openvax / topiary / 25872231615

90%

master: 90%

Build Type

Pull #167

github

Committed by iskandr

Commit Message

DSL: class_i / class_ii shortcuts + ranking docs catch-up

- `topiary.class_i` and `topiary.class_ii` — pre-built `IsIn` nodes
  referencing the `mhc_class` column.  Compose with any other DSL node:
  `apply_filter(df, class_i & (Affinity.value <= 500))`.
- `topiary.derive_mhc_class(allele_series)` — public utility for
  stamping `mhc_class` onto a DataFrame produced outside the pvacseq
  loader (e.g. a fresh `TopiaryPredictor` result that carries class
  only at the model-level `kind_support`).  Returns a Series of
  `"I"` / `"II"` / `pd.NA`.

`docs/ranking.md` — `Column()` section gets an "Equality / membership
on any dtype" subsection covering:

- `Column("x").eq(value)` / `.ne(value)` / `.isin(values)` and why
  `Column("x") == "y"` doesn't compose (`__eq__` deliberately not
  overridden — nodes stay hashable).
- The `parse('mhc_class == "I"')` string-DSL form and the
  `<` / `<=` / `>` / `>=` rejection on string operands.
- The `class_i` / `class_ii` shortcuts and the `derive_mhc_class`
  utility for non-pvacseq DataFrames.

Also tightens the "Only numeric columns" wording — it's still true
for arithmetic and ordered comparisons, but no longer for equality /
membership.

Tests: 4 new `TestIsIn` cases for the shortcuts and their composition
with numeric clauses; `test_derive_mhc_class_public_utility` covers
the new package-level export across class I / class II / heterodimer /
unknown / None inputs.

143 dsl + pvacseq tests pass; 701 across pvacseq / io / io_lens / dsl /
ranking / wide.

Pull Request Pull Request #167: Add pVACseq report loader (closes #94)

Coverage Stats

267 of 302 new or added lines in 5 files covered. (88.41%)

59 existing lines in 2 files now uncovered.

4552 of 5067 relevant lines covered (89.84%)

2.7 hits per line