feat(scanner): Skip testdata directories by default - #179
Conversation
Fixture repositories under testdata are not the project's source. Scanning them inflates file counts, and once a fixture is written in a language the file-level edge model does not cover, it changes the project's reported coverage: a three-file Swift fixture is enough to make an entire Go project report partial. Go's own toolchain ignores the directory for the same reason, so this matches the convention users already expect, alongside vendor and node_modules. A fixture is still scannable when it is itself the scan root, which is how fixture tests use them; only testdata encountered during a walk is skipped. Relates to #172 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PEUvjGsJemDFSV8nbxvxBo
Two failures on this PR, neither from the testdata exclusion — scanner passed on both legs. codemap/mcp: the cargo-metadata deadline failed the whole graph build instead of falling back. Ported from @reneleonhardt's #171. codemap root: TestRunWatchModeRunDaemonAndWatchStart hit "reading daemon readiness: unexpected end of JSON input" — the same window #177 fixes, in a different test than the one that led me to open it. waitWatchReadiness gave up on the first unparseable read of a file still being written. Both no-op once main carries them. Relates to #172 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PEUvjGsJemDFSV8nbxvxBo
|
The review findings on this PR's description are correct and both errors are mine. Not pushing while the branch is being patched — recording the acknowledgement. The before/after in my description cannot be reproduced on That My "scan a fixture directly by passing it as the root" claim was under-tested. I verified The underlying change stands: fixture repositories aren't project source, and Go's toolchain ignores the directory for the same reason. It's the description and the test's coverage that were wrong, not the exclusion. I'll re-verify the restated body and the root-directory fix against a clean Generated by Claude Code |
…olves `--min-importers 1` hid every hazard on this repository, and the reason was not the threshold. Go resolves imports at package level, and BuildFileGraph deliberately drops an import that resolves to more than one file rather than fanning it into an edge per file, so a file inside a multi-file Go package has zero file-level importers by construction. `scanner/filegraph.go` scored 0 and read as harmless. So did every same-package collision, including all six pairs issue #134 verified by hand. A shared file whose language resolves at package granularity is now weighted as the files outside its package that import the package, plus the package's other files, and the count is labelled `package importers` so it is not read as a file-level number. Languages whose imports name files keep the file-level count and the plain label. FileGraph.Packages is populated for Go and nothing else, which is exactly the set this is correct for. The cross-package term cannot come from the graph — the edges are the ones that were dropped — so collide now keeps the scan outcome it was already paying for and counts the raw import strings. ScanForDeps plus BuildFileGraphFromOutcome is the same single scan BuildFileGraph was doing. Real effect on this repository, where the default previously printed nothing: 6 PRs scanner/filegraph.go 73 package importers <- #171, #174, #175, ... 3 PRs config/config.go 38 package importers <- #171, #181, #182 3 PRs main.go 6 package importers <- #175, #179, #180 --min-importers now defaults to 0. A file two open PRs both change is a hazard whatever its weight, and a default that hides hazards answers "no collisions" on a repository full of them. The flag stays for narrowing a long list. Three tests added: a Go fixture where two same-package files collide and carry a non-zero package weight (with the self-import and third-party cases held out of the count), a file-resolved language keeping file scope and its plain label, and #134's six measured pairs proved unchanged by the weighting — reordering them is allowed, adding or dropping one is not. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TcyheQmM3HCvxF5wRL3s5t
IgnoredDirs is matched on a directory's base name, and the walk root is matched along with everything under it. So `codemap testdata/`, or `codemap .` from inside vendor/ or node_modules/, returned Files: 0 — no error, no message, just an empty result for the directory the user explicitly named. Adding testdata to the list widened the set of paths that hit this. Exempt the walk root from the fast-path check in ScanFiles, and from the same check in ReadExternalDeps so a manifest at that root is still read. Nested directories with those names are still skipped, at any depth. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TcyheQmM3HCvxF5wRL3s5t
