openvax / topiary / 25876363815

90%

master: 90%

push

github

DSL: harden walker, fix IsIn empty-df fragility, fill api.md gaps

Three fixes from the review nudge after the prior "clean stopping
point" claim — all real, not nits.

1. `IsIn.eval` empty-DataFrame branch built a Series from a
   zero-length boolean array indexed by `ctx.group_index`.  Crashes
   if the context happened to carry a non-empty index alongside an
   empty df ("Length of values (0) does not match length of
   index (N)").  Unreachable on current callers but fragile, and
   inconsistent with `Column.eval`'s empty path.  Switch to
   `ctx.empty_series().astype("boolean")` for index alignment and
   pattern parity.

2. `_collect_column_names` walker grew an `isinstance + elif` chain
   when `IsIn` landed.  The docstring promised "implement
   child_nodes() and you're done" — duck-type the leaf check via
   `getattr(n, "col_name", None)` so future column-bearing leaves
   work without touching the walker.  Drops the now-unused `Column`
   and `IsIn` imports from `apply.py`.

3. `docs/api.md` was missing four real public exports: `IsIn`,
   `class_i`, `class_ii`, `derive_mhc_class`.  Add `IsIn` to the
   DSL leaves table; new "Categorical equality and membership"
   subsection under Column covering the methods, the string-DSL
   `==` / `!=` form, and NaN semantics; pre-built shortcuts table
   for `class_i` / `class_ii`; row for `derive_mhc_class` in the
   input-functions table.

Tests: new `test_empty_dataframe` exercising IsIn against an
empty input (regression guard); 145 dsl + pvacseq tests pass; 705
across pvacseq / io / io_lens / dsl / ranking / wide / dataframe.

3 of 4 new or added lines in 2 files covered. (75.0%)

4551 of 5064 relevant lines covered (89.87%)

2.7 hits per line