Added more documentation

Reviewed-on: #4

cmd/main.go

@@ -129,6 +129,8 @@ func main() {
 			matchIndices = regComp.FindAllSubmatch(test_str)
 		}
 
+		test_str_runes := []rune(test_str) // Converting to runes preserves unicode characters
+
 		if *printMatchesFlag {
 			// if we are in single line mode, print the line on which
 			// the matches occur
@@ -158,10 +160,10 @@ func main() {
 			oldIndices := indicesToPrint.values()
 			indicesToPrint = new_uniq_arr[int]()
 			// Explanation:
-			// Find all numbers from 0 to len(test_str) that are NOT in oldIndices.
+			// Find all numbers from 0 to len(test_str_runes) that are NOT in oldIndices.
 			// These are the values we want to print, now that we have inverted the match.
 			// Re-initialize indicesToPrint and add all of these values to it.
-			indicesToPrint.add(setDifference(genRange(0, len(test_str)), oldIndices)...)
+			indicesToPrint.add(setDifference(genRange(0, len(test_str_runes)), oldIndices)...)
 
 		}
 		// If lineFlag is enabled, we should only print something if:
@@ -182,7 +184,7 @@ func main() {
 		//			the corresponding end index.
 		// 		3. If not, just print the character.
 		if substituteFlagEnabled {
-			for i := range test_str {
+			for i := range test_str_runes {
 				inMatchIndex := false
 				for _, m := range matchIndices {
 					if i == m[0].StartIdx {
@@ -193,11 +195,11 @@ func main() {
 					}
 				}
 				if !inMatchIndex {
-					fmt.Fprintf(out, "%c", test_str[i])
+					fmt.Fprintf(out, "%c", test_str_runes[i])
 				}
 			}
 		} else {
-			for i, c := range test_str {
+			for i, c := range test_str_runes {
 				if indicesToPrint.contains(i) {
 					color.New(color.FgRed).Fprintf(out, "%c", c)
 					// Newline after every match - only if -o is enabled and -v is disabled.

regex/compile.go

@@ -108,6 +108,48 @@ func getPOSIXClass(str []rune) (bool, string) {
 	return true, rtv
 }
 
+// isUnicodeCharClassLetter returns whether or not the given letter represents a unicode character class.
+func isUnicodeCharClassLetter(c rune) bool {
+	return slices.Contains([]rune{'L', 'M', 'S', 'N', 'P', 'C', 'Z'}, c)
+}
+
+// rangeTableToRuneSlice converts the given range table into a rune slice and returns it.
+func rangeTableToRuneSlice(rangetable *unicode.RangeTable) []rune {
+	var rtv []rune
+	for _, r := range rangetable.R16 {
+		for c := r.Lo; c <= r.Hi; c += r.Stride {
+			rtv = append(rtv, rune(c))
+		}
+	}
+	for _, r := range rangetable.R32 {
+		for c := r.Lo; c <= r.Hi; c += r.Stride {
+			rtv = append(rtv, rune(c))
+		}
+	}
+	return rtv
+}
+
+// unicodeCharClassToRange converts the given unicode character class name into a list of characters in that class.
+// This class could also be a single letter eg. 'C'.
+func unicodeCharClassToRange(class string) ([]rune, error) {
+	if len(class) == 0 {
+		return nil, fmt.Errorf("empty unicode character class")
+	}
+	if len(class) == 1 || len(class) == 2 {
+		if rangeTable, ok := unicode.Categories[class]; ok {
+			return rangeTableToRuneSlice(rangeTable), nil
+		} else {
+			return nil, fmt.Errorf("invalid short unicode character class")
+		}
+	} else {
+		if rangeTable, ok := unicode.Scripts[class]; ok {
+			return rangeTableToRuneSlice(rangeTable), nil
+		} else {
+			return nil, fmt.Errorf("invalid long unicode character class")
+		}
+	}
+}
+
 // Stores whether the case-insensitive flag has been enabled.
 var caseInsensitive bool
 
@@ -309,17 +351,44 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 					}
 				} else if isHex(re_runes[i]) {
 					re_postfix = append(re_postfix, re_runes[i:i+2]...)
-					i += 2
+					i += 1 // I don't skip forward 2 steps, because the second step will happen with the loop increment
 				} else {
 					return nil, fmt.Errorf("invalid hex value in expression")
 				}
-			} else if isOctal(re_runes[i]) {
+			} else if re_runes[i] == 'p' || re_runes[i] == 'P' { // Unicode character class (P is negated unicode charclass)
+				re_postfix = append(re_postfix, re_runes[i])
+				i++
+				if i >= len(re_runes) {
+					return nil, fmt.Errorf("error parsing unicode character class in expression")
+				}
+				if re_runes[i] == '{' { // Full name charclass
+					for re_runes[i] != '}' {
+						re_postfix = append(re_postfix, re_runes[i])
+						i++
+					}
+					re_postfix = append(re_postfix, re_runes[i])
+					i++
+				} else if isUnicodeCharClassLetter(re_runes[i]) {
+					re_postfix = append(re_postfix, re_runes[i])
+					i++
+				} else {
+					return nil, fmt.Errorf("error parsing unicode character class in expression")
+				}
+				i-- // The loop increment at the top will move us forward
+			} else if re_runes[i] == '0' { // Start of octal value
 				numDigits := 1
-				for i+numDigits < len(re_runes) && numDigits < 3 && isOctal(re_runes[i+numDigits]) { // Skip while we see an octal character (max of 3)
+				for i+numDigits < len(re_runes) && numDigits < 4 && isOctal(re_runes[i+numDigits]) { // Skip while we see an octal character (max of 4, starting with 0)
 					numDigits++
 				}
 				re_postfix = append(re_postfix, re_runes[i:i+numDigits]...)
 				i += (numDigits - 1) // I have to move back a step, so that I can add a concatenation operator if necessary, and so that the increment at the bottom of the loop works as intended
+			} else if unicode.IsDigit(re_runes[i]) { // Any other number - backreference
+				numDigits := 1
+				for i+numDigits < len(re_runes) && unicode.IsDigit(re_runes[i+numDigits]) { // Skip while we see a digit
+					numDigits++
+				}
+				re_postfix = append(re_postfix, re_runes[i:i+numDigits]...)
+				i += (numDigits - 1) // Move back a step to add concatenation operator
 			} else {
 				re_postfix = append(re_postfix, re_runes[i])
 			}
@@ -364,7 +433,9 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 	outQueue := make([]postfixNode, 0) // Output queue
 
