Commit bea39e63 authored by Aliaksandr Valialkin's avatar Aliaksandr Valialkin Committed by Brad Fitzpatrick

regexp: reduce mallocs in Regexp.Find* and Regexp.ReplaceAll*.

This improves Regexp.Find* and Regexp.ReplaceAll* speed:

name                  old time/op    new time/op    delta
Find-4                   345ns ± 1%     314ns ± 1%    -8.94%    (p=0.000 n=9+8)
FindString-4             341ns ± 1%     308ns ± 0%    -9.85%   (p=0.000 n=10+9)
FindSubmatch-4           440ns ± 1%     404ns ± 0%    -8.27%   (p=0.000 n=10+8)
FindStringSubmatch-4     426ns ± 0%     387ns ± 0%    -9.07%   (p=0.000 n=10+9)
ReplaceAll-4            1.75µs ± 1%    1.67µs ± 0%    -4.45%   (p=0.000 n=9+10)

name                  old alloc/op   new alloc/op   delta
Find-4                   16.0B ± 0%     0.0B ±NaN%  -100.00%  (p=0.000 n=10+10)
FindString-4             16.0B ± 0%     0.0B ±NaN%  -100.00%  (p=0.000 n=10+10)
FindSubmatch-4           80.0B ± 0%     48.0B ± 0%   -40.00%  (p=0.000 n=10+10)
FindStringSubmatch-4     64.0B ± 0%     32.0B ± 0%   -50.00%  (p=0.000 n=10+10)
ReplaceAll-4              152B ± 0%      104B ± 0%   -31.58%  (p=0.000 n=10+10)

name                  old allocs/op  new allocs/op  delta
Find-4                    1.00 ± 0%     0.00 ±NaN%  -100.00%  (p=0.000 n=10+10)
FindString-4              1.00 ± 0%     0.00 ±NaN%  -100.00%  (p=0.000 n=10+10)
FindSubmatch-4            2.00 ± 0%      1.00 ± 0%   -50.00%  (p=0.000 n=10+10)
FindStringSubmatch-4      2.00 ± 0%      1.00 ± 0%   -50.00%  (p=0.000 n=10+10)
ReplaceAll-4              8.00 ± 0%      5.00 ± 0%   -37.50%  (p=0.000 n=10+10)

