = src_layer.hash_batch(&kmers);
let mut cols: Vec<Vec<u32>> = (0..src_n)
.map(|_| Vec::with_capacity(kmers.len()))
.collect();
src_layer.fill_sub_matrix(&slots, &mut cols);
// Membership against dst, grouped by dst layer
// instead of interleaved kmer by kmer — same
// grouping `PartitionCache::find_presence_batch`
// does in `obikphylo`, via the shared primitive.
let (by_dst_layer, not_found) = find_batch_grouped(&dst_layers_t2, &kmers);
// Genome-major within each layer group (not kmer-major)
// so every `(layer, col)` pair's values land together —
// one entry per column, not one per `(kmer, genome)`.
let mut ops: Vec<(Option<usize>, usize, Vec<(usize, u32)>)> =
Vec::with_capacity(src_n * (by_dst_layer.len() + 1));
for (dst_layer, hits) in by_dst_layer.into_iter().enumerate() {
if hits.is_empty() {
continue;
}
for g in 0..src_n {
let col_vals: Vec<(usize, u32)> = hits
.iter()
.map(|&(idx, slot)| (slot, cols[g][idx]))
.collect();
ops.push((Some(dst_layer), col_offset + g, col_vals));
}
}
if let (Some(mphf), false) = (&new_mphf_t2, not_found.is_empty()) {
let not_found_kmers: Vec<CanonicalKmer> =
not_found.iter().map(|&idx| kmers[idx]).collect();
let new_slots = mphf.index_batch(&not_found_kmers);
for g in 0..src_n {
let col_vals: Vec<(usize, u32)> = not_found
.iter()
.zip(new_slots.iter())
.map(|(&idx, &slot)| (slot, cols[g][idx]))
.collect();
ops.push((None, col_offset + g, col_vals));
}
}
ops
}
},
RawBatch,
WriteBatch
),
],
make_sink!(
Pass2Data,
{
move |ops: Vec<(Option<usize>, usize, Vec<(usize, u32)>)>| {
for (layer_opt, col, col_vals) in ops {
let builder = match layer_opt {
Some(l) => &exist_sink[l][col],
None => &new_sink[col],
};
let mut b = builder.lock().unwrap();
for (slot, val) in col_vals {
b.set_val(slot, val);
}
}
}
},
WriteBatch
),
);
WorkerPool::new(pipeline2, n_workers, capacity).run();
if let Some(msg) = Arc::try_unwrap(pass2_err)
.unwrap_or_else(|_| panic!("pass2: pass2_err not uniquely owned"))
.into_inner()
.unwrap_or_else(|e| e.into_inner())
{
return Err(OKIError::InvalidData {
context: "merge pass2",
detail: msg,
});
}
let unlock = |b: Arc<Mutex<ColBuilder>>| -> ColBuilder {
Arc::try_unwrap(b)
.unwrap_or_else(|_| panic!("pass2: column builder not uniquely owned"))
.into_inner()
.unwrap_or_else(|e| e.into_inner())
};
let exist = exist_locked
.into_iter()
.map(|layer| layer.into_iter().map(unlock).collect())
.collect();
let new = new_locked.into_iter().map(unlock).collect();
Ok((exist, new))
}
-1
View File
@@ -10,7 +10,6 @@ obiskio = { path = "../obiskio" }
obisys = { path = "../obisys", default-features = false } obisys = { path = "../obisys", default-features = false }
obicompactvec = { path = "../obicompactvec" } obicompactvec = { path = "../obicompactvec" }
obikidxcache = { path = "../obikidxcache" } obikidxcache = { path = "../obikidxcache" }
obiskbuilder = { path = "../obiskbuilder" }
obipipeline = { path = "../obipipeline" } obipipeline = { path = "../obipipeline" }
kodama = "0.3.0" kodama = "0.3.0"
memmap2 = "0.9" memmap2 = "0.9"
+54 -69
View File
@@ -51,20 +51,6 @@ pub struct DistanceOutput {
pub shared_kmers: Option<Array2<u64>>, pub shared_kmers: Option<Array2<u64>>,
} }
impl DistanceMetric {
pub fn requires_counts(self) -> bool {
matches!(
self,
DistanceMetric::BrayCurtis
| DistanceMetric::RelfreqBrayCurtis
| DistanceMetric::Euclidean
| DistanceMetric::RelfreqEuclidean
| DistanceMetric::Hellinger
| DistanceMetric::HellingerEuclidean
)
}
}
// ── Aggregators — bounded parallel reduction of every cached layer's own // ── Aggregators — bounded parallel reduction of every cached layer's own
// partial (of the wanted content kind) into one global partial, then handed // partial (of the wanted content kind) into one global partial, then handed
// off to `CountPartials`/`BitPartials`'s own provided finalisation methods. // off to `CountPartials`/`BitPartials`'s own provided finalisation methods.
@@ -188,10 +174,33 @@ impl BitPartials for AggregatedBit<'_> {
/// extension trait in this codebase. /// extension trait in this codebase.
pub trait Metrics { pub trait Metrics {
fn distance(&self, metric: DistanceMetric, shared_kmers: bool, presence_threshold: u32) -> OKIResult<DistanceOutput>; fn distance(&self, metric: DistanceMetric, shared_kmers: bool, presence_threshold: u32) -> OKIResult<DistanceOutput>;
/// Genome-by-genome matrix of the number of k-mers shared by each pair of
/// genomes — on a count index a k-mer is present when its count reaches
/// `presence_threshold`. Independent of any distance: nothing else is
/// computed.
fn shared_kmers(&self, presence_threshold: u32) -> OKIResult<Array2<u64>>;
} }
impl Metrics for IndexCache { impl Metrics for IndexCache {
fn shared_kmers(&self, presence_threshold: u32) -> OKIResult<Array2<u64>> {
let n_genomes = self.meta().genomes().len();
if n_genomes < 2 {
return Err(OKIError::InvalidInput(
"shared k-mer counts require at least 2 genomes in the index".into(),
));
}
let (inter, _union) = if self.meta().config().with_counts {
AggregatedCount { cache: self, n_genomes }.partial_threshold_jaccard(presence_threshold)
} else {
AggregatedBit { cache: self, n_genomes }.partial_jaccard()
};
Ok(inter)
}
fn distance(&self, metric: DistanceMetric, shared_kmers: bool, presence_threshold: u32) -> OKIResult<DistanceOutput> { fn distance(&self, metric: DistanceMetric, shared_kmers: bool, presence_threshold: u32) -> OKIResult<DistanceOutput> {
use DistanceMetric::*;
let n_genomes = self.meta().genomes().len(); let n_genomes = self.meta().genomes().len();
if n_genomes < 2 { if n_genomes < 2 {
return Err(OKIError::InvalidInput( return Err(OKIError::InvalidInput(
@@ -199,64 +208,40 @@ impl Metrics for IndexCache {
)); ));
} }
let use_counts = self.meta().config().with_counts;
if metric.requires_counts() && !use_counts {
return Err(OKIError::InvalidInput(format!(
"{metric:?} requires a count index (with_counts = true)"
)));
}
let kmer_size = self.meta().config().kmer_size; let kmer_size = self.meta().config().kmer_size;
let use_counts = self.meta().config().with_counts;
let count = AggregatedCount { cache: self, n_genomes };
let bit = AggregatedBit { cache: self, n_genomes };
if use_counts { // One construction decides both what the index can serve and how:
let agg = AggregatedCount { cache: self, n_genomes }; // the six count-based distances need a count index; Jaccard, Mash and
// Hamming work on either (on counts, through the presence threshold).
let matrix = match (use_counts, metric) {
(true, BrayCurtis) => count.bray_dist_matrix(),
(true, RelfreqBrayCurtis) => count.relfreq_bray_dist_matrix(),
(true, Euclidean) => count.euclidean_dist_matrix(),
(true, RelfreqEuclidean) => count.relfreq_euclidean_dist_matrix(),
(true, Hellinger) => count.hellinger_dist_matrix(),
(true, HellingerEuclidean) => count.hellinger_euclidean_dist_matrix(),
(true, Jaccard) => count.threshold_jaccard_dist_matrix(presence_threshold),
(true, Mash) => count.threshold_mash_dist_matrix(kmer_size, presence_threshold),
(true, Hamming) => count.threshold_hamming_dist_matrix(presence_threshold).mapv(|v| v as f64),
(false, Jaccard) => bit.jaccard_dist_matrix(),
(false, Mash) => bit.mash_dist_matrix(kmer_size),
(false, Hamming) => bit.hamming_dist_matrix().mapv(|v| v as f64),
(
false,
BrayCurtis | RelfreqBrayCurtis | Euclidean | RelfreqEuclidean | Hellinger | HellingerEuclidean,
) => {
return Err(OKIError::InvalidInput(format!(
"{metric:?} requires a count index (with_counts = true); \
on an index without counts use --distance jaccard, mash or hamming"
)));
}
};
let matrix = match metric { let shared = shared_kmers.then(|| self.shared_kmers(presence_threshold)).transpose()?;
DistanceMetric::BrayCurtis => agg.bray_dist_matrix(),
DistanceMetric::RelfreqBrayCurtis => agg.relfreq_bray_dist_matrix(),
DistanceMetric::Euclidean => agg.euclidean_dist_matrix(),
DistanceMetric::RelfreqEuclidean => agg.relfreq_euclidean_dist_matrix(),
DistanceMetric::Hellinger => agg.hellinger_dist_matrix(),
DistanceMetric::HellingerEuclidean => agg.hellinger_euclidean_dist_matrix(),
DistanceMetric::Jaccard => agg.threshold_jaccard_dist_matrix(presence_threshold),
DistanceMetric::Mash => agg.threshold_mash_dist_matrix(kmer_size, presence_threshold),
DistanceMetric::Hamming => {
return Err(OKIError::InvalidInput(
"Hamming is only available for presence/absence indexes".into(),
));
}
};
let shared = if shared_kmers { Ok(DistanceOutput { matrix, shared_kmers: shared })
let (inter, _) = agg.partial_threshold_jaccard(presence_threshold);
Some(inter)
} else {
None
};
Ok(DistanceOutput { matrix, shared_kmers: shared })
} else {
let agg = AggregatedBit { cache: self, n_genomes };
let matrix = match metric {
DistanceMetric::Jaccard => agg.jaccard_dist_matrix(),
DistanceMetric::Mash => agg.mash_dist_matrix(kmer_size),
DistanceMetric::Hamming => agg.hamming_dist_matrix().mapv(|v| v as f64),
other => {
return Err(OKIError::InvalidInput(format!(
"{other:?} requires a count index; use --metric jaccard or --metric hamming"
)));
}
};
let shared = if shared_kmers {
let (inter, _) = agg.partial_jaccard();
Some(inter)
} else {
None
};
Ok(DistanceOutput { matrix, shared_kmers: shared })
}
} }
} }
@@ -5,12 +5,26 @@
//! masking ([`super::masking::survives_masking`]). //! masking ([`super::masking::survives_masking`]).
use obikidxcache::index_cache::IndexCache; use obikidxcache::index_cache::IndexCache;
use obikseq::RawKmer;
use rkyv::{Archive, Deserialize, Serialize}; use rkyv::{Archive, Deserialize, Serialize};
use super::masking::iupac_code; use super::masking::iupac_code;
use super::scan::LayerReducer; use super::scan::LayerReducer;
use super::subsample::SurvivingFamily; use super::subsample::SurvivingFamily;
/// Where an alignment column comes from: enough to give the same family the
/// same virtual coordinates every time for a given index (VCF `CHROM`/`POS`)
/// and an index-independent label (`family_id`).
#[derive(Clone, Copy, Archive, Serialize, Deserialize)]
pub struct SiteLocus {
pub partition: usize,
pub layer: usize,
/// `scan_layer_families`'s iteration-order index within `(partition, layer)`.
pub family_idx: usize,
/// Raw encoding of [`obikseq::CanonicalKmer::family_id`].
pub family_id: RawKmer,
}
/// `sequences[i]` is `genome_indices[i]`'s IUPAC-coded row — every row the /// `sequences[i]` is `genome_indices[i]`'s IUPAC-coded row — every row the
/// same length (one byte per sampled family, in the order /// same length (one byte per sampled family, in the order
/// [`super::scan::sampled_scan`] produced them: layer by layer, /// [`super::scan::sampled_scan`] produced them: layer by layer,
@@ -27,6 +41,8 @@ pub struct SnpAlignment {
/// order — the caller's authoritative mapping back to genome labels, so /// order — the caller's authoritative mapping back to genome labels, so
/// it never has to re-derive which genomes were excluded itself. /// it never has to re-derive which genomes were excluded itself.
pub genome_indices: Vec<usize>, pub genome_indices: Vec<usize>,
/// `sites[j]` describes column `j` of every row.
pub sites: Vec<SiteLocus>,
} }
/// Accumulates a [`SnpAlignment`] over a [`super::scan::sampled_scan`] /// Accumulates a [`SnpAlignment`] over a [`super::scan::sampled_scan`]
@@ -40,6 +56,7 @@ pub struct SnpAlignment {
pub struct AlignmentReducer { pub struct AlignmentReducer {
genome_indices: Vec<usize>, genome_indices: Vec<usize>,
sequences: Vec<Vec<u8>>, sequences: Vec<Vec<u8>>,
sites: Vec<SiteLocus>,
} }
impl AlignmentReducer { impl AlignmentReducer {
@@ -49,16 +66,22 @@ impl AlignmentReducer {
.filter(|&g| !excluded.get(g).copied().unwrap_or(false)) .filter(|&g| !excluded.get(g).copied().unwrap_or(false))
.collect(); .collect();
let sequences = vec![Vec::new(); genome_indices.len()]; let sequences = vec![Vec::new(); genome_indices.len()];
Self { genome_indices, sequences } Self { genome_indices, sequences, sites: Vec::new() }
} }
pub fn finish(self) -> SnpAlignment { pub fn finish(self) -> SnpAlignment {
SnpAlignment { sequences: self.sequences, genome_indices: self.genome_indices } SnpAlignment { sequences: self.sequences, genome_indices: self.genome_indices, sites: self.sites }
} }
} }
impl LayerReducer for AlignmentReducer { impl LayerReducer for AlignmentReducer {
fn reduce(&mut self, _partition: usize, _layer: usize, survivors: &[SurvivingFamily]) { fn reduce(&mut self, partition: usize, layer: usize, survivors: &[SurvivingFamily]) {
self.sites.extend(survivors.iter().map(|f| SiteLocus {
partition,
layer,
family_idx: f.family_idx,
family_id: f.kmer.family_id().raw(),
}));
reduce_alignment(survivors, &self.genome_indices, &mut self.sequences); reduce_alignment(survivors, &self.genome_indices, &mut self.sequences);
} }
} }
@@ -0,0 +1,270 @@
//! VCF rendering of a [`SnpAlignment`].
