test: raise coverage to ~98% via grammar drift guard; bump to 2.0.1#998
Merged
Conversation
The generated `From<u16>` / `From<Enum> for &'static str` tables in `language_*.rs` are pure lookup tables, so most arms only execute when a fixture contains a node of that exact kind — leaving them at a few percent line coverage and letting a grammar-bump renumbering drift silently. Add a hand-written crate-root test that, per language, round-trips every `0..node_kind_count` id through both tables and asserts the enum name agrees with the live grammar for each visible kind. It lives at the crate root, not under `src/languages/`, because that directory is owned by the `enums/` codegen and the drift gate rejects hand-written files there. This pins the mapping against drift (the same fear behind the `grammar_version` guard) and lifts the generated tables from ~1-9% to ~99-100% line coverage, moving the root crate from ~86.6% to ~98%.
`FromPathError`'s `Display`, `Error::source` chaining, and the `From<MetricsError>` conversion are part of the stable error surface but were only reachable on `from_path` failure paths no test drives, leaving the impls at ~22% line coverage. Pin the message shape (substring) and the `source` chaining contract callers rely on.
Eight languages (Objc, Mozcpp, Csharp, Lua, Tcl, iRules, Ruby, Elixir) had no AST-dump test reaching their `Alterator::alterate` impl, leaving the string-flattening arms uncovered. Add a behavioural test that dumps a minimal snippet per language and asserts the string-like literal survives as a single verbatim leaf — the observable effect of the flatten arm, and non-vacuous since an unflattened literal splits into delimiter/content children.
The class-based metrics wmc, npm, and npa each carry a `Display` impl and per-space `class_*` / `interface_*` accessors that no test reached, unlike the sibling metrics (nom, nargs, halstead) which all pin their `Display`. Add a `Display`-plus-accessor test per metric driven by a real Java fixture (a class and an interface) with empirically verified values. Per lesson #311, each test asserts both halves: the structural side (the named subspace opens with the expected `SpaceKind`) and the value side. The singular per-space accessors are asserted on the owning class / interface subspace, where they are nonzero, rather than on the always-zero file-unit root — an accessor that always returned 0 or read the wrong field is caught (verified by reverting `class_wmc` to a constant 0). Adds a shared `child_space` test helper in `tools.rs` next to `assert_child_space_kind`; its +6 SLOC nudges tools.rs into the soft headroom band, so the baseline is refreshed in the same commit.
The python flag measures only the pure-Python surface; the PyO3 Rust layer is exercised by pytest but never instrumented by cargo-llvm-cov (the crate has no cargo tests), so its standalone Rust-line number understates how well it is tested. Document this so a future reader does not paper over the gap with redundant cargo tests.
The eight per-language flattening tests were near-identical `assert_flattened` calls wrapped in `#[test]` boilerplate. Collapse them into a `flatten_cases!` macro (one line per case), mirroring the `roundtrip_tests!` macro in `src/language_enum_roundtrip.rs` and hoisting the iRules rationale above its case per the macro-comments rule. Names and behavior are unchanged (`#[track_caller]` on `assert_flattened` keeps panic locations at the case); −35 lines.
Lockstep version bump 2.0.0 → 2.0.1 across every owned crate and the references that must track it, per RELEASING.md's lockstep policy: - [workspace.package] version + enums/ and the five bca-tree-sitter-* leaf [package] versions, and every internal =2.0.0 path-dep pin (root + enums grammar pins, the cli/web big-code-analysis pins). - The vendored grammar_version literals in src/langs.rs. - User-facing install/version references in README, STABILITY, and the book (quick-start, cargo-features, stability, the CI recipe pins); the non-gated CI cache key and historical/semver-spec references are left as-is. - Regenerated derived artifacts: Cargo.lock, the man pages (cargo run -p xtask), and the SARIF driver-version snapshots. check-versions.py passes (every owned crate at 2.0.1). This is the number bump only — no tag, no publish, no CHANGELOG release section.
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #998 +/- ##
===========================================
+ Coverage 85.91% 98.03% +12.11%
===========================================
Files 266 267 +1
Lines 65831 65971 +140
Branches 65401 65541 +140
===========================================
+ Hits 56560 64673 +8113
+ Misses 8831 857 -7974
- Partials 440 441 +1
Flags with carried forward coverage won't be shown. Click here to find out more.
🚀 New features to boost your workflow:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Raises code coverage from 86.6% → 98.6% (root-crate line coverage) and bumps the version to 2.0.1. No production bugs were found — the codebase is well-maintained; the gaps were untested-but-correct code, plus generated lookup tables that dominated the "missing" lines.
The key insight
The reported 86% was largely a generated-code artifact: ~88% of the missed root-crate lines were in the auto-generated
From<u16>/From<Lang> for &strenum tables insrc/languages/language_*.rs. Those arms only execute when a fixture contains a node of that exact kind, so most never ran. Hand-written code was already ~98% covered.What's in here
Coverage
test(languages): a grammar drift-guard test — per language, round-trips every node-kind idu16 → Lang → &strand asserts the enum name matches the live grammar for visible kinds. Lifts the generated tables from 1–9% → 99–100% and doubles as regression protection against grammar-bump renumbering.test(error): coversFromPathErrorDisplay/source/From(~22% → covered).test(alterator): per-language string-flattening for 8 languages (Objc, Mozcpp, Csharp, Lua, Tcl, iRules, Ruby, Elixir) that had no AST-dump test.test(metrics):Display+ per-space accessor tests for wmc/npm/npa, asserting nonzero values on the owning class/interface subspace and pinningSpaceKind(per lesson test(npm,npa): tighten interface tests against is_func_space revert #311). A first==0-on-root version was vacuous — caught during review and strengthened; verified non-vacuous by test-via-revert.refactor(test): macro-ized the 8 alterator cases to match theroundtrip_tests!pattern.Docs
docs(codecov): documents thebig-code-analysis-py49% measurement artifact — its PyO3 Rust layer is exercised by pytest but never instrumented bycargo-llvm-cov nextest, so don't "fix" it with redundant cargo tests.Release
chore(release): lockstep version bump 2.0.0 → 2.0.1 (manifests,langs.rsgrammar literals, doc pins, all 7Cargo.locks, man pages, SARIF snapshots).check-versions.pypasses. Not tagged/published.Validation
make pre-commitpasses on every commit (fmt, clippy ×2, full workspace tests, doc, self-scan + headroom, enums-codegen-drift, man-drift, snapshot-anchors, actionlint, Python stack). Reviewed across/code-review,/audit-tests,/review, and/rust-optimize.🤖 Generated with Claude Code