Flesh out README

This commit switches back to using the regexp engine for
case-insensitive literal string matches.

This is slower, but at least case-insensitive matches for string
literals will function now. The code is a tiny bit shorter and simpler
too.

Given the aim of the tool is to be useful for ad-hoc searches,
efficiency isn't the concern but rather just getting the job done with
the minimum of fuss / unexpected behaviour.

README.md

@@ -1,3 +1,44 @@
 # gg
 
 Recursive grep written in Go, for everyday ease of use.
+
+`gg` is a recursive grep. Given a regexp (or fixed pattern) it will search for
+the pattern recursively in the current working directory. It will print a
+coloured header per file along with the matching line and pattern.
+
+It is possible to scan specific files or directories, rather than the default
+current working directory. To do this, simply specify the path(s) as arguments
+following the pattern.
+
+It is possible to scan for multiple patterns using the `-e` (or `-Q`) argument,
+which can be repeated multiple times. `-e` specifies a regular expression and
+`-Q` a fixed pattern. When using either flag, any non-flag arguments are treated
+as paths to scan.
+
+Search defaults to case-sensitive but the `-i` flag may be passed to make all
+search terms case-insensitive. Alternatively, the `"(?i)"` construct may be added
+to a regular expression to make that specific expression case insensitive.
+
+Files and directories can be excluded with the `-x` option. This supports bash-style
+globs with `'*'`, `'?'`, `'[a-z]'`, `'{this,that}'`, or `'/**/'` to match zero or more
+directories. By default, `.git` and vim swap files are ignored. Similarly, `-I`
+filters files to include. Examples:
+
+```
+	# ignore files/dirs with .js or .css suffix
+	gg -x '*.js' -x '*.css' pattern
+
+	# only match files with .go suffix (any subdir)
+	gg -I '*.go' pattern
+
+	# only match files whose parent dir is "stuff", but ignore "foo" subdir
+	gg -x ./foo -I 'stuff/*' pattern
+
+	# only match .js files with a directory "things" in the path, but ignore
+	# .min.js (e.g. will match "foo/things/bar/my.js")
+	gg -I 'things/**/*.js' -x '*.min.js' pattern
+```
+
+Symlinks named on the command line are followed, but by default symlinks are
+not followed when recursing into directories. `-L` allows them to be
+dereferenced.

file.go

@@ -12,6 +12,20 @@ import (
 	"unicode/utf8"
 )
 
+const (
+	// if a matching line is longer than longLine runes then we will print
+	// a truncated form
+	longLine = 256
+
+	// if any line in the file is detected to be longer than this many runes
+	// we will count the file as being minified
+	minifiedLine = 1024
+
+	// number of bytes at the start / end of each file to examine for binary
+	// or minified file detection.
+	bytesToExamine = 8192
+)
+
 type notPlainTextBehaviour int
 
 const (
@@ -54,10 +68,20 @@ func (nptf *notPlainTextFlag) Type() string {
 	return "short|full|skip" // ???
 }
 
+var (
+	// printedFull is set by file() if it prints a full file's matches
+	// (i.e. a header line, then the match lines). It is cleared if we
+	// just printed a "Binary file <foo> matches" line. It lets us have a
+	// nice one-line separator between full files, but show the binary
+	// matches compactly.
+	printedFull bool
+)
+
 func file(path string, data []byte) {
 	var short string
+	isBinary, isMinified := notPlainText(data)
 	switch {
-	case isBinary(data):
+	case isBinary:
 		switch binaryFile.b {
 		case notPlainTextShort:
 			short = "Binary"
@@ -65,7 +89,7 @@ func file(path string, data []byte) {
 			return
 		}
 
-	case isMinified(data):
+	case isMinified:
 		switch minifiedFile.b {
 		case notPlainTextShort:
 			short = "Minified"
@@ -194,55 +218,62 @@ func file(path string, data []byte) {
 	}
 }
 
-func isBinary(data []byte) bool {
+func notPlainText(data []byte) (isBinary, isMinified bool) {
-	bytesToExamine := 4096
+	// examine bytes at the start of the file for binary data
+	b, m := notPlainTextAux(data)
+	if b || len(data) < bytesToExamine*2 {
+		return b, m
+	}
 
-	for bytesToExamine > 0 {
+	// some files, like .a files, have a header which passes plaintext
+	// detection but we can expect the trailer to be binary
+	data = data[len(data)-bytesToExamine:]
+	for i := 0; i < 5; i++ { // attempt to align to UTF-8 char boundary
+		if utf8.RuneStart(data[0]) {
+			break
+		}
+		data = data[1:]
+	}
+	b2, m2 := notPlainTextAux(data)
+	return (b || b2), (m || m2)
+}
+
+func notPlainTextAux(data []byte) (isBinary, isMinified bool) {
+	n := bytesToExamine
+	var lineLength int
+	for n > 0 {
 		r, s := utf8.DecodeRune(data)
 		switch {
 		case s == 0:
 			// end of string
-			return false
+			return
+
 		case s == 1 && r == utf8.RuneError:
 			// invalid UTF-8
-			return true
+			isBinary = true
-		case r == '\r', r == '\n', r == '\t':
+			return
-			// valid control chars
-		case r < ' ':
-			// invalid control chars
-			return true
-		}
-		data = data[s:]
-		bytesToExamine -= s
-	}
 
-	return false
-}
-
-const longLine = 256
-
-func isMinified(data []byte) bool {
-	bytesToExamine := 4096
-	var lineLength int
-
-	for bytesToExamine > 0 {
-		r, s := utf8.DecodeRune(data)
-		switch {
-		case s == 0:
-			// end of string
-			return false
 		case r == '\n':
-			lineLength = 0
+			// newline
-		default:
+			lineLength = -1
-			lineLength++
+
-			if lineLength >= longLine {
+		case r == '\r', r == '\t', r == '\v':
-				return true
+			// valid control chars often present in text
-			}
+
+		case r < ' ':
+			// control chars not expected in plain text
+			isBinary = true
+			return
 		}
+
 		data = data[s:]
-		bytesToExamine -= s
+		n -= s
+		lineLength++
+		if lineLength >= minifiedLine {
+			isMinified = true
 		}
-	return false
+	}
+	return
 }
 
 func findMatches(data []byte) (loc []int) {
@@ -252,12 +283,6 @@ func findMatches(data []byte) (loc []int) {
 			return loc
 		}
 	}
-	for _, s := range searchBytes {
-		pos := bytes.Index(data, s)
-		if pos != -1 {
-			return []int{pos, pos + len(s)}
-		}
-	}
 	return nil
 }
 

main.go

@@ -9,6 +9,7 @@ import (
 	"os"
 	"path/filepath"
 	"regexp"
+	"strings"
 
 	"github.com/bmatcuk/doublestar/v4"
 	"github.com/spf13/cobra"
@@ -16,7 +17,6 @@ import (
 )
 
 // TODO:
-//  - it would be better to make fixed patterns case insensitive too.
 //  - configurable defaults for exclude.
 
 func main() {
@@ -41,46 +41,73 @@ which can be repeated multiple times. -e specifies a regular expression and
 -Q a fixed pattern. When using either flag, any non-flag arguments are treated
 as paths to scan.
 
-Search defaults to case-sensitive but the -i flag may be passed to make regular
+Search defaults to case-sensitive but the -i flag may be passed to make all
-expression searches case-insensitive. Alternatively, the "(?i)" construct may be
+search terms case-insensitive. Alternatively, the "(?i)" construct may be added
-added to a regular expression to make that specific expression case insensitive.
+to a regular expression to make that specific expression case insensitive.
-Fixed pattern matches are always case-sensitive.
 
 Files and directories can be excluded with the -x option. This supports bash-style
 globs with '*', '?', '[a-z]', '{this,that}', or '/**/' to match zero or more
-directories. By default, .git and vim swap files are ignored.`,
+directories. By default, .git and vim swap files are ignored. Similarly, -I
+filters files to include. Examples:
+
+	# ignore files/dirs with .js or .css suffix
+	gg -x '*.js' -x '*.css' pattern
+
+	# only match files with .go suffix (any subdir)
+	gg -I '*.go' pattern
+
+	# only match files whose parent dir is "stuff", but ignore "foo" subdir
+	gg -x ./foo -I 'stuff/*' pattern
+
+	# only match .js files with a directory "things" in the path, but ignore
+	# .min.js (e.g. will match "foo/things/bar/my.js")
+	gg -I 'things/**/*.js' -x '*.min.js' pattern
+
+Symlinks named on the command line are followed, but by default symlinks are
+not followed when recursing into directories. -L allows them to be
+dereferenced.`,
 
 	RunE: run,
 }
 
 var (
+	// flags
 	searchRegexp   []string
-	regexps      []*regexp.Regexp
 	searchFixed    []string
-	searchBytes  [][]byte
 	searchPath     []string
 	excludeList    []string
-	binaryFile   notPlainTextFlag
+	includeList    []string
-	minifiedFile notPlainTextFlag
 	ignoreCase     bool
 	noColour       bool
+	binaryFile     notPlainTextFlag
+	minifiedFile   notPlainTextFlag
+	followSymlinks bool
+
+	// computed from searchRegexp, searchFixed. Each regexp here will be
+	// matched against each line of each input file.
+	regexps []*regexp.Regexp
+
+	// formats output
 	display *Display
-	printedFull  bool
 )
 
 func init() {
 	rootCmd.Flags().StringSliceVarP(&searchRegexp, "grep", "e", nil, "pattern to match (regular expression)")
 	rootCmd.Flags().StringSliceVarP(&searchFixed, "fixed", "Q", nil, "pattern to match (fixed string)")
 	rootCmd.Flags().StringSliceVarP(&excludeList, "exclude", "x", []string{".git", ".*.swp"}, "files/directories to exclude")
+	rootCmd.Flags().StringSliceVarP(&includeList, "include", "I", nil, "files/directories to include")
 	rootCmd.Flags().BoolVarP(&ignoreCase, "ignore-case", "i", false, "make all searches case insensitive")
 	rootCmd.Flags().BoolVarP(&noColour, "no-colour", "C", false, "disable colour output")
 	rootCmd.Flags().Var(&binaryFile, "binary", "what to do with binary files")
 	rootCmd.Flags().Var(&minifiedFile, "minified", "what to do with minified text files")
+	rootCmd.Flags().BoolVarP(&followSymlinks, "dereference", "L", false, "follow symlinks when recursing")
 }
 
 func run(c *cobra.Command, args []string) error {
 	display = NewDisplay(noColour)
 