 	// Actual algorithm
-	numOpenParens := 0 // Number of open parentheses
+	numOpenParens := 0                               // Number of open parentheses
+	parenIndices := make([]Group, 0)                 // I really shouldn't be using Group here, because that's strictly for matching purposes, but its a convenient way to store the indices of the opening and closing parens.
+	parenIndices = append(parenIndices, Group{0, 0}) // I append a weird value here, because the 0-th group doesn't have any parens. This way, the 1st group will be at index 1, 2nd at 2 ...
 	for i := 0; i < len(re_postfix); i++ {
 		/* Two cases:
 		1. Current character is alphanumeric - send to output queue
@@ -420,11 +491,44 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 				} else {
 					return nil, fmt.Errorf("not enough hex characters found in expression")
 				}
-			} else if isOctal(re_postfix[i]) { // Octal value
+			} else if re_postfix[i] == 'p' || re_postfix[i] == 'P' {
+				charClassInverted := (re_postfix[i] == 'P')
+				charsInClass := []rune{}
+				i++
+				if isUnicodeCharClassLetter(re_postfix[i]) {
+					var err error
+					charsInClass, err = unicodeCharClassToRange(string(re_postfix[i]))
+					if err != nil {
+						return nil, err
+					}
+				} else if re_postfix[i] == '{' {
+					i++ // Skip opening bracket
+					unicodeCharClassStr := ""
+					for re_postfix[i] != '}' {
+						unicodeCharClassStr += string(re_postfix[i])
+						i++
+					}
+					var err error
+					charsInClass, err = unicodeCharClassToRange(unicodeCharClassStr)
+					if err != nil {
+						return nil, err
+					}
+				} else {
+					return nil, fmt.Errorf("error parsing unicode character class in expression")
+				}
+				var toAppend postfixNode
+				if !charClassInverted { // \p
+					toAppend = newPostfixNode(charsInClass...)
+				} else { // \P
+					toAppend = newPostfixDotNode()
+					toAppend.except = append([]postfixNode{}, newPostfixNode(charsInClass...))
+				}
+				outQueue = append(outQueue, toAppend)
+			} else if re_postfix[i] == '0' { // Octal value
 				var octVal int64
 				var octValStr string
 				numDigitsParsed := 0
-				for (i+numDigitsParsed) < len(re_postfix) && isOctal(re_postfix[i+numDigitsParsed]) && numDigitsParsed <= 3 {
+				for (i+numDigitsParsed) < len(re_postfix) && isOctal(re_postfix[i+numDigitsParsed]) && numDigitsParsed <= 4 {
 					octValStr += string(re_postfix[i+numDigitsParsed])
 					numDigitsParsed++
 				}
@@ -437,6 +541,20 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 				}
 				i += numDigitsParsed - 1 // Shift forward by the number of digits that were parsed. Move back one character, because the loop increment will move us back to the next character automatically
 				outQueue = append(outQueue, newPostfixCharNode(rune(octVal)))
+			} else if unicode.IsDigit(re_postfix[i]) { // Backreference
+				var num int64
+				var numStr string
+				numDigitsParsed := 0
+				for (i+numDigitsParsed) < len(re_postfix) && unicode.IsDigit(re_postfix[i+numDigitsParsed]) {
+					numStr += string(re_postfix[i+numDigitsParsed])
+					numDigitsParsed++
+				}
+				num, err := strconv.ParseInt(numStr, 10, 32)
+				if err != nil {
+					return nil, fmt.Errorf("error parsing backreference in expresion")
+				}
+				i += numDigitsParsed - 1
+				outQueue = append(outQueue, newPostfixBackreferenceNode(int(num)))
 			} else {
 				escapedNode, err := newEscapedNode(re_postfix[i], false)
 				if err != nil {
@@ -588,11 +706,44 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 						} else {
 							return nil, fmt.Errorf("not enough hex characters found in character class")
 						}
-					} else if isOctal(re_postfix[i]) { // Octal value
+					} else if re_postfix[i] == 'p' || re_postfix[i] == 'P' {
+						charClassInverted := (re_postfix[i] == 'P')
+						charsInList := []rune{}
+						i++
+						if isUnicodeCharClassLetter(re_postfix[i]) {
+							var err error
+							charsInList, err = unicodeCharClassToRange(string(re_postfix[i]))
+							if err != nil {
+								return nil, err
+							}
+						} else if re_postfix[i] == '{' {
+							i++ // Skip opening bracket
+							unicodeCharClassStr := ""
+							for re_postfix[i] != '}' {
+								unicodeCharClassStr += string(re_postfix[i])
+								i++
+							}
+							var err error
+							charsInList, err = unicodeCharClassToRange(unicodeCharClassStr)
+							if err != nil {
+								return nil, err
+							}
+						} else {
+							return nil, fmt.Errorf("error parsing unicode character class in expression")
+						}
+						if !charClassInverted {
+							chars = append(chars, newPostfixNode(charsInList...))
+						} else {
+							toAppend := newPostfixDotNode()
+							toAppend.except = append([]postfixNode{}, newPostfixNode(charsInList...))
+							chars = append(chars, toAppend)
+						}
+					} else if re_postfix[i] == '0' { // Octal value
+
 						var octVal int64
 						var octValStr string
 						numDigitsParsed := 0
-						for (i+numDigitsParsed) < len(re_postfix)-1 && isOctal(re_postfix[i+numDigitsParsed]) && numDigitsParsed <= 3 { // The '-1' exists, because even in the worst case (the character class extends till the end), the last character must be a closing bracket (and nothing else)
+						for (i+numDigitsParsed) < len(re_postfix)-1 && isOctal(re_postfix[i+numDigitsParsed]) && numDigitsParsed <= 4 { // The '-1' exists, because even in the worst case (the character class extends till the end), the last character must be a closing bracket (and nothing else)
 							octValStr += string(re_postfix[i+numDigitsParsed])
 							numDigitsParsed++
 						}
@@ -796,6 +947,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 				outQueue = append(outQueue, newPostfixNode(c))
 			}
 			numOpenParens++
+			parenIndices = append(parenIndices, Group{StartIdx: len(outQueue) - 1}) // Push the index of the lparen into parenIndices
 		}
 		if c == ')' {
 			// Keep popping from opStack until we encounter an opening parantheses or a NONCAPLPAREN_CHAR. Throw error if we reach the end of the stack.
@@ -812,6 +964,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 			if val == '(' {       // Whatever was inside the parentheses was a _capturing_ group, so we append the closing parentheses as well
 				outQueue = append(outQueue, newPostfixNode(')')) // Add closing parentheses
 			}
+			parenIndices[numOpenParens].EndIdx = len(outQueue) - 1
 			numOpenParens--
 		}
 	}
@@ -826,6 +979,11 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 		return nil, fmt.Errorf("imbalanced parantheses")
 	}
 
