Merge pull request #14090 from colega/improve-zeroOrOneCharacterStringMatcher-Matches

Improve `zeroOrOneCharacterStringMatcher` by using `utf8.DecodeRuneInString`
This commit is contained in:
Bryan Boreham 2024-05-16 09:28:53 +01:00 committed by GitHub
commit 1e0b0e250a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 98 additions and 34 deletions

View file

@ -828,7 +828,12 @@ type zeroOrOneCharacterStringMatcher struct {
}
func (m *zeroOrOneCharacterStringMatcher) Matches(s string) bool {
if moreThanOneRune(s) {
// If there's more than one rune in the string, then it can't match.
if r, size := utf8.DecodeRuneInString(s); r == utf8.RuneError {
// Size is 0 for empty strings, 1 for invalid rune.
// Empty string matches, invalid rune matches if there isn't anything else.
return size == len(s)
} else if size < len(s) {
return false
}
@ -840,27 +845,6 @@ func (m *zeroOrOneCharacterStringMatcher) Matches(s string) bool {
return s[0] != '\n'
}
// moreThanOneRune returns true if there are more than one runes in the string.
// It doesn't check whether the string is valid UTF-8.
// The return value should be always equal to utf8.RuneCountInString(s) > 1,
// but the function is optimized for the common case where the string prefix is ASCII.
func moreThanOneRune(s string) bool {
// If len(s) is exactly one or zero, there can't be more than one rune.
// Exit through this path quickly.
if len(s) <= 1 {
return false
}
// There's one or more bytes:
// If first byte is ASCII then there are multiple runes if there are more bytes after that.
if s[0] < utf8.RuneSelf {
return len(s) > 1
}
// Less common case: first is a multibyte rune.
return utf8.RuneCountInString(s) > 1
}
// trueMatcher is a stringMatcher which matches any string (always returns true).
type trueMatcher struct{}

View file

@ -19,6 +19,7 @@ import (
"strings"
"testing"
"time"
"unicode/utf8"
"github.com/grafana/regexp"
"github.com/grafana/regexp/syntax"
@ -36,6 +37,7 @@ var (
".*foo",
"^.*foo$",
"^.+foo$",
".?",
".*",
".+",
"foo.+",
@ -88,6 +90,12 @@ var (
// Values matching / not matching the test regexps on long alternations.
"zQPbMkNO", "zQPbMkNo", "jyyfj00j0061", "jyyfj00j006", "jyyfj00j00612", "NNSPdvMi", "NNSPdvMiXXX", "NNSPdvMixxx", "nnSPdvMi", "nnSPdvMiXXX",
// Invalid utf8
"\xfefoo",
"foo\xfe",
"\xfd",
"\xff\xff",
}
)
@ -926,19 +934,91 @@ func BenchmarkOptimizeEqualStringMatchers(b *testing.B) {
}
func TestZeroOrOneCharacterStringMatcher(t *testing.T) {
t.Run("match newline", func(t *testing.T) {
matcher := &zeroOrOneCharacterStringMatcher{matchNL: true}
require.True(t, matcher.Matches(""))
require.True(t, matcher.Matches("x"))
require.True(t, matcher.Matches("\n"))
require.False(t, matcher.Matches("xx"))
require.False(t, matcher.Matches("\n\n"))
})
matcher = &zeroOrOneCharacterStringMatcher{matchNL: false}
t.Run("do not match newline", func(t *testing.T) {
matcher := &zeroOrOneCharacterStringMatcher{matchNL: false}
require.True(t, matcher.Matches(""))
require.True(t, matcher.Matches("x"))
require.False(t, matcher.Matches("\n"))
require.False(t, matcher.Matches("xx"))
require.False(t, matcher.Matches("\n\n"))
})
t.Run("unicode", func(t *testing.T) {
// Just for documentation purposes, emoji1 is 1 rune, emoji2 is 2 runes.
// Having this in mind, will make future readers fixing tests easier.
emoji1 := "😀"
emoji2 := "❤️"
require.Equal(t, 1, utf8.RuneCountInString(emoji1))
require.Equal(t, 2, utf8.RuneCountInString(emoji2))
matcher := &zeroOrOneCharacterStringMatcher{matchNL: true}
require.True(t, matcher.Matches(emoji1))
require.False(t, matcher.Matches(emoji2))
require.False(t, matcher.Matches(emoji1+emoji1))
require.False(t, matcher.Matches("x"+emoji1))
require.False(t, matcher.Matches(emoji1+"x"))
require.False(t, matcher.Matches(emoji1+emoji2))
})
t.Run("invalid unicode", func(t *testing.T) {
// Just for reference, we also compare to what `^.?$` regular expression matches.
re := regexp.MustCompile("^.?$")
matcher := &zeroOrOneCharacterStringMatcher{matchNL: true}
requireMatches := func(s string, expected bool) {
t.Helper()
require.Equal(t, expected, matcher.Matches(s))
require.Equal(t, re.MatchString(s), matcher.Matches(s))
}
requireMatches("\xff", true)
requireMatches("x\xff", false)
requireMatches("\xffx", false)
requireMatches("\xff\xfe", false)
})
}
func BenchmarkZeroOrOneCharacterStringMatcher(b *testing.B) {
type benchCase struct {
str string
matches bool
}
emoji1 := "😀"
emoji2 := "❤️"
cases := []benchCase{
{"", true},
{"x", true},
{"\n", true},
{"xx", false},
{"\n\n", false},
{emoji1, true},
{emoji2, false},
{emoji1 + emoji1, false},
{strings.Repeat("x", 100), false},
{strings.Repeat(emoji1, 100), false},
{strings.Repeat(emoji2, 100), false},
}
matcher := &zeroOrOneCharacterStringMatcher{matchNL: true}
b.ResetTimer()
for n := 0; n < b.N; n++ {
c := cases[n%len(cases)]
got := matcher.Matches(c.str)
if got != c.matches {
b.Fatalf("unexpected result for %q: got %t, want %t", c.str, got, c.matches)
}
}
}
func TestLiteralPrefixStringMatcher(t *testing.T) {