godoc: full text index for whitelisted non-Go files

R=rsc
CC=golang-dev
https://golang.org/cl/4029046

src/cmd/godoc/godoc.go

@@ -25,7 +25,6 @@ import (
 	"strings"
 	"template"
 	"time"
-	"utf8"
 )
 
 
@@ -56,7 +55,7 @@ var (
 	// TODO(gri) consider the invariant that goroot always end in '/'
 	goroot      = flag.String("goroot", runtime.GOROOT(), "Go root directory")
 	testDir     = flag.String("testdir", "", "Go root subdirectory - for testing only (faster startups)")
-	path        = flag.String("path", "", "additional package directories (colon-separated)")
+	pkgPath     = flag.String("path", "", "additional package directories (colon-separated)")
 	filter      = flag.String("filter", "", "filter file containing permitted package directory paths")
 	filterMin   = flag.Int("filter_minutes", 0, "filter file update interval in minutes; disabled if <= 0")
 	filterDelay delayTime // actual filter update interval in minutes; usually filterDelay == filterMin, but filterDelay may back off exponentially
@@ -80,7 +79,7 @@ var (
 
 
 func initHandlers() {
-	fsMap.Init(*path)
+	fsMap.Init(*pkgPath)
 	fileServer = http.FileServer(*goroot, "")
 	cmdHandler = httpHandler{"/cmd/", pathutil.Join(*goroot, "src/cmd"), false}
 	pkgHandler = httpHandler{"/pkg/", pathutil.Join(*goroot, "src/pkg"), true}
@@ -768,53 +767,6 @@ func redirect(w http.ResponseWriter, r *http.Request) (redirected bool) {
 }
 
 
-// TODO(gri): Should have a mapping from extension to handler, eventually.
-
-// textExt[x] is true if the extension x indicates a text file, and false otherwise.
-var textExt = map[string]bool{
-	".css": false, // must be served raw
-	".js":  false, // must be served raw
-}
-
-
-func isTextFile(path string) bool {
-	// if the extension is known, use it for decision making
-	if isText, found := textExt[pathutil.Ext(path)]; found {
-		return isText
-	}
-
-	// the extension is not known; read an initial chunk of
-	// file and check if it looks like correct UTF-8; if it
-	// does, it's probably a text file
-	f, err := os.Open(path, os.O_RDONLY, 0)
-	if err != nil {
-		return false
-	}
-	defer f.Close()
-
-	var buf [1024]byte
-	n, err := f.Read(buf[0:])
-	if err != nil {
-		return false
-	}
-
-	s := string(buf[0:n])
-	n -= utf8.UTFMax // make sure there's enough bytes for a complete unicode char
-	for i, c := range s {
-		if i > n {
-			break
-		}
-		if c == 0xFFFD || c < ' ' && c != '\n' && c != '\t' {
-			// decoding error or control character - not a text file
-			return false
-		}
-	}
-
-	// likely a text file
-	return true
-}
-
-
 func serveTextFile(w http.ResponseWriter, r *http.Request, abspath, relpath, title string) {
 	src, err := ioutil.ReadFile(abspath)
 	if err != nil {
@@ -922,7 +874,7 @@ type PageInfoMode uint
 
 const (
 	exportsOnly PageInfoMode = 1 << iota // only keep exported stuff
-	genDoc                               // generate documentation
+	genDoc      // generate documentation
 )
 
 

src/cmd/godoc/index.go

@@ -47,7 +47,7 @@ import (
 	"index/suffixarray"
 	"io/ioutil"
 	"os"
-	pathutil "path"
+	"path"
 	"regexp"
 	"sort"
 	"strings"
@@ -430,8 +430,9 @@ func (a *AltWords) filter(s string) *AltWords {
 // Indexer
 
 // Adjust these flags as seems best.
-const excludeMainPackages = false
-const excludeTestFiles = false
+const includeNonGoFiles = true
+const includeMainPackages = true
+const includeTestFiles = true
 
 
 type IndexResult struct {
@@ -619,11 +620,14 @@ func pkgName(filename string) string {
 }
 