Fixes #15643

Change-Id: I594fe51172373e2adb98d1d25c76ca2cde54ff48
Reviewed-on: https://go-review.googlesource.com/23030Reviewed-by: default avatarBrad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
parent 5923df1a
......@@ -564,6 +564,72 @@ func TestSwitchBacktrack(t *testing.T) {
re.Match(long[:1]) // triggers backtracker
}
func BenchmarkFind(b *testing.B) {
b.StopTimer()
re := MustCompile("a+b+")
wantSubs := "aaabb"
s := []byte("acbb" + wantSubs + "dd")
b.StartTimer()
b.ReportAllocs()
for i := 0; i < b.N; i++ {
subs := re.Find(s)
if string(subs) != wantSubs {
b.Fatalf("Find(%q) = %q; want %q", s, subs, wantSubs)
}
}
}
func BenchmarkFindString(b *testing.B) {
b.StopTimer()
re := MustCompile("a+b+")
wantSubs := "aaabb"
s := "acbb" + wantSubs + "dd"
b.StartTimer()
b.ReportAllocs()
for i := 0; i < b.N; i++ {
subs := re.FindString(s)
if subs != wantSubs {
b.Fatalf("FindString(%q) = %q; want %q", s, subs, wantSubs)
}
}
}
func BenchmarkFindSubmatch(b *testing.B) {
b.StopTimer()
re := MustCompile("a(a+b+)b")
wantSubs := "aaabb"
s := []byte("acbb" + wantSubs + "dd")
b.StartTimer()
b.ReportAllocs()
for i := 0; i < b.N; i++ {
subs := re.FindSubmatch(s)
if string(subs[0]) != wantSubs {
b.Fatalf("FindSubmatch(%q)[0] = %q; want %q", s, subs[0], wantSubs)
}
if string(subs[1]) != "aab" {
b.Fatalf("FindSubmatch(%q)[1] = %q; want %q", s, subs[1], "aab")
}
}
}
func BenchmarkFindStringSubmatch(b *testing.B) {
b.StopTimer()
re := MustCompile("a(a+b+)b")
wantSubs := "aaabb"
s := "acbb" + wantSubs + "dd"
b.StartTimer()
b.ReportAllocs()
for i := 0; i < b.N; i++ {
subs := re.FindStringSubmatch(s)
if subs[0] != wantSubs {
b.Fatalf("FindStringSubmatch(%q)[0] = %q; want %q", s, subs[0], wantSubs)
}
if subs[1] != "aab" {
b.Fatalf("FindStringSubmatch(%q)[1] = %q; want %q", s, subs[1], "aab")
}
}
}
func BenchmarkLiteral(b *testing.B) {
x := strings.Repeat("x", 50) + "y"
b.StopTimer()
......
......@@ -405,14 +405,16 @@ func (m *machine) onepass(i input, pos int) bool {
return m.matched
}
// empty is a non-nil 0-element slice,
// so doExecute can avoid an allocation
// when 0 captures are requested from a successful match.
var empty = make([]int, 0)
// doMatch reports whether either r, b or s match the regexp.
func (re *Regexp) doMatch(r io.RuneReader, b []byte, s string) bool {
return re.doExecute(r, b, s, 0, 0, nil) != nil
}
// doExecute finds the leftmost match in the input and returns
// the position of its subexpressions.
func (re *Regexp) doExecute(r io.RuneReader, b []byte, s string, pos int, ncap int) []int {
// doExecute finds the leftmost match in the input, appends the position
// of its subexpressions to dstCap and returns dstCap.
//
// nil is returned if no matches are found and non-nil if matches are found.
func (re *Regexp) doExecute(r io.RuneReader, b []byte, s string, pos int, ncap int, dstCap []int) []int {
m := re.get()
var i input
var size int
......@@ -445,12 +447,15 @@ func (re *Regexp) doExecute(r io.RuneReader, b []byte, s string, pos int, ncap i
return nil
}
}
if ncap == 0 {
re.put(m)
return empty // empty but not nil
dstCap = append(dstCap, m.matchcap...)
if dstCap == nil {
// Keep the promise of returning non-nil value on match.
dstCap = arrayNoInts[:0]
}
cap := make([]int, len(m.matchcap))
copy(cap, m.matchcap)
re.put(m)
return cap
return dstCap
}
// arrayNoInts is returned by doExecute match if nil dstCap is passed
// to it with ncap=0.
var arrayNoInts [0]int
......@@ -408,17 +408,17 @@ func (re *Regexp) LiteralPrefix() (prefix string, complete bool) {
// MatchReader reports whether the Regexp matches the text read by the
// RuneReader.
func (re *Regexp) MatchReader(r io.RuneReader) bool {
return re.doExecute(r, nil, "", 0, 0) != nil
return re.doMatch(r, nil, "")
}
// MatchString reports whether the Regexp matches the string s.
func (re *Regexp) MatchString(s string) bool {
return re.doExecute(nil, nil, s, 0, 0) != nil
return re.doMatch(nil, nil, s)
}
// Match reports whether the Regexp matches the byte slice b.
func (re *Regexp) Match(b []byte) bool {
return re.doExecute(nil, b, "", 0, 0) != nil
return re.doMatch(nil, b, "")
}
// MatchReader checks whether a textual regular expression matches the text
......@@ -502,8 +502,9 @@ func (re *Regexp) replaceAll(bsrc []byte, src string, nmatch int, repl func(dst
nmatch = re.prog.NumCap
}
var dstCap [2]int