+	//	outQueue, _, err := rewriteBackreferences(outQueue, parenIndices)
+	//	if err != nil {
+	//		return nil, err
+	//	}
+
 	return outQueue, nil
 }
 
@@ -1037,6 +1195,21 @@ func thompson(re []postfixNode) (Reg, error) {
 			})
 			nfa = append(nfa, toAdd)
 		}
+		if c.nodetype == backreferenceNode {
+			if c.referencedGroup > numGroups {
+				return Reg{}, fmt.Errorf("invalid backreference")
+			}
+			stateToAdd := &nfaState{}
+			stateToAdd.assert = noneAssert
+			stateToAdd.content = newContents(epsilon)
+			stateToAdd.isEmpty = true
+			stateToAdd.isBackreference = true
+			stateToAdd.output = make([]*nfaState, 0)
+			stateToAdd.output = append(stateToAdd.output, stateToAdd)
+			stateToAdd.referredGroup = c.referencedGroup
+			stateToAdd.threadBackref = 0
+			nfa = append(nfa, stateToAdd)
+		}
 		// Must be an operator if it isn't a character
 		switch c.nodetype {
 		case concatenateNode:

regex/doc.go

@@ -18,7 +18,7 @@ Single characters:
 	[^abc]			Negated character class - match any character except a, b and c
 	[^a-z]			Negated character range - do not match any character from a to z
 	\[				Match a literal '['. Backslashes can escape any character with special meaning, including another backslash.
-	\452			Match the character with the octal value 452 (up to 3 digits)
+	\0452			Match the character with the octal value 452 (up to 4 digits, first digit must be 0)
 	\xFF			Match the character with the hex value FF (exactly 2 characters)
 	\x{0000FF}		Match the character with the hex value 0000FF (exactly 6 characters)
 	\n				Newline
@@ -93,6 +93,10 @@ Lookarounds:
 	(?<=x)y			Positive lookbehind - Match y if preceded by x
 	(?<!x)y			Negative lookbehind - Match y if NOT preceded by x
 
+Backreferences:
+
+	(xy)\1			Match 'xy' followed by the text most recently captured by group 1 (in this case, 'xy')
+
 Numeric ranges:
 
 	<x-y>			Match any number from x to y (inclusive) (x and y must be positive numbers)
@@ -149,13 +153,14 @@ returns the 0-group.
 The following features from [regexp] are (currently) NOT supported:
  1. Named capturing groups
  2. Non-greedy operators
- 3. Unicode character classes
+ 3. Negated POSIX classes
  4. Embedded flags (flags are instead passed as arguments to [Compile])
  5. Literal text with \Q ... \E
 
 The following features are not available in [regexp], but are supported in my engine:
  1. Lookarounds
  2. Numeric ranges
+ 3. Backreferences
 
 I hope to shorten the first list, and expand the second.
 */

regex/matching.go

@@ -236,14 +236,14 @@ func addStateToList(str []rune, idx int, list []nfaState, state nfaState, thread
 
 	if state.isKleene || state.isQuestion {
 		copyThread(state.splitState, state)
-		list = addStateToList(str, idx, list, *state.splitState, threadGroups, visited, preferLongest)
+		list := addStateToList(str, idx, list, *state.splitState, threadGroups, visited, preferLongest)
 		copyThread(state.next, state)
 		list = addStateToList(str, idx, list, *state.next, threadGroups, visited, preferLongest)
 		return list
 	}
 	if state.isAlternation {
 		copyThread(state.next, state)
-		list = addStateToList(str, idx, list, *state.next, threadGroups, visited, preferLongest)
+		list := addStateToList(str, idx, list, *state.next, threadGroups, visited, preferLongest)
 		copyThread(state.splitState, state)
 		list = addStateToList(str, idx, list, *state.splitState, threadGroups, visited, preferLongest)
 		return list
@@ -257,10 +257,12 @@ func addStateToList(str []rune, idx int, list []nfaState, state nfaState, thread
 	}
 	if state.groupBegin {
 		state.threadGroups[state.groupNum].StartIdx = idx
+		copyThread(state.next, state)
 		return addStateToList(str, idx, list, *state.next, state.threadGroups, visited, preferLongest)
 	}
 	if state.groupEnd {
 		state.threadGroups[state.groupNum].EndIdx = idx
+		copyThread(state.next, state)
 		return addStateToList(str, idx, list, *state.next, state.threadGroups, visited, preferLongest)
 	}
 	return append(list, state)
@@ -313,11 +315,25 @@ func findAllSubmatchHelper(start *nfaState, str []rune, offset int, numGroups in
 				if !preferLongest {
 					break
 				}
-			} else if !currentState.isAlternation && !currentState.isKleene && !currentState.isQuestion && !currentState.groupBegin && !currentState.groupEnd && currentState.assert == noneAssert { // Normal character
+			} else if !currentState.isAlternation && !currentState.isKleene && !currentState.isQuestion && !currentState.isBackreference && !currentState.groupBegin && !currentState.groupEnd && currentState.assert == noneAssert { // Normal character
 				if currentState.contentContains(str, idx, preferLongest) {
 					nextStates = addStateToList(str, idx+1, nextStates, *currentState.next, currentState.threadGroups, nil, preferLongest)
 				}
+			} else if currentState.isBackreference && currentState.threadGroups[currentState.referredGroup].IsValid() {
+				groupLength := currentState.threadGroups[currentState.referredGroup].EndIdx - currentState.threadGroups[currentState.referredGroup].StartIdx
+				if currentState.threadBackref == groupLength {
+					currentState.threadBackref = 0
+					copyThread(currentState.next, currentState)
+					currentStates = addStateToList(str, idx, currentStates, *currentState.next, currentState.threadGroups, nil, preferLongest)
+				} else {
+					idxInReferredGroup := currentState.threadGroups[currentState.referredGroup].StartIdx + currentState.threadBackref
+					if idxInReferredGroup < len(str) && idx < len(str) && str[idxInReferredGroup] == str[idx] {
+						currentState.threadBackref += 1
+						nextStates = append(nextStates, currentState)
+					}
+				}
 			}
+
 		}
 		currentStates = append([]nfaState{}, nextStates...)
 		nextStates = nil

regex/nfa.go

@@ -45,8 +45,10 @@ type nfaState struct {
 	groupEnd                   bool       // Whether or not the node ends a capturing group
 	groupNum                   int        // Which capturing group the node starts / ends
 	// The following properties depend on the current match - I should think about resetting them for every match.
-	zeroMatchFound bool    // Whether or not the state has been used for a zero-length match - only relevant for zero states
-	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.
+	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
 }
 
 // Clones the NFA starting from the given state.
@@ -76,7 +78,6 @@ func cloneStateHelper(stateToClone *nfaState, cloneMap map[*nfaState]*nfaState)
 		isQuestion:      stateToClone.isQuestion,
 		isAlternation:   stateToClone.isAlternation,
 		assert:          stateToClone.assert,
-		zeroMatchFound:  stateToClone.zeroMatchFound,
 		allChars:        stateToClone.allChars,
 		except:          append([]rune{}, stateToClone.except...),
 		lookaroundRegex: stateToClone.lookaroundRegex,
@@ -122,6 +123,7 @@ func resetThreadsHelper(state *nfaState, visitedMap map[*nfaState]bool) {
 	}
 	// Assuming it hasn't been visited
 	state.threadGroups = nil
+	state.threadBackref = 0
 	visitedMap[state] = true
 	if state.isAlternation {
 		resetThreadsHelper(state.next, visitedMap)
@@ -428,7 +430,8 @@ func (s nfaState) equals(other nfaState) bool {
 		s.groupBegin == other.groupBegin &&
 		s.groupEnd == other.groupEnd &&
 		s.groupNum == other.groupNum &&
-		slices.Equal(s.threadGroups, other.threadGroups)
+		slices.Equal(s.threadGroups, other.threadGroups) &&
+		s.threadBackref == other.threadBackref
 }
 
 func stateExists(list []nfaState, s nfaState) bool {

regex/postfixNode.go

@@ -1,6 +1,8 @@
 package regex
 
-import "fmt"
+import (
+	"fmt"
+)
 
 type nodeType int
 
@@ -20,6 +22,7 @@ const (
 	assertionNode
 	lparenNode
 	rparenNode
+	backreferenceNode
 )
 
 // Helper constants for lookarounds
@@ -31,15 +34,16 @@ const lookbehind = -1
 var infinite_reps int = -1 // Represents infinite reps eg. the end range in {5,}
 // This represents a node in the postfix representation of the expression
 type postfixNode struct {
-	nodetype       nodeType
-	contents       []rune        // Contents of the node
-	startReps      int           // Minimum number of times the node should be repeated - used with numeric specifiers
-	endReps        int           // Maximum number of times the node should be repeated - used with numeric specifiers
-	allChars       bool          // Whether or not the current node represents all characters (eg. dot metacharacter)
-	except         []postfixNode // For inverted character classes, we match every unicode character _except_ a few. In this case, allChars is true and the exceptions are placed here.
-	lookaroundSign int           // ONLY USED WHEN nodetype == ASSERTION. Whether we have a positive or negative lookaround.
-	lookaroundDir  int           // Lookbehind or lookahead
-	nodeContents   []postfixNode // ONLY USED WHEN nodetype == CHARCLASS. Holds all the nodes inside the given CHARCLASS node.
+	nodetype        nodeType
+	contents        []rune        // Contents of the node
+	startReps       int           // Minimum number of times the node should be repeated - used with numeric specifiers
+	endReps         int           // Maximum number of times the node should be repeated - used with numeric specifiers
+	allChars        bool          // Whether or not the current node represents all characters (eg. dot metacharacter)
+	except          []postfixNode // For inverted character classes, we match every unicode character _except_ a few. In this case, allChars is true and the exceptions are placed here.
+	lookaroundSign  int           // ONLY USED WHEN nodetype == ASSERTION. Whether we have a positive or negative lookaround.
+	lookaroundDir   int           // Lookbehind or lookahead
+	nodeContents    []postfixNode // ONLY USED WHEN nodetype == CHARCLASS. Holds all the nodes inside the given CHARCLASS node.
+	referencedGroup int           // ONLY USED WHEN nodetype == backreferenceNode. Holds the group which this one refers to. After parsing is done, the expression will be rewritten eg. (a)\1 will become (a)(a). So the return value of ShuntingYard() shouldn't contain a backreferenceNode.
 }
 
 // Converts the given list of postfixNodes to one node of type CHARCLASS.
@@ -208,3 +212,44 @@ func newPostfixCharNode(contents ...rune) postfixNode {
 	toReturn.contents = append(toReturn.contents, contents...)
 	return toReturn
 }
+
+// newPostfixBackreferenceNode creates and returns a backreference node, referring to the given group
+func newPostfixBackreferenceNode(referred int) postfixNode {
+	toReturn := postfixNode{}
+	toReturn.startReps = 1
+	toReturn.endReps = 1
+	toReturn.nodetype = backreferenceNode
+	toReturn.referencedGroup = referred
+	return toReturn
+}
+
+// rewriteBackreferences rewrites any backreferences in the given postfixNode slice, into their respective groups.
+// It stores the relation in a map, and returns it as the second return value.
+// It uses parenIndices to determine where a group starts and ends in nodes.
+// For example, \1(a) will be rewritten into (a)(a), and 1 -> 2 will be the hashmap value.
+// It returns an error if a backreference points to an invalid group.
+// func rewriteBackreferences(nodes []postfixNode, parenIndices []Group) ([]postfixNode, map[int]int, error) {
+// 	rtv := make([]postfixNode, 0)
+// 	referMap := make(map[int]int)
+// 	numGroups := 0
+// 	groupIncrement := 0 // If we have a backreference before the group its referring to, then the group its referring to will have its group number incremented.
+// 	for i, node := range nodes {
+// 		if node.nodetype == backreferenceNode {
+// 			if node.referencedGroup >= len(parenIndices) {
+// 				return nil, nil, fmt.Errorf("invalid backreference")
+// 			}
+// 			rtv = slices.Concat(rtv, nodes[parenIndices[node.referencedGroup].StartIdx:parenIndices[node.referencedGroup].EndIdx+1]) // Add all the nodes in the group to rtv
+// 			numGroups += 1
+// 			if i < parenIndices[node.referencedGroup].StartIdx {
+// 				groupIncrement += 1
+// 			}
+// 			referMap[numGroups] = node.referencedGroup + groupIncrement
+// 		} else {
+// 			rtv = append(rtv, node)
+// 			if node.nodetype == lparenNode {
+// 				numGroups += 1
+// 			}
+// 		}
+// 	}
+// 	return rtv, referMap, nil
+// }

regex/re_test.go

@@ -179,7 +179,7 @@ var reTests = []struct {
 	{"[[:graph:]]+", nil, "abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPRQSTUVWXYZ0123456789!@#$%^&*", []Group{{0, 70}}},
 
 	// Test cases from Python's RE test suite
-	{`[\1]`, nil, "\x01", []Group{{0, 1}}},
+	{`[\01]`, nil, "\x01", []Group{{0, 1}}},
 
 	{`\0`, nil, "\x00", []Group{{0, 1}}},
 	{`[\0a]`, nil, "\x00", []Group{{0, 1}}},
@@ -194,7 +194,7 @@ var reTests = []struct {
 	{`\x00ffffffffffffff`, nil, "\xff", []Group{}},
 	{`\x00f`, nil, "\x0f", []Group{}},
 	{`\x00fe`, nil, "\xfe", []Group{}},
-	{`^\w+=(\\[\000-\277]|[^\n\\])*`, nil, "SRC=eval.c g.c blah blah blah \\\\\n\tapes.c", []Group{{0, 32}}},
+	{`^\w+=(\\[\000-\0277]|[^\n\\])*`, nil, "SRC=eval.c g.c blah blah blah \\\\\n\tapes.c", []Group{{0, 32}}},
 
