Testing and Logging

[emiliorighi]

The job seems to run successfully, but are we sure that each assembly has at least 1 row with aliases? How are the assemblies without any found sequence/alias list handled?

• Add not found logging/tracking (tsv)
• Create test/clean up assemblies with less then 2 rows
• Add smoke test on commit action (we want to be sure that what we commit on main is still what we expect)

[Max25R]

Status update on the three sub-items:

• Not-found logging: done. collect_aliases.py now writes a structured failures.tsv alongside the output, with columns ACCESSION, ASSEMBLY_NAME, STAGE, REASON, DETAIL. End-of-run summary buckets failures by reason. Two new failure categories caught that were previously silent: empty_report (zero rows after parse) and no_valid_lengths. The workflow now uploads failures.tsv as a Release asset alongside the TSV and DB; Sunday's run should be the first with all three.

• Assemblies with <2 rows:*investigated. 184 single-row assemblies in the production DB; all legitimate single-molecule submissions (Y chromosomes, single-chromosome deposits, bacterial / endosymbiont genomes, mitochondria, phages). Findings written up in docs/DESIGN.md. No cleanup needed.

• CI smoke test: still open. Will tackle separately.

[emiliorighi]

Good job.

1. The smoke tests should run through a github action (either on push of targeted folders or on version release or both)
2. Failed fetching should be further categorized and skip only 404/410 not found paths (making sure that the fpt parent directory (the one containing only the assembly accession has been correctly scanned https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/900/963/305
3. Failed assemblies with errors different from not found should be retried