for searchPos <= endPos {
a := re.doExecute(nil, bsrc, src, searchPos, nmatch)
a := re.doExecute(nil, bsrc, src, searchPos, nmatch, dstCap[:0])
if len(a) == 0 {
break // no more matches
}
......@@ -641,7 +642,7 @@ func (re *Regexp) allMatches(s string, b []byte, n int, deliver func([]int)) {
}
for pos, i, prevMatchEnd := 0, 0, -1; i < n && pos <= end; {
matches := re.doExecute(nil, b, s, pos, re.prog.NumCap)
matches := re.doExecute(nil, b, s, pos, re.prog.NumCap, nil)
if len(matches) == 0 {
break
}
......@@ -681,7 +682,8 @@ func (re *Regexp) allMatches(s string, b []byte, n int, deliver func([]int)) {
// Find returns a slice holding the text of the leftmost match in b of the regular expression.
// A return value of nil indicates no match.
func (re *Regexp) Find(b []byte) []byte {
a := re.doExecute(nil, b, "", 0, 2)
var dstCap [2]int
a := re.doExecute(nil, b, "", 0, 2, dstCap[:0])
if a == nil {
return nil
}
......@@ -693,7 +695,7 @@ func (re *Regexp) Find(b []byte) []byte {
// b[loc[0]:loc[1]].
// A return value of nil indicates no match.
func (re *Regexp) FindIndex(b []byte) (loc []int) {
a := re.doExecute(nil, b, "", 0, 2)
a := re.doExecute(nil, b, "", 0, 2, nil)
if a == nil {
return nil
}
......@@ -706,7 +708,8 @@ func (re *Regexp) FindIndex(b []byte) (loc []int) {
// an empty string. Use FindStringIndex or FindStringSubmatch if it is
// necessary to distinguish these cases.
func (re *Regexp) FindString(s string) string {
a := re.doExecute(nil, nil, s, 0, 2)
var dstCap [2]int
a := re.doExecute(nil, nil, s, 0, 2, dstCap[:0])
if a == nil {
return ""
}
......@@ -718,7 +721,7 @@ func (re *Regexp) FindString(s string) string {
// itself is at s[loc[0]:loc[1]].
// A return value of nil indicates no match.
func (re *Regexp) FindStringIndex(s string) (loc []int) {
a := re.doExecute(nil, nil, s, 0, 2)
a := re.doExecute(nil, nil, s, 0, 2, nil)
if a == nil {
return nil
}
......@@ -731,7 +734,7 @@ func (re *Regexp) FindStringIndex(s string) (loc []int) {
// byte offset loc[0] through loc[1]-1.
// A return value of nil indicates no match.
func (re *Regexp) FindReaderIndex(r io.RuneReader) (loc []int) {
a := re.doExecute(r, nil, "", 0, 2)
a := re.doExecute(r, nil, "", 0, 2, nil)
if a == nil {
return nil
}
......@@ -744,7 +747,8 @@ func (re *Regexp) FindReaderIndex(r io.RuneReader) (loc []int) {
// comment.
// A return value of nil indicates no match.
func (re *Regexp) FindSubmatch(b []byte) [][]byte {
a := re.doExecute(nil, b, "", 0, re.prog.NumCap)
var dstCap [4]int
a := re.doExecute(nil, b, "", 0, re.prog.NumCap, dstCap[:0])
if a == nil {
return nil
}
......@@ -891,7 +895,7 @@ func extract(str string) (name string, num int, rest string, ok bool) {
// in the package comment.
// A return value of nil indicates no match.
func (re *Regexp) FindSubmatchIndex(b []byte) []int {
return re.pad(re.doExecute(nil, b, "", 0, re.prog.NumCap))
return re.pad(re.doExecute(nil, b, "", 0, re.prog.NumCap, nil))
}
// FindStringSubmatch returns a slice of strings holding the text of the
......@@ -900,7 +904,8 @@ func (re *Regexp) FindSubmatchIndex(b []byte) []int {
// package comment.
// A return value of nil indicates no match.
func (re *Regexp) FindStringSubmatch(s string) []string {
a := re.doExecute(nil, nil, s, 0, re.prog.NumCap)
var dstCap [4]int
a := re.doExecute(nil, nil, s, 0, re.prog.NumCap, dstCap[:0])
if a == nil {
return nil
}
......@@ -919,7 +924,7 @@ func (re *Regexp) FindStringSubmatch(s string) []string {
// 'Index' descriptions in the package comment.
// A return value of nil indicates no match.
func (re *Regexp) FindStringSubmatchIndex(s string) []int {
return re.pad(re.doExecute(nil, nil, s, 0, re.prog.NumCap))
return re.pad(re.doExecute(nil, nil, s, 0, re.prog.NumCap, nil))
}
// FindReaderSubmatchIndex returns a slice holding the index pairs
......@@ -928,7 +933,7 @@ func (re *Regexp) FindStringSubmatchIndex(s string) []int {
// by the 'Submatch' and 'Index' descriptions in the package comment. A
// return value of nil indicates no match.
func (re *Regexp) FindReaderSubmatchIndex(r io.RuneReader) []int {
return re.pad(re.doExecute(r, nil, "", 0, re.prog.NumCap))
return re.pad(re.doExecute(r, nil, "", 0, re.prog.NumCap, nil))
}
const startSize = 10 // The size at which to start a slice in the 'All' routines.
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment