openvax / pyensembl / 35021350287

89%

master: 83%

push

github

Fix name-mangled other.__class typo crashing Locus/Exon comparisons (#391)

* Fix name-mangled other.__class typo crashing Locus/Exon comparisons

Locus.__eq__, __lt__, and __gt__ (locus.py) and Exon.__eq__ (exon.py)
build their TypeError message with other.__class.__name__ instead of
other.__class__.__name__. Written inside a class body, the identifier
other.__class (two leading underscores, zero trailing) gets name-mangled
to other._Locus__class / other._Exon__class, which essentially no
object has. So comparing a Locus or Exon with anything that isn't one
raised AttributeError before the intended, more informative TypeError
was ever constructed:

    >>> Locus("1", 10, 20, "+") == None
    AttributeError: 'NoneType' object has no attribute '_Locus__class'

Fixed the typo in all four spots (adding the missing trailing __).
This is the narrow, uncontroversial half of #367: the reporter also
raised a separate design question about whether __eq__ should return
NotImplemented instead of raising for foreign types (to support things
like `locus in [None, ...]` and mixed-type set/dict lookups), but
explicitly asked for maintainer input before changing that behavior
since other classes in the hierarchy (Gene, Transcript, Protein)
currently disagree on this. This PR only fixes the crash; it leaves
that broader semantics question untouched.

Added regression tests in tests/test_locus.py covering Locus == None,
Locus < None, Locus > None, Locus == 42, and Exon == None / Exon ==
"not an exon" -- all previously raised AttributeError, now correctly
raise TypeError. Verified via git stash that both new tests fail with
the exact reported AttributeError against the unpatched code, and pass
after the fix. Ran tests/test_locus.py (10 tests) and tests/test_exon_id.py
in full: no regressions (the 3 pre-existing test_exon_id.py failures
are unrelated -- they require `pyensembl install` to fetch Ensembl
release data, and fail identically on unpatched main).

Fixes #367 (t... (continued)

1 of 1 new or added line in 1 file covered. (100.0%)

1720 of 1970 relevant lines covered (87.31%)

5.24 hits per line