♻️ refactor rope implementation to use obikrope
- rename `obirope` → `obikroper`
- replace legacy rope with new in-place, Cell-based implementation
- add ForwardCursor/Backward Cursor & SeekMode support (no more BytesMut)
- update all dependents:
- obiread: switch to Rope + cursors, remove tape.rs
• chunk iterator yields `Rope` instead of Vec<Bytes>
- obiskbuilder: use ForwardCursor over Rope
- remove bytes dependency from affected crates
This commit is contained in:
Eric Coissac
@@ -4,6 +4,6 @@ version = "0.1.0"
|
||||
edition = "2024"
|
||||
|
||||
[dependencies]
|
||||
bytes = "1"
|
||||
obikrope = { path = "../obikrope" }
|
||||
niffler = { version = "2", default-features = false, features = ["gz", "bz2", "lzma", "zstd"] }
|
||||
ureq = "2"
|
||||
|
||||
+28
-118
@@ -1,54 +1,40 @@
|
||||
//! Chunk iterator: yields rope slices that each end on a complete sequence record.
|
||||
//! Chunk iterator: yields Rope slices that each end on a complete sequence record.
|
||||
//!
|
||||
//! Each `Vec<Bytes>` yielded by [`SeqChunkIter`] contains one or more reference-counted
|
||||
//! byte slices that together form a self-contained block of complete sequence records.
|
||||
//! The slices are NOT contiguous in memory — the consumer iterates over them in order.
|
||||
//!
|
||||
//! The splitter operates directly on the rope via [`RopeCursor`], so no packing
|
||||
//! (flattening into a contiguous buffer) is ever required — even for sequences
|
||||
//! longer than the read block size.
|
||||
//! Each `Rope` yielded by [`SeqChunkIter`] contains one or more blocks that
|
||||
//! together form a self-contained block of complete sequence records.
|
||||
|
||||
use std::io::{self, Read};
|
||||
|
||||
use bytes::{Bytes, BytesMut};
|
||||
use obikrope::Rope;
|
||||
|
||||
/// A splitter function: given the accumulated rope, returns the absolute byte
|
||||
/// offset at which to cut, or `None` if no complete-record boundary was found.
|
||||
pub type Splitter = fn(&[Bytes]) -> Option<usize>;
|
||||
pub type Splitter = fn(&Rope) -> Option<usize>;
|
||||
|
||||
/// Iterator that reads from `R` in blocks and yields `Vec<Bytes>` chunks,
|
||||
/// Iterator that reads from `R` in blocks and yields `Rope` chunks,
|
||||
/// each ending on a complete sequence record boundary.
|
||||
pub struct SeqChunkIter<R> {
|
||||
source: R,
|
||||
rope: Vec<Bytes>,
|
||||
rope: Rope,
|
||||
block_size: usize,
|
||||
splitter: Splitter,
|
||||
probe: &'static [u8],
|
||||
eof: bool,
|
||||
}
|
||||
|
||||
impl<R: Read> SeqChunkIter<R> {
|
||||
/// Create a new iterator.
|
||||
///
|
||||
/// - `block_size`: bytes per read call (default 1 MiB).
|
||||
/// - `splitter`: format-specific backward boundary detector working on the rope.
|
||||
/// - `probe`: short byte string whose presence is necessary (not sufficient)
|
||||
/// for a boundary to exist (e.g. `b"\n>"` for FASTA, `b"\n@"` for FASTQ).
|
||||
pub fn new(source: R, block_size: usize, splitter: Splitter, probe: &'static [u8]) -> Self {
|
||||
pub fn new(source: R, block_size: usize, splitter: Splitter) -> Self {
|
||||
Self {
|
||||
source,
|
||||
rope: Vec::with_capacity(4),
|
||||
rope: Rope::new(),
|
||||
block_size,
|
||||
splitter,
|
||||
probe,
|
||||
eof: false,
|
||||
}
|
||||
}
|
||||
|
||||
/// Read one block from source into a fresh `Bytes`.
|
||||
/// Returns `None` on EOF (zero bytes read).
|
||||
fn read_block(&mut self) -> io::Result<Option<Bytes>> {
|
||||
let mut buf = BytesMut::zeroed(self.block_size);
|
||||
fn read_block(&mut self) -> io::Result<Option<Vec<u8>>> {
|
||||
let mut buf = vec![0u8; self.block_size];
|
||||
let mut filled = 0;
|
||||
loop {
|
||||
match self.source.read(&mut buf[filled..]) {
|
||||
@@ -67,46 +53,12 @@ impl<R: Read> SeqChunkIter<R> {
|
||||
return Ok(None);
|
||||
}
|
||||
buf.truncate(filled);
|
||||
Ok(Some(buf.freeze()))
|
||||
}
|
||||
|
||||
/// Check whether the boundary probe might appear in the newly added block
|
||||
/// or at the seam between the last two blocks.
|
||||
///
|
||||
/// This is a fast O(block_size) heuristic: if the probe is absent, the
|
||||
/// splitter is not called.
|
||||
fn probe_in_last_chunk(&self) -> bool {
|
||||
let last = match self.rope.last() {
|
||||
Some(b) => b,
|
||||
None => return false,
|
||||
};
|
||||
|
||||
// Within the last block.
|
||||
if last.windows(self.probe.len()).any(|w| w == self.probe) {
|
||||
return true;
|
||||
}
|
||||
|
||||
// At the seam between the previous block and this one.
|
||||
if self.rope.len() >= 2 {
|
||||
let prev = &self.rope[self.rope.len() - 2];
|
||||
let overlap = self.probe.len() - 1;
|
||||
let from = prev.len().saturating_sub(overlap);
|
||||
let seam: Vec<u8> = prev[from..]
|
||||
.iter()
|
||||
.chain(last.iter().take(overlap))
|
||||
.copied()
|
||||
.collect();
|
||||
if seam.windows(self.probe.len()).any(|w| w == self.probe) {
|
||||
return true;
|
||||
}
|
||||
}
|
||||
|
||||
false
|
||||
Ok(Some(buf))
|
||||
}
|
||||
}
|
||||
|
||||
impl<R: Read> Iterator for SeqChunkIter<R> {
|
||||
type Item = io::Result<Vec<Bytes>>;
|
||||
type Item = io::Result<Rope>;
|
||||
|
||||
fn next(&mut self) -> Option<Self::Item> {
|
||||
loop {
|
||||
@@ -114,7 +66,7 @@ impl<R: Read> Iterator for SeqChunkIter<R> {
|
||||
if self.rope.is_empty() {
|
||||
return None;
|
||||
}
|
||||
return Some(Ok(std::mem::take(&mut self.rope)));
|
||||
return Some(Ok(std::mem::replace(&mut self.rope, Rope::new())));
|
||||
}
|
||||
|
||||
match self.read_block() {
|
||||
@@ -128,54 +80,16 @@ impl<R: Read> Iterator for SeqChunkIter<R> {
|
||||
}
|
||||
}
|
||||
|
||||
if self.probe_in_last_chunk() {
|
||||
if let Some(abs_offset) = (self.splitter)(&self.rope) {
|
||||
return Some(Ok(self.split_at(abs_offset)));
|
||||
}
|
||||
if let Some(abs_offset) = (self.splitter)(&self.rope) {
|
||||
let tail = self.rope.split_off(abs_offset)
|
||||
.expect("splitter returned valid offset");
|
||||
let chunk = std::mem::replace(&mut self.rope, tail);
|
||||
return Some(Ok(chunk));
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
impl<R> SeqChunkIter<R> {
|
||||
/// Split the rope at absolute byte offset `abs_offset`.
|
||||
///
|
||||
/// Returns the chunk (`rope[..abs_offset]`) as a `Vec<Bytes>` and stores
|
||||
/// the remainder (`rope[abs_offset..]`) in `self.rope`.
|
||||
fn split_at(&mut self, abs_offset: usize) -> Vec<Bytes> {
|
||||
let mut chunk = Vec::with_capacity(self.rope.len());
|
||||
let mut remaining = abs_offset;
|
||||
|
||||
let mut old_rope = std::mem::take(&mut self.rope);
|
||||
let mut remainder_rope: Vec<Bytes> = Vec::with_capacity(2);
|
||||
|
||||
for piece in old_rope.drain(..) {
|
||||
if remaining == 0 {
|
||||
remainder_rope.push(piece);
|
||||
} else if remaining >= piece.len() {
|
||||
remaining -= piece.len();
|
||||
chunk.push(piece);
|
||||
} else {
|
||||
// The cut falls inside this piece.
|
||||
// Copy both halves so each gets its own Arc (unique ownership),
|
||||
// which is required by RopeTape for in-place writing.
