Merge pull request #2311 from vincentbernat/feature/negative-pattern

filter: ability to use negative patterns

changelog/unreleased/issue-233

@@ -0,0 +1,31 @@
+Enhancement: Add negative patterns for include/exclude
+
+If a pattern starts with an exclamation mark and it matches a file that
+was previously matched by a regular pattern, the match is cancelled.
+Notably, this can be used with `--exclude-file` to cancel the
+exclusion of some files.
+
+It works similarly to `gitignore`, with the same limitation: once a
+directory is excluded, it is not possible to include files inside the
+directory.
+
+Example of use (as an exclude pattern for backup):
+
+    $HOME/**/*
+    !$HOME/Documents
+    !$HOME/code
+    !$HOME/.emacs.d
+    !$HOME/games
+    # [...]
+    node_modules
+    *~
+    *.o
+    *.lo
+    *.pyc
+    # [...]
+    $HOME/code/linux/*
+    !$HOME/code/linux/.git
+    # [...]
+
+https://github.com/restic/restic/issues/233
+https://github.com/restic/restic/pull/2311

doc/040_backup.rst

@@ -289,6 +289,28 @@ On most Unixy shells, you can either quote or use backslashes. For example:
  * ``--exclude="foo bar star/foo.txt"``
  * ``--exclude=foo\ bar\ star/foo.txt``
 
+If a pattern starts with exclamation mark and matches a file that
+was previously matched by a regular pattern, the match is cancelled.
+It works similarly to ``gitignore``, with the same limitation: once a
+directory is excluded, it is not possible to include files inside the
+directory. Here is a complete example to backup a selection of
+directories inside the home directory. It works by excluding any
+directory, then selectively add back some of them.
+
+::
+
+    $HOME/**/*
+    !$HOME/Documents
+    !$HOME/code
+    !$HOME/.emacs.d
+    !$HOME/games
+    # [...]
+    node_modules
+    *~
+    *.o
+    *.lo
+    *.pyc
+
 By specifying the option ``--one-file-system`` you can instruct restic
 to only backup files from the file systems the initially specified files
 or directories reside on. In other words, it will prevent restic from crossing

internal/filter/filter.go

@@ -17,7 +17,10 @@ type patternPart struct {
 }
 
 // Pattern represents a preparsed filter pattern
-type Pattern []patternPart
+type Pattern struct {
+	parts     []patternPart
+	isNegated bool
+}
 
 func prepareStr(str string) ([]string, error) {
 	if str == "" {
@@ -26,20 +29,26 @@ func prepareStr(str string) ([]string, error) {
 	return splitPath(str), nil
 }
 
-func preparePattern(pattern string) Pattern {
+func preparePattern(patternStr string) Pattern {
-	parts := splitPath(filepath.Clean(pattern))
+	var negate bool
-	patterns := make([]patternPart, len(parts))
+	if patternStr[0] == '!' {
-	for i, part := range parts {
+		negate = true
+		patternStr = patternStr[1:]
+	}
+
+	pathParts := splitPath(filepath.Clean(patternStr))
+	parts := make([]patternPart, len(pathParts))
+	for i, part := range pathParts {
 		isSimple := !strings.ContainsAny(part, "\\[]*?")
 		// Replace "**" with the empty string to get faster comparisons
 		// (length-check only) in hasDoubleWildcard.
 		if part == "**" {
 			part = ""
 		}
-		patterns[i] = patternPart{part, isSimple}
+		parts[i] = patternPart{part, isSimple}
 	}
 
-	return patterns
+	return Pattern{parts, negate}
 }
 
 // Split p into path components. Assuming p has been Cleaned, no component
@@ -62,19 +71,19 @@ func splitPath(p string) []string {
 // In addition patterns suitable for filepath.Match, pattern accepts a
 // recursive wildcard '**', which greedily matches an arbitrary number of
 // intermediate directories.
-func Match(pattern, str string) (matched bool, err error) {
+func Match(patternStr, str string) (matched bool, err error) {
-	if pattern == "" {
+	if patternStr == "" {
 		return true, nil
 	}
 
-	patterns := preparePattern(pattern)
+	pattern := preparePattern(patternStr)
 	strs, err := prepareStr(str)
 
 	if err != nil {
 		return false, err
 	}
 
-	return match(patterns, strs)
+	return match(pattern, strs)
 }
 
 // ChildMatch returns true if children of str can match the pattern. When the pattern is
@@ -87,28 +96,28 @@ func Match(pattern, str string) (matched bool, err error) {
 // In addition patterns suitable for filepath.Match, pattern accepts a
 // recursive wildcard '**', which greedily matches an arbitrary number of
 // intermediate directories.
-func ChildMatch(pattern, str string) (matched bool, err error) {
+func ChildMatch(patternStr, str string) (matched bool, err error) {
-	if pattern == "" {
+	if patternStr == "" {
 		return true, nil
 	}
 