#180) * feat: codemap collide, rank open PRs by shared-file merge-order hazard CI structurally cannot see cross-PR collisions: every PR is built against main and never against its siblings. Issue #134 measured that blind spot by merging six worktree pairs by hand, and #117/#118 shipped a miscompile through it. `codemap collide` reads open PRs through `gh pr list --json files`, intersects their changed paths, and weights each shared file by the importer count from the graph on the current checkout. The intersection is glue; the weighting is the part that needs codemap, because only the graph knows that a collision on a 23-importer hub is a different severity from one on a test fixture. Honesty rules, per the design principle in #134 (a composite inherits the honesty of its primitives and states it with more authority): - Importer counts are stated as facts only while graph coverage is complete. Degraded coverage prints "unknown importers", drops the verdict to TRUST LOW, and says ranking fell back to shared-file count. - A "no collisions" answer from a degraded graph is TRUST LOW too: a negative finding from a partial graph is as unreliable as a positive one. - Coverage attribution is whole-graph, not per-language. Narrowing "partial" to a subset of languages by matching free-text notes would hand back confidence the graph never claimed. #174's ResolvesFileLevelImports is the supported seam for per-language attribution; collideImportersKnown is the single function it belongs in. - --min-importers never hides a file whose count is unknown, and never drops a hazard silently: the hidden count and the way to see them are printed. - A file the graph carries no edges for at all (a YAML rule, a fixture) reports "not in graph" rather than a zero that reads as "nothing imports it". Tests cover the pair/shared-file computation against issue #134's measured 4-PR matrix (6 of 6 pairs, including the 2-vs-3 distinction), the ranking order, degraded coverage yielding TRUST LOW with unknown counts, and a golden human output. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TcyheQmM3HCvxF5wRL3s5t * fix(scanner): Keep the Rust fallback when cargo metadata times out Ported verbatim from @reneleonhardt's open PR #171, which fixes this already. Carrying it here so this PR can go green rather than waiting on that one to merge; it becomes a no-op once main has it. buildRustWorkspaceIndex shadowed its caller's ctx with the cargo-metadata deadline, so once that deadline passed ctx.Err() returned DeadlineExceeded and the whole graph build failed with a bare "context deadline exceeded" instead of falling back to the manually derived Rust workspace. On a cold or loaded runner three seconds is not always enough for cargo metadata, and mcp/TestRustGraphContextHandlersDisclosePartialCoverage has now failed this way on three separate pull requests. Separating the metadata context from the caller's lets an expired deadline break out of the loop and keep the fallback index, which is what the test asserts and what a consumer needs: partial coverage disclosed, not a failed graph. Relates to #147, #172 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PEUvjGsJemDFSV8nbxvxBo (cherry picked from commit 24af8fb) * fix(watch): Treat an unparseable readiness file as not-ready-yet waitWatchReadiness returned on the first successful read, so a readiness file caught mid-write — existing but empty or partial — failed json.Unmarshal and aborted the wait immediately, reporting a startup failure for a daemon that had not finished writing. Only os.ErrNotExist counted as "not ready". Measured against the real function: an empty file returns "reading daemon readiness: unexpected end of JSON input" after 0s, without waiting out any part of the 30s timeout. Keep polling on a parse failure until the deadline, and surface the last parse error when the deadline passes, so a file that never becomes valid still says why rather than only that it timed out. publishWatchReadiness already renames its payload into place atomically, so codemap's own daemon does not open this window; the reader was brittle to any writer that is not atomic. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PEUvjGsJemDFSV8nbxvxBo (cherry picked from commit 35d2af3) * fix(collide): Weight Go collisions at the granularity Go actually resolves `--min-importers 1` hid every hazard on this repository, and the reason was not the threshold. Go resolves imports at package level, and BuildFileGraph deliberately drops an import that resolves to more than one file rather than fanning it into an edge per file, so a file inside a multi-file Go package has zero file-level importers by construction. `scanner/filegraph.go` scored 0 and read as harmless. So did every same-package collision, including all six pairs issue #134 verified by hand. A shared file whose language resolves at package granularity is now weighted as the files outside its package that import the package, plus the package's other files, and the count is labelled `package importers` so it is not read as a file-level number. Languages whose imports name files keep the file-level count and the plain label. FileGraph.Packages is populated for Go and nothing else, which is exactly the set this is correct for. The cross-package term cannot come from the graph — the edges are the ones that were dropped — so collide now keeps the scan outcome it was already paying for and counts the raw import strings. ScanForDeps plus BuildFileGraphFromOutcome is the same single scan BuildFileGraph was doing. Real effect on this repository, where the default previously printed nothing: 6 PRs scanner/filegraph.go 73 package importers <- #171, #174, #175, ... 