go/doc: better headscan

- scan all comments not just the package documentation
- declutter output so that false positives are more easily spotted
- count the number of headings to quickly see differences
- minor tweaks

R=golang-dev, r, r
CC=golang-dev
https://golang.org/cl/5450061

3 files changed, 82 insertions, 24 deletions

diff --git a/src/pkg/go/doc/comment.go b/src/pkg/go/doc/comment.go
index 44a047588d..d7bb384ed0 100644
--- a/src/pkg/go/doc/comment.go
+++ b/src/pkg/go/doc/comment.go
@@ -303,9 +303,8 @@ func heading(line []byte) []byte {
 		return nil
 	}
 
-	// allow ' for possessive 's only
-	b := line
-	for {
+	// allow "'" for possessive "'s" only
+	for b := line; ; {
 		i := bytes.IndexRune(b, '\'')
 		if i < 0 {
 			break
@@ -339,7 +338,7 @@ func heading(line []byte) []byte {
 func ToHTML(w io.Writer, s []byte, words map[string]string) {
 	inpara := false
 	lastWasBlank := false
-	lastNonblankWasHeading := false
+	lastWasHeading := false
 
 	close := func() {
 		if inpara {
@@ -389,10 +388,11 @@ func ToHTML(w io.Writer, s []byte, words map[string]string) {
 				emphasize(w, line, nil, false) // no nice text formatting
 			}
 			w.Write(html_endpre)
+			lastWasHeading = false
 			continue
 		}
 
-		if lastWasBlank && !lastNonblankWasHeading && i+2 < len(lines) &&
+		if lastWasBlank && !lastWasHeading && i+2 < len(lines) &&
 			isBlank(lines[i+1]) && !isBlank(lines[i+2]) && indentLen(lines[i+2]) == 0 {
 			// current line is non-blank, sourounded by blank lines
 			// and the next non-blank line is not indented: this
@@ -403,7 +403,7 @@ func ToHTML(w io.Writer, s []byte, words map[string]string) {
 				template.HTMLEscape(w, head)
 				w.Write(html_endh)
 				i += 2
-				lastNonblankWasHeading = true
+				lastWasHeading = true
 				continue
 			}
 		}
@@ -411,7 +411,7 @@ func ToHTML(w io.Writer, s []byte, words map[string]string) {
 		// open paragraph
 		open()
 		lastWasBlank = false
-		lastNonblankWasHeading = false
+		lastWasHeading = false
 		emphasize(w, lines[i], words, true) // nice text formatting
 		i++
 	}
diff --git a/src/pkg/go/doc/comment_test.go b/src/pkg/go/doc/comment_test.go
index 9e77ae2cde..870660ad62 100644
--- a/src/pkg/go/doc/comment_test.go
+++ b/src/pkg/go/doc/comment_test.go
@@ -19,7 +19,7 @@ var headingTests = []struct {
 	{"", false},
 	{"section", false},
 	{"A typical usage:", true},
-	{"δ is Greek", false}, // TODO: consider allowing this 
+	{"δ is Greek", false},
 	{"Foo §", false},
 	{"Fermat's Last Sentence", true},
 	{"Fermat's", true},
diff --git a/src/pkg/go/doc/headscan.go b/src/pkg/go/doc/headscan.go
index 95953b3bdc..83f24627c9 100644
--- a/src/pkg/go/doc/headscan.go
+++ b/src/pkg/go/doc/headscan.go
@@ -1,53 +1,111 @@
+// Copyright 2011 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.
+
+/*
+	The headscan command extracts comment headings from package files;
+	it is used to detect false positives which may require an adjustment
+	to the comment formatting heuristics in comment.go.
+
+	Usage: headscan [-root root_directory]
+
+	By default, the $GOROOT/src directory is scanned.
+*/
 package main
 
 import (
 	"bytes"
 	"flag"
+	"fmt"
 	"go/doc"
 	"go/parser"
 	"go/token"
-	"log"
 	"os"
 	"path/filepath"
+	"runtime"
 	"strings"
 )
 
+var (
+	root    = flag.String("root", filepath.Join(runtime.GOROOT(), "src"), "root of filesystem tree to scan")
+	verbose = flag.Bool("v", false, "verbose mode")
+)
+
+const (
+	html_h    = "<h3>"
+	html_endh = "</h3>\n"
+)
+
 func isGoFile(fi os.FileInfo) bool {
 	return strings.HasSuffix(fi.Name(), ".go") &&
 		!strings.HasSuffix(fi.Name(), "_test.go")
 }
 
+func appendHeadings(list []string, comment string) []string {
+	var buf bytes.Buffer
+	doc.ToHTML(&buf, []byte(comment), nil)
+	for s := buf.String(); ; {
+		i := strings.Index(s, html_h)
+		if i < 0 {
+			break
+		}
+		i += len(html_h)
+		j := strings.Index(s, html_endh)
+		if j < 0 {
+			list = append(list, s[i:]) // incorrect HTML
+			break
+		}
+		list = append(list, s[i:j])
+		s = s[j+len(html_endh):]
+	}
+	return list
+}
+
 func main() {
-	fset := token.NewFileSet()
-	rootDir := flag.String("root", "./", "root of filesystem tree to scan")
 	flag.Parse()
-	err := filepath.Walk(*rootDir, func(path string, fi os.FileInfo, err error) error {
+	fset := token.NewFileSet()
+	nheadings := 0
+	err := filepath.Walk(*root, func(path string, fi os.FileInfo, err error) error {
 		if !fi.IsDir() {
 			return nil
 		}
 		pkgs, err := parser.ParseDir(fset, path, isGoFile, parser.ParseComments)
 		if err != nil {
-			log.Println(path, err)
+			if *verbose {
+				fmt.Fprintln(os.Stderr, err)
+			}
 			return nil
 		}
 		for _, pkg := range pkgs {
 			d := doc.NewPackageDoc(pkg, path)
-			buf := new(bytes.Buffer)
-			doc.ToHTML(buf, []byte(d.Doc), nil)
-			b := buf.Bytes()
-			for {
-				i := bytes.Index(b, []byte("<h3>"))
-				if i == -1 {
-					break
+			list := appendHeadings(nil, d.Doc)
+			for _, d := range d.Consts {
+				list = appendHeadings(list, d.Doc)
+			}
+			for _, d := range d.Types {
+				list = appendHeadings(list, d.Doc)
+			}
+			for _, d := range d.Vars {
+				list = appendHeadings(list, d.Doc)
+			}
+			for _, d := range d.Funcs {
+				list = appendHeadings(list, d.Doc)
+			}
+			if len(list) > 0 {
+				// directories may contain multiple packages;
+				// print path and package name
+				fmt.Printf("%s (package %s)\n", path, pkg.Name)
+				for _, h := range list {
+					fmt.Printf("\t%s\n", h)
 				}
-				line := bytes.SplitN(b[i:], []byte("\n"), 2)[0]
-				log.Printf("%s: %s", path, line)
-				b = b[i+len(line):]
+				nheadings += len(list)
 			}
 		}
 		return nil
 	})
 	if err != nil {
-		log.Fatal(err)
+		fmt.Fprintln(os.Stderr, err)
+		os.Exit(1)
 	}
+	fmt.Println(nheadings, "headings found")
 }