 	{`a.b`, nil, `acb`, []Group{{0, 3}}},
 	{`a.b`, nil, "a\nb", []Group{}},
@@ -312,11 +312,7 @@ var reTests = []struct {
 	{`a[-]?c`, nil, `ac`, []Group{{0, 2}}},
 	{`^(.+)?B`, nil, `AB`, []Group{{0, 2}}},
 	{`\0009`, nil, "\x009", []Group{{0, 2}}},
-	{`\141`, nil, "a", []Group{{0, 1}}},
-
-	// At this point, the python test suite has a bunch
-	// of backreference tests. Since my engine doesn't
-	// implement backreferences, I've skipped those tests.
+	{`\0141`, nil, "a", []Group{{0, 1}}},
 
 	{`*a`, nil, ``, nil},
 	{`(*)b`, nil, ``, nil},
@@ -433,7 +429,8 @@ var reTests = []struct {
 	{`a[-]?c`, []ReFlag{RE_CASE_INSENSITIVE}, `AC`, []Group{{0, 2}}},
 	{`^(.+)?B`, []ReFlag{RE_CASE_INSENSITIVE}, `ab`, []Group{{0, 2}}},
 	{`\0009`, []ReFlag{RE_CASE_INSENSITIVE}, "\x009", []Group{{0, 2}}},
-	{`\141`, []ReFlag{RE_CASE_INSENSITIVE}, "A", []Group{{0, 1}}},
+	{`\0141`, []ReFlag{RE_CASE_INSENSITIVE}, "A", []Group{{0, 1}}},
+	{`\0141\0141`, []ReFlag{RE_CASE_INSENSITIVE}, "AA", []Group{{0, 2}}},
 
 	{`a[-]?c`, []ReFlag{RE_CASE_INSENSITIVE}, `AC`, []Group{{0, 2}}},
 
@@ -464,8 +461,10 @@ var reTests = []struct {
 	{`[\D5]+`, nil, `1234abc5678`, []Group{{4, 8}}},
 	{`[\da-fA-F]+`, nil, `123abc`, []Group{{0, 6}}},
 	{`\xff`, nil, "\u00ff", []Group{{0, 1}}},
+	{`\xff+`, nil, "\u00ff\u00ff", []Group{{0, 2}}},
 	{`\xFF`, nil, "\u00ff", []Group{{0, 1}}},
 	{`\x00ff`, nil, "\u00ff", []Group{}},
+	{`\x{0000ff}+`, nil, "\u00ff\u00ff", []Group{{0, 2}}},
 	{`\x{0000ff}`, nil, "\u00ff", []Group{{0, 1}}},
 	{`\x{0000FF}`, nil, "\u00ff", []Group{{0, 1}}},
 	{"\t\n\v\r\f\a", nil, "\t\n\v\r\f\a", []Group{{0, 6}}},
@@ -473,7 +472,7 @@ var reTests = []struct {
 	{`[\t][\n][\v][\r][\f][\b]`, nil, "\t\n\v\r\f\b", []Group{{0, 6}}},
 	{`.*d`, nil, "abc\nabd", []Group{{4, 7}}},
 	{`(`, nil, "-", nil},
-	{`[\41]`, nil, `!`, []Group{{0, 1}}},
+	{`[\041]`, nil, `!`, []Group{{0, 1}}},
 	{`(?<!abc)(d.f)`, nil, `abcdefdof`, []Group{{6, 9}}},
 	{`[\w-]+`, nil, `laser_beam`, []Group{{0, 10}}},
 	{`M+`, []ReFlag{RE_CASE_INSENSITIVE}, `MMM`, []Group{{0, 3}}},
@@ -520,6 +519,14 @@ var reTests = []struct {
 	{`<389-400`, nil, `-`, nil},
 	{`<389-400>`, nil, `391`, []Group{{0, 3}}},
 	{`\b<1-10000>\b`, nil, `America declared independence in 1776.`, []Group{{33, 37}}},
+
+	{`\p{Tamil}+`, nil, `உயிரெழுத்து`, []Group{{0, 11}}}, // Each letter and matra is counted as a separate rune, so 'u', 'ya', 'e (matra), 'ra', 'e (matra)', 'zha', (oo (matra), 'tha', 'ith', 'tha', 'oo (matra)'.
+	{`\P{Tamil}+`, nil, `vowel=உயிரெழுத்து`, []Group{{0, 6}}},
+	{`\P`, nil, `உயிரெழுத்து`, nil},
+	{`\PM\pM*`, nil, `உயிரெழுத்து`, []Group{{0, 1}, {1, 3}, {3, 5}, {5, 7}, {7, 9}, {9, 11}}},
+	{`\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}}},
 }
 
 var groupTests = []struct {
@@ -581,13 +588,37 @@ var groupTests = []struct {
 	{`(bc+d$|ef*g.|h?i(j|k))`, nil, `bcdd`, []Match{}},
 	{`(bc+d$|ef*g.|h?i(j|k))`, nil, `reffgz`, []Match{[]Group{{1, 6}, {1, 6}}}},
 	{`(((((((((a)))))))))`, nil, `a`, []Match{[]Group{{0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}}}},
-	{`(((((((((a)))))))))\41`, nil, `a!`, []Match{[]Group{{0, 2}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}}}},
+	{`(((((((((a)))))))))\041`, nil, `a!`, []Match{[]Group{{0, 2}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}}}},
 	{`(.*)c(.*)`, nil, `abcde`, []Match{[]Group{{0, 5}, {0, 2}, {3, 5}}}},
 	{`\((.*), (.*)\)`, nil, `(a, b)`, []Match{[]Group{{0, 6}, {1, 2}, {4, 5}}}},
 