 
-func (x *Indexer) addFile(filename string) *ast.File {
+// addFile adds a file to the index if possible and returns the file set file
+// and the file's AST if it was successfully parsed as a Go file. If addFile
+// failed (that is, if the file was not added), it returns file == nil.
+func (x *Indexer) addFile(filename string, goFile bool) (file *token.File, ast *ast.File) {
 	// open file
 	f, err := os.Open(filename, os.O_RDONLY, 0)
 	if err != nil {
-		return nil
+		return
 	}
 	defer f.Close()
 
@@ -643,59 +647,126 @@ func (x *Indexer) addFile(filename string) *ast.File {
 		panic("internal error - file base incorrect")
 	}
 
-	// append file contents to x.sources
-	if _, err := x.sources.ReadFrom(f); err != nil {
-		x.sources.Truncate(base) // discard possibly added data
-		return nil               // ignore files with I/O errors
+	// append file contents (src) to x.sources
+	if _, err := x.sources.ReadFrom(f); err == nil {
+		src := x.sources.Bytes()[base:]
+
+		if goFile {
+			// parse the file and in the process add it to the file set
+			if ast, err = parser.ParseFile(x.fset, filename, src, parser.ParseComments); err == nil {
+				file = x.fset.File(ast.Pos()) // ast.Pos() is inside the file
+				return
+			}
+			// file has parse errors, and the AST may be incorrect -
+			// set lines information explicitly and index as ordinary
+			// text file (cannot fall through to the text case below
+			// because the file has already been added to the file set
+			// by the parser)
+			file = x.fset.File(token.Pos(base)) // token.Pos(base) is inside the file
+			file.SetLinesForContent(src)
+			ast = nil
+			return
+		}
+
+		if isText(src) {
+			// only add the file to the file set (for the full text index)
+			file = x.fset.AddFile(filename, x.fset.Base(), len(src))
+			file.SetLinesForContent(src)
+			return
+		}
 	}
 
-	// parse the file and in the process add it to the file set
-	src := x.sources.Bytes()[base:] // no need to reread the file
-	file, err := parser.ParseFile(x.fset, filename, src, parser.ParseComments)
-	if err != nil {
-		// do not discard the added source code in this case
-		// because the file has been added to the file set and
-		// the source size must match the file set base
-		// TODO(gri): given a FileSet.RemoveFile() one might be
-		//            able to discard the data here (worthwhile?)
-		return nil // ignore files with (parse) errors
-	}
+	// discard possibly added data
+	x.sources.Truncate(base - 1) // -1 to remove added byte 0 since no file was added
+	return
+}
 
-	return file
+
+// Design note: Using an explicit white list of permitted files for indexing
+// makes sure that the important files are included and massively reduces the
+// number of files to index. The advantage over a blacklist is that unexpected
+// (non-blacklisted) files won't suddenly explode the index.
+//
+// TODO(gri): We may want to make this list customizable, perhaps via a flag.
+
+// Files are whitelisted if they have a file name or extension
+// present as key in whitelisted.
+var whitelisted = map[string]bool{
+	".bash":        true,
+	".c":           true,
+	".css":         true,
+	".go":          true,
+	".goc":         true,
+	".h":           true,
+	".html":        true,
+	".js":          true,
+	".out":         true,
+	".py":          true,
+	".s":           true,
+	".sh":          true,
+	".txt":         true,
+	".xml":         true,
+	"AUTHORS":      true,
+	"CONTRIBUTORS": true,
+	"LICENSE":      true,
+	"Makefile":     true,
+	"PATENTS":      true,
+	"README":       true,
+}
+
+
+// isWhitelisted returns true if a file is on the list
+// of "permitted" files for indexing.
+func isWhitelisted(filename string) bool {
+	key := path.Ext(filename)
+	if key == "" {
+		// file has no extension - use entire filename
+		key = filename
+	}
+	return whitelisted[key]
 }
 
 
 func (x *Indexer) visitFile(dirname string, f *os.FileInfo) {
-	if !isGoFile(f) {
+	if !f.IsRegular() {
 		return
 	}
 
