Updated TODO

README.md

@@ -2,8 +2,8 @@
 
 Kleingrep is a regular expression engine, providing a library and command-line tool written in Go.
 
-It aims to provide a more featureful engine, compared to the one in
-[Go's standard library](https://pkg.go.dev/regexp), while retaining some semblance of efficiency.
+It aims to provide a more featureful engine, compared to the one in Go's
+[regexp](https://pkg.go.dev/regexp), while retaining some semblance of efficiency.
 
 The engine does __not__ use backtracking, relying on the NFA-based method described in
 [Russ Cox's articles](https://swtch.com/~rsc/regexp). As such, it is immune to catastrophic backtracking.

cmd/main.go

@@ -64,18 +64,30 @@ func main() {
 	// 2. Build NFA from postfix representation (Thompson's algorithm)
 	// 3. Run the string against the NFA
 
-	if len(flag.Args()) != 1 { // flag.Args() also strips out program name
+	if len(flag.Args()) < 1 || len(flag.Args()) > 2 { // flag.Args() also strips out program name
 		fmt.Println("ERROR: Missing cmdline args")
 		os.Exit(22)
 	}
 	var re string
 	re = flag.Args()[0]
+	var inputFile *os.File
+	if len(flag.Args()) == 1 || flag.Args()[1] == "-" { // Either no file argument, or file argument is "-"
+		inputFile = os.Stdin
+	} else {
+		var err error
+		inputFile, err = os.Open(flag.Args()[1])
+		if err != nil {
+			fmt.Printf("%s: No such file or directory\n", flag.Args()[1])
+			os.Exit(2)
+		}
+	}
+
 	var test_str string
 	var err error
 	var linesRead bool // Whether or not we have read the lines in the file
 	lineNum := 0       // Current line number
 	// Create reader for stdin and writer for stdout
-	reader := bufio.NewReader(os.Stdin)
+	reader := bufio.NewReader(inputFile)
 	out := bufio.NewWriter(os.Stdout)
 
 	regComp, err := reg.Compile(re, flagsToCompile...)
@@ -204,10 +216,12 @@ func main() {
 					color.New(color.FgRed).Fprintf(out, "%c", c)
 					// Newline after every match - only if -o is enabled and -v is disabled.
 					if *onlyFlag && !(*invertFlag) {
-						for _, idx := range matchIndices {
-							if i+1 == idx[0].EndIdx { // End index is one more than last index of match
-								fmt.Fprintf(out, "\n")
-								break
+						for matchIdxNum, idx := range matchIndices {
+							if matchIdxNum < len(matchIndices)-1 { // Only print a newline afte printing a match, if there are multiple matches on the line, and we aren't on the last one. This is because the newline that gets added at the end will take care of that.
+								if i+1 == idx[0].EndIdx { // End index is one more than last index of match
+									fmt.Fprintf(out, "\n")
+									break
+								}
 							}
 						}
 					}
@@ -222,6 +236,10 @@ func main() {
 		if err != nil {
 			panic(err)
 		}
-		fmt.Println()
+		// If the last character in the string wasn't a newline, AND we either have don't -o set or we do (and we've matched something), then print a newline
+		if (len(test_str_runes) > 0 && test_str_runes[len(test_str_runes)-1] != '\n') &&
+			(!*onlyFlag || indicesToPrint.len() > 0) {
+			fmt.Println()
+		}
 	}
 }

cmd/unique_array.go

@@ -36,3 +36,7 @@ func (s uniq_arr[T]) values() []T {
 	}
 	return toRet
 }
+
+func (s uniq_arr[T]) len() int {
+	return len(s.backingMap)
+}

regex/compile.go

@@ -410,10 +410,10 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 				if i >= len(re_runes) {
 					return nil, fmt.Errorf("unclosed lookaround")
 				}
-				if re_runes[i] == '(' || re_runes[i] == nonCapLparenRune {
+				if (re_runes[i] == '(' && re_runes[i-1] != '\\') || re_runes[i] == nonCapLparenRune {
 					numOpenParens++
 				}
-				if re_runes[i] == ')' {
+				if re_runes[i] == ')' && re_runes[i-1] != '\\' {
 					numOpenParens--
 					if numOpenParens == 0 {
 						break
@@ -498,7 +498,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 				}
 			} else if re_postfix[i] == 'p' || re_postfix[i] == 'P' {
 				charClassInverted := (re_postfix[i] == 'P')
-				charsInClass := []rune{}
+				var charsInClass []rune
 				i++
 				if isUnicodeCharClassLetter(re_postfix[i]) {
 					var err error
@@ -589,10 +589,10 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 				if i >= len(re_postfix) {
 					return nil, fmt.Errorf("unclosed lookaround")
 				}
-				if re_postfix[i] == '(' || re_postfix[i] == nonCapLparenRune {
+				if (re_postfix[i] == '(' && re_postfix[i-1] != '\\') || re_postfix[i] == nonCapLparenRune {
 					numOpenParens++
 				}
-				if re_postfix[i] == ')' {
+				if re_postfix[i] == ')' && re_postfix[i-1] != '\\' {
 					numOpenParens--
 					if numOpenParens == 0 {
 						break
@@ -713,7 +713,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 						}
 					} else if re_postfix[i] == 'p' || re_postfix[i] == 'P' {
 						charClassInverted := (re_postfix[i] == 'P')
-						charsInList := []rune{}
+						var charsInList []rune
 						i++
 						if isUnicodeCharClassLetter(re_postfix[i]) {
 							var err error

regex/doc.go

@@ -161,6 +161,7 @@ The following features from [regexp] are (currently) NOT supported:
  2. Negated POSIX classes
  3. Embedded flags (flags are instead passed as arguments to [Compile])
  4. Literal text with \Q ... \E
+ 5. Finite repetition with no start (defaulting at 0)
 