-	// At this point, the python test suite has a bunch
-	// of backreference tests. Since my engine doesn't
-	// implement backreferences, I've skipped those tests.
+	// Backreference tests
+	{`(abc)\1`, nil, `abcabc`, []Match{[]Group{{0, 6}, {0, 3}}}},
+	{`([a-c]+)\1`, nil, `abcabc`, []Match{[]Group{{0, 6}, {0, 3}}}},
+	{`([a-c]*)\1`, nil, `abcabc`, []Match{[]Group{{0, 6}, {0, 3}}, []Group{{6, 6}, {6, 6}}}},
+	{`^(.+)?B`, nil, `AB`, []Match{[]Group{{0, 2}, {0, 1}}}},
+	{`(a+).\1$`, nil, `aaaaa`, []Match{[]Group{{0, 5}, {0, 2}}}},
+	{`^(a+).\1$`, nil, `aaaa`, []Match{}},
+	{`(a)\1`, nil, `aa`, []Match{[]Group{{0, 2}, {0, 1}}}},
+	{`(a+)\1`, nil, `aa`, []Match{[]Group{{0, 2}, {0, 1}}}},
+	{`(a+)+\1`, nil, `aa`, []Match{[]Group{{0, 2}, {0, 1}}}},
+	{`(a).+\1`, nil, `aba`, []Match{[]Group{{0, 3}, {0, 1}}}},
+	{`(a)ba*\1`, nil, `aba`, []Match{[]Group{{0, 3}, {0, 1}}}},
+	{`(aa|a)a\1$`, nil, `aaa`, []Match{[]Group{{0, 3}, {0, 1}}}},
+	{`(a|aa)a\1$`, nil, `aaa`, []Match{[]Group{{0, 3}, {0, 1}}}},
+	{`(a+)a\1$`, nil, `aaa`, []Match{[]Group{{0, 3}, {0, 1}}}},
+	{`([abc]*)\1`, nil, `abcabc`, []Match{[]Group{{0, 6}, {0, 3}}, []Group{{6, 6}, {6, 6}}}},
+	{`(a)(?:b)\1`, nil, `aba`, []Match{[]Group{{0, 3}, {0, 1}}}},
+	{`(a)(?:b)\1`, nil, `abb`, []Match{}},
+	{`(?:a)(b)\1`, nil, `aba`, []Match{}},
+	{`(?:a)(b)\1`, nil, `abb`, []Match{[]Group{{0, 3}, {1, 2}}}},
+	{`(?:(cat)|(dog))\2`, nil, `catdog`, []Match{}},
+	{`(?:a)\1`, nil, `aa`, nil},
+	{`((cat)|(dog)|(cow)|(bat))\4`, nil, `cowcow`, []Match{[]Group{{0, 6}, {0, 3}, {-1, -1}, {-1, -1}, {0, 3}, {-1, -1}}}},
+	{`(a|b)*\1`, nil, `abb`, []Match{[]Group{{0, 3}, {1, 2}}}},
+	{`(a|b)*\1`, nil, `aba`, []Match{}},
+	{`(a|b)*\1`, nil, `bab`, []Match{}},
+	{`(a|b)*\1`, nil, `baa`, []Match{[]Group{{0, 3}, {1, 2}}}},
 
