tecnickcom / gogen / 28879721455

100%

push

github

perf(stringkey): single-pass normalization with fewer allocations, plus uhex reuse, docs, and fuzz tests

- Rewrite New to normalize each field in a single pass via a new appendNormalizedField helper (collapse Unicode whitespace and lowercase together) instead of chaining strings.Fields, strings.Join, and strings.ToLower.
- Use norm.NFC.Bytes on a pre-grown buffer instead of transform.Bytes, avoiding an allocation and error return when the bytes are already normalized (the common ASCII case).
- Output keys are byte-identical: verified across all existing vectors and 200k random field combinations, so no stored keys change.
- Cut BenchmarkNew from 9 to 2 allocs/op (272 to 72 B/op) and ~30% faster; add BenchmarkNewASCII for the common fast path.
- Reimplement Hex via pkg/uhex.Hex64, dropping the manual strings.Repeat padding and the literal width, consistent with pkg/random's use of uhex.
- Document the NFC (not NFKC) behavior, the locale-independent unicode.ToLower lowercasing, and a Stability section covering the golang.org/x/text Unicode-version dependency for persisted keys.
- Add FuzzNew asserting determinism, case- and whitespace-insensitivity, and the Hex/String representation round-trip; add regression tests for composed-vs-decomposed equivalence and field-boundary distinctness.

33 of 33 new or added lines in 1 file covered. (100.0%)

13906 of 13910 relevant lines covered (99.97%)

981.84 hits per line