pirl-unc / tcrsift / 25821338153
84%

DEFAULT BRANCH: main
Ran
Jobs 4
Files 25
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: 71.373% (+0.1%) from 71.229%
25821338153

push

github

web-flow 
Fix #45 bug 1: load_vdjdb resolves real release-zip directory correctly (#53)

* Fix #45 bug 1: load_vdjdb against the real release-zip directory

After `tcrsift data download --db vdjdb`, the cache directory expands
to ~18 files. The old loader did `path.glob("vdjdb*.txt")[0]`, which
picked `vdjdb.meta.txt` alphabetically — a metadata-about-fields file
with no CDR3 data — and downstream `match_clonotypes` then died with
KeyError: cdr3_alpha.

Two complementary changes:

1. `_pick_vdjdb_file` resolves the directory to a canonical file by
   priority: `vdjdb_full.txt` (paired αβ, the one we actually want)
   first, then `vdjdb.txt` (slim/long β-only), then a filtered glob
   that skips meta/slim/broken/motif/cluster/summary/scored sidecars.
   The fallback path also raises with a clear hint when the directory
   contains only sidecars, instead of silently picking one.

2. Column mapping fixes:
   - Add `cdr3.beta` → `cdr3_beta`. The full export uses paired-chain
     rows with `cdr3.alpha` + `cdr3.beta`; the slim file uses `cdr3`.
     The old mapping only handled `cdr3`, so the full-format file
     produced a dataframe with no `cdr3_beta` column at all.
   - Drop the donor `species` column before renaming
     `antigen.species` → `species`. The full export carries both
     fields (donor T-cell origin vs. epitope source organism); we
     use the latter, and dropping the former first guards against
     any pandas same-name collision.

Verified locally against ~/.cache/tcrsift/vdjdb/ (139,744 rows;
102K with α, 124K with β, 83K viral by antigen species).

Bumps to 0.8.12.

Closes #45 (bug 1; IEDB half was already closed by #51).

* Fix schema conflation between paired/long VDJdb formats

Code review (and a closer look at the real on-disk shapes) caught
that the original PR conflated VDJdb's two formats:

  - vdjdb_full.txt: 1 row per clonotype, paired columns
    `cdr3.alpha` / `cdr3.beta`.
  - vdjdb.txt:      1 row per **chain**, with `cdr3` holding... (continued)

4767 of 6679 relevant lines covered (71.37%)

2.85 hits per line

Coverage Regressions

Lines Coverage ∆ File
9
96.86
 0.76% annotate.py
Jobs
ID Job ID Ran Files Coverage
1 python-3.10 - 25821338153.1 25
71.36
 GitHub Action Run
2 python-3.9 - 25821338153.2 25
71.34
 GitHub Action Run
3 python-3.11 - 25821338153.3 25
71.36
 GitHub Action Run
4 python-3.12 - 25821338153.4 25
71.36
 GitHub Action Run
Source Files on build 25821338153