[gwct]

stats: per-species source scaffolds

• New <prefix>.species_scaffolds.tsv: one row per species/scaffold present in the MAF, with its length (srcSize), blocks, aligned bases and aligned fraction.
• species.tsv gains scaffolds_in_maf and scaffold_bases_in_maf, appended as the last columns; overall.tsv gains scaffold_size_parse_errors.
• These describe scaffolds present in the MAF, not the whole assembly. Unaligned scaffolds never appear.
• Dashboard: a collapsible "largest scaffolds" list per species (--dashboard-top-scaffolds, default 10).
• A scaffold reporting two different srcSize values stops the run with an error.
• Fixed the "Processed N blocks in M tasks" log line undercounting tasks.

Progress reporting (index, stats, gc, fetch)

• Progress lines with rate and estimated time left, every 60 s by default.
• New --progress-interval SECONDS on each command: 60 is the default and the minimum, and 0 turns progress off.
• fetch workers report through a shared-memory counter: no locks and no messages between processes.
• Parameter logs now align their colons.

fetch

• BED pre-check before any work: unusable lines, -b id regions without names, and regions that would overwrite each other's output files. Problems are reported in one error, then it exits 1.
• maf_fetch_summary.tsv is written in batch (genomic) order, so it is identical across runs and -p values.
• Scaffold mode logs each scaffold as it finishes.

Behaviour changes

• fetch now stops on BEDs it used to half-process:
  • malformed lines, which previously crashed it;
  • inverted or empty intervals, which were previously reported only as "no overlap";
  • -b id with missing or duplicate names;
  • duplicate regions under -b coords.
• index now prints a start line, progress lines and a final summary, to the screen only.
• Test fixture fix: chimp.chr4 in tests/example.maf had inconsistent srcSize values (1000 and 998) since the first commit. It is now 1000, with the compressed copies and indexes regenerated.

Testing

• 257 tests, up from 148.
• Checked against the 42.7 GB hamster MAF:
  • the scaffold table matches an independent awk pass exactly;
  • index, gc and fetch show no measurable slowdown in timed comparisons with the previous code, and their outputs are byte-identical.

🤖 Generated with Claude Code