//!
//! One contig per `(partition, layer)` (`obikmer_<partition>_<layer>`), `POS`
//! = the family's `family_idx + 1` inside it: for a given index and sibling
//! annex the same family always gets the same virtual coordinates, and
//! sorting by `(partition, layer, family_idx)` is what makes the file a
//! well-formed (sorted) VCF. `ID` is the index-independent
//! [`obikseq::CanonicalKmer::family_id`] label (centre base masked as `N`).
//!
//! `REF`/`ALT` are the unambiguous bases (`A`/`C`/`G`/`T` cells of the
//! alignment) observed at the site, in the alignment's own convention: the
//! central base of the family member's *canonical* form, which is not
//! necessarily the strand of the `ID` k-mer. A `-` cell (family absent from
//! the genome — a real state, unlike `?`, which `--free-loss` produces and
//! which means missing) is the symbolic allele `<ABS>`, declared by a `##ALT`
//! line and listed after the bases. (`*` is not used: in VCF it means an
//! allele missing because of an overlapping deletion.)
//!
//! Ploidy of the genotypes is a caller choice. Haploid: one allele per
//! genome; an ambiguous cell is missing. Diploid: a single base `X` is
//! `X/X`, a two-base code is the heterozygote, `<ABS>` is `<ABS>/<ABS>`; codes with
//! three or four bases (and `?`) are missing (`./.`). In both modes a
//! genome gets the allele indices in ascending order.
use std::io::{self, Write};
use obikseq::CanonicalKmer;
use super::alignment::SnpAlignment;
use super::masking::iupac_state;
const BASES: [u8; 4] = *b"ACGT";
/// Internal marker of the absence allele (the alignment's own `-`), written
/// as [`ABSENT_ALT`].
const ABSENT: u8 = b'-';
/// How the absence allele appears in the `ALT` column.
const ABSENT_ALT: &[u8] = b"<ABS>";
fn base_index(cell: u8) -> Option<usize> {
BASES.iter().position(|&b| b == cell)
}
/// One cell as alleles: `None` = missing; else its (one or, diploid, two)
/// allele characters — a base, or [`ABSENT`] for absence. A cell with more forms
/// than the ploidy allows, or `?`, is missing.
fn cell_alleles(cell: u8, diploid: bool) -> Option<[u8; 2]> {
let state = iupac_state(cell)?;
let allele = |bit: u8| if bit == 0 { ABSENT } else { BASES[bit.trailing_zeros() as usize] };
match state.count_ones() {
0 => Some([ABSENT, ABSENT]),
1 => Some([allele(state), allele(state)]),
2 if diploid => {
let low = state & state.wrapping_neg();
Some([allele(low), allele(state ^ low)])
}
_ => None,
}
}
/// `REF` base (index into `ACGT`) of one site: the most frequent unambiguous
/// base, lexicographic minimum on ties. `None` when no cell carries one.
fn majority_base(counts: &[usize; 4]) -> Option<usize> {
(0..4)
.filter(|&b| counts[b] > 0)
.max_by_key(|&b| (counts[b], std::cmp::Reverse(b)))
}
impl SnpAlignment {
/// Writes this alignment as a VCF (haploid `GT` per genome). `labels` is
/// indexed by the index's own genome numbering ([`Self::genome_indices`]
/// values). `diploid` selects the genotype ploidy (see the module docs).
/// `REF` is the per-site majority allele (each allele copy counted once).
/// Returns the number of sites written: a site with no base among its
/// calls has no `REF` to report and is skipped.
pub fn write_vcf<W: Write>(&self, labels: &[String], diploid: bool, out: &mut W) -> io::Result<usize> {
let n_sites = self.sequences.first().map_or(0, Vec::len);
if self.sites.len() != n_sites {
return Err(io::Error::new(
io::ErrorKind::InvalidInput,
format!("alignment has {n_sites} column(s) but {} site locus record(s)", self.sites.len()),
));
}
let mut order: Vec<usize> = (0..n_sites).collect();
order.sort_unstable_by_key(|&j| {
let s = &self.sites[j];
(s.partition, s.layer, s.family_idx)
});
writeln!(out, "##fileformat=VCFv4.5")?;
writeln!(out, "##source=obikmer")?;
let mut last = None;
for &j in &order {
let s = &self.sites[j];
if last != Some((s.partition, s.layer)) {
writeln!(out, "##contig=<ID=obikmer_{}_{}>", s.partition, s.layer)?;
last = Some((s.partition, s.layer));
}
}
writeln!(out, "##ALT=<ID=ABS,Description=\"Family absent from the genome\">")?;
writeln!(out, "##FORMAT=<ID=GT,Number=1,Type=String,Description=\"Genotype ({})\">", if diploid { "diploid" } else { "haploid" })?;
write!(out, "#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\tFORMAT")?;
for &g in &self.genome_indices {
write!(out, "\t{}", labels[g])?;
}
writeln!(out)?;
let mut written = 0usize;
let mut locus = Vec::new();
let mut calls: Vec<Option<[u8; 2]>> = Vec::with_capacity(self.sequences.len());
let copies = if diploid { 2 } else { 1 };
for &j in &order {
calls.clear();
calls.extend(self.sequences.iter().map(|seq| cell_alleles(seq[j], diploid)));
let mut counts = [0usize; 4];
let mut absent = false;
for call in calls.iter().flatten() {
for &a in &call[..copies] {
match base_index(a) {
Some(b) => counts[b] += 1,
None => absent = true,
}
}
}
let Some(ref_base) = majority_base(&counts) else {
continue;
};
let mut alleles = vec![BASES[ref_base]];
alleles.extend((0..4).filter(|&b| b != ref_base && counts[b] > 0).map(|b| BASES[b]));
if absent {
alleles.push(ABSENT);
}
let s = &self.sites[j];
locus.clear();
CanonicalKmer::from_raw_unchecked(s.family_id).write_family_id_ascii(&mut locus)?;
write!(out, "obikmer_{}_{}\t{}\t", s.partition, s.layer, s.family_idx + 1)?;
out.write_all(&locus)?;
write!(out, "\t{}\t", alleles[0] as char)?;
if alleles.len() == 1 {
out.write_all(b".")?;
} else {
for (i, &a) in alleles[1..].iter().enumerate() {
if i > 0 {
out.write_all(b",")?;
}
if a == ABSENT {
out.write_all(ABSENT_ALT)?;
} else {
out.write_all(&[a])?;
}
}
}
out.write_all(b"\t.\t.\t.\tGT")?;
for call in &calls {
let idx = |a: u8| alleles.iter().position(|&x| x == a);
match call.map(|c| (idx(c[0]), idx(c[1]))) {
Some((Some(x), Some(y))) if diploid => write!(out, "\t{}/{}", x.min(y), x.max(y))?,
Some((Some(x), _)) => write!(out, "\t{x}")?,
_ => out.write_all(if diploid { b"\t./." } else { b"\t." })?,
}
}
out.write_all(b"\n")?;
written += 1;
}
Ok(written)
}
}
#[cfg(test)]
mod tests {
use super::*;
use crate::siblings::SiteLocus;
fn alignment(rows: &[&[u8]], sites: Vec<SiteLocus>) -> SnpAlignment {
SnpAlignment {
sequences: rows.iter().map(|r| r.to_vec()).collect(),
genome_indices: (0..rows.len()).collect(),
sites,
}
}
fn site(partition: usize, layer: usize, family_idx: usize) -> SiteLocus {
// Same global k as the crate's other tests (`siblingannex.rs`): the
// global is process-wide here, so it must not conflict.
obikseq::params::set_k(11);
SiteLocus { partition, layer, family_idx, family_id: 0 }
}
fn first_record(a: &SnpAlignment) -> Vec<String> {
first_record_ploidy(a, false)
}
fn first_record_ploidy(a: &SnpAlignment, diploid: bool) -> Vec<String> {
let labels: Vec<String> = (0..a.sequences.len()).map(|i| format!("g{i}")).collect();
let mut out = Vec::new();
a.write_vcf(&labels, diploid, &mut out).unwrap();
let text = String::from_utf8(out).unwrap();
let line = text.lines().find(|l| !l.starts_with('#')).unwrap();
line.split('\t').map(String::from).collect()
}
#[test]
fn reference_is_majority_then_lexicographic() {
assert_eq!(majority_base(&[1, 3, 0, 0]), Some(1));
assert_eq!(majority_base(&[2, 2, 0, 0]), Some(0));
assert_eq!(majority_base(&[0, 0, 0, 0]), None);
}
#[test]
fn genotypes_follow_ref_then_alt_order() {
// Column 0: A,G,G,R,? -> REF=G (majority), ALT=A; Column 1: all ambiguous -> skipped.
let a = alignment(
&[b"AY", b"GY", b"GR", b"RR", b"?R"],
vec![site(0, 0, 4), site(0, 0, 9)],
);
let f = first_record(&a);
assert_eq!(&f[..2], ["obikmer_0_0", "5"]);
assert_eq!(&f[3..5], ["G", "A"]);
assert_eq!(&f[9..], ["1", "0", "0", ".", "."]);
}
#[test]
fn absence_is_the_abs_allele_after_the_bases() {
// A, T, - (family absent), ? (missing): REF=A, ALT=T,<ABS>.
let a = alignment(&[b"A", b"T", b"-", b"?"], vec![site(0, 0, 0)]);
let f = first_record(&a);
assert_eq!(&f[3..5], ["A", "T,<ABS>"]);
assert_eq!(&f[9..], ["0", "1", "2", "."]);
// Only presence/absence differs: REF=A, ALT=<ABS>.
let a = alignment(&[b"A", b"-"], vec![site(0, 0, 0)]);
let f = first_record(&a);
assert_eq!(&f[3..5], ["A", "<ABS>"]);
assert_eq!(&f[9..], ["0", "1"]);
}
#[test]
fn iupac_state_inverts_iupac_code() {
for state in 0u8..16 {
assert_eq!(iupac_state(super::super::masking::iupac_code(state)), Some(state));
}
assert_eq!(iupac_state(b'?'), None);
}
#[test]
fn diploid_genotypes() {
// A, R (A/G), G, N (4 forms -> missing), - (absent), ? (missing).
let a = alignment(&[b"A", b"R", b"G", b"N", b"-", b"?"], vec![site(0, 0, 0)]);
let f = first_record_ploidy(&a, true);
// majority: A = 2 + 1 = 3, G = 1 + 2 = 3 -> tie -> A; ALT = G,<ABS>
assert_eq!(&f[3..5], ["A", "G,<ABS>"]);
assert_eq!(&f[9..], ["0/0", "0/1", "1/1", "./.", "2/2", "./."]);
// Same column, haploid: the ambiguous cells are missing.
let f = first_record_ploidy(&a, false);
assert_eq!(&f[9..], ["0", ".", "1", ".", "2", "."]);
}
#[test]
fn sites_sorted_by_contig_then_position() {
let a = alignment(&[b"AA", b"GG"], vec![site(1, 0, 0), site(0, 2, 7)]);
let labels = vec!["x".to_string(), "y".to_string()];
let mut out = Vec::new();
a.write_vcf(&labels, false, &mut out).unwrap();
let text = String::from_utf8(out).unwrap();
let body: Vec<&str> = text.lines().filter(|l| !l.starts_with('#')).collect();
assert!(body[0].starts_with("obikmer_0_2\t8\t"));
assert!(body[1].starts_with("obikmer_1_0\t1\t"));
}
}
@@ -16,10 +16,16 @@
/// real minorant, guarded against rather than dividing by zero). Second /// real minorant, guarded against rather than dividing by zero). Second
/// value: the denominator (`n_genomes_present`) used, for traceability. /// value: the denominator (`n_genomes_present`) used, for traceability.
