audit_test_subtests

Overview ¶

Command audit_test_subtests finds case loops in test functions that assert without opening a subtest, and can rewrite the mechanical ones.

The project's Go guidelines ask for table-driven tests whose cases run under t.Run: a failing case then names itself, a t.Fatal inside one case stops that case rather than the whole table, and `go test -run` can select a single case. The shape that violates the rule is a range over a table (a slice or map literal, inline or bound to a local variable) whose body calls t.Error, t.Errorf, t.Fatal, t.Fatalf, t.Fail or t.FailNow, or hands t to a helper, with no t.Run anywhere inside the loop body.

A loop that walks a sequence of dependent steps rather than independent cases is not a table. Mark it with a comment on the line before the loop or at the end of its first line:

// sequential: each step depends on the state the previous one left
for _, step := range []step{...} {

and the auditor records it as declared-sequential instead of reporting it. A loop inside a synctest.Test bubble is skipped too: the testing package panics on t.Run called inside a bubble, so the rule cannot apply there and the site is counted separately.

-fix rewrites the sites whose subtest name is unambiguous: a []string table names each case after its element, a struct table after a string field called name, desc, description, label, title or id, and a map[string]... table after its key. The body is wrapped in t.Run(name, func(t *testing.T) { ... }) and a bare continue becomes return; bodies that break, goto, or use a table without such a name are left for a hand rewrite and stay in the report.

Usage:

go run ./cmd/audit_test_subtests [-json out.json] [-check] [-fix] [dirs...]

With no directories the module's test files under ./cmd, ./internal and ./test are scanned. -check exits non-zero when any site remains, so the tool gates CI.