pirl-unc / hitlist / 26995946149
80%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 29
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: 78.728% (+1.7%) from 76.995%
26995946149

push

github

web-flow 
v1.30.64: CI build smoke test from packaged data (#176/#63) (#275)

Answers 'is there a smaller version we can ship for CI?' — turns out we
don't need to ship the 15 GB IEDB/CEDAR sources at all. build_observations
runs the full build *orchestration* (scanner -> classify_ms_row -> dedup ->
supplement merge -> categorical compression -> atomic parquet write, plus
the packaged bulk-proteomics + line-expression indexes) off PACKAGED data
plus a tiny synthetic IEDB CSV, with build_mappings=False skipping the only
proteome/pyensembl-dependent step. ~11s, no network.

New tests/test_build_smoke.py:
- Real end-to-end build into a tmp data dir; asserts a well-formed
  observations.parquet (>0 rows, core schema).
- Validates #261 (synthetic 'K562-Myeloid cell' -> line 'K562' +
  cell_type 'Myeloid cell') survives a true build, not just the scanner
  unit test.
- Validates #263 categorical dtypes on the built parquet.

Marked integration so it runs on the single build-heavy 3.11 job
(in the -n 2 pass), isolated from the downloaded corpus via
_override_data_dir. Gives the data build real CI coverage (#176/#63)
that previously required a workstation + the full source data.

NOTE: this covers the build CODE PATHS, not scale profiling — finding
what's slow on 4.4M rows still needs real-scale data. peptide_mappings
(proteome-dependent) build remains uncovered; a small FASTA fixture could
extend coverage there as a follow-up.

4952 of 6290 relevant lines covered (78.73%)

0.79 hits per line

Jobs
ID Job ID Ran Files Coverage
1 26995946149.1 29
78.73
 GitHub Action Run
Source Files on build 26995946149