pub(crate) fn family_entropy(genome_mask: &[u8]) -> Option<(f64, usize)> { pub(crate) fn family_entropy(genome_mask: &[u8]) -> Option<(f64, usize)> {
family_entropy_with(genome_mask, AmbiguityPolicy::Keep)
}
/// [`family_entropy`] over the genomes `policy` keeps as carriers
/// ([`is_carrier`]).
pub(crate) fn family_entropy_with(genome_mask: &[u8], policy: AmbiguityPolicy) -> Option<(f64, usize)> {
let mut freq = [0u32; 16]; // index 0 (∅) never incremented, kept for direct `m as usize` indexing let mut freq = [0u32; 16]; // index 0 (∅) never incremented, kept for direct `m as usize` indexing
let mut present = 0u32; let mut present = 0u32;
for &m in genome_mask { for &m in genome_mask {
if m != 0 { if is_carrier(m, policy) {
freq[m as usize] += 1; freq[m as usize] += 1;
present += 1; present += 1;
} }
@@ -70,6 +76,9 @@ pub(crate) fn family_entropy_4(genome_mask: &[u8]) -> Option<(f64, usize)> {
Some((h, n_single as usize)) Some((h, n_single as usize))
} }
use crate::siblings::AmbiguityPolicy;
use crate::siblings::helpers::is_carrier;
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::*; use super::*;
@@ -104,10 +113,16 @@ mod tests {
#[test] #[test]
fn entropy_under_selects_the_counting_convention() { fn entropy_under_selects_the_counting_convention() {
let masks = [0b0001u8, 0b0001, 0b0011]; // the last genome carries two members let masks = [0b0001u8, 0b0001, 0b0011]; // the last genome carries two members
assert_eq!(family_entropy_under(&masks, true).unwrap().1, 2); use AmbiguityPolicy::{Forbid, Keep, OnlyDiploid};
assert_eq!(family_entropy_under(&masks, false).unwrap().1, 3); assert_eq!(family_entropy_under(&masks, Forbid).unwrap().1, 2);
assert_eq!(family_entropy_under(&masks, true).unwrap().0, 0.0); assert_eq!(family_entropy_under(&masks, Keep).unwrap().1, 3);
assert!(family_entropy_under(&masks, false).unwrap().0 > 0.0); assert_eq!(family_entropy_under(&masks, OnlyDiploid).unwrap().1, 3);
assert_eq!(family_entropy_under(&masks, Forbid).unwrap().0, 0.0);
assert!(family_entropy_under(&masks, Keep).unwrap().0 > 0.0);
// three members: kept by Keep, dropped by OnlyDiploid
let masks = [0b0001u8, 0b0001, 0b0111];
assert_eq!(family_entropy_under(&masks, Keep).unwrap().1, 3);
assert_eq!(family_entropy_under(&masks, OnlyDiploid).unwrap().1, 2);
} }
#[test] #[test]
@@ -117,21 +132,24 @@ mod tests {
} }
} }
/// Shannon entropy (bits) of one family under the counting convention /// Shannon entropy (bits) of one family under the counting convention an
/// `--no-ambiguity` selects: /// [`AmbiguityPolicy`] selects:
/// - `false`: [`family_entropy`] — every genome carrying the family counts, /// - `Keep`: [`family_entropy`] — every genome carrying the family counts,
/// a genome with several members being a state of its own; /// a genome with several members being a state of its own;
/// - `true`: [`family_entropy_4`] — only the genomes carrying exactly one /// - `Forbid` (`--no-ambiguity`): [`family_entropy_4`] — only the genomes
/// member count, the others being missing data, exactly as the alignment /// carrying exactly one member count, the others being missing data,
/// itself will see them. Over those genomes the 15-state entropy can only /// exactly as the alignment itself will see them. Over those genomes the
/// ever take the four single-base states, so the two definitions coincide. /// 15-state entropy can only ever take the four single-base states, so the
/// two definitions coincide;
/// - `OnlyDiploid`: the 15-state entropy over the genomes with one or two
/// members (ten states), the others being missing data.
/// ///
/// Second value: the number of genomes the entropy was computed over. /// Second value: the number of genomes the entropy was computed over.
pub(crate) fn family_entropy_under(genome_mask: &[u8], no_ambiguity: bool) -> Option<(f64, usize)> { pub(crate) fn family_entropy_under(genome_mask: &[u8], policy: AmbiguityPolicy) -> Option<(f64, usize)> {
if no_ambiguity { match policy {
family_entropy_4(genome_mask) AmbiguityPolicy::Keep => family_entropy(genome_mask),
} else { AmbiguityPolicy::Forbid => family_entropy_4(genome_mask),
family_entropy(genome_mask) AmbiguityPolicy::OnlyDiploid => family_entropy_with(genome_mask, policy),
} }
} }
@@ -67,7 +67,7 @@ pub(crate) fn scan_family_entropies(
n_genomes, n_genomes,
fast_mode, fast_mode,
&Selection::Some(&eligible), &Selection::Some(&eligible),
|family_idx, mask, genome_mask| { |family_idx, _kmer, mask, genome_mask| {
if failure.is_some() { if failure.is_some() {
return; return;
} }
@@ -10,7 +10,7 @@
use std::collections::HashSet; use std::collections::HashSet;
use std::io; use std::io;
use std::path::PathBuf; use std::path::{Path, PathBuf};
use obicompactvec::PersistentBitVecBuilder; use obicompactvec::PersistentBitVecBuilder;
use obikidxcache::index_cache::IndexCache; use obikidxcache::index_cache::IndexCache;
@@ -21,34 +21,66 @@ use super::family_scan::{Selection, scan_layer_families};
use super::masking::is_quartet_informative; use super::masking::is_quartet_informative;
use super::minorant_selection::non_monomorphic_reindex_layer; use super::minorant_selection::non_monomorphic_reindex_layer;
use super::pair_weight::pair_balance_weight; use super::pair_weight::pair_balance_weight;
use crate::siblings::quartet_annex::quartet_annex_file_name; use crate::siblings::AmbiguityPolicy;
use crate::siblings::quartet_annex::{QuartetConvention, quartet_annex_file_name};
use crate::siblings::{ use crate::siblings::{
FamilyOverlapAccumulator, FamilyScalarAnnexBuilder, ScalarKind, family_overlap_file_name, FamilyOverlapAccumulator, FamilyScalarAnnexBuilder, ScalarKind, family_overlap_file_name,
}; };
/// Which annexes a caller needs, each listed once per `--no-ambiguity` /// Which annexes a caller needs, each listed once per ambiguity policy (how
/// counting convention (`false` = genomes carrying several members count as a /// genomes carrying several members of a family are counted) — an empty list
/// state of their own, `true` = they are left out) — an empty list asks for /// asks for none (the quartet annex also depends on `--free-loss`, so its
/// none. A pair weight is computed from the family-overlap annex of its own /// convention is a [`QuartetConvention`]). A pair weight is computed from the family-overlap annex of
/// convention, so asking for one asks for the other. /// its own policy, so asking for one asks for the other.
#[derive(Clone, Debug, Default)] #[derive(Clone, Debug, Default)]
pub(crate) struct AnnexRequest { pub(crate) struct AnnexRequest {
pub entropy: Vec<bool>, pub entropy: Vec<AmbiguityPolicy>,
pub quartet: Vec<bool>, pub quartet: Vec<QuartetConvention>,
pub overlap: Vec<bool>, pub overlap: Vec<AmbiguityPolicy>,
pub pair_weight: Vec<bool>, pub pair_weight: Vec<AmbiguityPolicy>,
}
/// The derived annexes a run asks to have ready, all under the one ambiguity
/// `policy` of that run (`--sibling-annex` together with the options that
/// say which annexes, and under which convention, the following computations
/// will read). Nothing is built that is not asked for.
#[derive(Clone, Copy, Debug, Default)]
pub struct AnnexSelection {
pub policy: AmbiguityPolicy,
/// `--free-loss`: part of the quartet annex's convention (whether absence
/// counts as a state), irrelevant to the other annexes.
pub free_loss: bool,
pub entropy: bool,
pub quartet: bool,
pub overlap: bool,
pub pair_weight: bool,
}
impl AnnexSelection {
pub(crate) fn request(&self) -> AnnexRequest {
let one = |wanted: bool| if wanted { vec![self.policy] } else { Vec::new() };
AnnexRequest {
entropy: one(self.entropy),
quartet: if self.quartet {
vec![QuartetConvention { policy: self.policy, free_loss: self.free_loss }]
} else {
Vec::new()
},
overlap: one(self.overlap),
pair_weight: one(self.pair_weight),
}
}
/// Whether any derived annex is asked for.
pub fn is_empty(&self) -> bool {
!(self.entropy || self.quartet || self.overlap || self.pair_weight)
}
} }
impl AnnexRequest { impl AnnexRequest {
/// Everything under both conventions — what `--sibling-annex` precomputes.
pub(crate) fn everything() -> Self {
let both = vec![false, true];
Self { entropy: both.clone(), quartet: both.clone(), overlap: both.clone(), pair_weight: both }
}
/// The overlap conventions to ensure: the requested ones plus every one a /// The overlap conventions to ensure: the requested ones plus every one a
/// requested pair weight is computed from, each once. /// requested pair weight is computed from, each once.
pub(crate) fn overlap_conventions(&self) -> Vec<bool> { pub(crate) fn overlap_conventions(&self) -> Vec<AmbiguityPolicy> {
dedup(self.overlap.iter().chain(&self.pair_weight).copied()) dedup(self.overlap.iter().chain(&self.pair_weight).copied())
} }
@@ -59,25 +91,39 @@ impl AnnexRequest {
let overlap = self let overlap = self
.overlap_conventions() .overlap_conventions()
.into_iter() .into_iter()
.all(|nb| cache.index().dir().join(family_overlap_file_name(nb)).exists()); .all(|p| cache.index().dir().join(family_overlap_file_name(p)).exists());
let pair_weight = self.pair_weight_conventions(); let pair_weight = self.pair_weight_conventions();
overlap overlap
&& cache.iter_indexed().all(|(_, _, layer)| { && cache.iter_indexed().all(|(_, _, layer)| {
let has = |name: &str| layer.dir().join(name).exists(); let has = |name: &str| layer.dir().join(name).exists();
self.entropy.iter().all(|&nb| has(ScalarKind::Entropy.file_name(nb))) self.entropy.iter().all(|&p| has(ScalarKind::Entropy.file_name(p)))
&& self.quartet.iter().all(|&nb| has(quartet_annex_file_name(nb))) && self.quartet.iter().all(|&p| has(quartet_annex_file_name(p)))
&& pair_weight.iter().all(|&nb| has(ScalarKind::PairWeight.file_name(nb))) && pair_weight.iter().all(|&p| has(ScalarKind::PairWeight.file_name(p)))
}) })
} }
/// The pair-weight conventions requested, each once. /// The pair-weight conventions requested, each once.
pub(crate) fn pair_weight_conventions(&self) -> Vec<bool> { pub(crate) fn pair_weight_conventions(&self) -> Vec<AmbiguityPolicy> {
dedup(self.pair_weight.iter().copied()) dedup(self.pair_weight.iter().copied())
} }
} }
fn dedup(variants: impl IntoIterator<Item = bool>) -> Vec<bool> { /// The annexes among `conventions` (each once) whose file does not exist yet
let mut v: Vec<bool> = variants.into_iter().collect(); /// in `dir`, with their path.
