chore: release v0.6.4

[MagicalTux]

🤖 New release

• compcol: 0.6.3 -> 0.6.4 (✓ API compatible changes)

Changelog

0.6.4 - 2026-06-15

Fixed

• (cli) drain decoder's internal block buffer before finish

Other

• changelog for CLI decode-drain fix
• changelog for iterative optimal brotli parse
• (enc) enable optimal parse at q9 (2 passes)
• (enc) cargo fmt encoder_optimal
• (enc) price ring-reuse short codes for explicit DP matches
• (enc) iterative zopfli-style optimal parse at q10/q11
• changelog for round-2 ratio + speed work
• (enc) cross-block matching + two-pass statistics-driven optimal parse
• (enc) single-pass scan-and-shift move-to-front
• tighten encode/decode hot loops
• (enc) replace prefix-doubling rotation sort with SA-IS
• test linked-block cross-boundary references and cross-tool decode
• implement frame linked-block mode (cross-block match window)
• emit continue-dict chunks; feed uncompressed chunks to the dict
• continuous cross-chunk dictionary in the shared chunk encoder

Fixed

• (cli) compcol -d no longer truncates highly-compressible large inputs.
  The streaming decode loop stopped once the compressed input was consumed,
  leaving output a block-buffering decoder (notably bzip2) still held
  internally — finish does not flush it, so compcol -t bzip2 -d cut output
  at 64 KiB. A drain loop now pulls the decoder's buffered output before
  finishing. (Library decoders were already correct; this was CLI-only.)

Added

• (brotli enc) iterative, statistics-driven optimal LZ77 parse
  (zopfli-style forward DP) at quality 9–11. The cost model is rebuilt from
  the previous pass's command/literal/distance histograms each round;
  candidate matches are precomputed once and shared across passes. Improves
  the max-quality ratio on the 2.9 MB corpus from 707558 to 669632 bytes
  (1.473 → 1.394 vs brotli -q 11) and q9 from 709198 to 680156, with
  reference cross-decode verified.

Performance

• Round 2 of encoder ratio + codec speed work (encoder-only for ratio;
  decoders unchanged and every format still decodes byte-for-byte with its
  reference tool). Ratios on a 2.9 MB real-source corpus, our max level vs the
  reference's max (ours/ref, lower is better):

  • xz / lzma2: 1.51 → 1.10 — the LZMA2 chunk encoder now keeps the LZ
    dictionary continuous across chunks (emits 0xC0 continue-dict control
    bytes after the first 0xE0, with a single match-finder spanning the whole
    input) instead of resetting every 64 KiB. Closes nearly all of the gap to
    the .lzma path (1.07). Also fixes the raw-LZMA2 decoder to feed
    uncompressed (stored) chunks into the dictionary.
  • zstd: 1.40 → 1.04 vs zstd -19 at max level (now beats zstd -12)
    — cross-block matching over a retained sliding window (≤8 MiB, within the
    advertised window) plus a two-pass, statistics-driven optimal parse
    (btultra2-style repricing) and repeat-offset-aware DP pricing.
  • lz4: 1.18 → 1.02 (frame, -l 12 beats lz4 -9) — implemented LZ4
    frame block-linked mode so matches reference up to 64 KiB of prior
    blocks' output, not just the current block.

• Standalone-codec encode throughput (output byte-identical):

  • bwt encode ~3× faster — replaced the prefix-doubling rotation sort with
    linear-time SA-IS suffix-array construction.
  • mtf encode ~2.3× faster (single-pass scan-and-shift); rangecoder
    encode/decode ~+15% (tightened hot loops).

  Note: the xz/lzma2 and zstd encoders now buffer the whole input to drive a
  single continuous match-finder (same memory profile the .lzma path already
  had); higher levels trade encode time for ratio, decode speed is unchanged.

---

This PR was generated with release-plz.