pirl-unc / hitlist / 33401174352
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: 81.161% (+0.07%) from 81.096%
33401174352

push

github

web-flow 
v1.51.2: stop all-NA entries driving concat dtypes (#377) (#384)

pandas deprecated inferring result dtypes from empty or all-NA entries.
Both reported call sites feed a parquet write, where an object column
either changes the on-disk schema or is rejected by pyarrow outright.

Two things the issue got slightly wrong, found while fixing it:

There is a third site. #377 named bulk_proteomics.py:427 and
builder.py:1447; build_line_expression has an identical concat line, and
my first patch replaced the wrong one of the two — build_bulk_proteomics
got the fix, with _LINE_EXPRESSION_COLUMNS as its empty-frame schema,
which is the wrong contract for that function. Caught by running the
suite with -W error::FutureWarning rather than by reading.

Dropping empty / all-NA frames is not sufficient. The deprecation is
about all-NA *columns*, and a frame can be perfectly good apart from
one, so excluding it would lose rows. Instead of excluding data,
_concat_non_empty removes the ambiguity: each column's dtype is taken
from the first frame that has real values for it, and any frame
contributing only NA is cast to that dtype before the concat. Output is
identical today and stays identical when pandas flips the default.

tests/test_line_expression.py and tests/test_bulk_proteomics.py now pass
under -W error::FutureWarning (139 tests). Added unit tests covering the
dtype rule, that a partially-NA frame is not dropped, the all-empty
column contract, and the deprecation itself, so a new concat site on
these paths is caught.

Closes #377

Claude-Session: https://claude.ai/code/session_014cBpazLoFvftLSLavT7T9X

6139 of 7564 relevant lines covered (81.16%)

0.81 hits per line

Coverage Regressions

Lines Coverage ∆ File
146
77.04
 0.72% builder.py
Jobs
ID Job ID Ran Files Coverage
1 33401174352.1 30
81.16
 GitHub Action Run
Source Files on build 33401174352