fn missing_files<T: Ord + Copy>(
dir: &Path,
name: impl Fn(T) -> &'static str,
conventions: &[T],
) -> Vec<(T, PathBuf)> {
dedup(conventions.iter().copied())
.into_iter()
.map(|convention| (convention, dir.join(name(convention))))
.filter(|(_, path)| !path.exists())
.collect()
}
fn dedup<T: Ord>(variants: impl IntoIterator<Item = T>) -> Vec<T> {
let mut v: Vec<T> = variants.into_iter().collect();
v.sort_unstable(); v.sort_unstable();
v.dedup(); v.dedup();
v v
@@ -102,20 +148,14 @@ pub(crate) fn ensure_layer_family_annexes(
n_genomes: usize, n_genomes: usize,
fast_mode: bool, fast_mode: bool,
request: &AnnexRequest, request: &AnnexRequest,
overlap_accs: &mut [(bool, FamilyOverlapAccumulator)], overlap_accs: &mut [(AmbiguityPolicy, FamilyOverlapAccumulator)],
) -> OKIResult<()> { ) -> OKIResult<()> {
let layer = cache let layer = cache
.get_layer(partition, layer_idx) .get_layer(partition, layer_idx)
.expect("cache-consistent: caller already bounds layer_idx by n_layer(partition)"); .expect("cache-consistent: caller already bounds layer_idx by n_layer(partition)");
let missing = |name: &dyn Fn(bool) -> &'static str, variants: &[bool]| -> Vec<(bool, PathBuf)> { let missing_entropy =
dedup(variants.iter().copied()) missing_files(layer.dir(), |p| ScalarKind::Entropy.file_name(p), &request.entropy);
.into_iter() let missing_quartet = missing_files(layer.dir(), quartet_annex_file_name, &request.quartet);
.map(|no_ambiguity| (no_ambiguity, layer.dir().join(name(no_ambiguity))))
.filter(|(_, path)| !path.exists())
.collect()
};
let missing_entropy = missing(&|nb| ScalarKind::Entropy.file_name(nb), &request.entropy);
let missing_quartet = missing(&quartet_annex_file_name, &request.quartet);
if missing_entropy.is_empty() && missing_quartet.is_empty() && overlap_accs.is_empty() { if missing_entropy.is_empty() && missing_quartet.is_empty() && overlap_accs.is_empty() {
return Ok(()); return Ok(());
} }
@@ -124,16 +164,16 @@ pub(crate) fn ensure_layer_family_annexes(
let eligible: HashSet<usize> = reindex.keys().copied().collect(); let eligible: HashSet<usize> = reindex.keys().copied().collect();
let mut entropy_builders = missing_entropy let mut entropy_builders = missing_entropy
.iter() .iter()
.map(|(no_ambiguity, path)| { .map(|(policy, path)| {
FamilyScalarAnnexBuilder::new(reindex.len(), path, ScalarKind::Entropy, *no_ambiguity) FamilyScalarAnnexBuilder::new(reindex.len(), path, ScalarKind::Entropy, *policy)
.map(|b| (*no_ambiguity, b)) .map(|b| (*policy, b))
}) })
.collect::<io::Result<Vec<_>>>() .collect::<io::Result<Vec<_>>>()
.map_err(OKIError::Io)?; .map_err(OKIError::Io)?;
let mut quartet_builders = missing_quartet let mut quartet_builders = missing_quartet
.iter() .iter()
.map(|(no_ambiguity, path)| { .map(|(policy, path)| {
PersistentBitVecBuilder::new(reindex.len(), path).map(|b| (*no_ambiguity, b)) PersistentBitVecBuilder::new(reindex.len(), path).map(|b| (*policy, b))
}) })
.collect::<io::Result<Vec<_>>>() .collect::<io::Result<Vec<_>>>()
.map_err(OKIError::Io)?; .map_err(OKIError::Io)?;
@@ -145,19 +185,22 @@ pub(crate) fn ensure_layer_family_annexes(
n_genomes, n_genomes,
fast_mode, fast_mode,
&Selection::Some(&eligible), &Selection::Some(&eligible),
|family_idx, mask, genome_mask| { |family_idx, _kmer, mask, genome_mask| {
debug_assert!( debug_assert!(
mask.family_size() >= 2, mask.family_size() >= 2,
"Selection::Some(eligible) must only visit non-monomorphic minorants" "Selection::Some(eligible) must only visit non-monomorphic minorants"
); );
let compact = reindex[&family_idx]; let compact = reindex[&family_idx];
for (no_ambiguity, builder) in entropy_builders.iter_mut() { for (policy, builder) in entropy_builders.iter_mut() {
if let Some((h, _)) = family_entropy_under(genome_mask, *no_ambiguity) { if let Some((h, _)) = family_entropy_under(genome_mask, *policy) {
builder.set(compact, h as f32); builder.set(compact, h as f32);
} }
} }
for (no_ambiguity, builder) in quartet_builders.iter_mut() { for (convention, builder) in quartet_builders.iter_mut() {
builder.set(compact, is_quartet_informative(genome_mask, *no_ambiguity)); builder.set(
compact,
is_quartet_informative(genome_mask, convention.policy, convention.free_loss),
);
} }
for (_, acc) in overlap_accs.iter_mut() { for (_, acc) in overlap_accs.iter_mut() {
acc.add_family(genome_mask); acc.add_family(genome_mask);
@@ -175,7 +218,7 @@ pub(crate) fn ensure_layer_family_annexes(
} }
/// Builds the pair-weight annexes of one layer that don't exist yet, one per /// Builds the pair-weight annexes of one layer that don't exist yet, one per
/// `(no_ambiguity, inverse overlap matrix)` in `matrices` (see /// `(policy, inverse overlap matrix)` in `matrices` (see
/// [`super::pair_weight::inverse_overlap_matrix`]), in one scan. A no-op (no /// [`super::pair_weight::inverse_overlap_matrix`]), in one scan. A no-op (no
/// scan) when they all exist. Needs the family-overlap annex of each /// scan) when they all exist. Needs the family-overlap annex of each
/// convention to be complete: run after the sweep that builds it. /// convention to be complete: run after the sweep that builds it.
@@ -185,14 +228,14 @@ pub(crate) fn ensure_layer_pair_weight_annexes(
layer_idx: usize, layer_idx: usize,
n_genomes: usize, n_genomes: usize,
fast_mode: bool, fast_mode: bool,
matrices: &[(bool, Vec<f32>)], matrices: &[(AmbiguityPolicy, Vec<f32>)],
) -> OKIResult<()> { ) -> OKIResult<()> {
let layer = cache let layer = cache
.get_layer(partition, layer_idx) .get_layer(partition, layer_idx)
.expect("cache-consistent: caller already bounds layer_idx by n_layer(partition)"); .expect("cache-consistent: caller already bounds layer_idx by n_layer(partition)");
let pending: Vec<&(bool, Vec<f32>)> = matrices let pending: Vec<&(AmbiguityPolicy, Vec<f32>)> = matrices
.iter() .iter()
.filter(|(no_ambiguity, _)| !layer.dir().join(ScalarKind::PairWeight.file_name(*no_ambiguity)).exists()) .filter(|(policy, _)| !layer.dir().join(ScalarKind::PairWeight.file_name(*policy)).exists())
.collect(); .collect();
if pending.is_empty() { if pending.is_empty() {
return Ok(()); return Ok(());
@@ -202,10 +245,10 @@ pub(crate) fn ensure_layer_pair_weight_annexes(
let eligible: HashSet<usize> = reindex.keys().copied().collect(); let eligible: HashSet<usize> = reindex.keys().copied().collect();
let mut builders = pending let mut builders = pending
.iter() .iter()
.map(|(no_ambiguity, inverse)| { .map(|(policy, inverse)| {
let path = layer.dir().join(ScalarKind::PairWeight.file_name(*no_ambiguity)); let path = layer.dir().join(ScalarKind::PairWeight.file_name(*policy));
FamilyScalarAnnexBuilder::new(reindex.len(), &path, ScalarKind::PairWeight, *no_ambiguity) FamilyScalarAnnexBuilder::new(reindex.len(), &path, ScalarKind::PairWeight, *policy)
.map(|b| (*no_ambiguity, inverse, b)) .map(|b| (*policy, inverse, b))
}) })
.collect::<io::Result<Vec<_>>>() .collect::<io::Result<Vec<_>>>()
.map_err(OKIError::Io)?; .map_err(OKIError::Io)?;
@@ -217,10 +260,10 @@ pub(crate) fn ensure_layer_pair_weight_annexes(
n_genomes, n_genomes,
fast_mode, fast_mode,
&Selection::Some(&eligible), &Selection::Some(&eligible),
|family_idx, _mask, genome_mask| { |family_idx, _kmer, _mask, genome_mask| {
let compact = reindex[&family_idx]; let compact = reindex[&family_idx];
for (no_ambiguity, inverse, builder) in builders.iter_mut() { for (policy, inverse, builder) in builders.iter_mut() {
builder.set(compact, pair_balance_weight(genome_mask, *no_ambiguity, inverse)); builder.set(compact, pair_balance_weight(genome_mask, *policy, inverse));
} }
}, },
)?; )?;
@@ -125,6 +125,8 @@ struct SourceBatch {
struct GeneratedBatch { struct GeneratedBatch {
start_family_idx: usize, start_family_idx: usize,
masks: Vec<FamilyMask>, masks: Vec<FamilyMask>,
/// Each family's minorant k-mer, same order as `masks`.
kmers: Vec<CanonicalKmer>,
/// Flat `slots.len() * n_genomes` — `genome_mask[i * n_genomes + g]`. /// Flat `slots.len() * n_genomes` — `genome_mask[i * n_genomes + g]`.
genome_mask: Vec<u8>, genome_mask: Vec<u8>,
/// `outgoing[dest_partition]` = `(variant, family_idx_in_batch, base, /// `outgoing[dest_partition]` = `(variant, family_idx_in_batch, base,
@@ -146,8 +148,8 @@ enum FamData {
/// generation and resolution use different concurrency. `on_family` is /// generation and resolution use different concurrency. `on_family` is
/// called once per /// called once per
/// family, in iteration order, with that family's iteration-order index /// family, in iteration order, with that family's iteration-order index
/// (the same numbering [`Selection`] indices are drawn from), its own annex /// (the same numbering [`Selection`] indices are drawn from), its minorant
/// mask, and its per-genome base-presence (`genome_mask[g]`: bit `b` set /// k-mer, its own annex mask, and its per-genome base-presence (`genome_mask[g]`: bit `b` set
/// iff genome `g` carries the member whose own canonical central base is /// iff genome `g` carries the member whose own canonical central base is
/// `b`), backed by a scratch buffer reused across every call — callers that /// `b`), backed by a scratch buffer reused across every call — callers that
/// need to keep data past the call must copy it themselves. /// need to keep data past the call must copy it themselves.
@@ -164,7 +166,7 @@ pub(crate) fn scan_layer_families(
n_genomes: usize, n_genomes: usize,
fast_mode: bool, fast_mode: bool,
selection: &Selection, selection: &Selection,
mut on_family: impl FnMut(usize, FamilyMask, &[u8]), mut on_family: impl FnMut(usize, CanonicalKmer, FamilyMask, &[u8]),
) -> OKIResult<()> { ) -> OKIResult<()> {
let n_parts = cache.index().n_partitions(); let n_parts = cache.index().n_partitions();
let k = cache.index().kmer_size(); let k = cache.index().kmer_size();
@@ -225,6 +227,7 @@ pub(crate) fn scan_layer_families(
let ctx = &worker_ctx; let ctx = &worker_ctx;
let n = batch.entries.len(); let n = batch.entries.len();
let mut masks = Vec::with_capacity(n); let mut masks = Vec::with_capacity(n);
let mut kmers = Vec::with_capacity(n);
let mut bases = Vec::with_capacity(n); let mut bases = Vec::with_capacity(n);
let mut genome_mask = vec![0u8; n * ctx.n_genomes]; let mut genome_mask = vec![0u8; n * ctx.n_genomes];
let mut outgoing: Vec<Vec<(CanonicalKmer, usize, u8, u8)>> = let mut outgoing: Vec<Vec<(CanonicalKmer, usize, u8, u8)>> =
@@ -248,6 +251,7 @@ pub(crate) fn scan_layer_families(
for (i, entry) in batch.entries.iter().enumerate() { for (i, entry) in batch.entries.iter().enumerate() {
let (kmer, mask) = (entry.kmer, entry.mask); let (kmer, mask) = (entry.kmer, entry.mask);
masks.push(mask); masks.push(mask);
kmers.push(kmer);
let base = central_base(kmer, ctx.k); let base = central_base(kmer, ctx.k);
bases.push(base); bases.push(base);
let family_idx = batch.start_family_idx + i; let family_idx = batch.start_family_idx + i;
@@ -297,6 +301,7 @@ pub(crate) fn scan_layer_families(
GeneratedBatch { GeneratedBatch {
start_family_idx: batch.start_family_idx, start_family_idx: batch.start_family_idx,
masks, masks,
kmers,
genome_mask, genome_mask,
outgoing, outgoing,
_permit: batch._permit, _permit: batch._permit,
@@ -344,7 +349,7 @@ pub(crate) fn scan_layer_families(
for (g, dst) in scratch.iter_mut().enumerate() { for (g, dst) in scratch.iter_mut().enumerate() {
*dst = genome_mask[i * n_genomes + g].load(Ordering::Relaxed); *dst = genome_mask[i * n_genomes + g].load(Ordering::Relaxed);
} }
on_family(family_idx, batch.masks[i], &scratch); on_family(family_idx, batch.kmers[i], batch.masks[i], &scratch);
} }
next_expected += n; next_expected += n;
} }
+111 -41
View File
@@ -17,20 +17,21 @@
//! waste sample quota on sites that die anyway once the excluded genome's //! waste sample quota on sites that die anyway once the excluded genome's
//! data is dropped. //! data is dropped.
use crate::siblings::helpers::is_carrier; use crate::siblings::AmbiguityPolicy;
/// This genome's masked character state, or `None` for "?" (missing). `m` /// This genome's masked character state, or `None` for "?" (missing). `m`
/// is one `genome_mask` entry (bit `b` set iff this genome carries the /// is one `genome_mask` entry (bit `b` set iff this genome carries the
/// family member whose own canonical central base is `b`; `0` = no member /// family member whose own canonical central base is `b`; `0` = no member
/// present in this genome, `∅` — a real state on its own unless /// present in this genome, `∅` — a real state on its own unless
/// `free_loss` says otherwise). /// `free_loss` says otherwise). `policy` makes missing the genomes carrying
/// too many members ([`AmbiguityPolicy::drops`]).
#[inline] #[inline]
pub(crate) fn masked_state(m: u8, free_loss: bool, no_ambiguity: bool) -> Option<u8> { pub(crate) fn masked_state(m: u8, free_loss: bool, policy: AmbiguityPolicy) -> Option<u8> {
if free_loss && m == 0 { if free_loss && m == 0 {
return None; // absence -> missing, not a real state return None; // absence -> missing, not a real state
} }
if no_ambiguity && m.count_ones() > 1 { if policy.drops(m) {
return None; // ambiguous (>1 member in this genome) -> missing return None; // ambiguous beyond what the policy keeps -> missing
} }
Some(m) Some(m)
} }
@@ -67,39 +68,43 @@ pub(crate) fn survives_masking(masked: &[Option<u8>], excluded: &[bool]) -> bool
} }
/// Whether `counts` — the number of genomes per *called* state, indexed by /// Whether `counts` — the number of genomes per *called* state, indexed by
/// the `0..=15` state value, index `0` (absence) never counted — shows a /// the `0..=15` state value — shows a bipartition: at least two distinct
/// bipartition: at least two distinct states each carried by at least two /// states each carried by at least two genomes. That is the pattern that can
/// genomes. That is the pattern that can separate two pairs of genomes from /// separate two pairs of genomes from each other (`AB|CD`); a singleton
/// each other (`AB|CD`); a singleton variant, or a family carried by fewer /// variant, or a family carried by fewer than four genomes, fits every
/// than four genomes, fits every topology and informs none. /// topology and informs none.