-	patterns := preparePattern(pattern)
+	pattern := preparePattern(patternStr)
 	strs, err := prepareStr(str)
 
 	if err != nil {
 		return false, err
 	}
 
-	return childMatch(patterns, strs)
+	return childMatch(pattern, strs)
 }
 
-func childMatch(patterns Pattern, strs []string) (matched bool, err error) {
+func childMatch(pattern Pattern, strs []string) (matched bool, err error) {
-	if patterns[0].pattern != "/" {
+	if pattern.parts[0].pattern != "/" {
 		// relative pattern can always be nested down
 		return true, nil
 	}
 
-	ok, pos := hasDoubleWildcard(patterns)
+	ok, pos := hasDoubleWildcard(pattern)
 	if ok && len(strs) >= pos {
 		// cut off at the double wildcard
 		strs = strs[:pos]
@@ -116,16 +125,16 @@ func childMatch(patterns Pattern, strs []string) (matched bool, err error) {
 
 	// match path against absolute pattern prefix
 	l := 0
-	if len(strs) > len(patterns) {
+	if len(strs) > len(pattern.parts) {
-		l = len(patterns)
+		l = len(pattern.parts)
 	} else {
 		l = len(strs)
 	}
-	return match(patterns[0:l], strs)
+	return match(Pattern{pattern.parts[0:l], pattern.isNegated}, strs)
 }
 
 func hasDoubleWildcard(list Pattern) (ok bool, pos int) {
-	for i, item := range list {
+	for i, item := range list.parts {
 		if item.pattern == "" {
 			return true, i
 		}
@@ -134,22 +143,22 @@ func hasDoubleWildcard(list Pattern) (ok bool, pos int) {
 	return false, 0
 }
 
-func match(patterns Pattern, strs []string) (matched bool, err error) {
+func match(pattern Pattern, strs []string) (matched bool, err error) {
-	if ok, pos := hasDoubleWildcard(patterns); ok {
+	if ok, pos := hasDoubleWildcard(pattern); ok {
 		// gradually expand '**' into separate wildcards
-		newPat := make(Pattern, len(strs))
+		newPat := make([]patternPart, len(strs))
 		// copy static prefix once
-		copy(newPat, patterns[:pos])
+		copy(newPat, pattern.parts[:pos])
-		for i := 0; i <= len(strs)-len(patterns)+1; i++ {
+		for i := 0; i <= len(strs)-len(pattern.parts)+1; i++ {
 			// limit to static prefix and already appended '*'
 			newPat := newPat[:pos+i]
 			// in the first iteration the wildcard expands to nothing
 			if i > 0 {
 				newPat[pos+i-1] = patternPart{"*", false}
 			}
-			newPat = append(newPat, patterns[pos+1:]...)
+			newPat = append(newPat, pattern.parts[pos+1:]...)
 
-			matched, err := match(newPat, strs)
+			matched, err := match(Pattern{newPat, pattern.isNegated}, strs)
 			if err != nil {
 				return false, err
 			}
@@ -162,20 +171,20 @@ func match(patterns Pattern, strs []string) (matched bool, err error) {
 		return false, nil
 	}
 
-	if len(patterns) == 0 && len(strs) == 0 {
+	if len(pattern.parts) == 0 && len(strs) == 0 {
 		return true, nil
 	}
 
 	// an empty pattern never matches a non-empty path
-	if len(patterns) == 0 {
+	if len(pattern.parts) == 0 {
 		return false, nil
 	}
 
-	if len(patterns) <= len(strs) {
+	if len(pattern.parts) <= len(strs) {
 		minOffset := 0
-		maxOffset := len(strs) - len(patterns)
+		maxOffset := len(strs) - len(pattern.parts)
 		// special case absolute patterns
-		if patterns[0].pattern == "/" {
+		if pattern.parts[0].pattern == "/" {
 			maxOffset = 0
 		} else if strs[0] == "/" {
 			// skip absolute path marker if pattern is not rooted
@@ -184,12 +193,12 @@ func match(patterns Pattern, strs []string) (matched bool, err error) {
 	outer:
 		for offset := maxOffset; offset >= minOffset; offset-- {
 
-			for i := len(patterns) - 1; i >= 0; i-- {
+			for i := len(pattern.parts) - 1; i >= 0; i-- {
 				var ok bool
-				if patterns[i].isSimple {
+				if pattern.parts[i].isSimple {
-					ok = patterns[i].pattern == strs[offset+i]
+					ok = pattern.parts[i].pattern == strs[offset+i]
 				} else {
-					ok, err = filepath.Match(patterns[i].pattern, strs[offset+i])
+					ok, err = filepath.Match(pattern.parts[i].pattern, strs[offset+i])
 					if err != nil {
 						return false, errors.Wrap(err, "Match")
 					}
@@ -208,9 +217,9 @@ func match(patterns Pattern, strs []string) (matched bool, err error) {
 }
 
