rm-hull / jasengo / 30310472180

78%

Build Type

push

github

Committed by web-flow

Commit Message

perf: Reduce string and slice allocations (#21) (#27)

* Add comprehensive benchmark test suite for performance analysis

Add benchmark tests covering:
- Reader operations (Read, Checkpoint, Rollback, Slice, Remaining)
- Basic parsers (Char, Digit, Satisfy, OneOf, Whitespace, StringP)
- Combinators (Many, Many1, Choice, Map, Bind, Sequence, Optional, SepBy, ChainL)
- Regex parsing (simple and complex patterns)
- Buffer operations (RingBuffer and UnboundedBuffer write/read/slice)
- Location tracking overhead
- Error creation and pickBestError
- Integration tests (syslog parsing, arithmetic expression evaluation)

These benchmarks establish baseline performance metrics for identifying
and measuring the impact of future optimizations.

* Add benchmark workflow to CI

Add a benchmark job to the build workflow that:
- Runs all benchmarks across parser, buffer, and parser_test packages
- Stores results using benchmark-action/github-action-benchmark
- Generates charts for tracking performance over time
- Comments on PRs with benchmark results
- Alerts on significant performance regressions (150% threshold)

This mirrors the benchmark workflow from the dot-block project.

* Fix errcheck lint errors in benchmark tests

- Add explicit error handling for Read/Rollback calls in benchmarks
- Add explicit return value handling for parser.Run calls
- Add explicit return value handling for pickBestError calls

* Update benchmarks to use b.Loop() (Go 1.24+)

Replace traditional for i := 0; i < b.N; i++ loops with b.Loop() for:
- Better timer management (automatic reset on first call, stop on exit)
- More robust benchmark execution
- Alignment with modern Go benchmarking practices

Files updated:
- internal/buffer/bench_test.go
- parser/bench_test.go
- parser_test/bench_test.go

* Remove redundant b.ResetTimer() calls

b.Loop() automatically resets the timer on first call, making explicit
b.ResetTimer() calls before the loop redundant. Removed from all benchmarks
that ... (continued)

Coverage Stats

2 of 7 new or added lines in 1 file covered. (28.57%)

1 existing line in 1 file now uncovered.

475 of 593 relevant lines covered (80.1%)

174704.44 hits per line