openvax / topiary / 24421423053

90%

push

github

CachedPredictor vaxrank-review polish (#131)

Follow-up to #130 addressing concrete concerns from the vaxrank-consumer review (comment 4246617648):

- Partial-allele cache preservation: _fallback_resolve filters fallback output to keys not already in the index before merging, so pre-loaded rows aren't overwritten when the fallback covers overlapping alleles.
- Docstring notes on silent peptide-length lock-in and non-thread-safety.
- save() raises on empty never-queried caches so users don't write schema-only files that can't be round-tripped.

3 new regression tests (41 total for CachedPredictor, up from 38), full suite 1093 passing.

Concern #3 (_index memory footprint on whole-proteome caches) is deferred to a benchmark-driven follow-up.

11 of 12 new or added lines in 1 file covered. (91.67%)

3222 of 3649 relevant lines covered (88.3%)

2.65 hits per line