refactor: replace single spectrum file with per-partition outputs

Replace the single `kmer_spectrum_raw.json` output with per-partition JSON files in a `spectrums/` directory. Add a `keep_intermediate` parameter to control intermediate file cleanup, and introduce a `write_spectrum` helper for serialization. Update the completion sentinel to `count.done` and align state documentation accordingly.

src/obikpartitionner/src/lib.rs

@@ -2,4 +2,4 @@ mod index_layer;
 mod kmer_sort;
 mod partition;
 
-pub use partition::{KmerPartition, PARTITIONS_SUBDIR};
+pub use partition::{KmerPartition, KmerSpectrum, PARTITIONS_SUBDIR};

src/obikpartitionner/src/partition.rs

@@ -27,6 +27,12 @@ use crate::kmer_sort::{chunk_size_from_ram, sort_unique_kmers};
 
 type Mphf = PtrHash<u64, CubicEps, CachelineEfVec<Vec<CachelineEf>>, Xx64, Vec<u8>>;
 
+pub struct KmerSpectrum {
+    pub f0:     u64,
+    pub f1:     u64,
+    pub counts: BTreeMap<u32, u64>,
+}
+
 const SK_EXT: &str = "skmer.zst";
 pub const PARTITIONS_SUBDIR: &str = "partitions";
 
@@ -238,11 +244,13 @@ impl KmerPartition {
     /// 3. Writes a flat binary count file (`counts1.bin`, one `u32` per slot,
     ///    memory-mapped) accumulating kmer abundances from the superkmer counts.
     /// 4. Persists the MPHF to `mphf1.bin` for downstream use.
-    /// 5. Writes a global `kmer_spectrum_raw.json` at the partition root.
+    ///
+    /// Returns the aggregated `KmerSpectrum`. Per-partition spectrum files are
+    /// deleted after aggregation unless `keep_partial` is true.
     ///
     /// Partitions are processed in parallel via Rayon (one task per thread).
     /// Peak memory per partition is ~80 MB, so n_threads partitions run simultaneously.
-    pub fn count_kmer(&self) -> SKResult<()> {
+    pub fn count_kmer(&self, keep_partial: bool) -> SKResult<KmerSpectrum> {
         let sys = System::new_all();
         let available = match sys.available_memory() {
             0 => sys.total_memory() / 2,
@@ -282,10 +290,10 @@ impl KmerPartition {
             r?;
         }
 
-        // Aggregate per-partition spectra into a global one at the root.
-        let mut global_spectrum: BTreeMap<u32, u64> = BTreeMap::new();
-        let mut global_f0: u64 = 0;
-        let mut global_f1: u64 = 0;
+        // Aggregate per-partition spectra.
+        let mut counts: BTreeMap<u32, u64> = BTreeMap::new();
+        let mut f0: u64 = 0;
+        let mut f1: u64 = 0;
 
         for i in 0..self.n_partitions {
             let path = self.part_dir(i).join("kmer_spectrum_raw.json");
@@ -294,28 +302,21 @@ impl KmerPartition {
             }
             let v: serde_json::Value =
                 serde_json::from_str(&fs::read_to_string(&path)?).map_err(io::Error::other)?;
-            global_f0 += v["f0"].as_u64().unwrap_or(0);
-            global_f1 += v["f1"].as_u64().unwrap_or(0);
+            f0 += v["f0"].as_u64().unwrap_or(0);
+            f1 += v["f1"].as_u64().unwrap_or(0);
             if let Some(obj) = v["spectrum"].as_object() {
                 for (c_str, freq) in obj {
                     if let (Ok(c), Some(f)) = (c_str.parse::<u32>(), freq.as_u64()) {
-                        *global_spectrum.entry(c).or_insert(0) += f;
+                        *counts.entry(c).or_insert(0) += f;
                     }
                 }
             }
+            if !keep_partial {
+                let _ = fs::remove_file(&path);
+            }
         }
 
-        let global_spectrum_map: BTreeMap<String, u64> = global_spectrum
-            .iter()
-            .map(|(&c, &f)| (format!("{c:010}"), f))
-            .collect();
-        serde_json::to_writer_pretty(
-            fs::File::create(self.root_path.join("kmer_spectrum_raw.json"))?,
-            &serde_json::json!({ "f0": global_f0, "f1": global_f1, "spectrum": &global_spectrum_map }),
-        )
-        .map_err(io::Error::other)?;
-
-        Ok(())
+        Ok(KmerSpectrum { f0, f1, counts })
     }
 
     // ── private ───────────────────────────────────────────────────────────────