-	path := pathutil.Join(dirname, f.Name)
-	if excludeTestFiles && (!isPkgFile(f) || strings.HasPrefix(path, "test/")) {
+	filename := path.Join(dirname, f.Name)
+	goFile := false
+
+	switch {
+	case isGoFile(f):
+		if !includeTestFiles && (!isPkgFile(f) || strings.HasPrefix(filename, "test/")) {
+			return
+		}
+		if !includeMainPackages && pkgName(filename) == "main" {
+			return
+		}
+		goFile = true
+
+	case !includeNonGoFiles || !isWhitelisted(filename):
 		return
 	}
 
-	if excludeMainPackages && pkgName(path) == "main" {
-		return
-	}
-
-	file := x.addFile(path)
+	file, fast := x.addFile(filename, goFile)
 	if file == nil {
-		return
+		return // addFile failed
 	}
 
-	// we've got a file to index
-	x.current = x.fset.File(file.Pos()) // file.Pos is in the current file
-	dir, _ := pathutil.Split(path)
-	pak := Pak{dir, file.Name.Name}
-	x.file = &File{path, pak}
-	ast.Walk(x, file)
+	if fast != nil {
+		// we've got a Go file to index
+		x.current = file
+		dir, _ := path.Split(filename)
+		pak := Pak{dir, fast.Name.Name}
+		x.file = &File{filename, pak}
+		ast.Walk(x, fast)
+	}
 
 	// update statistics
-	// (count real file size as opposed to using the padded x.sources.Len())
-	x.stats.Bytes += x.current.Size()
+	x.stats.Bytes += file.Size()
 	x.stats.Files++
-	x.stats.Lines += x.current.LineCount()
+	x.stats.Lines += file.LineCount()
 }
 
 

src/cmd/godoc/utils.go

@@ -15,11 +15,13 @@ import (
 	"strings"
 	"sync"
 	"time"
+	"utf8"
 )
 
 
 // An RWValue wraps a value and permits mutually exclusive
 // access to it and records the time the value was last set.
+//
 type RWValue struct {
 	mutex     sync.RWMutex
 	value     interface{}
@@ -107,3 +109,63 @@ func writeFileAtomically(filename string, data []byte) os.Error {
 	}
 	return os.Rename(f.Name(), filename)
 }
+
+
+// isText returns true if a significant prefix of s looks like correct UTF-8;
+// that is, if it is likely that s is human-readable text.
+//
+func isText(s []byte) bool {
+	const max = 1024 // at least utf8.UTFMax
+	if len(s) > max {
+		s = s[0:max]
+	}
+	for i, c := range string(s) {
+		if i+utf8.UTFMax > len(s) {
+			// last char may be incomplete - ignore
+			break
+		}
+		if c == 0xFFFD || c < ' ' && c != '\n' && c != '\t' {
+			// decoding error or control character - not a text file
+			return false
+		}
+	}
+	return true
+}
+
+
+// TODO(gri): Should have a mapping from extension to handler, eventually.
+
+// textExt[x] is true if the extension x indicates a text file, and false otherwise.
+var textExt = map[string]bool{
+	".css": false, // must be served raw
+	".js":  false, // must be served raw
+}
+
+
+// isTextFile returns true if the file has a known extension indicating
+// a text file, or if a significant chunk of the specified file looks like
+// correct UTF-8; that is, if it is likely that the file contains human-
+// readable text.
+//
+func isTextFile(filename string) bool {
+	// if the extension is known, use it for decision making
+	if isText, found := textExt[pathutil.Ext(filename)]; found {
+		return isText
+	}
+
+	// the extension is not known; read an initial chunk
+	// of the file and check if it looks like text
+	f, err := os.Open(filename, os.O_RDONLY, 0)
+	if err != nil {
+		return false
+	}
+	defer f.Close()
+
+	var buf [1024]byte
+	n, err := f.Read(buf[0:])
+	if err != nil {
+		return false
+	}
+
+	return isText(buf[0:n])
+}