 // ParsePatterns prepares a list of patterns for use with List.
-func ParsePatterns(patterns []string) []Pattern {
+func ParsePatterns(pattern []string) []Pattern {
 	patpat := make([]Pattern, 0)
-	for _, pat := range patterns {
+	for _, pat := range pattern {
 		if pat == "" {
 			continue
 		}
@@ -232,7 +241,9 @@ func ListWithChild(patterns []Pattern, str string) (matched bool, childMayMatch
 	return list(patterns, true, str)
 }
 
-// List returns true if str matches one of the patterns. Empty patterns are ignored.
+// list returns true if str matches one of the patterns. Empty patterns are ignored.
+// Patterns prefixed by "!" are negated: any matching file excluded by a previous pattern
+// will become included again.
 func list(patterns []Pattern, checkChildMatches bool, str string) (matched bool, childMayMatch bool, err error) {
 	if len(patterns) == 0 {
 		return false, false, nil
@@ -242,6 +253,12 @@ func list(patterns []Pattern, checkChildMatches bool, str string) (matched bool,
 	if err != nil {
 		return false, false, err
 	}
+
+	hasNegatedPattern := false
+	for _, pat := range patterns {
+		hasNegatedPattern = hasNegatedPattern || pat.isNegated
+	}
+
 	for _, pat := range patterns {
 		m, err := match(pat, strs)
 		if err != nil {
@@ -258,11 +275,17 @@ func list(patterns []Pattern, checkChildMatches bool, str string) (matched bool,
 			c = true
 		}
 
+		if pat.isNegated {
+			matched = matched && !m
+			childMayMatch = childMayMatch && !m
+		} else {
 			matched = matched || m
 			childMayMatch = childMayMatch || c
 
-		if matched && childMayMatch {
+			if matched && childMayMatch && !hasNegatedPattern {
-			return true, true, nil
+				// without negative patterns the result cannot change any more
+				break
+			}
 		}
 	}
 

internal/filter/filter_test.go

@@ -248,6 +248,7 @@ var filterListTests = []struct {
 }{
 	{[]string{}, "/foo/bar/test.go", false, false},
 	{[]string{"*.go"}, "/foo/bar/test.go", true, true},
+	{[]string{"*.go"}, "/foo/bar", false, true},
 	{[]string{"*.c"}, "/foo/bar/test.go", false, true},
 	{[]string{"*.go", "*.c"}, "/foo/bar/test.go", true, true},
 	{[]string{"*"}, "/foo/bar/test.go", true, true},
@@ -255,8 +256,25 @@ var filterListTests = []struct {
 	{[]string{"?"}, "/foo/bar/test.go", false, true},
 	{[]string{"?", "x"}, "/foo/bar/x", true, true},
 	{[]string{"/*/*/bar/test.*"}, "/foo/bar/test.go", false, false},
+	{[]string{"/*/*/bar/test.*"}, "/foo/bar/bar", false, true},
 	{[]string{"/*/*/bar/test.*", "*.go"}, "/foo/bar/test.go", true, true},
 	{[]string{"", "*.c"}, "/foo/bar/test.go", false, true},
+	{[]string{"!**", "*.go"}, "/foo/bar/test.go", true, true},
+	{[]string{"!**", "*.c"}, "/foo/bar/test.go", false, true},
+	{[]string{"/foo/*/test.*", "!*.c"}, "/foo/bar/test.c", false, false},
+	{[]string{"/foo/*/test.*", "!*.c"}, "/foo/bar/test.go", true, true},
+	{[]string{"/foo/*/*", "!test.*", "*.c"}, "/foo/bar/test.go", false, true},
+	{[]string{"/foo/*/*", "!test.*", "*.c"}, "/foo/bar/test.c", true, true},
+	{[]string{"/foo/*/*", "!test.*", "*.c"}, "/foo/bar/file.go", true, true},
+	{[]string{"/**/*", "!/foo", "/foo/*", "!/foo/bar"}, "/foo/other/test.go", true, true},
+	{[]string{"/**/*", "!/foo", "/foo/*", "!/foo/bar"}, "/foo/bar", false, false},
+	{[]string{"/**/*", "!/foo", "/foo/*", "!/foo/bar"}, "/foo/bar/test.go", false, false},
+	{[]string{"/**/*", "!/foo", "/foo/*", "!/foo/bar"}, "/foo/bar/test.go/child", false, false},
+	{[]string{"/**/*", "!/foo", "/foo/*", "!/foo/bar", "/foo/bar/test*"}, "/foo/bar/test.go/child", true, true},
+	{[]string{"/foo/bar/*"}, "/foo", false, true},
+	{[]string{"/foo/bar/*", "!/foo/bar/[a-m]*"}, "/foo", false, true},
+	{[]string{"/foo/**/test.c"}, "/foo/bar/foo/bar/test.c", true, true},
+	{[]string{"/foo/*/test.c"}, "/foo/bar/foo/bar/test.c", false, false},
 }
 
 func TestList(t *testing.T) {