pirl-unc / mhcgnomes / 33465277695
90%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 41
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 89.273% (+0.08%) from 89.198%
33465277695

push

github

web-flow 
Import the evidenced prefix aliases (#154)

The mhcseqs registry has 339 species/prefix pairs attested in an
external database or the literature, every one with an evidence URL.
118 already resolved here; this imports the rest.

evidenced_prefix_aliases.yaml carries them -- 183 species, 196 rows,
each with its own status and evidence URL, per (species, alias) rather
than per species, since one species can have current, historical and
database spellings from different sources.

Global vs context-only is computed rather than curated: an alias claimed
by one species and unclaimed elsewhere becomes a global alias, anything
claimed by two or more or already owned in species.yaml becomes
context-only. So the decision cannot go stale as species are added.
141 global, 32 context-only.

Three things the first cut got wrong, all caught by existing tests:

It overrode a deliberate holdback -- Otel-DAB, Phtr-UA and Phco-UA
started parsing, and test_birds.py asserts they must not.
underrepresented_taxa_source_registry.yaml marks all three blocked or
registry_only, the holding area docs/curation.md describes for source
signal not stable enough to parse with. An attested spelling does not
outrank that; the loader now reads the registry.

It made "B" a species prefix. Real chicken nomenclature, but as a bare
prefix it shadows the mouse haplotype b and b/d stopped parsing.
Aliases shorter than three characters are not imported globally.

A refactor broke the claimant count by keying on the alias instead of
the species, silently turning contested aliases global.

Recounted against 3.47.1 first, since the issue's numbers were measured
against 3.40. Seven apparent collisions turned out to be the genus
umbrella working correctly and are left alone.

0 of 11,558 corpus names change. Ignoring the holdback fails 8 tests;
allowing short aliases fails 3.

Closes #136

Claude-Session: https://claude.ai/code/session_012s4siLj2Vm33eNMGjNSazH

4186 of 4689 relevant lines covered (89.27%)

0.89 hits per line

Coverage Regressions

Lines Coverage ∆ File
91
89.58
 0.34% species.py
2
96.3
 0.14% data.py
Jobs
ID Job ID Ran Files Coverage
1 33465277695.1 41
89.27
 GitHub Action Run
Source Files on build 33465277695