pirl-unc / hitlist / 33389320176
81%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 30
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: 81.096% (+0.06%) from 81.034%
33389320176

push

github

web-flow 
v1.51.0: adopt mhcgnomes' species API; surface both species axes (#383)

* v1.51.0: adopt mhcgnomes' species API; surface both species axes (#4 from review)

mhcgnomes 3.39.0 shipped the two helpers this work had been hand-rolling
(mhcgnomes #116, #118), so the local versions are gone.

Species.compatible_with replaces curation.species_compatible. The
semantics are identical — same species, or one a direct ancestor of the
other — and the library version is a strict superset: it also accepts a
Species, an MHC prefix, a common name or a latin name, where ours took
only name strings and only worked correctly after yesterday's fix to its
string-comparison bug.

species_source replaces the trap-pinning tests. Those asserted the
*wrong* answers (classify_mhc_species("BLB2*02") == "Coturnix japonica")
purely so an upstream fix would fail loudly, which it duly did. The
invariant is now stated directly: no curated mhc token may resolve with
species_source == "inferred". "default" stays allowed — a bare
"A*03:01" meaning the human allele is conventional, and 69 curated
tokens rely on it.

That guard found 4 more tokens: chicken BF2*2101 / *1301 / *1501 /
*1901 across PMIDs 18612635 and 36695776 were resolving by cross-species
inference, the same shape as the BLB2*02 quail bug. Pinned with the
Gaga- prefix; inferred-species tokens now 0.

Review finding #4: the source-vs-MHC species invariant lived only in the
test suite, so a consumer reading the exported table got no flag. The
samples table now carries mhc_species (derived from the sample's own
alleles) and species_axes_agree, which is what docs/source-classification.md
has described as a tracked axis all along and only observations exposed.
644 agree, 42 undeterminable, 2 disagree — both engineered chimeras
(#46), correct as curated and now visible rather than implicit.

Claude-Session: https://claude.ai/code/session_014cBpazLoFvftLSLavT7T9X

* v1.51.1: address review — one species derivation, not two (#4 fol... (continued)

6113 of 7538 relevant lines covered (81.1%)

0.81 hits per line

Coverage Regressions

Lines Coverage ∆ File
138
86.25
 0.02% export.py
33
93.06
 0.42% curation.py
Jobs
ID Job ID Ran Files Coverage
1 33389320176.1 30
81.1
 GitHub Action Run
Source Files on build 33389320176