v0.2.1 — DFA backend with SIMD prefilter

Major performance release: DFA compilation with SIMD-accelerated prefilter.

Highlights

• Up to 7.4 GB/s throughput on IsMatch (14x improvement over v0.1.0)
• Up to 4.0 GB/s on Find (13x improvement)
• Zero heap allocations on IsMatch hot path
• 71% fewer allocations in FindAll

What changed

The search backend is now a fully compiled DFA with a flat []uint32 transition table and premultiplied state IDs. All failure transitions are pre-computed at build time, so search is a single table lookup per byte.

A SIMD-accelerated prefilter using bytes.IndexByte (SSE2/AVX2 on amd64) scans for pattern start bytes before and during DFA traversal, skipping entire non-matching regions in bulk.

Benchmarks

Intel i7-1255U, 64KB haystack, 4-7 patterns:

Method	v0.1.0	v0.2.1	Improvement
Find	300 MB/s	4.0 GB/s	13x
IsMatch (no match)	260 MB/s	6.5 GB/s	25x
IsMatch (match@32KB)	545 MB/s	7.4 GB/s	14x
FindAll (77B, 10 matches)	40 MB/s, 14 alloc	125 MB/s, 4 alloc	3x

Breaking changes

None. Public API is fully backward-compatible with v0.1.0.

Full changelog

https://github.com/coregx/ahocorasick/blob/v0.2.1/CHANGELOG.md

v0.1.0: Initial Release

ahocorasick v0.1.0

High-performance Aho-Corasick multi-pattern string matching for Go.

Highlights

• 1+ GB/s throughput — comparable to Rust's aho-corasick
• Zero dependencies — pure Go, no cgo
• Zero allocations — IsMatch() hot path allocates nothing

Performance

Benchmarks on Intel i7-1255U (64KB haystack, 4 patterns):

Method	Throughput	Allocations
IsMatch (with match)	1.6 GB/s	0
Find	1.1 GB/s	1
IsMatch (no match)	780 MB/s	0

Installation

go get github.com/coregx/ahocorasick@v0.1.0

Quick Example

ac, _ := ahocorasick.NewBuilder().
    AddStrings([]string{"error", "warning", "fatal"}).
    Build()

if ac.IsMatch([]byte("an error occurred")) {
    fmt.Println("Found match!")
}

What's Included

• Builder pattern — fluent API for configuration
• Multiple match methods — Find, FindAll, FindAllOverlapping, IsMatch, Count
• Match semantics — LeftmostFirst (Perl) and LeftmostLongest (POSIX)
• Optimizations — dense transitions, byte classes, precomputed root

Testing

• 27 unit tests + 2 fuzz tests
• 93% code coverage
• CI on Linux, Windows, macOS with race detector

---

See CHANGELOG.md for full details.