pirl-unc / hitlist / 34297077114
84%

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: 83.644% (+0.2%) from 83.484%
34297077114

push

github

web-flow 
Audit unattributed curated arms, guard study-level keys, fix #373 review findings (#445)

* Audit unattributed curated arms, guard study-level keys, fix #373 review findings

Three strands, all about curation that describes something no consumer sees.

qc.sample_attribution_audit() reports the 232 profiled ms_samples arms that
reach zero observation rows (#442), bucketed by whether the study attributes
anything at all: 137 label_mismatch_candidate, where the join demonstrably
works in that study so the label is the suspect, and 95 study_unattributed,
a different failure triaged per study rather than per label. Not wired into
run_all or curation_plan -- sample_label is synthesized by the export join,
so this is the only check needing the full enriched table, and either rollup
would make every `hitlist qc` pay that build.

PMID_ENTRY_FIELDS is the study-level twin of MS_SAMPLE_FIELDS, declaring all
37 permitted top-level keys against their readers, with the loader rejecting
the rest. Its absence is why four study-level keys were curated and read by
nothing -- most seriously exclude_from_ms, documented as excluding a study
from the MS index, set on 11 studies curated as non-MS, honored by no code,
leaving 40,355 rows in the corpus (#444). Those keys are declared as UNREAD
citing the issue rather than described as if they worked; removing the rows
is a deliberate corpus change and ships separately. samples/tissues renamed
to n_samples/n_tissues per the count-suffix rule.

Code review found 8 real defects in the #373 provenance work already
published as 1.58.8:

- effective_override skipped the rules level. PMID 27846572's primary
  fibroblasts exported cell_line/study while the study's rule sends every
  Direct Ex Vivo fibroblast row to healthy -- exported provenance
  contradicting the classification the build applied, on 131,252 rows across
  the 14 studies carrying both rules and ms_samples. Rules match per row, so
  a sample-level value cannot resolve... (continued)

6914 of 8266 relevant lines covered (83.64%)

0.84 hits per line

Coverage Regressions

Lines Coverage ∆ File
127
56.35
 0.99% cli.py
110
90.75
 0.07% export.py
50
94.45
 0.02% curation.py
19
94.74
 0.6% qc.py
Jobs
ID Job ID Ran Files Coverage
1 34297077114.1 30
83.64
 GitHub Action Run
Source Files on build 34297077114