///
/// Index `0` is absence (`∅`). It holds genomes only when absence is a real
/// state of the alignment (no `--free-loss`), and is then a state like any
/// other; under `--free-loss` absent genomes are missing data, are never
/// counted, and the index stays empty.
pub(crate) fn has_bipartition(counts: &[u32; 16]) -> bool { pub(crate) fn has_bipartition(counts: &[u32; 16]) -> bool {
counts[1..].iter().filter(|&&c| c >= 2).count() >= 2 counts.iter().filter(|&&c| c >= 2).count() >= 2
} }
/// Quartet-informativeness of a family from its raw per-genome masks alone, /// Quartet-informativeness of a family from its raw per-genome masks alone,
/// under the `--no-ambiguity` counting convention: absent genomes (`0`) /// under one counting convention: each genome is read through
/// never count, and with `no_ambiguity` a genome carrying several members /// [`masked_state`] exactly as in the alignment — absence is a state of its
/// doesn't either — otherwise its combination of bases is a state of its /// own unless `free_loss`, and `policy` makes the genomes carrying too many
/// own. Ignores `--exclude-genome` and `--free-loss` on purpose: absence is /// members missing. Ignores `--exclude-genome` on purpose: removing genomes
/// never a called state here, and removing genomes can only destroy a /// can only destroy a bipartition, never create one — so `false` is final, and
/// bipartition, never create one — so `false` is final, and `true` is only /// `true` is only a necessary condition, re-checked exactly by
/// a necessary condition, re-checked exactly by [`survives_quartet`] once /// [`survives_quartet`] once the family is resolved. This is what the
/// the family is resolved. This is what the persisted quartet annex stores. /// persisted quartet annex stores.
pub(crate) fn is_quartet_informative(genome_mask: &[u8], no_ambiguity: bool) -> bool { pub(crate) fn is_quartet_informative(genome_mask: &[u8], policy: AmbiguityPolicy, free_loss: bool) -> bool {
let mut counts = [0u32; 16]; let mut counts = [0u32; 16];
for &m in genome_mask { for &m in genome_mask {
if !is_carrier(m, no_ambiguity) { if let Some(state) = masked_state(m, free_loss, policy) {
continue; counts[(state & 0b1111) as usize] += 1;
} }
counts[(m & 0b1111) as usize] += 1;
} }
has_bipartition(&counts) has_bipartition(&counts)
} }
/// Exact quartet check on an already-[`masked_state`]d family, over the /// Exact quartet check on an already-[`masked_state`]d family, over the
/// *included* genomes (`excluded` as in [`survives_masking`]). Only called /// *included* genomes (`excluded` as in [`survives_masking`]). Missing
/// states count: a genome whose masked state is `0` (absence kept as a real /// genomes (`None`) count for nothing; absence kept as a real state (no
/// state, i.e. no `--free-loss`) is not a carrier of anything. /// `--free-loss`, masked state `0`) is a state like any other.
pub(crate) fn survives_quartet(masked: &[Option<u8>], excluded: &[bool]) -> bool { pub(crate) fn survives_quartet(masked: &[Option<u8>], excluded: &[bool]) -> bool {
let mut counts = [0u32; 16]; let mut counts = [0u32; 16];
for (g, &state) in masked.iter().enumerate() { for (g, &state) in masked.iter().enumerate() {
@@ -117,33 +122,48 @@ pub(crate) fn survives_quartet(masked: &[Option<u8>], excluded: &[bool]) -> bool
mod quartet_tests { mod quartet_tests {
use super::*; use super::*;
const KEEP: AmbiguityPolicy = AmbiguityPolicy::Keep;
const FORBID: AmbiguityPolicy = AmbiguityPolicy::Forbid;
#[test] #[test]
fn two_pairs_of_distinct_states_is_informative() { fn two_pairs_of_distinct_states_is_informative() {
assert!(is_quartet_informative(&[0b0001, 0b0001, 0b0100, 0b0100], false)); for free_loss in [true, false] {
assert!(is_quartet_informative(&[0b0001, 0b0001, 0b0100, 0b0100], true)); assert!(is_quartet_informative(&[0b0001, 0b0001, 0b0100, 0b0100], KEEP, free_loss));
assert!(is_quartet_informative(&[0b0001, 0b0001, 0b0100, 0b0100], FORBID, free_loss));
}
} }
#[test] #[test]
fn singleton_variant_is_not_informative() { fn singleton_variant_is_not_informative() {
// three genomes in one state plus one lone variant: fits any topology for free_loss in [true, false] {
assert!(!is_quartet_informative(&[0b0001, 0b0001, 0b0001, 0b0100], false)); // three genomes in one state plus one lone variant: fits any topology
// fewer than four carriers can never be informative assert!(!is_quartet_informative(&[0b0001, 0b0001, 0b0001, 0b0100], KEEP, free_loss));
assert!(!is_quartet_informative(&[0b0001, 0b0100, 0, 0], false)); // fewer than four genomes with a state can never be informative
assert!(!is_quartet_informative(&[0b0001, 0b0001, 0b0100], false)); assert!(!is_quartet_informative(&[0b0001, 0b0100, 0, 0], KEEP, free_loss));
assert!(!is_quartet_informative(&[0b0001, 0b0001, 0b0100], KEEP, free_loss));
}
} }
#[test] #[test]
fn absent_genomes_are_not_a_state() { fn under_free_loss_absence_is_missing_and_forms_no_pair() {
// two carriers of one base, two absences: absence must not form the 2nd pair // two carriers of one base, two absences: absence is missing data
assert!(!is_quartet_informative(&[0b0001, 0b0001, 0, 0], false)); assert!(!is_quartet_informative(&[0b0001, 0b0001, 0, 0], KEEP, true));
}
#[test]
fn without_free_loss_absence_is_a_state_that_can_form_a_pair() {
// A A ∅ ∅ separates two pairs of genomes once ∅ is a character state
assert!(is_quartet_informative(&[0b0001, 0b0001, 0, 0], KEEP, false));
// ... but a single absent genome is a lone state, like any singleton
assert!(!is_quartet_informative(&[0b0001, 0b0001, 0b0001, 0], KEEP, false));
} }
#[test] #[test]
fn no_ambiguity_drops_multi_member_genomes_from_the_count() { fn no_ambiguity_drops_multi_member_genomes_from_the_count() {
// the two-member genomes form a pair of their own state without the flag // the two-member genomes form a pair of their own state without the flag
let masks = [0b0001u8, 0b0001, 0b0011, 0b0011]; let masks = [0b0001u8, 0b0001, 0b0011, 0b0011];
assert!(is_quartet_informative(&masks, false)); assert!(is_quartet_informative(&masks, KEEP, true));
assert!(!is_quartet_informative(&masks, true)); assert!(!is_quartet_informative(&masks, FORBID, true));
} }
#[test] #[test]
@@ -152,8 +172,19 @@ mod quartet_tests {
assert!(survives_quartet(&masked, &[])); assert!(survives_quartet(&masked, &[]));
// excluding one genome of the second pair breaks the bipartition // excluding one genome of the second pair breaks the bipartition
assert!(!survives_quartet(&masked, &[false, false, true])); assert!(!survives_quartet(&masked, &[false, false, true]));
// an absence kept as state 0 (no --free-loss) is not a called state // an absence kept as state 0 (no --free-loss) is a state: ∅ ∅ is a pair
assert!(!survives_quartet(&[Some(1), Some(1), Some(0), Some(0)], &[])); assert!(survives_quartet(&[Some(1), Some(1), Some(0), Some(0)], &[]));
// the same genomes under --free-loss are missing: no second pair
assert!(!survives_quartet(&[Some(1), Some(1), None, None], &[]));
}
#[test]
fn the_annex_rule_and_the_exact_check_agree_on_the_same_family() {
let raw = [0b0001u8, 0b0001, 0, 0];
for free_loss in [true, false] {
let masked: Vec<_> = raw.iter().map(|&m| masked_state(m, free_loss, KEEP)).collect();
assert_eq!(is_quartet_informative(&raw, KEEP, free_loss), survives_quartet(&masked, &[]));
}
} }
} }
@@ -188,6 +219,32 @@ pub(crate) fn iupac_code(state: u8) -> u8 {
} }
} }
/// Inverse of [`iupac_code`]: the `0..=15` state a cell character encodes
/// (`-` -> 0, a base -> its single bit, an ambiguity code -> its bits), or
/// `None` for anything else — notably `?`, a missing cell.
#[inline]
pub(crate) fn iupac_state(cell: u8) -> Option<u8> {
Some(match cell {
b'-' => 0b0000,
b'A' => 0b0001,
b'C' => 0b0010,
b'G' => 0b0100,
b'T' => 0b1000,
b'R' => 0b0101,
b'Y' => 0b1010,
b'S' => 0b0110,
b'W' => 0b1001,
b'K' => 0b1100,
b'M' => 0b0011,
b'B' => 0b1110,
b'D' => 0b1101,
b'H' => 0b1011,
b'V' => 0b0111,
b'N' => 0b1111,
_ => return None,
})
}
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::*; use super::*;
@@ -196,7 +253,20 @@ mod tests {
/// `subsample::sample_layer` does once, up front, before calling /// `subsample::sample_layer` does once, up front, before calling
/// [`survives_masking`]. /// [`survives_masking`].
fn masked(raw: &[u8], free_loss: bool, no_ambiguity: bool) -> Vec<Option<u8>> { fn masked(raw: &[u8], free_loss: bool, no_ambiguity: bool) -> Vec<Option<u8>> {
raw.iter().map(|&m| masked_state(m, free_loss, no_ambiguity)).collect() let policy = AmbiguityPolicy::from_flags(no_ambiguity, false);
raw.iter().map(|&m| masked_state(m, free_loss, policy)).collect()
}
#[test]
fn only_diploid_keeps_up_to_two_members() {
let m = |raw: u8| masked_state(raw, false, AmbiguityPolicy::OnlyDiploid);
assert_eq!(m(0b0000), Some(0)); // absence stays a state (no --free-loss)
assert_eq!(m(0b0100), Some(0b0100));
assert_eq!(m(0b0101), Some(0b0101)); // R
assert_eq!(m(0b0111), None); // V
assert_eq!(m(0b1111), None); // N
// `Forbid` is stricter: a heterozygote is missing there
assert_eq!(masked_state(0b0101, false, AmbiguityPolicy::Forbid), None);
} }
#[test] #[test]
@@ -11,7 +11,7 @@ use std::path::Path;
use obicompactvec::TempBitVecBuilder; use obicompactvec::TempBitVecBuilder;
use obikindex::{OKIError, OKIResult}; use obikindex::{OKIError, OKIResult};
use crate::siblings::{FamilyMask, FamilyScalarAnnex, ScalarKind, SiblingAnnex}; use crate::siblings::{AmbiguityPolicy, FamilyMask, FamilyScalarAnnex, ScalarKind, SiblingAnnex};
pub(crate) fn is_non_monomorphic_minorant(mask: FamilyMask) -> bool { pub(crate) fn is_non_monomorphic_minorant(mask: FamilyMask) -> bool {
mask.is_minorant() && mask.siblings() >= 1 mask.is_minorant() && mask.siblings() >= 1
@@ -92,7 +92,7 @@ pub(crate) fn eligible_bitset_layer(layer_dir: &Path) -> OKIResult<TempBitVecBui
/// at each site. /// at each site.
