pirl-unc / hitlist / 33456779337
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: 81.13% (-0.02%) from 81.148%
33456779337

push

github

web-flow 
v1.53.3: harden the public peptide-allele API against its own review (#393)

Code review of #389 found twelve defects in the API I made public there.
All twelve reproduced; this fixes them.

Two were live correctness bugs:

The functions returned the `lru_cache` dict itself. Verified: popping a key
from the returned map made `attribute_peptide_to_sample_alleles` return an
empty frozenset for that peptide from then on, and since the per-donor map
feeds the scanner's row emission (#236), a caller pruning it in place would
change which observation rows the next build writes, with no error anywhere.
Both now return `MappingProxyType`. A docstring asking to treat it as
read-only is not a guard.

They were typed `pmid: int` with no coercion, so a string PMID returned an
empty map -- read, correctly per the docstring, as "this study deposited no
attributions". That is the precise misreading the docstrings were written to
prevent, reachable by pulling a PMID out of a dataframe column. Both sibling
accessors already coerced; the logic is now in `_coerce_pmid` and the
accessors delegate to it rather than repeating it.

The Examples block I wrote was fabricated. `SLYNTVATL` is the HIV-1 Gag
epitope, not a peptide in the Sarkizova cohort -- it raises KeyError -- and
showing `['HLA-A*02:01']` presented the value as one presenting allele when
it is the union of matched donors' whole genotypes (5-34 alleles, median 6).
Replaced with a verified example that says what the value actually is. The
`Returns` section also claimed canonical HLA form, which contradicts the
non-human support `_parse_sample_mhc_field` was fixed to have; a test was
baking that in as `assert all(a.startswith("HLA-"))`.

The caveat prose sat directly under `Returns`, so numpydoc parsed each line
as a separate return value: nine phantom entries on one function, six on the
other. Moved to `Notes`, section order corrected.

Structurally: the merged map is now derived from the per-donor one rather
th... (continued)

6131 of 7557 relevant lines covered (81.13%)

0.81 hits per line

Coverage Regressions

Lines Coverage ∆ File
39
93.04
 -0.16% curation.py
Jobs
ID Job ID Ran Files Coverage
1 33456779337.1 30
81.13
 GitHub Action Run
Source Files on build 33456779337