|
||||
let head = BytesMut::from(&piece[..remaining]).freeze();
|
||||
let tail = BytesMut::from(&piece[remaining..]).freeze();
|
||||
remaining = 0;
|
||||
if !head.is_empty() {
|
||||
chunk.push(head);
|
||||
}
|
||||
if !tail.is_empty() {
|
||||
remainder_rope.push(tail);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
self.rope = remainder_rope;
|
||||
chunk
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
@@ -183,17 +97,15 @@ mod tests {
|
||||
use crate::fastq::end_of_last_fastq_entry;
|
||||
|
||||
fn fasta_iter(data: &'static [u8], block_size: usize) -> SeqChunkIter<&'static [u8]> {
|
||||
SeqChunkIter::new(data, block_size, end_of_last_fasta_entry, b"\n>")
|
||||
SeqChunkIter::new(data, block_size, end_of_last_fasta_entry)
|
||||
}
|
||||
|
||||
fn fastq_iter(data: &'static [u8], block_size: usize) -> SeqChunkIter<&'static [u8]> {
|
||||
SeqChunkIter::new(data, block_size, end_of_last_fastq_entry, b"\n@")
|
||||
SeqChunkIter::new(data, block_size, end_of_last_fastq_entry)
|
||||
}
|
||||
|
||||
fn collect_fasta(chunks: Vec<Vec<Bytes>>) -> Vec<Vec<u8>> {
|
||||
chunks.into_iter().map(|rope| {
|
||||
rope.into_iter().flat_map(|b| b.to_vec()).collect()
|
||||
}).collect()
|
||||
fn rope_to_vec(rope: &Rope) -> Vec<u8> {
|
||||
rope.fw_cursor().collect()
|
||||
}
|
||||
|
||||
// ── FASTA ─────────────────────────────────────────────────────────────────
|
||||
@@ -203,16 +115,14 @@ mod tests {
|
||||
let data: &[u8] = b">s1\nACGT\n";
|
||||
let chunks: Vec<_> = fasta_iter(data, 64).collect::<Result<_, _>>().unwrap();
|
||||
assert_eq!(chunks.len(), 1);
|
||||
let flat: Vec<u8> = chunks.into_iter().flatten().flat_map(|b| b.to_vec()).collect();
|
||||
assert_eq!(flat, b">s1\nACGT\n");
|
||||
assert_eq!(rope_to_vec(&chunks[0]), b">s1\nACGT\n");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_two_records_split_across_chunks() {
|
||||
let data: &[u8] = b">s1\nACGT\n>s2\nTTTT\n";
|
||||
let chunks: Vec<_> = fasta_iter(data, 10).collect::<Result<_, _>>().unwrap();
|
||||
let flat = collect_fasta(chunks);
|
||||
let all: Vec<u8> = flat.into_iter().flatten().collect();
|
||||
let all: Vec<u8> = chunks.iter().flat_map(|r| rope_to_vec(r)).collect();
|
||||
assert_eq!(all, b">s1\nACGT\n>s2\nTTTT\n");
|
||||
}
|
||||
|
||||
@@ -222,7 +132,7 @@ mod tests {
|
||||
for block in [8, 12, 20, 100] {
|
||||
let chunks: Vec<_> = fasta_iter(data, block).collect::<Result<_, _>>().unwrap();
|
||||
for rope in &chunks {
|
||||
let flat: Vec<u8> = rope.iter().flat_map(|b| b.to_vec()).collect();
|
||||
let flat = rope_to_vec(rope);
|
||||
assert_eq!(flat[0], b'>', "block={block}: chunk doesn't start with '>'");
|
||||
assert_eq!(*flat.last().unwrap(), b'\n', "block={block}: chunk doesn't end with newline");
|
||||
}
|
||||
@@ -258,7 +168,7 @@ mod tests {
|
||||
(b"TTTTTTTT", b"HHHHHHHH"),
|
||||
]).into_boxed_slice());
|
||||
let chunks: Vec<_> = fastq_iter(data, 16).collect::<Result<_, _>>().unwrap();
|
||||
let all: Vec<u8> = chunks.into_iter().flatten().flat_map(|b| b.to_vec()).collect();
|
||||
let all: Vec<u8> = chunks.iter().flat_map(|r| rope_to_vec(r)).collect();
|
||||
assert_eq!(all, *data);
|
||||
}
|
||||
|
||||
@@ -273,7 +183,7 @@ mod tests {
|
||||
for block in [18, 30, 60] {
|
||||
let chunks: Vec<_> = fastq_iter(data, block).collect::<Result<_, _>>().unwrap();
|
||||
for rope in &chunks {
|
||||
let first_byte = rope.iter().flat_map(|b| b.iter().copied()).next().unwrap();
|
||||
let first_byte = rope_to_vec(rope)[0];
|
||||
assert_eq!(first_byte, b'@', "block={block}: chunk doesn't start with '@'");
|
||||
}
|
||||
}
|
||||
|
||||
+42
-46
@@ -1,8 +1,6 @@
|
||||
//! Backward boundary detection for FASTA chunks.
|
||||
|
||||
use bytes::Bytes;
|
||||
|
||||
use crate::tape::RopeCursor;
|
||||
use obikrope::{Rope, RopeCursor};
|
||||
|
||||
/// Scan the rope backward for the start of the last complete FASTA entry.
|
||||
///
|
||||
@@ -10,57 +8,51 @@ use crate::tape::RopeCursor;
|
||||
/// record, so that `rope[..offset]` is a self-contained chunk and
|
||||
/// `rope[offset..]` is the remainder carried into the next chunk.
|
||||
/// Returns `None` if no valid boundary is found (need more data).
|
||||
///
|
||||
/// Port of Go's `EndOfLastFastaEntry`, now working directly on a rope
|
||||
/// via [`RopeCursor`] — no contiguous packing required.
|
||||
pub fn end_of_last_fasta_entry(rope: &[Bytes]) -> Option<usize> {
|
||||
let total_len: usize = rope.iter().map(|b| b.len()).sum();
|
||||
if total_len == 0 {
|
||||
return None;
|
||||
}
|
||||
|
||||
let mut cursor = RopeCursor::end(rope)?;
|
||||
pub fn end_of_last_fasta_entry(rope: &Rope) -> Option<usize> {
|
||||
let cursor = rope.bw_cursor();
|
||||
let mut state: u8 = 0;
|
||||
let mut last: usize = 0;
|
||||
let mut i = total_len as isize - 1;
|
||||
|
||||
while i >= 0 && state < 2 {
|
||||
let c = cursor.peek(rope).unwrap();
|
||||
for c in cursor.iter() {
|
||||
match state {
|
||||
0 if c == b'>' => {
|
||||
last = cursor.tell()?;
|
||||
state = 1;
|
||||
last = i as usize;
|
||||
}
|
||||
1 if c == b'\n' || c == b'\r' => {
|
||||
state = 2;
|
||||
if last > 0 {
|
||||
return Some(last);
|
||||
}
|
||||
return None;
|
||||
}
|
||||
_ => {
|
||||
1 => {
|
||||
state = 0;
|
||||
}
|
||||
}
|
||||
i -= 1;
|
||||
if i >= 0 {
|
||||
cursor.previous(rope);
|
||||
_ => {}
|
||||
}
|
||||
}
|
||||
|
||||
if i <= 0 || state != 2 {
|
||||
return None;
|
||||
}
|
||||
Some(last)
|
||||
None
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
use bytes::BytesMut;
|
||||
|
||||
fn rope(data: &[u8]) -> Vec<Bytes> {
|
||||
vec![BytesMut::from(data).freeze()]
|
||||
fn rope(data: &[u8]) -> Rope {
|
||||
let mut r = Rope::new();
|
||||
r.push(data.to_vec());
|
||||
r
|
||||
}
|
||||
|
||||
fn rope2(a: &[u8], b: &[u8]) -> Vec<Bytes> {
|
||||
vec![BytesMut::from(a).freeze(), BytesMut::from(b).freeze()]
|
||||
fn rope2(a: &[u8], b: &[u8]) -> Rope {
|
||||
let mut r = Rope::new();
|
||||
r.push(a.to_vec());
|
||||
r.push(b.to_vec());
|
||||
r
|
||||
}
|
||||
|
||||
fn flat(r: &Rope) -> Vec<u8> {
|
||||
r.fw_cursor().collect()
|
||||
}
|
||||
|
||||
#[test]
|
||||
@@ -71,39 +63,43 @@ mod tests {
|
||||
#[test]
|
||||
fn two_entries_cuts_at_second_header() {
|
||||
let data = b">seq1\nACGT\n>seq2\nTTTT\n";
|
||||
let pos = end_of_last_fasta_entry(&rope(data)).unwrap();
|
||||
assert_eq!(&data[pos..], b">seq2\nTTTT\n");
|
||||
assert_eq!(&data[..pos], b">seq1\nACGT\n");
|
||||
let r = rope(data);
|
||||
let pos = end_of_last_fasta_entry(&r).unwrap();
|
||||
assert_eq!(&flat(&r)[pos..], b">seq2\nTTTT\n");
|
||||
assert_eq!(&flat(&r)[..pos], b">seq1\nACGT\n");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn three_entries_cuts_at_last_header() {
|
||||
let data = b">s1\nAA\n>s2\nCC\n>s3\nGG\n";
|
||||
let pos = end_of_last_fasta_entry(&rope(data)).unwrap();
|
||||
assert_eq!(&data[pos..], b">s3\nGG\n");
|
||||
let r = rope(data);
|
||||
let pos = end_of_last_fasta_entry(&r).unwrap();
|
||||
assert_eq!(&flat(&r)[pos..], b">s3\nGG\n");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn multiline_sequence() {
|
||||
let data = b">s1\nACGT\nACGT\n>s2\nTTTT\n";
|
||||
let pos = end_of_last_fasta_entry(&rope(data)).unwrap();
|
||||
assert_eq!(&data[pos..], b">s2\nTTTT\n");
|
||||
let r = rope(data);
|
||||
let pos = end_of_last_fasta_entry(&r).unwrap();
|
||||
assert_eq!(&flat(&r)[pos..], b">s2\nTTTT\n");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn crlf_line_endings() {
|
||||
let data = b">s1\r\nACGT\r\n>s2\r\nTTTT\r\n";
|
||||
let pos = end_of_last_fasta_entry(&rope(data)).unwrap();
|
||||
assert_eq!(&data[pos..], b">s2\r\nTTTT\r\n");
|
||||
let r = rope(data);
|
||||
let pos = end_of_last_fasta_entry(&r).unwrap();
|
||||
assert_eq!(&flat(&r)[pos..], b">s2\r\nTTTT\r\n");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn boundary_spans_two_blocks() {
|
||||
// Split so that "\n" is at end of first block and ">" at start of second.
