pirl-unc / hitlist / 34384646076
84%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 31
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: 83.932% (+0.2%) from 83.723%
34384646076

push

github

web-flow 
Flat experimental-condition columns on samples, observations and training rows (#450) (#453)

* Add the flat condition-column spec and task plan

Records the implementation handoff for flat experimental-condition columns:
the column registry, missing/mixed-value rules, paper-extraction contract,
peptide-attribution rules, migration sequence, and acceptance criteria.

Supersedes the nested `line -> sample -> conditions -> modifiers` proposal
on `spec/condition-model`. No package code or data changes yet.

Claude-Session: https://claude.ai/code/session_01PTfM9KnQoVFwVZ6TcgFdYM

* Add the condition column registry, vocabulary and curated annotations (#450)

`hitlist/conditions.py` declares the 23 flat condition columns once, with
their vocabularies, missing-value contract and validation. `curation` splices
the registry into MS_SAMPLE_FIELDS so the loader accepts the keys, and runs
the per-record and study-scoped guards: unique condition_id, valid statuses,
sorted/unique multi-value cells, canonical tokens, no dangling
condition_control_for.

The loader also rejects duplicate YAML mapping keys. PyYAML keeps the last
one silently, so a second `mhc:` on one record discarded a genotype with no
error — the duplicate-PMID guard (#438) is the same failure a level up.

`scripts/migrate_condition_columns.py` annotates all 761 arms from the 167
distinct curated strings and writes them into the YAML as reviewed data.
Every record is `condition_evidence: curated_text`: this normalizes wording
somebody already curated and inherits whatever that curation got right.
Records whose string describes alternatives it does not separate are `mixed`
with empty agent columns — "EZH2i + decitabine + IFNg (various combinations)"
establishes no agent for every contributing condition, and
"unperturbed — some patients post-ipilimumab" is not an untreated arm.

All 761 records survive with every pre-existing value byte-identical.

Claude-Session: https://claude.ai/code/session_01PTfM... (continued)

7146 of 8514 relevant lines covered (83.93%)

0.84 hits per line

Coverage Regressions

Lines Coverage ∆ File
117
90.82
 -0.06% export.py
49
94.62
 0.06% curation.py
Jobs
ID Job ID Ran Files Coverage
1 34384646076.1 31
83.93
 GitHub Action Run
Source Files on build 34384646076