go/doc: compile regexps lazily

Compile go/doc's 4 regexps lazily, on demand.

Also, add a test for the one that had no test coverage.

This reduces init-time CPU as well as heap by ~20KB when they're not
used, which seems to be common enough. As an example, cmd/doc only
seems to use 1 of them. (as noted by temporary print statements)

Updates #26775

Change-Id: I85df89b836327a53fb8e1ace3f92480374270368
Reviewed-on: https://go-review.googlesource.com/127875
Run-TryBot: Brad Fitzpatrick <[email protected]>
TryBot-Result: Gobot Gobot <[email protected]>
Reviewed-by: Ian Lance Taylor <[email protected]>

src/go/build/deps_test.go

@@ -204,7 +204,7 @@ var pkgDeps = map[string][]string{
 
 	// Go parser.
 	"go/ast":     {"L4", "OS", "go/scanner", "go/token"},
-	"go/doc":     {"L4", "go/ast", "go/token", "regexp", "text/template"},
+	"go/doc":     {"L4", "OS", "go/ast", "go/token", "regexp", "text/template"},
 	"go/parser":  {"L4", "OS", "go/ast", "go/scanner", "go/token"},
 	"go/printer": {"L4", "OS", "go/ast", "go/scanner", "go/token", "text/tabwriter"},
 	"go/scanner": {"L4", "OS", "go/token"},

src/go/doc/comment.go

@@ -8,7 +8,6 @@ package doc
 
 import (
 	"io"
-	"regexp"
 	"strings"
 	"text/template" // for HTMLEscape
 	"unicode"
@@ -63,7 +62,7 @@ const (
 	urlRx = protoPart + `://` + hostPart + pathPart
 )
 
-var matchRx = regexp.MustCompile(`(` + urlRx + `)|(` + identRx + `)`)
+var matchRx = newLazyRE(`(` + urlRx + `)|(` + identRx + `)`)
 
 var (
 	html_a      = []byte(`<a href="`)
@@ -276,7 +275,7 @@ type block struct {
 	lines []string
 }
 
-var nonAlphaNumRx = regexp.MustCompile(`[^a-zA-Z0-9]`)
+var nonAlphaNumRx = newLazyRE(`[^a-zA-Z0-9]`)
 
 func anchorID(line string) string {
 	// Add a "hdr-" prefix to avoid conflicting with IDs used for package symbols.

src/go/doc/doc_test.go

@@ -144,3 +144,12 @@ func Test(t *testing.T) {
 	test(t, AllDecls)
 	test(t, AllMethods)
 }
+
+func TestAnchorID(t *testing.T) {
+	const in = "Important Things 2 Know & Stuff"
+	const want = "hdr-Important_Things_2_Know___Stuff"
+	got := anchorID(in)
+	if got != want {
+		t.Errorf("anchorID(%q) = %q; want %q", in, got, want)
+	}
+}

src/go/doc/lazyre.go

@@ -0,0 +1,51 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package doc
+
+import (
+	"os"
+	"regexp"
+	"strings"
+	"sync"
+)
+
+type lazyRE struct {
+	str  string
+	once sync.Once
+	rx   *regexp.Regexp
+}
+
+func (r *lazyRE) re() *regexp.Regexp {
+	r.once.Do(r.build)
+	return r.rx
+}
+
+func (r *lazyRE) build() {
+	r.rx = regexp.MustCompile(r.str)
+	r.str = ""
+}
+
+func (r *lazyRE) FindStringSubmatchIndex(s string) []int {
+	return r.re().FindStringSubmatchIndex(s)
+}
+
+func (r *lazyRE) ReplaceAllString(src, repl string) string {
+	return r.re().ReplaceAllString(src, repl)
+}
+
+func (r *lazyRE) MatchString(s string) bool {
+	return r.re().MatchString(s)
+}
+
+var inTest = len(os.Args) > 0 && strings.HasSuffix(strings.TrimSuffix(os.Args[0], ".exe"), ".test")
+
+func newLazyRE(str string) *lazyRE {
+	lr := &lazyRE{str: str}
+	if inTest {
+		// In tests, always compile the regexps early.
+		lr.re()
+	}
+	return lr
+}

src/go/doc/reader.go

@@ -7,7 +7,6 @@ package doc
 import (
 	"go/ast"
 	"go/token"
-	"regexp"
 	"sort"
 	"strconv"
 )
@@ -425,9 +424,9 @@ func (r *reader) readFunc(fun *ast.FuncDecl) {
 }
 
 var (
-	noteMarker    = `([A-Z][A-Z]+)\(([^)]+)\):?`                    // MARKER(uid), MARKER at least 2 chars, uid at least 1 char
-	noteMarkerRx  = regexp.MustCompile(`^[ \t]*` + noteMarker)      // MARKER(uid) at text start
-	noteCommentRx = regexp.MustCompile(`^/[/*][ \t]*` + noteMarker) // MARKER(uid) at comment start
+	noteMarker    = `([A-Z][A-Z]+)\(([^)]+)\):?`           // MARKER(uid), MARKER at least 2 chars, uid at least 1 char
+	noteMarkerRx  = newLazyRE(`^[ \t]*` + noteMarker)      // MARKER(uid) at text start
+	noteCommentRx = newLazyRE(`^/[/*][ \t]*` + noteMarker) // MARKER(uid) at comment start
 )
 
 // readNote collects a single note from a sequence of comments.