|
||||
let a = b">s1\nACGT\n";
|
||||
let b = b">s2\nTTTT\n";
|
||||
let total: Vec<u8> = a.iter().chain(b.iter()).copied().collect();
|
||||
let pos = end_of_last_fasta_entry(&rope2(a, b)).unwrap();
|
||||
assert_eq!(&total[pos..], b">s2\nTTTT\n");
|
||||
let r = rope2(a, b);
|
||||
let all: Vec<u8> = flat(&r);
|
||||
let pos = end_of_last_fasta_entry(&r).unwrap();
|
||||
assert_eq!(&all[pos..], b">s2\nTTTT\n");
|
||||
}
|
||||
}
|
||||
|
||||
+42
-67
@@ -9,13 +9,8 @@
|
||||
//! The `@` in quality lines (Phred 31 = ASCII 64) makes forward heuristics
|
||||
//! unreliable. The backward scanner identifies a genuine record start by
|
||||
//! verifying the structural context around each `@` candidate.
|
||||
//!
|
||||
//! Port of Go's `EndOfLastFastqEntry` (7-state machine), now working directly
|
||||
//! on a rope via [`RopeCursor`] — no contiguous packing required.
|
||||
|
||||
use bytes::Bytes;
|
||||
|
||||
use crate::tape::RopeCursor;
|
||||
use obikrope::{Rope, RopeCursor, SeekMode};
|
||||
|
||||
#[inline]
|
||||
fn is_eol(c: u8) -> bool {
|
||||
@@ -38,42 +33,28 @@ fn is_seq_char(c: u8) -> bool {
|
||||
/// record, so that `rope[..offset]` is a self-contained chunk and
|
||||
/// `rope[offset..]` is the remainder for the next chunk.
|
||||
/// Returns `None` if no valid boundary is found (need more data).
|
||||
pub fn end_of_last_fastq_entry(rope: &[Bytes]) -> Option<usize> {
|
||||
let total_len: usize = rope.iter().map(|b| b.len()).sum();
|
||||
if total_len == 0 {
|
||||
return None;
|
||||
}
|
||||
|
||||
let mut cursor = RopeCursor::end(rope)?;
|
||||
pub fn end_of_last_fastq_entry(rope: &Rope) -> Option<usize> {
|
||||
let mut cursor = rope.bw_cursor();
|
||||
let mut state: u8 = 0;
|
||||
let mut restart: isize = total_len as isize - 1;
|
||||
let mut restart_cursor = cursor;
|
||||
let mut cut: usize = total_len;
|
||||
let mut i: isize = total_len as isize - 1;
|
||||
|
||||
while i >= 0 && state < 7 {
|
||||
let c = cursor.peek(rope).unwrap();
|
||||
let mut restart: usize = 0;
|
||||
let mut cut: usize = rope.len();
|
||||
|
||||
while let Some(c) = cursor.next() {
|
||||
match state {
|
||||
// Looking for `+` separator line content
|
||||
0 => {
|
||||
if c == b'+' {
|
||||
state = 1;
|
||||
restart = i;
|
||||
restart_cursor = cursor;
|
||||
restart = cursor.tell()?;
|
||||
}
|
||||
}
|
||||
// Found `+` — expect end-of-line immediately before it (going backward)
|
||||
1 => {
|
||||
if is_eol(c) {
|
||||
state = 2;
|
||||
} else {
|
||||
state = 0;
|
||||
i = restart;
|
||||
cursor = restart_cursor;
|
||||
cursor.seek(restart as isize, SeekMode::Absolute).ok();
|
||||
}
|
||||
}
|
||||
// After `\n+`: skip separators, then expect sequence characters
|
||||
2 => {
|
||||
if is_sep(c) {
|
||||
// stay
|
||||
@@ -81,11 +62,9 @@ pub fn end_of_last_fastq_entry(rope: &[Bytes]) -> Option<usize> {
|
||||
state = 3;
|
||||
} else {
|
||||
state = 0;
|
||||
i = restart;
|
||||
cursor = restart_cursor;
|
||||
cursor.seek(restart as isize, SeekMode::Absolute).ok();
|
||||
}
|
||||
}
|
||||
// Scanning sequence characters backward
|
||||
3 => {
|
||||
if is_eol(c) {
|
||||
state = 4;
|
||||
@@ -93,60 +72,48 @@ pub fn end_of_last_fastq_entry(rope: &[Bytes]) -> Option<usize> {
|
||||
// stay
|
||||
} else {
|
||||
state = 0;
|
||||
i = restart;
|
||||
cursor = restart_cursor;
|
||||
cursor.seek(restart as isize, SeekMode::Absolute).ok();
|
||||
}
|
||||
}
|
||||
// Found end-of-line before sequence — skip any extra newlines
|
||||
4 => {
|
||||
if is_eol(c) {
|
||||
// stay
|
||||
} else {
|
||||
if !is_eol(c) {
|
||||
state = 5;
|
||||
}
|
||||
}
|
||||
// Scanning header content — looking for `@` at start of line
|
||||
5 => {
|
||||
if is_eol(c) {
|
||||
state = 0;
|
||||
i = restart;
|
||||
cursor = restart_cursor;
|
||||
cursor.seek(restart as isize, SeekMode::Absolute).ok();
|
||||
} else if c == b'@' {
|
||||
cut = cursor.tell()?;
|
||||
state = 6;
|
||||
cut = i as usize;
|
||||
}
|
||||
// else: stay
|
||||
// else stay
|
||||
}
|
||||
// Found `@` — expect end-of-line before it
|
||||
6 => {
|
||||
if is_eol(c) {
|
||||
state = 7; // success
|
||||
if cut > 0 {
|
||||
return Some(cut);
|
||||
}
|
||||
return None;
|
||||
} else {
|
||||
state = 5;
|
||||
}
|
||||
}
|
||||
_ => unreachable!(),
|
||||
}
|
||||
|
||||
i -= 1;
|
||||
if i >= 0 {
|
||||
cursor.previous(rope);
|
||||
}
|
||||
}
|
||||
|
||||
if i <= 0 || state != 7 {
|
||||
return None;
|
||||
}
|
||||
Some(cut)
|
||||
None
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
use bytes::BytesMut;
|
||||
|
||||
fn rope(data: &[u8]) -> Vec<Bytes> {
|
||||
vec![BytesMut::from(data).freeze()]
|
||||
fn rope(data: &[u8]) -> Rope {
|
||||
let mut r = Rope::new();
|
||||
r.push(data.to_vec());
|
||||
r
|
||||
}
|
||||
|
||||
fn make_fastq(records: &[(&[u8], &[u8])]) -> Vec<u8> {
|
||||
@@ -162,6 +129,10 @@ mod tests {
|
||||
buf
|
||||
}
|
||||
|
||||
fn flat(r: &Rope) -> Vec<u8> {
|
||||
r.fw_cursor().collect()
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn single_record_no_boundary() {
|
||||
let buf = make_fastq(&[(b"ACGT", b"IIII")]);
|
||||
@@ -171,9 +142,10 @@ mod tests {
|
||||
#[test]
|
||||
fn two_records_cuts_at_second() {
|
||||
let buf = make_fastq(&[(b"ACGT", b"IIII"), (b"TTTT", b"HHHH")]);
|
||||
let pos = end_of_last_fastq_entry(&rope(&buf)).unwrap();
|
||||
assert_eq!(buf[pos], b'@');
|
||||
assert_eq!(&buf[pos..], make_fastq(&[(b"TTTT", b"HHHH")]).as_slice());
|
||||
let r = rope(&buf);
|
||||
let pos = end_of_last_fastq_entry(&r).unwrap();
|
||||
assert_eq!(flat(&r)[pos], b'@');
|
||||
assert_eq!(&flat(&r)[pos..], make_fastq(&[(b"TTTT", b"HHHH")]).as_slice());
|
||||
}
|
||||
|
||||
#[test]
|
||||
@@ -183,8 +155,9 @@ mod tests {
|
||||
(b"CCCC", b"JJJJ"),
|
||||
(b"GGGG", b"KKKK"),
|
||||
]);
|
||||
let pos = end_of_last_fastq_entry(&rope(&buf)).unwrap();
|
||||
assert_eq!(&buf[pos..], make_fastq(&[(b"GGGG", b"KKKK")]).as_slice());
|
||||
let r = rope(&buf);
|
||||
let pos = end_of_last_fastq_entry(&r).unwrap();
|
||||
assert_eq!(&flat(&r)[pos..], make_fastq(&[(b"GGGG", b"KKKK")]).as_slice());
|
||||
}
|
||||
|
||||
#[test]
|
||||
@@ -193,15 +166,17 @@ mod tests {
|
||||
(b"ACGTACGT", b"@@@@IIII"),
|
||||
(b"TTTT", b"HHHH"),
|
||||
]);
|
||||
let pos = end_of_last_fastq_entry(&rope(&buf)).unwrap();
|
||||
assert_eq!(&buf[pos..], make_fastq(&[(b"TTTT", b"HHHH")]).as_slice());
|
||||
let r = rope(&buf);
|
||||
let pos = end_of_last_fastq_entry(&r).unwrap();
|
||||
assert_eq!(&flat(&r)[pos..], make_fastq(&[(b"TTTT", b"HHHH")]).as_slice());
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn crlf_line_endings() {
|
||||
let buf = b"@h\r\nACGT\r\n+\r\nIIII\r\n@h\r\nTTTT\r\n+\r\nHHHH\r\n";
|
||||
let pos = end_of_last_fastq_entry(&rope(buf)).unwrap();
|
||||
assert_eq!(buf[pos], b'@');
|
||||
assert_eq!(&buf[pos..], b"@h\r\nTTTT\r\n+\r\nHHHH\r\n");
|
||||
let data = b"@h\r\nACGT\r\n+\r\nIIII\r\n@h\r\nTTTT\r\n+\r\nHHHH\r\n";
|
||||
let r = rope(data);
|
||||
let pos = end_of_last_fastq_entry(&r).unwrap();
|
||||
assert_eq!(flat(&r)[pos], b'@');
|
||||
assert_eq!(&flat(&r)[pos..], b"@h\r\nTTTT\r\n+\r\nHHHH\r\n");
|
||||
}
|
||||
}
|
||||
|
||||
+4
-15
@@ -1,7 +1,7 @@
|
||||
//! Streaming sequence file reader for obikmer.
|
||||
//!
|
||||
//! Yields rope chunks (`Vec<Bytes>`) from FASTA or FASTQ files, each ending
|
||||
//! on a complete sequence record boundary. Zero allocation in the common case.
|
||||
//! Yields rope chunks from FASTA or FASTQ files, each ending
|
||||
//! on a complete sequence record boundary.
|
||||
|
||||
#![deny(missing_docs)]
|
||||
|
||||
@@ -9,7 +9,6 @@ mod fasta;
|
||||
mod fastq;
|
||||
pub mod chunk;
|
||||
pub mod normalize;
|
||||
pub mod tape;
|
||||
pub mod xopen;
|
||||
|
||||
use std::io::Read;
|
||||
@@ -22,20 +21,10 @@ pub const DEFAULT_BLOCK_SIZE: usize = 1024 * 1024;
|
||||
|
||||
/// Create a FASTA chunk iterator over `source`.
|
||||
pub fn fasta_chunks<R: Read>(source: R) -> SeqChunkIter<R> {
|
||||
SeqChunkIter::new(
|
||||
source,
|
||||
DEFAULT_BLOCK_SIZE,
|
||||
fasta::end_of_last_fasta_entry,
|
||||
b"\n>",
|
||||
)
|
||||
SeqChunkIter::new(source, DEFAULT_BLOCK_SIZE, fasta::end_of_last_fasta_entry)
|
||||
}
|
||||
|
||||
/// Create a FASTQ chunk iterator over `source`.
|
||||
pub fn fastq_chunks<R: Read>(source: R) -> SeqChunkIter<R> {
|
||||
SeqChunkIter::new(
|
||||
source,
|
||||
DEFAULT_BLOCK_SIZE,
|
||||
fastq::end_of_last_fastq_entry,
|
||||
b"\n@",
|
||||
)
|
||||
SeqChunkIter::new(source, DEFAULT_BLOCK_SIZE, fastq::end_of_last_fastq_entry)
|
||||
}
|
||||
|
||||
+191
-270
@@ -1,141 +1,114 @@
|
||||
//! Sequence normalisation automata for FASTA and FASTQ rope chunks.
|
||||
//!
|
||||
//! Both automata operate on a [`RopeTape`] in place (two-cursor, zero
|
||||
//! reallocation in the common case) and produce a compact byte stream:
|
||||
//! ACGT segments of length ≥ k separated by `0x00`, uppercased.
|
||||
//!
|
||||
//! Ambiguous bases terminate the current segment. Segments shorter than k
|
||||
//! are silently discarded by retreating the write cursor.
|
||||
//! Both automata operate on a Rope in place (two-cursor, zero reallocation)
|
||||
//! and produce a compact byte stream: ACGT segments of length ≥ k separated
|
||||
//! by `0x00`, uppercased. Ambiguous bases terminate the current segment.
|
||||
//! Segments shorter than k are silently discarded.
|
||||
|
||||
use bytes::Bytes;
|
||||
|
||||
use crate::tape::{RopeCursor, RopeTape};
|
||||
use obikrope::{ForwardCursor, Rope, RopeCursor};
|
||||
|
||||
// ── public entry points ───────────────────────────────────────────────────────
|
||||
|
||||
/// Normalise a FASTA chunk: skip headers; copy, filter and concatenate
|
||||
/// sequence lines across multi-line records.
|
||||
///
|
||||
/// Returns the written region of the tape as a rope of `Bytes`.
|
||||
pub fn normalize_fasta_chunk(chunks: Vec<Bytes>, k: usize) -> Vec<Bytes> {
|
||||
let mut tape = RopeTape::new(chunks);
|
||||
normalize_fasta(&mut tape, k);
|
||||
tape.into_output()
|
||||
/// Normalise a FASTA chunk into a compact ACGT\x00-separated rope.
|
||||
pub fn normalize_fasta_chunk(mut rope: Rope, k: usize) -> Rope {
|
||||
let write_pos = {
|
||||
let read = rope.fw_cursor();
|
||||
let write = rope.fw_cursor();
|
||||
normalize_fasta(&read, &write, k);
|
||||
write.tell().unwrap_or(0)
|
||||
};
|
||||
let _ = rope.split_off(write_pos);
|
||||
rope
|
||||
}
|
||||
|
||||
/// Normalise a FASTQ chunk: skip headers, `+` lines and quality lines;
|
||||
/// copy and filter sequence lines.
|
||||
///
|
||||
/// Returns the written region of the tape as a rope of `Bytes`.
|
||||
pub fn normalize_fastq_chunk(chunks: Vec<Bytes>, k: usize) -> Vec<Bytes> {
|
||||
let mut tape = RopeTape::new(chunks);
|
||||
normalize_fastq(&mut tape, k);
|
||||
tape.into_output()
|
||||
/// Normalise a FASTQ chunk into a compact ACGT\x00-separated rope.
|
||||
pub fn normalize_fastq_chunk(mut rope: Rope, k: usize) -> Rope {
|
||||
let write_pos = {
|
||||
let read = rope.fw_cursor();
|
||||
let write = rope.fw_cursor();
|
||||
normalize_fastq(&read, &write, k);
|
||||
write.tell().unwrap_or(0)
|
||||
};
|
||||
let _ = rope.split_off(write_pos);
|
||||
rope
|
||||
}
|
||||
|
||||
// ── FASTA automaton ───────────────────────────────────────────────────────────
|
||||
|
||||
/// Drive the FASTA normalisation automaton on `tape`.
|
||||
///
|
||||
/// After the initial header skip, the automaton reads sequence characters
|
||||
/// one by one. At each newline run it peeks at the next character to decide:
|
||||
/// `>` → new record (close segment, skip header); anything else → line
|
||||
/// continuation within the same sequence.
|
||||
fn normalize_fasta(tape: &mut RopeTape, k: usize) {
|
||||
// Skip the first header line (includes the leading `>`).
|
||||
skip_line(tape);
|
||||
fn normalize_fasta(read: &ForwardCursor<'_>, write: &ForwardCursor<'_>, k: usize) {
|
||||
skip_line(read);
|
||||
|
||||
loop {
|
||||
let mut seg_start = tape.write_snapshot();
|
||||
let mut seg_start = write.tell().unwrap_or(0);
|
||||
|
||||
'seq: loop {
|
||||
let Some(c) = tape.read_next() else {
|
||||
end_segment(tape, &mut seg_start, k);
|
||||
let Some(c) = read.read_next().ok() else {
|
||||
end_segment(write, &mut seg_start, k);
|
||||
return;
|
||||
};
|
||||
|
||||
if is_newline(c) {
|
||||
// Peek at the very next character — no newline run to skip.
|
||||
// \r\n is handled naturally: \r → peek \n → continue; \n → peek content.
|
||||
match tape.peek() {
|
||||
match read.read_ahead(1).ok() {
|
||||
None => {
|
||||
end_segment(tape, &mut seg_start, k);
|
||||
end_segment(write, &mut seg_start, k);
|
||||
return;
|
||||
}
|
||||
Some(b'>') => {
|
||||
// New record: close current segment, skip next header.
|
||||
end_segment(tape, &mut seg_start, k);
|
||||
skip_line(tape);
|
||||
break 'seq; // restart outer loop
|
||||
}
|
||||
Some(_) => {
|
||||
// Line continuation or \r before \n: next char is a nucleotide.
|
||||
continue 'seq;
|
||||
end_segment(write, &mut seg_start, k);
|
||||
skip_line(read);
|
||||
break 'seq;
|
||||
}
|
||||
Some(_) => continue 'seq,
|
||||
}
|
||||
}
|
||||
|
||||
let upper = c & !0x20u8;
|
||||
if is_acgt(upper) {
|
||||
tape.write_next(upper);
|
||||
write.write(upper).ok();
|
||||
} else {
|
||||
// Ambiguous base: close segment, skip the non-ACGT run.
|
||||
end_segment(tape, &mut seg_start, k);
|
||||
skip_until_acgt_or_newline(tape);
|
||||
seg_start = tape.write_snapshot();
|
||||
end_segment(write, &mut seg_start, k);
|
||||
skip_until_acgt_or_newline(read);
|
||||
seg_start = write.tell().unwrap_or(0);
|
||||
}
|
||||
}
|
||||
// Outer loop restarts for the next record.
|
||||
}
|
||||
}
|
||||
|
||||
// ── FASTQ automaton ───────────────────────────────────────────────────────────
|
||||
|
||||
/// Drive the FASTQ normalisation automaton on `tape`.
|
||||
///
|
||||
/// The FASTQ 4-line structure (`@header`, sequence, `+`, quality) is rigid,
|
||||
/// so the automaton cycles through four fixed phases without backtracking.
|
||||
fn normalize_fastq(tape: &mut RopeTape, k: usize) {
|
||||
fn normalize_fastq(read: &ForwardCursor<'_>, write: &ForwardCursor<'_>, k: usize) {
|
||||
loop {
|
||||
// ── Phase 1: skip header line (@…\n) ─────────────────────────────
|
||||
skip_line(tape);
|
||||
skip_line(read); // skip header
|
||||
|
||||
// ── Phase 2: copy sequence ────────────────────────────────────────
|
||||
let mut seg_start = tape.write_snapshot();
|
||||
let mut seg_start = write.tell().unwrap_or(0);
|
||||
|
||||
'seq: loop {
|
||||
let Some(c) = tape.read_next() else {
|
||||
// EOF inside a sequence: flush whatever we have.
|
||||
end_segment(tape, &mut seg_start, k);
|
||||
let Some(c) = read.read_next().ok() else {
|
||||
end_segment(write, &mut seg_start, k);
|
||||
return;
|
||||
};
|
||||
|
||||
if is_newline(c) {
|
||||
skip_newlines(tape);
|
||||
skip_newlines(read);
|
||||
break 'seq;
|
||||
}
|
||||
|
||||
let upper = c & !0x20u8; // ASCII uppercase trick
|
||||
let upper = c & !0x20u8;
|
||||
if is_acgt(upper) {
|
||||
tape.write_next(upper);
|
||||
write.write(upper).ok();
|
||||
} else {
|
||||
// Ambiguous base: close the current segment, skip non-ACGT run.
|
||||
end_segment(tape, &mut seg_start, k);
|
||||
skip_until_acgt_or_newline(tape);
|
||||
seg_start = tape.write_snapshot();
|
||||
end_segment(write, &mut seg_start, k);
|
||||
skip_until_acgt_or_newline(read);
|
||||
seg_start = write.tell().unwrap_or(0);
|
||||
}
|
||||
}
|
||||
|
||||
// Sequence line ended: close the last segment.
|
||||
end_segment(tape, &mut seg_start, k);
|
||||
end_segment(write, &mut seg_start, k);
|
||||
|
||||
// ── Phase 3: skip + line ──────────────────────────────────────────
|
||||
skip_line(tape);
|
||||
skip_line(read); // skip + line
|
||||
skip_line(read); // skip quality line
|
||||
|
||||
// ── Phase 4: skip quality line ────────────────────────────────────
|
||||
skip_line(tape);
|
||||
|
||||
if tape.peek().is_none() {
|
||||
if read.read_ahead(1).ok().is_none() {
|
||||
return;
|
||||
}
|
||||
}
|
||||
@@ -143,46 +116,38 @@ fn normalize_fastq(tape: &mut RopeTape, k: usize) {
|
||||
|
||||
// ── shared helpers ────────────────────────────────────────────────────────────
|
||||
|
||||
/// Skip to the end of the current line, consuming the newline run.
|
||||
fn skip_line(tape: &mut RopeTape) {
|
||||
while let Some(c) = tape.read_next() {
|
||||
fn skip_line(read: &ForwardCursor<'_>) {
|
||||
while let Some(c) = read.read_next().ok() {
|
||||
if is_newline(c) {
|
||||
skip_newlines(tape);
|
||||
skip_newlines(read);
|
||||
return;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// Consume a contiguous run of `\n` / `\r` characters.
|
||||
fn skip_newlines(tape: &mut RopeTape) {
|
||||
while matches!(tape.peek(), Some(c) if is_newline(c)) {
|
||||
tape.read_next();
|
||||
fn skip_newlines(read: &ForwardCursor<'_>) {
|
||||
while matches!(read.read_ahead(1).ok(), Some(c) if is_newline(c)) {
|
||||
read.read_next().ok();
|
||||
}
|
||||
}
|
||||
|
||||
/// Consume characters until the next ACGT base or newline (leaving it unread).
|
||||
fn skip_until_acgt_or_newline(tape: &mut RopeTape) {
|
||||
while let Some(c) = tape.peek() {
|
||||
fn skip_until_acgt_or_newline(read: &ForwardCursor<'_>) {
|
||||
while let Some(c) = read.read_ahead(1).ok() {
|
||||
if is_newline(c) || is_acgt(c & !0x20u8) {
|
||||
return;
|
||||
}
|
||||
tape.read_next();
|
||||
read.read_next().ok();
|
||||
}
|
||||
}
|
||||
|
||||
/// Close the current segment.
|
||||
///
|
||||
/// - If `seg_len >= k`: write `0x00` terminator and advance `seg_start`.
|
||||
/// - If `0 < seg_len < k`: erase by retreating the write cursor.
|
||||
/// - If `seg_len == 0`: nothing to do.
|
||||
fn end_segment(tape: &mut RopeTape, seg_start: &mut RopeCursor, k: usize) {
|
||||
let seg_len = tape.written_since(*seg_start);
|
||||
fn end_segment(write: &ForwardCursor<'_>, seg_start: &mut usize, k: usize) {
|
||||
let seg_len = write.tell().unwrap_or(0) - *seg_start;
|
||||
if seg_len >= k {
|
||||
tape.write_next(0x00);
|
||||
write.write(0x00).ok();
|
||||
} else if seg_len > 0 {
|
||||
tape.write_retreat(seg_len);
|
||||
write.rewind(seg_len).ok();
|
||||
}
|
||||
*seg_start = tape.write_snapshot();
|
||||
*seg_start = write.tell().unwrap_or(0);
|
||||
}
|
||||
|
||||
#[inline] fn is_newline(c: u8) -> bool { c == b'\n' || c == b'\r' }
|
||||
@@ -193,14 +158,23 @@ fn end_segment(tape: &mut RopeTape, seg_start: &mut RopeCursor, k: usize) {
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
use bytes::BytesMut;
|
||||
|
||||
fn run(data: &[u8], k: usize) -> Vec<u8> {
|
||||
let chunks = vec![BytesMut::from(data).freeze()];
|
||||
normalize_fastq_chunk(chunks, k)
|
||||
.into_iter()
|
||||
.flat_map(|b| b.to_vec())
|
||||
.collect()
|
||||
fn make_rope(data: &[u8]) -> Rope {
|
||||
let mut r = Rope::new();
|
||||
r.push(data.to_vec());
|
||||
r
|
||||
}
|
||||
|
||||
fn flat(r: Rope) -> Vec<u8> {
|
||||
r.fw_cursor().collect()
|
||||
}
|
||||
|
||||
fn run_fastq(data: &[u8], k: usize) -> Vec<u8> {
|
||||
flat(normalize_fastq_chunk(make_rope(data), k))
|
||||
}
|
||||
|
||||
fn run_fasta(data: &[u8], k: usize) -> Vec<u8> {
|
||||
flat(normalize_fasta_chunk(make_rope(data), k))
|
||||
}
|
||||
|
||||
fn make_fastq(records: &[&[u8]]) -> Vec<u8> {
|
||||
@@ -216,107 +190,6 @@ mod tests {
|
||||
buf
|
||||
}
|
||||
|
||||
// ── basic output format ───────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn single_record_produces_seq_then_null() {
|
||||
let out = run(&make_fastq(&[b"ACGTACGT"]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn two_records_concatenated() {
|
||||
let out = run(&make_fastq(&[b"ACGTACGT", b"TTTTTTTT"]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00TTTTTTTT\x00");
|
||||
}
|
||||
|
||||
// ── uppercase normalisation ───────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn lowercase_input_uppercased() {
|
||||
let out = run(&make_fastq(&[b"acgtacgt"]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn mixed_case_uppercased() {
|
||||
let out = run(&make_fastq(&[b"AcGtAcGt"]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
// ── k filter ─────────────────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn sequence_shorter_than_k_discarded() {
|
||||
let out = run(&make_fastq(&[b"ACG"]), 4);
|
||||
assert_eq!(out, b"");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn sequence_exactly_k_kept() {
|
||||
let out = run(&make_fastq(&[b"ACGT"]), 4);
|
||||
assert_eq!(out, b"ACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn short_record_among_valid_ones_discarded() {
|
||||
let out = run(&make_fastq(&[b"ACGTACGT", b"AC", b"TTTTTTTT"]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00TTTTTTTT\x00");
|
||||
}
|
||||
|
||||
// ── ambiguous bases ───────────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn ambiguous_splits_into_two_segments() {
|
||||
// 'N' in the middle splits "ACGT" + "ACGT"
|
||||
let out = run(&make_fastq(&[b"ACGTNACGT"]), 4);
|
||||
assert_eq!(out, b"ACGT\x00ACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn segment_after_ambiguous_too_short_discarded() {
|
||||
// "ACGTACGT" + 'N' + "AC" (< k=4)
|
||||
let out = run(&make_fastq(&[b"ACGTACGTNAC"]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn consecutive_ambiguous_produce_no_empty_segment() {
|
||||
let out = run(&make_fastq(&[b"ACGTNNNNACGT"]), 4);
|
||||
assert_eq!(out, b"ACGT\x00ACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn ambiguous_at_start_skipped() {
|
||||
let out = run(&make_fastq(&[b"NNACGTACGT"]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn ambiguous_at_end_produces_no_trailing_empty() {
|
||||
let out = run(&make_fastq(&[b"ACGTACGTNN"]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
// ── CRLF line endings ─────────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn crlf_handled() {
|
||||
let data = b"@hdr\r\nACGTACGT\r\n+\r\nIIIIIIII\r\n";
|
||||
let out = run(data, 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
// ── FASTA ─────────────────────────────────────────────────────────────────
|
||||
|
||||
fn run_fasta(data: &[u8], k: usize) -> Vec<u8> {
|
||||
let chunks = vec![BytesMut::from(data).freeze()];
|
||||
normalize_fasta_chunk(chunks, k)
|
||||
.into_iter()
|
||||
.flat_map(|b| b.to_vec())
|
||||
.collect()
|
||||
}
|
||||
|
||||
fn make_fasta(records: &[(&[u8], &[u8])]) -> Vec<u8> {
|
||||
let mut buf = Vec::new();
|
||||
for (id, seq) in records {
|
||||
@@ -329,112 +202,160 @@ mod tests {
|
||||
buf
|
||||
}
|
||||
|
||||
// ── FASTQ basic ──────────────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn single_record_produces_seq_then_null() {
|
||||
assert_eq!(run_fastq(&make_fastq(&[b"ACGTACGT"]), 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn two_records_concatenated() {
|
||||
assert_eq!(
|
||||
run_fastq(&make_fastq(&[b"ACGTACGT", b"TTTTTTTT"]), 4),
|
||||
b"ACGTACGT\x00TTTTTTTT\x00"
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn lowercase_input_uppercased() {
|
||||
assert_eq!(run_fastq(&make_fastq(&[b"acgtacgt"]), 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn mixed_case_uppercased() {
|
||||
assert_eq!(run_fastq(&make_fastq(&[b"AcGtAcGt"]), 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn sequence_shorter_than_k_discarded() {
|
||||
assert_eq!(run_fastq(&make_fastq(&[b"ACG"]), 4), b"");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn sequence_exactly_k_kept() {
|
||||
assert_eq!(run_fastq(&make_fastq(&[b"ACGT"]), 4), b"ACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn short_record_among_valid_ones_discarded() {
|
||||
assert_eq!(
|
||||
run_fastq(&make_fastq(&[b"ACGTACGT", b"AC", b"TTTTTTTT"]), 4),
|
||||
b"ACGTACGT\x00TTTTTTTT\x00"
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn ambiguous_splits_into_two_segments() {
|
||||
assert_eq!(run_fastq(&make_fastq(&[b"ACGTNACGT"]), 4), b"ACGT\x00ACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn segment_after_ambiguous_too_short_discarded() {
|
||||
assert_eq!(run_fastq(&make_fastq(&[b"ACGTACGTNAC"]), 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn consecutive_ambiguous_produce_no_empty_segment() {
|
||||
assert_eq!(run_fastq(&make_fastq(&[b"ACGTNNNNACGT"]), 4), b"ACGT\x00ACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn ambiguous_at_start_skipped() {
|
||||
assert_eq!(run_fastq(&make_fastq(&[b"NNACGTACGT"]), 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn ambiguous_at_end_produces_no_trailing_empty() {
|
||||
assert_eq!(run_fastq(&make_fastq(&[b"ACGTACGTNN"]), 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn crlf_handled() {
|
||||
let data = b"@hdr\r\nACGTACGT\r\n+\r\nIIIIIIII\r\n";
|
||||
assert_eq!(run_fastq(data, 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn multi_slice_rope() {
|
||||
let data = make_fastq(&[b"ACGTACGT", b"TTTTTTTT"]);
|
||||
let mid = data.len() / 2;
|
||||
let mut rope = Rope::new();
|
||||
rope.push(data[..mid].to_vec());
|
||||
rope.push(data[mid..].to_vec());
|
||||
assert_eq!(flat(normalize_fastq_chunk(rope, 4)), b"ACGTACGT\x00TTTTTTTT\x00");
|
||||
}
|
||||
|
||||
// ── FASTA ─────────────────────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn fasta_single_record() {
|
||||
let out = run_fasta(&make_fasta(&[(b"s1", b"ACGTACGT")]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
assert_eq!(run_fasta(&make_fasta(&[(b"s1", b"ACGTACGT")]), 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_two_records() {
|
||||
let out = run_fasta(&make_fasta(&[(b"s1", b"ACGTACGT"), (b"s2", b"TTTTTTTT")]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00TTTTTTTT\x00");
|
||||
assert_eq!(
|
||||
run_fasta(&make_fasta(&[(b"s1", b"ACGTACGT"), (b"s2", b"TTTTTTTT")]), 4),
|
||||
b"ACGTACGT\x00TTTTTTTT\x00"
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_multiline_sequence_concatenated() {
|
||||
let data = b">s1\nACGT\nACGT\nACGT\n";
|
||||
let out = run_fasta(data, 4);
|
||||
assert_eq!(out, b"ACGTACGTACGT\x00");
|
||||
assert_eq!(run_fasta(b">s1\nACGT\nACGT\nACGT\n", 4), b"ACGTACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_lowercase_uppercased() {
|
||||
let out = run_fasta(&make_fasta(&[(b"s1", b"acgtacgt")]), 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
assert_eq!(run_fasta(&make_fasta(&[(b"s1", b"acgtacgt")]), 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_short_record_discarded() {
|
||||
let out = run_fasta(&make_fasta(&[(b"s1", b"ACG")]), 4);
|
||||
assert_eq!(out, b"");
|
||||
assert_eq!(run_fasta(&make_fasta(&[(b"s1", b"ACG")]), 4), b"");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_short_among_valid_discarded() {
|
||||
let out = run_fasta(
|
||||
&make_fasta(&[(b"s1", b"ACGTACGT"), (b"s2", b"AC"), (b"s3", b"TTTTTTTT")]),
|
||||
4,
|
||||
assert_eq!(
|
||||
run_fasta(&make_fasta(&[(b"s1", b"ACGTACGT"), (b"s2", b"AC"), (b"s3", b"TTTTTTTT")]), 4),
|
||||
b"ACGTACGT\x00TTTTTTTT\x00"
|
||||
);
|
||||
assert_eq!(out, b"ACGTACGT\x00TTTTTTTT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_ambiguous_splits_segments() {
|
||||
let data = b">s1\nACGTNACGT\n";
|
||||
let out = run_fasta(data, 4);
|
||||
assert_eq!(out, b"ACGT\x00ACGT\x00");
|
||||
assert_eq!(run_fasta(b">s1\nACGTNACGT\n", 4), b"ACGT\x00ACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_ambiguous_across_line_boundary() {
|
||||
// 'N' at start of second line — still ambiguous
|
||||
let data = b">s1\nACGT\nNACGT\n";
|
||||
let out = run_fasta(data, 4);
|
||||
assert_eq!(out, b"ACGT\x00ACGT\x00");
|
||||
assert_eq!(run_fasta(b">s1\nACGT\nNACGT\n", 4), b"ACGT\x00ACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_ambiguous_short_segment_discarded() {
|
||||
let data = b">s1\nACGTACGTNAC\n";
|
||||
let out = run_fasta(data, 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
assert_eq!(run_fasta(b">s1\nACGTACGTNAC\n", 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_no_trailing_newline() {
|
||||
let data = b">s1\nACGTACGT";
|
||||
let out = run_fasta(data, 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00");
|
||||
assert_eq!(run_fasta(b">s1\nACGTACGT", 4), b"ACGTACGT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_crlf_line_endings() {
|
||||
let data = b">s1\r\nACGT\r\nACGT\r\n>s2\r\nTTTT\r\n";
|
||||
let out = run_fasta(data, 4);
|
||||
assert_eq!(out, b"ACGTACGT\x00TTTT\x00");
|
||||
assert_eq!(run_fasta(b">s1\r\nACGT\r\nACGT\r\n>s2\r\nTTTT\r\n", 4), b"ACGTACGT\x00TTTT\x00");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn fasta_multi_slice_rope() {
|
||||
let data = make_fasta(&[(b"s1", b"ACGTACGT"), (b"s2", b"TTTTTTTT")]);
|
||||
let mid = data.len() / 2;
|
||||
let chunks = vec![
|
||||
BytesMut::from(&data[..mid]).freeze(),
|
||||
BytesMut::from(&data[mid..]).freeze(),
|
||||
];
|
||||
let out: Vec<u8> = normalize_fasta_chunk(chunks, 4)
|
||||
.into_iter()
|
||||
.flat_map(|b| b.to_vec())
|
||||
.collect();
|
||||
assert_eq!(out, b"ACGTACGT\x00TTTTTTTT\x00");
|
||||
}
|
||||
|
||||
// ── multi-record rope (multiple Bytes slices) ─────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn multi_slice_rope() {
|
||||
let data = make_fastq(&[b"ACGTACGT", b"TTTTTTTT"]);
|
||||
// Split into two slices at an arbitrary boundary.
|
||||
let mid = data.len() / 2;
|
||||
let chunks = vec![
|
||||
BytesMut::from(&data[..mid]).freeze(),
|
||||
BytesMut::from(&data[mid..]).freeze(),
|
||||
];
|
||||
let out: Vec<u8> = normalize_fastq_chunk(chunks, 4)
|
||||
.into_iter()
|
||||
.flat_map(|b| b.to_vec())
|
||||
.collect();
|
||||
assert_eq!(out, b"ACGTACGT\x00TTTTTTTT\x00");
|
||||
let mut rope = Rope::new();
|
||||
rope.push(data[..mid].to_vec());
|
||||
rope.push(data[mid..].to_vec());
|
||||
assert_eq!(flat(normalize_fasta_chunk(rope, 4)), b"ACGTACGT\x00TTTTTTTT\x00");
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,427 +0,0 @@
|
||||
//! Two-cursor tape over a rope of `BytesMut` slices.
|
||||
//!
|
||||
//! A [`RopeCursor`] is a position `(slice, byte)` within a rope.
|
||||
//! Read and write cursors are the same type — the only difference is which
|
||||
//! primitive they call: [`RopeCursor::peek`] vs [`RopeCursor::poke`].
|
||||
//!
|
||||
//! The read-only methods (`peek`, `advance`, `retreat`, `distance`) are generic
|
||||
//! over `T: std::ops::Deref<Target = [u8]>` and work on both `Vec<Bytes>` and
|
||||
//! `Vec<BytesMut>`. The write method (`poke`) is specific to `BytesMut`.
|
||||
//!
|
||||
//! [`RopeTape`] owns the rope and two cursors, and exposes the four Turing-machine
|
||||
//! primitives used by the sequence normalisation automata.
|
||||
|
||||
use std::ops::Deref;
|
||||
use bytes::{Bytes, BytesMut};
|
||||
|
||||
// ── RopeCursor ────────────────────────────────────────────────────────────────
|
||||
|
||||
/// A position within a rope (`Vec<BytesMut>`), usable as either a read or write head.
|
||||
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
|
||||
pub struct RopeCursor {
|
||||
slice: usize,
|
||||
byte: usize,
|
||||
}
|
||||
|
||||
impl RopeCursor {
|
||||
/// Position at the very start of the rope.
|
||||
#[inline]
|
||||
pub fn new() -> Self {
|
||||
Self { slice: 0, byte: 0 }
|
||||
}
|
||||
|
||||
/// Read the byte at the current position without moving.
|
||||
/// Returns `None` at end of rope.
|
||||
#[inline]
|
||||
pub fn peek<T: Deref<Target = [u8]>>(&self, rope: &[T]) -> Option<u8> {
|
||||
rope.get(self.slice)?.get(self.byte).copied()
|
||||
}
|
||||
|
||||
/// Write `b` at the current position without moving.
|
||||
#[inline]
|
||||
pub fn poke(&self, rope: &mut [BytesMut], b: u8) {
|
||||
rope[self.slice][self.byte] = b;
|
||||
}
|
||||
|
||||
/// Advance by one byte, crossing slice boundaries transparently.
|
||||
/// Returns `false` when the end of the rope is reached.
|
||||
#[inline]
|
||||
pub fn advance<T: Deref<Target = [u8]>>(&mut self, rope: &[T]) -> bool {
|
||||
if self.slice >= rope.len() {
|
||||
return false;
|
||||
}
|
||||
self.byte += 1;
|
||||
if self.byte >= rope[self.slice].len() {
|
||||
self.slice += 1;
|
||||
self.byte = 0;
|
||||
}
|
||||
self.slice < rope.len()
|
||||
}
|
||||
|
||||
/// Position at the last byte of the rope.
|
||||
///
|
||||
/// Returns `None` if the rope is empty or all slices are empty.
|
||||
pub fn end<T: Deref<Target = [u8]>>(rope: &[T]) -> Option<Self> {
|
||||
for (i, slice) in rope.iter().enumerate().rev() {
|
||||
if !slice.is_empty() {
|
||||
return Some(Self { slice: i, byte: slice.len() - 1 });
|
||||
}
|
||||
}
|
||||
None
|
||||
}
|
||||
|
||||
/// Retreat by one byte and return the byte now under the cursor.
|
||||
///
|
||||
/// Returns `None` if already at the beginning of the rope.
|
||||
/// This is the inverse of the read-advance pattern used by `RopeTape::read_next`.
|
||||
pub fn previous<T: Deref<Target = [u8]>>(&mut self, rope: &[T]) -> Option<u8> {
|
||||
if self.slice == 0 && self.byte == 0 {
|
||||
return None;
|
||||
}
|
||||
self.retreat(1, rope);
|
||||
self.peek(rope)
|
||||
}
|
||||
|
||||
/// Retreat by `n` bytes, crossing slice boundaries transparently.
|
||||
///
|
||||
/// Panics in debug mode if retreating past the beginning of the rope.
|
||||
pub fn retreat<T: Deref<Target = [u8]>>(&mut self, mut n: usize, rope: &[T]) {
|
||||
while n > 0 {
|
||||
if self.byte >= n {
|
||||
self.byte -= n;
|
||||
return;
|
||||
}
|
||||
// Consume all bytes in current slice and cross boundary.
|
||||
n -= self.byte + 1;
|
||||
debug_assert!(self.slice > 0, "retreat past beginning of rope");
|
||||
self.slice -= 1;
|
||||
self.byte = rope[self.slice].len() - 1;
|
||||
}
|
||||
}
|
||||
|
||||
/// Number of bytes from `from` (inclusive) to `to` (exclusive).
|
||||
///
|
||||
/// Requires `from` ≤ `to` in rope order.
|
||||
pub fn distance<T: Deref<Target = [u8]>>(from: &Self, to: &Self, rope: &[T]) -> usize {
|
||||
if from.slice == to.slice {
|
||||
to.byte - from.byte
|
||||
} else {
|
||||
let mut d = rope[from.slice].len() - from.byte;
|
||||
for s in (from.slice + 1)..to.slice {
|
||||
d += rope[s].len();
|
||||
}
|
||||
d += to.byte;
|
||||
d
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
impl Default for RopeCursor {
|
||||
fn default() -> Self { Self::new() }
|
||||
}
|
||||
|
||||
// ── RopeTape ──────────────────────────────────────────────────────────────────
|
||||
|
||||
/// A mutable rope with independent read and write cursors.
|
||||
///
|
||||
/// The write cursor always stays at or behind the read cursor — this invariant
|
||||
/// is guaranteed structurally by the formats (FASTA headers and FASTQ
|
||||
/// non-sequence lines are consumed before any sequence byte reaches the same
|
||||
/// position).
|
||||
pub struct RopeTape {
|
||||
rope: Vec<BytesMut>,
|
||||
read: RopeCursor,
|
||||
write: RopeCursor,
|
||||
}
|
||||
|
||||
impl RopeTape {
|
||||
/// Build a tape from a chunk produced by [`crate::chunk::SeqChunkIter`].
|
||||
///
|
||||
/// Each `Bytes` is converted to `BytesMut` in O(1): succeeds because the
|
||||
/// chunk reader yields uniquely-owned, non-shared buffers.
|
||||
pub fn new(chunks: Vec<Bytes>) -> Self {
|
||||
let rope = chunks
|
||||
.into_iter()
|
||||
.map(|b| b.try_into_mut().expect("Bytes not uniquely owned"))
|
||||
.collect();
|
||||
Self { rope, read: RopeCursor::new(), write: RopeCursor::new() }
|
||||
}
|
||||
|
||||
/// Peek at the byte under the read cursor without advancing.
|
||||
#[inline]
|
||||
pub fn peek(&self) -> Option<u8> {
|
||||
self.read.peek(&self.rope)
|
||||
}
|
||||
|
||||
/// Read the byte under the read cursor and advance it.
|
||||
#[inline]
|
||||
pub fn read_next(&mut self) -> Option<u8> {
|
||||
let b = self.read.peek(&self.rope)?;
|
||||
self.read.advance(&self.rope);
|
||||
Some(b)
|
||||
}
|
||||
|
||||
/// Write `b` at the write cursor and advance it.
|
||||
#[inline]
|
||||
pub fn write_next(&mut self, b: u8) {
|
||||
self.write.poke(&mut self.rope, b);
|
||||
self.write.advance(&self.rope);
|
||||
}
|
||||
|
||||
/// Retreat the read cursor by one byte and return the byte now under it.
|
||||
///
|
||||
/// This is the inverse of `read_next`: move back one position and read.
|
||||
/// Returns `None` if already at the beginning of the tape.
|
||||
#[inline]
|
||||
pub fn previous(&mut self) -> Option<u8> {
|
||||
self.read.previous(&self.rope)
|
||||
}
|
||||
|
||||
/// Retreat the write cursor by `n` bytes (to erase a sequence shorter than k).
|
||||
#[inline]
|
||||
pub fn write_retreat(&mut self, n: usize) {
|
||||
self.write.retreat(n, &self.rope);
|
||||
}
|
||||
|
||||
/// Snapshot the current write position, typically used as `seq_start`.
|
||||
#[inline]
|
||||
pub fn write_snapshot(&self) -> RopeCursor {
|
||||
self.write
|
||||
}
|
||||
|
||||
/// Number of bytes written since `snapshot`.
|
||||
#[inline]
|
||||
pub fn written_since(&self, snapshot: RopeCursor) -> usize {
|
||||
RopeCursor::distance(&snapshot, &self.write, &self.rope)
|
||||
}
|
||||
|
||||
/// Consume the tape and return the written region as a `Vec<Bytes>`.
|
||||
///
|
||||
/// Slices fully overwritten are returned in full; the last partial slice is
|
||||
/// truncated to the write position; all remaining (unread) slices are dropped.
|
||||
pub fn into_output(mut self) -> Vec<Bytes> {
|
||||
let ws = self.write.slice;
|
||||
let wb = self.write.byte;
|
||||
|
||||
if ws < self.rope.len() {
|
||||
self.rope.truncate(ws + 1);
|
||||
if let Some(last) = self.rope.last_mut() {
|
||||
last.truncate(wb);
|
||||
}
|
||||
}
|
||||
|
||||
self.rope
|
||||
.into_iter()
|
||||
.filter(|s| !s.is_empty())
|
||||
.map(BytesMut::freeze)
|
||||
.collect()
|
||||
}
|
||||
}
|
||||
|
||||
// ── tests ─────────────────────────────────────────────────────────────────────
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
|
||||
fn rope(slices: &[&[u8]]) -> Vec<BytesMut> {
|
||||
slices.iter().map(|s| BytesMut::from(*s)).collect()
|
||||
}
|
||||
|
||||
/// Mimics what SeqChunkIter produces: Bytes allocated from BytesMut (uniquely owned).
|
||||
fn owned_bytes(data: &[u8]) -> Bytes {
|
||||
BytesMut::from(data).freeze()
|
||||
}
|
||||
|
||||
// ── RopeCursor::peek ──────────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn peek_first_byte() {
|
||||
let r = rope(&[b"ACGT"]);
|
||||
assert_eq!(RopeCursor::new().peek(&r), Some(b'A'));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn peek_past_end_returns_none() {
|
||||
let r = rope(&[b"A"]);
|
||||
let mut c = RopeCursor::new();
|
||||
c.advance(&r);
|
||||
assert_eq!(c.peek(&r), None);
|
||||
}
|
||||
|
||||
// ── RopeCursor::advance ───────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn advance_crosses_slice_boundary() {
|
||||
let r = rope(&[b"AC", b"GT"]);
|
||||
let mut c = RopeCursor::new();
|
||||
c.advance(&r); // → 'C'
|
||||
c.advance(&r); // → 'G' (slice 1)
|
||||
assert_eq!(c.peek(&r), Some(b'G'));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn advance_at_end_returns_false() {
|
||||
let r = rope(&[b"A"]);
|
||||
let mut c = RopeCursor::new();
|
||||
let still_in = c.advance(&r);
|
||||
assert!(!still_in);
|
||||
assert_eq!(c.peek(&r), None);
|
||||
}
|
||||
|
||||
// ── RopeCursor::poke ──────────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn poke_writes_without_advancing() {
|
||||
let mut r = rope(&[b"XXXX"]);
|
||||
let c = RopeCursor::new();
|
||||
c.poke(&mut r, b'A');
|
||||
assert_eq!(r[0][0], b'A');
|
||||
assert_eq!(r[0][1], b'X'); // unchanged
|
||||
// cursor position unchanged: peek still at 0
|
||||
assert_eq!(c.peek(&r), Some(b'A'));
|
||||
}
|
||||
|
||||
// ── RopeCursor::retreat ───────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn retreat_within_slice() {
|
||||
let r = rope(&[b"ACGT"]);
|
||||
let mut c = RopeCursor::new();
|
||||
c.advance(&r); c.advance(&r); c.advance(&r); // → 'T'
|
||||
c.retreat(2, &r);
|
||||
assert_eq!(c.peek(&r), Some(b'C'));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn retreat_crosses_slice_boundary() {
|
||||
let r = rope(&[b"ACGT", b"TTTT"]);
|
||||
let mut c = RopeCursor::new();
|
||||
for _ in 0..4 { c.advance(&r); } // → first 'T' of slice 1
|
||||
c.retreat(1, &r);
|
||||
assert_eq!(c.peek(&r), Some(b'T')); // last byte of "ACGT"
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn retreat_exactly_to_slice_start() {
|
||||
let r = rope(&[b"ACGT", b"TTTT"]);
|
||||
let mut c = RopeCursor::new();
|
||||
for _ in 0..5 { c.advance(&r); } // → second 'T' of slice 1
|
||||
c.retreat(5, &r); // back to very first byte
|
||||
assert_eq!(c.peek(&r), Some(b'A'));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn retreat_zero_is_noop() {
|
||||
let r = rope(&[b"ACGT"]);
|
||||
let mut c = RopeCursor::new();
|
||||
c.advance(&r); c.advance(&r); // → 'G'
|
||||
c.retreat(0, &r);
|
||||
assert_eq!(c.peek(&r), Some(b'G'));
|
||||
}
|
||||
|
||||
// ── RopeCursor::distance ──────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn distance_same_slice() {
|
||||
let r = rope(&[b"ACGTACGT"]);
|
||||
let mut a = RopeCursor::new();
|
||||
let mut b = RopeCursor::new();
|
||||
b.advance(&r); b.advance(&r); b.advance(&r); // at offset 3
|
||||
a.advance(&r); // at offset 1
|
||||
assert_eq!(RopeCursor::distance(&a, &b, &r), 2);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn distance_across_slices() {
|
||||
let r = rope(&[b"ACGT", b"TTTT"]);
|
||||
let from = RopeCursor::new(); // (0,0)
|
||||
let mut to = RopeCursor::new();
|
||||
for _ in 0..6 { to.advance(&r); } // (1,2)
|
||||
assert_eq!(RopeCursor::distance(&from, &to, &r), 6);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn distance_zero() {
|
||||
let r = rope(&[b"ACGT"]);
|
||||
let c = RopeCursor::new();
|
||||
assert_eq!(RopeCursor::distance(&c, &c, &r), 0);
|
||||
}
|
||||
|
||||
// ── RopeTape ──────────────────────────────────────────────────────────────
|
||||
|
||||
#[test]
|
||||
fn tape_read_skips_write_copies() {
|
||||
// Simulate: 5 bytes header, then 4 bytes sequence.
|
||||
// Read 5 without writing, then copy 4.
|
||||
let chunks = vec![owned_bytes(b">hdr\nACGT")];
|
||||
let mut tape = RopeTape::new(chunks);
|
||||
|
||||
for _ in 0..5 { tape.read_next(); } // skip ">hdr\n"
|
||||
for _ in 0..4 {
|
||||
let b = tape.read_next().unwrap();
|
||||
tape.write_next(b);
|
||||
}
|
||||
|
||||
let out: Vec<u8> = tape.into_output().into_iter().flat_map(|b| b.to_vec()).collect();
|
||||
assert_eq!(out, b"ACGT");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn tape_write_retreat_erases() {
|
||||
let chunks = vec![owned_bytes(b"XXXXXXACGT")];
|
||||
let mut tape = RopeTape::new(chunks);
|
||||
|
||||
// Write 3 bytes, then retreat — simulates a sequence < k being discarded.
|
||||
let snap = tape.write_snapshot();
|
||||
for _ in 0..3 {
|
||||
let b = tape.read_next().unwrap();
|
||||
tape.write_next(b);
|
||||
}
|
||||
assert_eq!(tape.written_since(snap), 3);
|
||||
tape.write_retreat(3); // erase the 3 bytes
|
||||
assert_eq!(tape.written_since(snap), 0);
|
||||
|
||||
// Now write 4 more bytes (the "ACGT").
|
||||
for _ in 0..3 { tape.read_next(); } // skip "XXX"
|
||||
for _ in 0..4 {
|
||||
let b = tape.read_next().unwrap();
|
||||
tape.write_next(b);
|
||||
}
|
||||
|
||||
let out: Vec<u8> = tape.into_output().into_iter().flat_map(|b| b.to_vec()).collect();
|
||||
assert_eq!(out, b"ACGT");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn tape_into_output_exact_boundary() {
|
||||
// Write exactly until the last byte of the first slice, nothing more.
|
||||
let chunks = vec![
|
||||
owned_bytes(b"ACGT"),
|
||||
owned_bytes(b"TTTT"),
|
||||
];
|
||||
let mut tape = RopeTape::new(chunks);
|
||||
for _ in 0..4 {
|
||||
let b = tape.read_next().unwrap();
|
||||
tape.write_next(b);
|
||||
}
|
||||
// Skip the rest
|
||||
let out: Vec<u8> = tape.into_output().into_iter().flat_map(|b| b.to_vec()).collect();
|
||||
assert_eq!(out, b"ACGT");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn tape_written_since_across_slices() {
|
||||
let chunks = vec![
|
||||
owned_bytes(b"XXXX"),
|
||||
owned_bytes(b"YYYY"),
|
||||
];
|
||||
let mut tape = RopeTape::new(chunks);
|
||||
let snap = tape.write_snapshot();
|
||||
for _ in 0..6 {
|
||||
let b = tape.read_next().unwrap();
|
||||
tape.write_next(b);
|
||||
}
|
||||
assert_eq!(tape.written_since(snap), 6);
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user