/// ///
/// Reads the `kind` annex (entropy, pair weight) built under the /// Reads the `kind` annex (entropy, pair weight) built under the
/// `no_ambiguity` counting convention; it must already exist /// `policy` counting convention; it must already exist
/// (`ensure_layer_family_annexes` / `ensure_layer_pair_weight_annexes` /// (`ensure_layer_family_annexes` / `ensure_layer_pair_weight_annexes`
/// first) — read positionally (`compact` advances only on non-monomorphic /// first) — read positionally (`compact` advances only on non-monomorphic
/// minorants, mirroring exactly how the annex was written), never /// minorants, mirroring exactly how the annex was written), never
@@ -101,15 +101,15 @@ pub(crate) fn eligible_bitset_layer(layer_dir: &Path) -> OKIResult<TempBitVecBui
pub(crate) fn iter_full_scalar( pub(crate) fn iter_full_scalar(
layer_dir: &Path, layer_dir: &Path,
kind: ScalarKind, kind: ScalarKind,
no_ambiguity: bool, policy: AmbiguityPolicy,
) -> OKIResult<impl Iterator<Item = f32> + use<>> { ) -> OKIResult<impl Iterator<Item = f32> + use<>> {
let annex = SiblingAnnex::open_in(layer_dir).map_err(OKIError::Io)?; let annex = SiblingAnnex::open_in(layer_dir).map_err(OKIError::Io)?;
let scalars = let scalars =
FamilyScalarAnnex::open(&layer_dir.join(kind.file_name(no_ambiguity)), kind).map_err(OKIError::Io)?; FamilyScalarAnnex::open(&layer_dir.join(kind.file_name(policy)), kind).map_err(OKIError::Io)?;
if scalars.no_ambiguity() != no_ambiguity { if scalars.policy() != policy {
return Err(OKIError::Io(std::io::Error::new( return Err(OKIError::Io(std::io::Error::new(
std::io::ErrorKind::InvalidData, std::io::ErrorKind::InvalidData,
format!("{kind:?} annex was built under the other --no-ambiguity convention"), format!("{kind:?} annex was built under {:?}, not {policy:?}", scalars.policy()),
))); )));
} }
+3 -1
View File
@@ -14,6 +14,7 @@
//! encoding. //! encoding.
mod alignment; mod alignment;
mod alignment_vcf;
mod annex; mod annex;
mod cardcomp; mod cardcomp;
mod entropy; mod entropy;
@@ -34,7 +35,7 @@ mod subsample;
use obikidxcache::index_cache::IndexCache; use obikidxcache::index_cache::IndexCache;
pub use alignment::{AlignmentReducer, SnpAlignment}; pub use alignment::{AlignmentReducer, SiteLocus, SnpAlignment};
pub use cardcomp::{ pub use cardcomp::{
cardinality_transition_probs, composition_transition_probs, pairwise_cost_matrix, cardinality_transition_probs, composition_transition_probs, pairwise_cost_matrix,
}; };
@@ -48,6 +49,7 @@ pub use subsample::{EntropyBias, LayerReport, SamplePlan, SamplingSpec, Survivin
pub(crate) use annex::build_layer_sibling_annex; pub(crate) use annex::build_layer_sibling_annex;
pub(crate) use entropy_scan::{EntropySummary, scan_family_entropies}; pub(crate) use entropy_scan::{EntropySummary, scan_family_entropies};
pub use family_annexes::AnnexSelection;
pub(crate) use family_annexes::{ pub(crate) use family_annexes::{
AnnexRequest, ensure_layer_family_annexes, ensure_layer_pair_weight_annexes, AnnexRequest, ensure_layer_family_annexes, ensure_layer_pair_weight_annexes,
}; };
@@ -13,7 +13,7 @@
//! Values are only meaningful relative to each other; the sampler rescales //! Values are only meaningful relative to each other; the sampler rescales
//! them per layer (`SamplingSpec::pair_balance`). //! them per layer (`SamplingSpec::pair_balance`).
use crate::siblings::FamilyOverlap; use crate::siblings::{AmbiguityPolicy, FamilyOverlap};
use crate::siblings::helpers::is_carrier; use crate::siblings::helpers::is_carrier;
/// `1 / O` for every pair of genomes, as a dense symmetric `n × n` matrix /// `1 / O` for every pair of genomes, as a dense symmetric `n × n` matrix
@@ -38,16 +38,16 @@ pub(crate) fn inverse_overlap_matrix(overlap: &FamilyOverlap) -> Vec<f32> {
/// The pair-balance weight of one family from its raw per-genome masks: /// The pair-balance weight of one family from its raw per-genome masks:
/// mean of `inv_overlap` over the pairs of carriers (carriers per /// mean of `inv_overlap` over the pairs of carriers (carriers per
/// [`is_carrier`], under `no_ambiguity`). `0.0` for a family with fewer than /// [`is_carrier`], under `policy`). `0.0` for a family with fewer than
/// two carriers — it covers no pair, so it can balance none. /// two carriers — it covers no pair, so it can balance none.
pub(crate) fn pair_balance_weight( pub(crate) fn pair_balance_weight(
genome_mask: &[u8], genome_mask: &[u8],
no_ambiguity: bool, policy: AmbiguityPolicy,
inv_overlap: &[f32], inv_overlap: &[f32],
) -> f32 { ) -> f32 {
let n = genome_mask.len(); let n = genome_mask.len();
let carriers: Vec<usize> = let carriers: Vec<usize> =
(0..n).filter(|&g| is_carrier(genome_mask[g], no_ambiguity)).collect(); (0..n).filter(|&g| is_carrier(genome_mask[g], policy)).collect();
let k = carriers.len(); let k = carriers.len();
if k < 2 { if k < 2 {
return 0.0; return 0.0;
@@ -83,30 +83,30 @@ mod tests {
#[test] #[test]
fn single_pair_weight_is_the_inverse_overlap() { fn single_pair_weight_is_the_inverse_overlap() {
let w = pair_balance_weight(&[1, 1, 0, 0], false, &matrix()); let w = pair_balance_weight(&[1, 1, 0, 0], AmbiguityPolicy::Keep, &matrix());
assert!((w - 0.5).abs() < 1e-6); assert!((w - 0.5).abs() < 1e-6);
} }
#[test] #[test]
fn weight_is_the_mean_not_the_sum_over_pairs() { fn weight_is_the_mean_not_the_sum_over_pairs() {
// carriers 0, 1, 2: pairs (0,1)=1/2, (0,2)=1/4, (1,2)=1 -> mean 7/12 // carriers 0, 1, 2: pairs (0,1)=1/2, (0,2)=1/4, (1,2)=1 -> mean 7/12
let w = pair_balance_weight(&[1, 1, 1, 0], false, &matrix()); let w = pair_balance_weight(&[1, 1, 1, 0], AmbiguityPolicy::Keep, &matrix());
assert!((w - 7.0 / 12.0).abs() < 1e-6); assert!((w - 7.0 / 12.0).abs() < 1e-6);
} }
#[test] #[test]
fn fewer_than_two_carriers_weigh_nothing() { fn fewer_than_two_carriers_weigh_nothing() {
assert_eq!(pair_balance_weight(&[1, 0, 0, 0], false, &matrix()), 0.0); assert_eq!(pair_balance_weight(&[1, 0, 0, 0], AmbiguityPolicy::Keep, &matrix()), 0.0);
assert_eq!(pair_balance_weight(&[0, 0, 0, 0], false, &matrix()), 0.0); assert_eq!(pair_balance_weight(&[0, 0, 0, 0], AmbiguityPolicy::Keep, &matrix()), 0.0);
} }
#[test] #[test]
fn no_ambiguity_leaves_multi_member_genomes_out_of_the_pairs() { fn no_ambiguity_leaves_multi_member_genomes_out_of_the_pairs() {
let masks = [1u8, 1, 0b0011, 0]; let masks = [1u8, 1, 0b0011, 0];
// without the flag genome 2 carries the family: pairs (0,1), (0,2), (1,2) // without the flag genome 2 carries the family: pairs (0,1), (0,2), (1,2)
let with = pair_balance_weight(&masks, false, &matrix()); let with = pair_balance_weight(&masks, AmbiguityPolicy::Keep, &matrix());
// with it, only the pair (0,1) remains // with it, only the pair (0,1) remains
let without = pair_balance_weight(&masks, true, &matrix()); let without = pair_balance_weight(&masks, AmbiguityPolicy::Forbid, &matrix());
assert!((with - 7.0 / 12.0).abs() < 1e-6); assert!((with - 7.0 / 12.0).abs() < 1e-6);
assert!((without - 0.5).abs() < 1e-6); assert!((without - 0.5).abs() < 1e-6);
} }
@@ -114,8 +114,8 @@ mod tests {
#[test] #[test]
fn a_rare_pair_outweighs_a_common_one() { fn a_rare_pair_outweighs_a_common_one() {
// genomes (0,2) share many families, (1,3) few // genomes (0,2) share many families, (1,3) few
let common = pair_balance_weight(&[1, 0, 1, 0], false, &matrix()); let common = pair_balance_weight(&[1, 0, 1, 0], AmbiguityPolicy::Keep, &matrix());
let rare = pair_balance_weight(&[0, 1, 0, 1], false, &matrix()); let rare = pair_balance_weight(&[0, 1, 0, 1], AmbiguityPolicy::Keep, &matrix());
assert!(rare > common); assert!(rare > common);
} }
} }
@@ -510,7 +510,7 @@ mod cardinality_masking_tests {
// `SurvivingFamily::masked`'s own docs) — the masking decision is baked // `SurvivingFamily::masked`'s own docs) — the masking decision is baked
// in here, not passed to `reduce_pairwise` itself anymore. // in here, not passed to `reduce_pairwise` itself anymore.
fn ambiguous_vs_single(masked: Vec<Option<u8>>) -> Vec<SurvivingFamily> { fn ambiguous_vs_single(masked: Vec<Option<u8>>) -> Vec<SurvivingFamily> {
vec![SurvivingFamily { family_idx: 0, mask: FamilyMask::EMPTY.with(0).with(1), masked }] vec![SurvivingFamily { family_idx: 0, kmer: obikseq::CanonicalKmer::from_raw_unchecked(0), mask: FamilyMask::EMPTY.with(0).with(1), masked }]
} }
#[test] #[test]
@@ -7,12 +7,12 @@ use obicompactvec::{PersistentBitVec, TempBitVecBuilder};
use obikindex::{OKIError, OKIResult}; use obikindex::{OKIError, OKIResult};
use super::minorant_selection::is_non_monomorphic_minorant; use super::minorant_selection::is_non_monomorphic_minorant;
use crate::siblings::quartet_annex::quartet_annex_file_name; use crate::siblings::quartet_annex::{QuartetConvention, quartet_annex_file_name};
use crate::siblings::SiblingAnnex; use crate::siblings::SiblingAnnex;
/// Clears, in `eligible` (indexed by `family_idx`, monomorphic minorants /// Clears, in `eligible` (indexed by `family_idx`, monomorphic minorants
/// included), every family the quartet annex flags as not quartet-informative /// included), every family the quartet annex flags as not quartet-informative
/// under the `no_ambiguity` counting convention. The annex is indexed by the /// under the `convention` counting convention. The annex is indexed by the
/// compact non-monomorphic ordinal, so the sibling annex is walked in the same /// compact non-monomorphic ordinal, so the sibling annex is walked in the same
/// order to translate one numbering into the other — annex-only, no /// order to translate one numbering into the other — annex-only, no
/// cross-partition resolution. The annex must already exist /// cross-partition resolution. The annex must already exist
@@ -20,10 +20,10 @@ use crate::siblings::SiblingAnnex;
pub(crate) fn restrict_to_quartet( pub(crate) fn restrict_to_quartet(
eligible: &mut TempBitVecBuilder, eligible: &mut TempBitVecBuilder,
layer_dir: &Path, layer_dir: &Path,
no_ambiguity: bool, convention: QuartetConvention,
) -> OKIResult<()> { ) -> OKIResult<()> {
let annex = SiblingAnnex::open_in(layer_dir).map_err(OKIError::Io)?; let annex = SiblingAnnex::open_in(layer_dir).map_err(OKIError::Io)?;
let flags = PersistentBitVec::open(&layer_dir.join(quartet_annex_file_name(no_ambiguity))) let flags = PersistentBitVec::open(&layer_dir.join(quartet_annex_file_name(convention)))
.map_err(OKIError::Io)?; .map_err(OKIError::Io)?;
let mut family_idx = 0usize; let mut family_idx = 0usize;
@@ -84,8 +84,12 @@ impl SampleStatsReducer {
let mut counts = [0u32; 16]; let mut counts = [0u32; 16];
let mut carriers: Vec<usize> = Vec::new(); let mut carriers: Vec<usize> = Vec::new();
for (g, &state) in masked.iter().enumerate() { for (g, &state) in masked.iter().enumerate() {
if let Some(state) = state.filter(|&s| s != 0 && !self.excluded[g]) { let Some(state) = state.filter(|_| !self.excluded[g]) else { continue };
counts[(state & 0b1111) as usize] += 1; // Absence (state 0, only present without `--free-loss`) is a state
// of the alignment, so it takes part in the bipartition — but its
// genome carries none of the family's members.
counts[(state & 0b1111) as usize] += 1;
if state != 0 {
carriers.push(g); carriers.push(g);
} }
} }
@@ -140,6 +144,19 @@ mod tests {
assert_eq!(s.overlap(3, 0), s.overlap(0, 3)); assert_eq!(s.overlap(3, 0), s.overlap(0, 3));
} }
#[test]
fn absence_as_a_state_makes_a_site_informative() {
let mut r = SampleStatsReducer::new(4, &[]);
// A A ∅ ∅ (no --free-loss): two pairs of states, though only two carriers
r.add_site(&[Some(1), Some(1), Some(0), Some(0)]);
// the same genomes under --free-loss: the absences are missing
r.add_site(&[Some(1), Some(1), None, None]);
let s = r.finish();
assert_eq!(s.sites_by_carriers[2], 2);
assert_eq!(s.informative_by_carriers[2], 1);
assert_eq!(s.calls, vec![2, 2, 0, 0]);
}
#[test] #[test]
fn excluded_genomes_count_nowhere() { fn excluded_genomes_count_nowhere() {
let mut r = SampleStatsReducer::new(4, &[false, false, false, true]); let mut r = SampleStatsReducer::new(4, &[false, false, false, true]);
@@ -279,13 +279,14 @@ mod tests {
/// something these tests need to exercise). /// something these tests need to exercise).