3 PRs config/config.go 38 package importers <- #171, #181, #182 3 PRs main.go 6 package importers <- #175, #179, #180 --min-importers now defaults to 0. A file two open PRs both change is a hazard whatever its weight, and a default that hides hazards answers "no collisions" on a repository full of them. The flag stays for narrowing a long list. Three tests added: a Go fixture where two same-package files collide and carry a non-zero package weight (with the self-import and third-party cases held out of the count), a file-resolved language keeping file scope and its plain label, and #134's six measured pairs proved unchanged by the weighting — reordering them is allowed, adding or dropping one is not. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TcyheQmM3HCvxF5wRL3s5t * fix(collide): rank a pair by its heaviest shared file, not the first one seen Shared files sort by PR count first, so a pair colliding on a hub could be reported by a fixture touched by more PRs and ranked below a lesser pair. Found by independent review; the new test reproduces it and fails without the change. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TcyheQmM3HCvxF5wRL3s5t --------- Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com> Co-authored-by: r <r@r>
Adds
testdatato the hardcoded ignore list, alongsidevendorandnode_modules. Relates to #172.Why
Fixture repositories under
testdataare not the project's source. They are checked-in inputs for tests: scanning them counts fixture files as project files, and a fixture written in a language the file-level edge model does not cover drags the project's reported coverage down with it.Go's own toolchain takes the same position —
go buildand thegotool ignoretestdataentirely, and have since the beginning — so this matches a convention Go users already have. It is the justification for the default; the coverage effect is the consequence.An earlier draft of this PR showed a coverage
partial→completebefore/after. That measurement came from a Swift fixture added by #174, which is not merged, so it does not reproduce onmain— it has been removed rather than left as an unverifiable claim.What actually changes on this repo
main's tracked tree contains exactly onetestdatadirectory:watch/testdata/, seven shell scripts (run_test.sh,test_mcp.sh, and five underoutput/). Measured with binaries built fromorigin/mainand from this branch, both scanning a clean export oforigin/main:So the visible effect today is modest and mostly forward-looking: seven shell fixtures leave the file count, and the bare
watch/testdata/directory node leaves the tree. The value is in what it prevents as fixture directories accumulate — this repo's own milestone work adds them steadily.Can it be turned off? No — and that is worth being explicit about
IgnoredDirsis a hardcoded map inscanner/walker.go, checked on the fast path before gitignore and before any project config.OnlyandExcludein.codemap/config.jsonare additive filters — they narrow what a scan returns, they cannot widen it — so there is no config key, flag, or gitignore negation that putstestdataback into a scan.The one escape hatch is scanning the fixture directly as the root, and this PR now guarantees that:
IgnoredDirsis no longer applied to the walk root, socodemap testdata/, orcodemap .from insidevendor/, scans the directory the user actually named. Before that fix the walker matched the root's own base name and returnedFiles: 0— no error, no message, just an empty result.ReadExternalDepsgets the same exemption so a manifest sitting at that root is still read.Anyone who deliberately keeps project source under a directory named
testdatawill see those files disappear from counts and from the tree, with no way to opt back in short of renaming the directory or scanning it directly. Flagging it plainly: it is a default-behaviour change with no override.Tests
TestScanSkipsTestdataDirectories— skipped at top level and nested (internal/testdata/nested/).TestTestdataFixtureScansWhenItIsTheRoot— a fixture passed as the scan root still scans, which is how the fixture tests inscannercallBuildFileGraph.TestIgnoredDirNameScansWhenItIsTheRoot— a root literally namedtestdata,vendor, ornode_modulesscans its files. Fails on the pre-fix walker withreturned [], want [app.go]for all three.TestIgnoredDirRootStillSkipsNestedIgnoredDirs— the root exemption covers the root only; a nestedtestdataunder atestdataroot is still skipped.Also on this branch
Two fixes ported from open PRs so this branch's CI can go green without waiting on them; both no-op once
maincarries them.scanner/rustcargo.go), from @reneleonhardt's perf: Bound repository analysis hot paths #171:buildRustWorkspaceIndexshadowed its caller'sctxwith the cargo-metadata deadline, so an expired deadline failed the whole graph build instead of falling back to the manually derived Rust workspace.main.go), the window fix(watch): Treat an unparseable readiness file as not-ready-yet #177 fixes:waitWatchReadinessgave up on the first unparseable read of a readiness file still being written, surfacing asreading daemon readiness: unexpected end of JSON input.Release note
testdatadirectories are now skipped by default, matching Go's own toolchain convention and joiningvendorandnode_modules. Fixture files no longer count as project files or affect a project's reported coverage. Passing such a directory as the scan root still scans it — and now does so for every name on the ignore list, where previously it returned no files.Verification
go vet ./...clean,gofmtclean,go test ./...fully green on this branch.🤖 Generated with Claude Code
https://claude.ai/code/session_01TcyheQmM3HCvxF5wRL3s5t