+	// if no -e or -Q flag is passed, then the first arg is taken to be
+	// the pattern to match
 	if len(searchRegexp) == 0 && len(searchFixed) == 0 {
 		if len(args) == 0 {
 			return errors.New("no pattern specified")
@@ -89,17 +116,38 @@ func run(c *cobra.Command, args []string) error {
 		args = args[1:]
 	}
 
+	// remaining arguments are treated as search paths; an empty list is
+	// taken to mean the CWD
 	searchPath = args
 	if len(searchPath) == 0 {
 		searchPath = append(searchPath, ".")
 	}
 
-	for _, x := range excludeList {
+	// if we got past argument passing, then returned errors are runtime
+	// things (like file not found) that shouldn't trigger a usage message.
+	c.SilenceUsage = true
+
+	// for -x and -I, an undecorated pattern is treated as a suffix match
+	for i, x := range excludeList {
+		if !strings.HasPrefix(x, "**/") && !strings.HasPrefix(x, "./") {
+			x = "**/" + x
+			excludeList[i] = x
+		}
 		if !doublestar.ValidatePattern(x) {
 			return fmt.Errorf("invalid exclude pattern %q", x)
 		}
 	}
+	for i, x := range includeList {
+		if !strings.HasPrefix(x, "**/") && !strings.HasPrefix(x, "./") {
+			x = "**/" + x
+			includeList[i] = x
+		}
+		if !doublestar.ValidatePattern(x) {
+			return fmt.Errorf("invalid include pattern %q", x)
+		}
+	}
 
+	// compile regular expressions for matching
 	for _, r := range searchRegexp {
 		if ignoreCase {
 			r = "(?i)" + r
@@ -110,14 +158,21 @@ func run(c *cobra.Command, args []string) error {
 		}
 		regexps = append(regexps, re)
 	}
-
+	for _, r := range searchFixed {
-	for _, s := range searchFixed {
+		r = regexp.QuoteMeta(r)
-		searchBytes = append(searchBytes, []byte(s))
+		if ignoreCase {
+			r = "(?i)" + r
+		}
+		re, err := regexp.Compile(r)
+		if err != nil {
+		}
+		regexps = append(regexps, re)
 	}
 
+	// search over named paths
 	var errs []error
 	for _, path := range searchPath {
-		if err := search(path); err != nil {
+		if err := search(path, true); err != nil {
 			errs = append(errs, err)
 		}
 	}
@@ -132,26 +187,63 @@ func recurse(path string) error {
 	}
 
 	var errs []error
-NextFile:
 	for _, de := range d {
-		name := de.Name()
+		fullPath := filepath.Join(path, de.Name())
-		fullPath := filepath.Join(path, name)
+		if !shouldSearch(fullPath, de.IsDir()) {
-
+			continue
-		for _, x := range excludeList {
-			if exclude, _ := doublestar.Match(x, fullPath); exclude {
-				continue NextFile
 		}
-		}
+		if err := search(fullPath, followSymlinks); err != nil {
-
-		if err := search(fullPath); err != nil {
 			errs = append(errs, err)
 		}
 	}
 	return errors.Join(errs...)
 }
 
-func search(path string) error {
+// shouldSearch matches the full path of the file against the include and
-	st, err := os.Stat(path)
+// exclude lists, returning true if we should consider the file/directory for
+// searching and false if not.
+func shouldSearch(fullPath string, isDir bool) bool {
+	// process the exclude list first
+	for _, x := range excludeList {
+		if exclude, _ := doublestar.Match(x, fullPath); exclude {
+			return false
+		}
+	}
+
+	// if the include list is empty, everything is included
+	if len(includeList) == 0 {
+		return true
+	}
+
+	for _, x := range includeList {
+		match, _ := doublestar.Match(x, fullPath)
+		fmt.Printf("[DEBUG] x=%q fullPath=%q isDir=%t match=%t\n", x, fullPath, isDir, match)
+
+		// if it's a directory, and we have at least one recursive
+		// matcher, then search
+		if isDir && strings.HasPrefix(x, "**/") {
+			return true
+		}
+
+		if include, _ := doublestar.Match(x, fullPath); include {
+			return true
+		}
+	}
+
+	return false
+}
+
+func search(path string, deref bool) error {
+	var (
+		st  os.FileInfo
+		err error
+	)
+
+	if deref {
+		st, err = os.Stat(path)
+	} else {
+		st, err = os.Lstat(path)
+	}
 	if err != nil {
 		return err
 	}