fn tally_2genomes(pairs: &[(u8, u8)], n_shared: usize) -> PairwiseTally { fn tally_2genomes(pairs: &[(u8, u8)], n_shared: usize) -> PairwiseTally {
// No masking requested (`free_loss = no_ambiguity = false`) — // No masking requested (`free_loss = no_ambiguity = false`) —
// `masked_state(m, false, false)` is always `Some(m)`, so the raw // `masked_state(m, false, AmbiguityPolicy::Keep)` is always `Some(m)`, so the raw
// mask is exactly what `subsample::sample_layer` would have // mask is exactly what `subsample::sample_layer` would have
// computed here. // computed here.
let mut families = Vec::new(); let mut families = Vec::new();
for &(a, b) in pairs { for &(a, b) in pairs {
families.push(SurvivingFamily { families.push(SurvivingFamily {
family_idx: families.len(), family_idx: families.len(),
kmer: obikseq::CanonicalKmer::from_raw_unchecked(0),
mask: FamilyMask::EMPTY.with(a.trailing_zeros() as u8).with(b.trailing_zeros() as u8), mask: FamilyMask::EMPTY.with(a.trailing_zeros() as u8).with(b.trailing_zeros() as u8),
masked: vec![Some(a), Some(b)], masked: vec![Some(a), Some(b)],
}); });
@@ -296,6 +297,7 @@ mod tests {
let other = BASES[(i + 1) % 4]; let other = BASES[(i + 1) % 4];
families.push(SurvivingFamily { families.push(SurvivingFamily {
family_idx: families.len(), family_idx: families.len(),
kmer: obikseq::CanonicalKmer::from_raw_unchecked(0),
mask: FamilyMask::EMPTY.with(b.trailing_zeros() as u8).with(other.trailing_zeros() as u8), mask: FamilyMask::EMPTY.with(b.trailing_zeros() as u8).with(other.trailing_zeros() as u8),
masked: vec![Some(b), Some(b)], masked: vec![Some(b), Some(b)],
}); });
@@ -106,7 +106,7 @@ pub(crate) fn sibling_annex_stats(cache: &IndexCache) -> OKIResult<SiblingAnnexS
n_genomes, n_genomes,
fast_mode, fast_mode,
&Selection::All, &Selection::All,
|_family_idx, mask, genome_mask| { |_family_idx, _kmer, mask, genome_mask| {
// "Genome g represents this family" means g carries // "Genome g represents this family" means g carries
// *any* of its members, not just the minorant's own — // *any* of its members, not just the minorant's own —
// `genome_mask[g] != 0` is exactly that. // `genome_mask[g] != 0` is exactly that.
@@ -49,6 +49,7 @@ use std::sync::Arc;
use obicompactvec::TempBitVecBuilder; use obicompactvec::TempBitVecBuilder;
use obikidxcache::index_cache::IndexCache; use obikidxcache::index_cache::IndexCache;
use obikindex::OKIResult; use obikindex::OKIResult;
use obikseq::CanonicalKmer;
use obisys::progress_bar; use obisys::progress_bar;
use rand::RngExt; use rand::RngExt;
@@ -56,7 +57,7 @@ use super::family_scan::{Selection, scan_layer_families};
use super::masking::{masked_state, survives_masking, survives_quartet}; use super::masking::{masked_state, survives_masking, survives_quartet};
use super::minorant_selection::{eligible_bitset_layer, iter_full_scalar}; use super::minorant_selection::{eligible_bitset_layer, iter_full_scalar};
use super::{AnnexRequest, is_fast_mode, restrict_to_quartet}; use super::{AnnexRequest, is_fast_mode, restrict_to_quartet};
use crate::siblings::{FamilyMask, ScalarKind}; use crate::siblings::{AmbiguityPolicy, FamilyMask, QuartetConvention, ScalarKind};
use crate::siblings::extensions::SiblingBuilder; use crate::siblings::extensions::SiblingBuilder;
/// Resolution rounds a single layer's sampling may take before giving up on /// Resolution rounds a single layer's sampling may take before giving up on
@@ -82,6 +83,8 @@ const MAX_TOPUP_ROUNDS: usize = 5;
/// batch, well after that callback returns. /// batch, well after that callback returns.
pub struct SurvivingFamily { pub struct SurvivingFamily {
pub family_idx: usize, pub family_idx: usize,
/// The family's minorant k-mer (feeds `CanonicalKmer::family_id`).
pub kmer: CanonicalKmer,
pub mask: FamilyMask, pub mask: FamilyMask,
pub masked: Vec<Option<u8>>, pub masked: Vec<Option<u8>>,
} }
@@ -106,7 +109,10 @@ pub struct EntropyBias {
#[derive(Clone, Copy)] #[derive(Clone, Copy)]
pub struct SamplingSpec { pub struct SamplingSpec {
pub free_loss: bool, pub free_loss: bool,
pub no_ambiguity: bool, /// How genomes carrying several members of a family are treated
/// (`--no-ambiguity` / `--only-diploid`): by the post-hoc masking, and as
/// the counting convention of the persisted annexes the draw reads.
pub ambiguity: AmbiguityPolicy,
pub quartet: bool, pub quartet: bool,
/// `--pair-balance`: weight each family by the mean, over the pairs of /// `--pair-balance`: weight each family by the mean, over the pairs of
/// genomes carrying it, of 1 / (number of families the pair shares), so /// genomes carrying it, of 1 / (number of families the pair shares), so
@@ -116,6 +122,15 @@ pub struct SamplingSpec {
pub entropy_bias: Option<EntropyBias>, pub entropy_bias: Option<EntropyBias>,
} }
impl SamplingSpec {
/// The counting convention of the quartet annex this draw reads: how
/// genomes with several members count, and whether absence is a state —
/// the same two choices the post-hoc masking makes.
pub(crate) fn quartet_convention(&self) -> QuartetConvention {
QuartetConvention { policy: self.ambiguity, free_loss: self.free_loss }
}
}
/// What one layer contributed to a draw — one row of `--sample-plan`. /// What one layer contributed to a draw — one row of `--sample-plan`.
#[derive(Clone, Debug)] #[derive(Clone, Debug)]
pub struct LayerReport { pub struct LayerReport {
@@ -156,17 +171,17 @@ impl SamplePlan {
/// [`sample_layer`]'s own circular index sequence walks, so the streams can be /// [`sample_layer`]'s own circular index sequence walks, so the streams can be
/// `.zip()`-ed directly. Two more streamed passes over the layer's (small) /// `.zip()`-ed directly. Two more streamed passes over the layer's (small)
/// annex files, not a materialised `Vec`/`HashMap` — `iter_full_scalar` /// annex files, not a materialised `Vec`/`HashMap` — `iter_full_scalar`
/// itself never allocates more than its own iterator state. `no_ambiguity` /// itself never allocates more than its own iterator state. `policy`
/// selects the annex built under the same counting convention as the masking /// selects the annex built under the same counting convention as the masking
/// the survivors go through afterwards. /// the survivors go through afterwards.
fn circular_scalar_values( fn circular_scalar_values(
layer_dir: &Path, layer_dir: &Path,
kind: ScalarKind, kind: ScalarKind,
start: usize, start: usize,
no_ambiguity: bool, policy: AmbiguityPolicy,
) -> OKIResult<impl Iterator<Item = f32> + use<>> { ) -> OKIResult<impl Iterator<Item = f32> + use<>> {
let head = iter_full_scalar(layer_dir, kind, no_ambiguity)?.skip(start); let head = iter_full_scalar(layer_dir, kind, policy)?.skip(start);
let tail = iter_full_scalar(layer_dir, kind, no_ambiguity)?.take(start); let tail = iter_full_scalar(layer_dir, kind, policy)?.take(start);
Ok(head.chain(tail)) Ok(head.chain(tail))
} }
@@ -192,7 +207,7 @@ struct LayerPlan {
/// per layer, calling `on_layer(partition, layer, survivors)` once per /// per layer, calling `on_layer(partition, layer, survivors)` once per
/// cached layer that contributed anything, with every family from that /// cached layer that contributed anything, with every family from that
/// layer that was both drawn *and* still carries real polymorphism after /// layer that was both drawn *and* still carries real polymorphism after
/// `free_loss`/`no_ambiguity`/`excluded` masking /// `free_loss`/`ambiguity`/`excluded` masking
/// (`algorithms::masking::survives_masking`) — i.e. every /// (`algorithms::masking::survives_masking`) — i.e. every
/// [`SurvivingFamily`] this produces is already a usable alignment site, no /// [`SurvivingFamily`] this produces is already a usable alignment site, no
/// second resolution pass needed by the caller. `excluded[g]` (see /// second resolution pass needed by the caller. `excluded[g]` (see
@@ -253,10 +268,10 @@ pub(crate) fn sample_index(
// layer, below, each under this run's `--no-ambiguity` counting // layer, below, each under this run's `--no-ambiguity` counting
// convention) — build any missing one now, once, rather than mid-walk. // convention) — build any missing one now, once, rather than mid-walk.
let request = AnnexRequest { let request = AnnexRequest {
entropy: spec.entropy_bias.map(|_| vec![spec.no_ambiguity]).unwrap_or_default(), entropy: spec.entropy_bias.map(|_| vec![spec.ambiguity]).unwrap_or_default(),
quartet: if spec.quartet { vec![spec.no_ambiguity] } else { Vec::new() }, quartet: if spec.quartet { vec![spec.quartet_convention()] } else { Vec::new() },
overlap: Vec::new(), overlap: Vec::new(),
pair_weight: if spec.pair_balance { vec![spec.no_ambiguity] } else { Vec::new() }, pair_weight: if spec.pair_balance { vec![spec.ambiguity] } else { Vec::new() },
}; };
if !request.entropy.is_empty() || !request.quartet.is_empty() || !request.pair_weight.is_empty() { if !request.entropy.is_empty() || !request.quartet.is_empty() || !request.pair_weight.is_empty() {
cache.ensure_family_annexes(&request)?; cache.ensure_family_annexes(&request)?;
@@ -276,12 +291,12 @@ pub(crate) fn sample_index(
.expect("cache-consistent: n_layer(partition) already bounds layer_idx"); .expect("cache-consistent: n_layer(partition) already bounds layer_idx");
let mut eligible = eligible_bitset_layer(layer.dir())?; let mut eligible = eligible_bitset_layer(layer.dir())?;
if spec.quartet { if spec.quartet {
restrict_to_quartet(&mut eligible, layer.dir(), spec.no_ambiguity)?; restrict_to_quartet(&mut eligible, layer.dir(), spec.quartet_convention())?;
} }
let n_minorants = eligible.view().len(); let n_minorants = eligible.view().len();
let eligible_count = eligible.view().count_ones(); let eligible_count = eligible.view().count_ones();
let mass = if spec.pair_balance { let mass = if spec.pair_balance {
iter_full_scalar(layer.dir(), ScalarKind::PairWeight, spec.no_ambiguity)? iter_full_scalar(layer.dir(), ScalarKind::PairWeight, spec.ambiguity)?
.enumerate() .enumerate()
.filter(|&(family_idx, _)| eligible.view().get(family_idx)) .filter(|&(family_idx, _)| eligible.view().get(family_idx))
.map(|(_, weight)| f64::from(weight)) .map(|(_, weight)| f64::from(weight))
@@ -426,7 +441,7 @@ fn sample_layer(
&layer_dir, &layer_dir,
ScalarKind::Entropy, ScalarKind::Entropy,
round_start, round_start,
spec.no_ambiguity, spec.ambiguity,
)?), )?),
None => Box::new(std::iter::repeat(0.0f32)), // never read: kernel forced to 1.0 below None => Box::new(std::iter::repeat(0.0f32)), // never read: kernel forced to 1.0 below
}; };
@@ -435,7 +450,7 @@ fn sample_layer(
&layer_dir, &layer_dir,
ScalarKind::PairWeight, ScalarKind::PairWeight,
round_start, round_start,
spec.no_ambiguity, spec.ambiguity,
)?) )?)
} else { } else {
Box::new(std::iter::repeat(1.0f32)) Box::new(std::iter::repeat(1.0f32))
@@ -478,19 +493,19 @@ fn sample_layer(
n_genomes, n_genomes,
fast_mode, fast_mode,
&Selection::Some(&accepted), &Selection::Some(&accepted),
|family_idx, mask, genome_mask| { |family_idx, kmer, mask, genome_mask| {
debug_assert!( debug_assert!(
mask.family_size() >= 2, mask.family_size() >= 2,
"eligible bitset must only accept non-monomorphic minorants" "eligible bitset must only accept non-monomorphic minorants"
); );
let masked: Vec<Option<u8>> = genome_mask let masked: Vec<Option<u8>> = genome_mask
.iter() .iter()
.map(|&m| masked_state(m, spec.free_loss, spec.no_ambiguity)) .map(|&m| masked_state(m, spec.free_loss, spec.ambiguity))
.collect(); .collect();
if survives_masking(&masked, excluded) if survives_masking(&masked, excluded)
&& (!spec.quartet || survives_quartet(&masked, excluded)) && (!spec.quartet || survives_quartet(&masked, excluded))
{ {
survivors.push(SurvivingFamily { family_idx, mask, masked }); survivors.push(SurvivingFamily { family_idx, kmer, mask, masked });
kept += 1; kept += 1;
} }
}, },
+96
View File
@@ -0,0 +1,96 @@
//! How a genome carrying several members of a family is treated — the one
//! parameter behind `--no-ambiguity` / `--only-diploid`, and the key of every
//! persisted annex whose values depend on which genomes count as carriers
//! (entropy, quartet, pair weight, family overlap).