 The following features are not available in [regexp], but are supported in my engine:
  1. Lookarounds

regex/nfa.go

@@ -45,11 +45,11 @@ type nfaState struct {
 	groupBegin                 bool       // Whether or not the node starts a capturing group
 	groupEnd                   bool       // Whether or not the node ends a capturing group
 	groupNum                   int        // Which capturing group the node starts / ends
+	isBackreference            bool       // Whether or not current node is backreference
+	referredGroup              int        // If current node is a backreference, the node that it points to
 	// The following properties depend on the current match - I should think about resetting them for every match.
-	threadGroups    []Group // Assuming that a state is part of a 'thread' in the matching process, this array stores the indices of capturing groups in the current thread. As matches are found for this state, its groups will be copied over.
-	isBackreference bool    // Whether or not current node is backreference
-	referredGroup   int     // If current node is a backreference, the node that it points to
-	threadBackref   int     // If current node is a backreference, how many characters to look forward into the referred group
+	threadGroups  []Group // Assuming that a state is part of a 'thread' in the matching process, this array stores the indices of capturing groups in the current thread. As matches are found for this state, its groups will be copied over.
+	threadBackref int     // If current node is a backreference, how many characters to look forward into the referred group
 }
 
 // Clones the NFA starting from the given state.
@@ -86,6 +86,8 @@ func cloneStateHelper(stateToClone *nfaState, cloneMap map[*nfaState]*nfaState)
 		groupEnd:        stateToClone.groupEnd,
 		groupBegin:      stateToClone.groupBegin,
 		groupNum:        stateToClone.groupNum,
+		isBackreference: stateToClone.isBackreference,
+		referredGroup:   stateToClone.referredGroup,
 	}
 	cloneMap[stateToClone] = clone
 	for i, s := range stateToClone.output {

regex/re_test.go

@@ -117,6 +117,7 @@ var reTests = []struct {
 	{`\d{3,4}`, nil, "ababab555", []Group{{6, 9}}},
 	{`\bpaint\b`, nil, "paints", []Group{}},
 	{`\b\w{5}\b`, nil, "paint", []Group{{0, 5}}},
+	{`\w{}`, nil, "test", nil},
 	{`[^\w]`, nil, "abcdef1230[]qq';;'", []Group{{10, 11}, {11, 12}, {14, 15}, {15, 16}, {16, 17}, {17, 18}}},
 	{`[^\W]`, nil, "abcdef1230[]qq';;'", []Group{{0, 1}, {1, 2}, {2, 3}, {3, 4}, {4, 5}, {5, 6}, {6, 7}, {7, 8}, {8, 9}, {9, 10}, {12, 13}, {13, 14}}},
 	{`[\[\]]`, nil, "a[b[l]]", []Group{{1, 2}, {3, 4}, {5, 6}, {6, 7}}},
@@ -545,6 +546,22 @@ var reTests = []struct {
 	{`\pN+`, nil, `123abc456def`, []Group{{0, 3}, {6, 9}}},
 	{`\PN+`, nil, `123abc456def`, []Group{{3, 6}, {9, 12}}},
 	{`[\p{Greek}\p{Cyrillic}]`, nil, `ΣωШД`, []Group{{0, 1}, {1, 2}, {2, 3}, {3, 4}}},
+
+	{`(?<=\().*?(?=\))`, nil, `(abc)`, []Group{{1, 4}}},
+
+	{`((a|b)\2)`, nil, `aa`, []Group{{0, 2}}},
+	{`((a|b)\2)`, nil, `bb`, []Group{{0, 2}}},
+	{`((a|b)\2)`, nil, `ab`, []Group{}},
+	{`((a|b)\2)`, nil, `ba`, []Group{}},
+
+	{`((a|b)\2){3}`, nil, `aaaaaa`, []Group{{0, 6}}},
+	{`((a|b)\2){3}`, nil, `bbbbbb`, []Group{{0, 6}}},
+	{`((a|b)\2){3}`, nil, `bbaaaa`, []Group{{0, 6}}},
+	{`((a|b)\2){3}`, nil, `aabbaa`, []Group{{0, 6}}},
+	{`((a|b)\2){3}`, nil, `aaaabb`, []Group{{0, 6}}},
+	{`((a|b)\2){3}`, nil, `bbaabb`, []Group{{0, 6}}},
+	{`((a|b)\2){3}`, nil, `baabab`, []Group{}},
+	{`((a|b)\2){3}`, nil, `bbabab`, []Group{}},
 }
 
 var groupTests = []struct {

regex/todo.txt

@@ -4,4 +4,5 @@
 Ideas for flags:
     -m <num> : Print <num>th match (-m 1 = first match, -m 2 = second match)
     -g <num> : Print the <num>th group
+    -r : Specify a directory instead of a file, reads recursively
 4. Refactor code for flags - make each flag's code a function, which modifies the result of findAllMatches