Module coverage 

Coverage rule (Python — issue #26; TypeScript — issue #31; Rust — issue #37; exemptions — issue #32).

Enforces the README’s Coverage rule: a library’s unit suite must meet the configured floor, with test files excluded from the denominator. This module is the deterministic core — given a parsed coverage report and the thresholds from config, an evaluate function decides pass/fail. Producing the report (shelling out to the language’s coverage tool) is a thin layer on top, kept separate so the guarantee is testable without that toolchain installed.

Python (#26) uses coverage.py: a single total, branch coverage on. Given a CoverageReport and Thresholds, evaluate decides pass/fail, and measure shells out to coverage. TypeScript (#31) is the twin: vitest reports four independent metrics (lines / branches / functions / statements), so it carries its own TypeScriptThresholds, VitestReport, and evaluate_typescript / measure_typescript pair — sharing only the Outcome type. Its subprocess layer shells out to vitest. Rust (#37) is the third twin: cargo llvm-cov reports regions/lines (branch coverage is experimental), so it carries RustThresholds, LlvmCovReport, and evaluate_rust / measure_rust; its subprocess layer shells out to cargo llvm-cov.

Files exempted from coverage in config (issue #32) are omitted from the denominator alongside the test files; the caller resolves them (crate::config::resolve_exempt) and passes their paths to measure / measure_typescript / measure_rust.

Structs

CoverageReport

A coverage.py JSON report (coverage json), pared to what the checks need: the totals (the floor) and the per-file files block (patch coverage, #132). Unmodeled fields (metadata, per-function/class data) are ignored.

FileCoverage

Per-file coverage detail from a coverage.py report (one files entry) — what patch coverage (#132) reads to decide whether a changed line is covered. Unmodeled fields (the summary, per-function/class data) are ignored.

LlvmCovData

One export entry — only its totals are needed (--summary-only omits the per-file and per-function detail).

LlvmCovMetric

One metric’s totals from an llvm-cov export, pared to what the check needs: the denominator size and the covered percent.

LlvmCovReport

A cargo llvm-cov --json export (LLVM’s llvm.coverage.json.export), pared to the totals the check needs. A single run produces one data entry; unmodeled fields (per-file/per-function detail, type, version) are ignored.

LlvmCovTotals

The totals block of an llvm-cov export — the two metrics this rule enforces. llvm-cov also reports functions, instantiations, and (experimental) branches, which the check ignores.

RustThresholds

The two cargo llvm-cov coverage floors, from a [rust].coverage table. Branch coverage is still experimental, so only regions and lines are enforced.

Thresholds

The coverage floor to enforce, from a [<language>].coverage table.

Totals

The totals block of a coverage.py report.

TypeScriptThresholds

The four vitest coverage floors, from a [typescript].coverage table. Each is an independent percent the unit suite must meet — vitest measures all four.

VitestMetric

One metric’s totals from a vitest json-summary block, pared to what the check needs: the covered percent and the denominator size.

VitestReport

A vitest coverage-summary.json report, pared to the total block the check needs. Per-file entries and unmodeled fields are ignored.

VitestTotals

The total block of a vitest json-summary report — the four metrics this rule enforces. vitest also emits branchesTrue, which the check ignores.

Enums

Outcome

The result of checking a report against the thresholds.

Functions

evaluate

Decide whether report meets thresholds.

evaluate_rust

Decide whether report meets both thresholds.

evaluate_typescript

Decide whether report meets every threshold in thresholds.

measure

Run the unit suite under coverage.py in root and check it against thresholds.

measure_patch_report

Run the Python unit suite under coverage.py in root with every source under root measured (coverage run --source=.) and return the parsed report — so an untested source shows in the files block as wholly uncovered rather than vanishing. The per-file detail is what patch coverage (#132) reads; omit is as in measure (an exempt file stays out of the run, so its changed lines are lifted).

measure_patch_rust

Run the Rust unit suite under cargo llvm-cov in root and return the uncovered lines per file — keyed by the absolute path llvm-cov reports, the caller re-keying to root-relative to match the diff. A line is uncovered when llvm-cov records no execution for it (an LCOV DA:<line>,0). What patch coverage (#136, crate::patch_coverage::check_rust) reads; ignore is the coverage-rule exemptions, dropped from the run so an exempt file’s changed lines are lifted. cargo-llvm-cov must be installed.

measure_patch_typescript

Run the TypeScript unit suite under vitest in root and return the uncovered lines per file — keyed by the absolute path vitest reports, the caller re-keying to root-relative to match the diff. A line is uncovered when it carries a statement the suite never executed, or the source of a branch a path of which the suite never took (the v8 analogue of the Python arm’s missing line / missing branch). exclude is the coverage-rule exemptions, dropped from the run so an exempt file’s changed lines are lifted. npx resolves the project-local vitest, so it and @vitest/coverage-v8 must be installed under root.

measure_rust

Run the unit suite under cargo llvm-cov in root and check it against thresholds.

measure_typescript

Run the unit suite under vitest coverage in root and check it against thresholds.

parse_llvm_cov_report

Parse a cargo llvm-cov --json export.

parse_report

Parse a coverage.py JSON report (the output of coverage json).

parse_vitest_report

Parse a vitest json-summary report (coverage-summary.json).