/// What becomes of a genome carrying more than one member of a family.
///
/// - [`Keep`](Self::Keep): nothing — its set of members is a state of its
/// own (an IUPAC ambiguity code in the alignment).
/// - [`OnlyDiploid`](Self::OnlyDiploid): kept up to two members (a
/// heterozygote); three or four members make it missing data.
/// - [`Forbid`](Self::Forbid) (`--no-ambiguity`): any genome with several
/// members is missing data.
///
/// The order is the on-disk header code (see [`header_flags`](Self::header_flags)),
/// kept so that annexes written when this was a `bool` stay valid.
#[derive(Clone, Copy, Debug, Default, PartialEq, Eq, PartialOrd, Ord, Hash)]
pub enum AmbiguityPolicy {
#[default]
Keep,
Forbid,
OnlyDiploid,
}
impl AmbiguityPolicy {
/// Every policy, in header-code order.
pub const ALL: [AmbiguityPolicy; 3] = [Self::Keep, Self::Forbid, Self::OnlyDiploid];
/// From the two (mutually exclusive) CLI flags.
pub fn from_flags(no_ambiguity: bool, only_diploid: bool) -> Self {
debug_assert!(!(no_ambiguity && only_diploid), "--no-ambiguity and --only-diploid are exclusive");
if no_ambiguity {
Self::Forbid
} else if only_diploid {
Self::OnlyDiploid
} else {
Self::Keep
}
}
/// Whether a genome whose family presence mask is `mask` becomes missing
/// data under this policy.
#[inline]
pub fn drops(self, mask: u8) -> bool {
match self {
Self::Keep => false,
Self::OnlyDiploid => mask.count_ones() > 2,
Self::Forbid => mask.count_ones() > 1,
}
}
/// The `flags` value persisted in an annex header: `0` for `Keep` and `1`
/// for `Forbid` are what the `no_ambiguity` bit used to be.
pub(crate) fn header_flags(self) -> u32 {
match self {
Self::Keep => 0,
Self::Forbid => 1,
Self::OnlyDiploid => 2,
}
}
/// Inverse of [`header_flags`](Self::header_flags); unknown codes read as
/// `Keep`, as a file from before the flag existed.
pub(crate) fn from_header_flags(flags: u32) -> Self {
match flags {
1 => Self::Forbid,
2 => Self::OnlyDiploid,
_ => Self::Keep,
}
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn drops_follows_the_member_count() {
for (mask, keep, diploid, forbid) in
[(0b0000u8, false, false, false), (0b0100, false, false, false), (0b0101, false, false, true), (0b0111, false, true, true), (0b1111, false, true, true)]
{
assert_eq!(AmbiguityPolicy::Keep.drops(mask), keep);
assert_eq!(AmbiguityPolicy::OnlyDiploid.drops(mask), diploid);
assert_eq!(AmbiguityPolicy::Forbid.drops(mask), forbid);
}
}
#[test]
fn header_flags_roundtrip_and_stay_compatible_with_the_old_bool() {
for p in AmbiguityPolicy::ALL {
assert_eq!(AmbiguityPolicy::from_header_flags(p.header_flags()), p);
}
assert_eq!(AmbiguityPolicy::Keep.header_flags(), 0);
assert_eq!(AmbiguityPolicy::Forbid.header_flags(), 1);
}
}
@@ -16,7 +16,7 @@ use crate::siblings::algorithms::{
ensure_layer_pair_weight_annexes, inverse_overlap_matrix, is_fast_mode, ensure_layer_pair_weight_annexes, inverse_overlap_matrix, is_fast_mode,
}; };
use crate::siblings::{ use crate::siblings::{
ANNEX_FILE_NAME, FAMILY_OVERLAP_FILE_NAMES, FamilyOverlap, FamilyOverlapAccumulator, ANNEX_FILE_NAME, AmbiguityPolicy, FAMILY_OVERLAP_FILE_NAMES, FamilyOverlap, FamilyOverlapAccumulator,
QUARTET_ANNEX_FILE_NAMES, ScalarKind, family_overlap_file_name, QUARTET_ANNEX_FILE_NAMES, ScalarKind, family_overlap_file_name,
}; };
@@ -154,14 +154,14 @@ impl SiblingBuilder for IndexCache {
fn build_family_annexes(&self, request: &AnnexRequest) -> OKIResult<()> { fn build_family_annexes(&self, request: &AnnexRequest) -> OKIResult<()> {
let n_genomes = self.meta().genomes().len(); let n_genomes = self.meta().genomes().len();
let fast_mode = is_fast_mode(self); let fast_mode = is_fast_mode(self);
let overlap_path = |no_ambiguity: bool| self.index().dir().join(family_overlap_file_name(no_ambiguity)); let overlap_path = |policy: AmbiguityPolicy| self.index().dir().join(family_overlap_file_name(policy));
// Sweep 1: entropy, quartet, and every family-overlap annex still missing. // Sweep 1: entropy, quartet, and every family-overlap annex still missing.
let mut overlap_accs: Vec<(bool, FamilyOverlapAccumulator)> = request let mut overlap_accs: Vec<(AmbiguityPolicy, FamilyOverlapAccumulator)> = request
.overlap_conventions() .overlap_conventions()
.into_iter() .into_iter()
.filter(|&no_ambiguity| !overlap_path(no_ambiguity).exists()) .filter(|&policy| !overlap_path(policy).exists())
.map(|no_ambiguity| (no_ambiguity, FamilyOverlapAccumulator::new(n_genomes, no_ambiguity))) .map(|policy| (policy, FamilyOverlapAccumulator::new(n_genomes, policy)))
.collect(); .collect();
let pb = progress_bar("family_annexes", self.n_partition() as u64, "partitions"); let pb = progress_bar("family_annexes", self.n_partition() as u64, "partitions");
for part in self.partitions() { for part in self.partitions() {
@@ -172,8 +172,8 @@ impl SiblingBuilder for IndexCache {
pb.inc(1); pb.inc(1);
} }
pb.finish_and_clear(); pb.finish_and_clear();
for (no_ambiguity, acc) in overlap_accs { for (policy, acc) in overlap_accs {
acc.write(&overlap_path(no_ambiguity)).map_err(OKIError::Io)?; acc.write(&overlap_path(policy)).map_err(OKIError::Io)?;
} }
// Sweep 2: pair weights, from the now-complete overlap annexes. // Sweep 2: pair weights, from the now-complete overlap annexes.
@@ -183,9 +183,9 @@ impl SiblingBuilder for IndexCache {
let matrices = request let matrices = request
.pair_weight_conventions() .pair_weight_conventions()
.into_iter() .into_iter()
.map(|no_ambiguity| { .map(|policy| {
FamilyOverlap::open(&overlap_path(no_ambiguity)) FamilyOverlap::open(&overlap_path(policy))
.map(|overlap| (no_ambiguity, inverse_overlap_matrix(&overlap))) .map(|overlap| (policy, inverse_overlap_matrix(&overlap)))
.map_err(OKIError::Io) .map_err(OKIError::Io)
}) })
.collect::<OKIResult<Vec<_>>>()?; .collect::<OKIResult<Vec<_>>>()?;
@@ -13,10 +13,10 @@ use obikidxcache::index_cache::IndexCache;
use obikindex::{OKIError, OKIResult}; use obikindex::{OKIError, OKIResult};
use crate::siblings::algorithms::{ use crate::siblings::algorithms::{
AnnexRequest, EntropySummary, SiblingAnnexStats, scan_family_entropies, sibling_annex_stats, sibling_family_size_histogram, AnnexRequest, AnnexSelection, EntropySummary, SiblingAnnexStats, scan_family_entropies, sibling_annex_stats, sibling_family_size_histogram,
}; };
use crate::siblings::extensions::SiblingBuilder; use crate::siblings::extensions::SiblingBuilder;
use crate::siblings::{FamilyOverlap, family_overlap_file_name}; use crate::siblings::{AmbiguityPolicy, FamilyOverlap, family_overlap_file_name};
pub trait SiblingExt { pub trait SiblingExt {
/// Build the sibling-count/minorant annex for every layer of every /// Build the sibling-count/minorant annex for every layer of every
@@ -39,7 +39,14 @@ pub trait SiblingExt {
/// Cross-partition/cross-layer lookups are required (a k-mer's siblings /// Cross-partition/cross-layer lookups are required (a k-mer's siblings
/// can live in any partition), but the layer loop itself — and thus the /// can live in any partition), but the layer loop itself — and thus the
/// annex file this produces — stays local to one layer at a time. /// annex file this produces — stays local to one layer at a time.
fn build_sibling_annex(&self) -> OKIResult<()>; ///
/// `derived` names the annexes derived from the sibling annex (entropy,
/// quartet, pair weight, family overlap) to have ready afterwards, all
/// under its one ambiguity policy — only those, so a routine rebuild
/// pays for nothing the following computations will not read. Derived
/// annexes that already existed are invalidated by the rebuild and are
/// rebuilt on demand, by the computation that reads them, if not listed.
fn build_sibling_annex(&self, derived: &AnnexSelection) -> OKIResult<()>;
/// Write a CSV (`partition,layer,family_idx,entropy15,entropy4, /// Write a CSV (`partition,layer,family_idx,entropy15,entropy4,
/// family_size,n_genomes_present,n_genomes_unambiguous`) of per-family /// family_size,n_genomes_present,n_genomes_unambiguous`) of per-family
@@ -91,9 +98,9 @@ pub trait SiblingExt {
/// The Family Overlap annex — number of shared *variable* families per /// The Family Overlap annex — number of shared *variable* families per
/// genome pair, index-wide (see [`FamilyOverlap`]'s own module docs), /// genome pair, index-wide (see [`FamilyOverlap`]'s own module docs),
/// under the `--no-ambiguity` counting convention `no_ambiguity` selects /// under the counting convention `policy` selects (which genomes
/// (genomes carrying several members of a family do not count as /// carrying several members of a family still count as carrying it, see
/// carrying it when `true`). Built if missing, same as every other /// [`AmbiguityPolicy`]). Built if missing, same as every other
/// annex-backed method here. Feeds `--family-overlap`'s own matrix /// annex-backed method here. Feeds `--family-overlap`'s own matrix
/// output directly, and `--min-shared-family`'s auto-exclusion /// output directly, and `--min-shared-family`'s auto-exclusion
/// (`FamilyOverlap::mean_row(g) < threshold` — same statistic /// (`FamilyOverlap::mean_row(g) < threshold` — same statistic
@@ -101,22 +108,22 @@ pub trait SiblingExt {
/// triggers adding `g` to the excluded-genome set consumed by /// triggers adding `g` to the excluded-genome set consumed by
/// `algorithms::scan::sampled_scan`'s callers; this needs no /// `algorithms::scan::sampled_scan`'s callers; this needs no
/// computation of its own beyond that check). /// computation of its own beyond that check).
fn family_overlap(&self, no_ambiguity: bool) -> OKIResult<FamilyOverlap>; fn family_overlap(&self, policy: AmbiguityPolicy) -> OKIResult<FamilyOverlap>;
} }
impl SiblingExt for IndexCache { impl SiblingExt for IndexCache {
fn build_sibling_annex(&self) -> OKIResult<()> { fn build_sibling_annex(&self, derived: &AnnexSelection) -> OKIResult<()> {
self.index().with_lock(|| { self.index().with_lock(|| {
self.build_base_sibling_annex()?; self.build_base_sibling_annex()?;
// Precompute the entropy, quartet, pair-weight and family-overlap // Then exactly the derived annexes asked for, now that every
// annexes too, now that every layer's fresh sibling annex is on // layer's fresh sibling annex is on disk
// disk (`non_monomorphic_reindex_layer`, which they depend on, // (`non_monomorphic_reindex_layer`, which they depend on, reads
// reads it) — same routine-rebuild footing as the sibling annex // it).
// itself, rather than leaving them to be built lazily on first if derived.is_empty() {
// `--shannon`/`--entropy`/`--quartet`/`--pair-balance`/ return Ok(());
// `--min-shared-family`/`--family-overlap` use. }
tracing::info!("building entropy, quartet, pair-weight (both --no-ambiguity conventions) and family-overlap annexes"); tracing::info!("building the requested derived annexes ({:?})", derived.policy);
self.ensure_family_annexes(&AnnexRequest::everything()) self.ensure_family_annexes(&derived.request())
}) })
} }
@@ -160,9 +167,9 @@ impl SiblingExt for IndexCache {
sibling_annex_stats(self) sibling_annex_stats(self)
} }
fn family_overlap(&self, no_ambiguity: bool) -> OKIResult<FamilyOverlap> { fn family_overlap(&self, policy: AmbiguityPolicy) -> OKIResult<FamilyOverlap> {
self.ensure_family_annexes(&AnnexRequest { overlap: vec![no_ambiguity], ..Default::default() })?; self.ensure_family_annexes(&AnnexRequest { overlap: vec![policy], ..Default::default() })?;
let path = self.index().dir().join(family_overlap_file_name(no_ambiguity)); let path = self.index().dir().join(family_overlap_file_name(policy));
FamilyOverlap::open(&path).map_err(OKIError::Io) FamilyOverlap::open(&path).map_err(OKIError::Io)
} }
} }
Loaded 100 of 114 files, more files were not shown because too many files have changed in this diff. Show more