 	{`(a)(b)c|ab`, nil, `ab`, []Match{[]Group{{0, 2}}}},
 	{`(a)+x`, nil, `aaax`, []Match{[]Group{{0, 4}, {2, 3}}}},
@@ -636,7 +667,7 @@ var groupTests = []struct {
 	{`(bc+d$|ef*g.|h?i(j|k))`, []ReFlag{RE_CASE_INSENSITIVE}, `BCDD`, []Match{}},
 	{`(bc+d$|ef*g.|h?i(j|k))`, []ReFlag{RE_CASE_INSENSITIVE}, `reffgz`, []Match{[]Group{{1, 6}, {1, 6}}}},
 	{`(((((((((a)))))))))`, []ReFlag{RE_CASE_INSENSITIVE}, `A`, []Match{[]Group{{0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}}}},
-	{`(((((((((a)))))))))\41`, []ReFlag{RE_CASE_INSENSITIVE}, `A!`, []Match{[]Group{{0, 2}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}}}},
+	{`(((((((((a)))))))))\041`, []ReFlag{RE_CASE_INSENSITIVE}, `A!`, []Match{[]Group{{0, 2}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}}}},
 	{`(.*)c(.*)`, []ReFlag{RE_CASE_INSENSITIVE}, `ABCDE`, []Match{[]Group{{0, 5}, {0, 2}, {3, 5}}}},
 	{`\((.*), (.*)\)`, []ReFlag{RE_CASE_INSENSITIVE}, `(A, B)`, []Match{[]Group{{0, 6}, {1, 2}, {4, 5}}}},
 	{`(a)(b)c|ab`, []ReFlag{RE_CASE_INSENSITIVE}, `AB`, []Match{[]Group{{0, 2}}}},
@@ -792,23 +823,24 @@ func TestFindSubmatch(t *testing.T) {
 				if test.result != nil {
 					panic(err)
 				}
-			}
-			match, err := regComp.FindSubmatch(test.str)
-			if err != nil {
-				if len(test.result) != 0 {
-					t.Errorf("Wanted %v got no match\n", test.result[0])
-				}
-			} else if len(test.result) == 0 {
-				t.Errorf("Wanted no match got %v\n", match)
-			}
-			for i := range match {
-				if match[i].IsValid() {
-					if test.result[0][i] != match[i] {
-						t.Errorf("Wanted %v	Got %v\n", test.result[0], match)
+			} else {
+				match, err := regComp.FindSubmatch(test.str)
+				if err != nil {
+					if len(test.result) != 0 {
+						t.Errorf("Wanted %v got no match\n", test.result[0])
 					}
-				} else {
-					if i < len(test.result) && test.result[0][i].IsValid() {
-						t.Errorf("Wanted %v	Got %v\n", test.result[0], match)
+				} else if len(test.result) == 0 {
+					t.Errorf("Wanted no match got %v\n", match)
+				}
+				for i := range match {
+					if match[i].IsValid() {
+						if test.result[0][i] != match[i] {
+							t.Errorf("Wanted %v	Got %v\n", test.result[0], match)
+						}
+					} else {
+						if i < len(test.result) && test.result[0][i].IsValid() {
+							t.Errorf("Wanted %v	Got %v\n", test.result[0], match)
+						}
 					}
 				}
 			}
@@ -823,10 +855,22 @@ func TestFindStringSubmatch(t *testing.T) {
 				if test.result != nil {
 					panic(err)
 				}
-			}
-			matchStr := regComp.FindStringSubmatch(test.str)
-			if matchStr == nil {
-				if len(test.result) != 0 {
+			} else {
+				matchStr := regComp.FindStringSubmatch(test.str)
+				if matchStr == nil {
+					if len(test.result) != 0 {
+						expectedStr := funcMap(test.result[0], func(g Group) string {
+							if g.IsValid() {
+								return test.str[g.StartIdx:g.EndIdx]
+							} else {
+								return ""
+							}
+						})
+						t.Errorf("Wanted %v got no match\n", expectedStr)
+					}
+				} else if len(test.result) == 0 {
+					t.Errorf("Wanted no match got %v\n", matchStr)
+				} else {
 					expectedStr := funcMap(test.result[0], func(g Group) string {
 						if g.IsValid() {
 							return test.str[g.StartIdx:g.EndIdx]
@@ -834,26 +878,15 @@ func TestFindStringSubmatch(t *testing.T) {
 							return ""
 						}
 					})
-					t.Errorf("Wanted %v got no match\n", expectedStr)
-				}
-			} else if len(test.result) == 0 {
-				t.Errorf("Wanted no match got %v\n", matchStr)
-			} else {
-				expectedStr := funcMap(test.result[0], func(g Group) string {
-					if g.IsValid() {
-						return test.str[g.StartIdx:g.EndIdx]
-					} else {
-						return ""
-					}
-				})
-				for i, groupStr := range matchStr {
-					if groupStr == "" {
-						if i < len(expectedStr) && expectedStr[i] != "" {
-							t.Errorf("Wanted %v	Got %v\n", expectedStr, matchStr)
-						}
-					} else {
-						if expectedStr[i] != groupStr {
-							t.Errorf("Wanted %v	Got %v\n", expectedStr, matchStr)
+					for i, groupStr := range matchStr {
+						if groupStr == "" {
+							if i < len(expectedStr) && expectedStr[i] != "" {
+								t.Errorf("Wanted %v	Got %v\n", expectedStr, matchStr)
+							}
+						} else {
+							if expectedStr[i] != groupStr {
+								t.Errorf("Wanted %v	Got %v\n", expectedStr, matchStr)
+							}
 						}
 					}
 				}
@@ -870,10 +903,24 @@ func TestFindAllStringSubmatch(t *testing.T) {
 				if test.result != nil {
 					panic(err)
 				}
-			}
-			matchStrs := regComp.FindAllStringSubmatch(test.str)
-			if matchStrs == nil {
-				if len(test.result) != 0 {
+			} else {
+				matchStrs := regComp.FindAllStringSubmatch(test.str)
+				if matchStrs == nil {
+					if len(test.result) != 0 {
+						expectedStrs := funcMap(test.result, func(m Match) []string {
+							return funcMap(m, func(g Group) string {
+								if g.IsValid() {
+									return test.str[g.StartIdx:g.EndIdx]
+								} else {
+									return ""
+								}
+							})
+						})
+						t.Errorf("Wanted %v got no match\n", expectedStrs)
+					}
+				} else if len(test.result) == 0 {
+					t.Errorf("Wanted no match got %v\n", matchStrs)
+				} else {
 					expectedStrs := funcMap(test.result, func(m Match) []string {
 						return funcMap(m, func(g Group) string {
 							if g.IsValid() {
@@ -883,29 +930,16 @@ func TestFindAllStringSubmatch(t *testing.T) {
 							}
 						})
 					})
-					t.Errorf("Wanted %v got no match\n", expectedStrs)
-				}
-			} else if len(test.result) == 0 {
-				t.Errorf("Wanted no match got %v\n", matchStrs)
-			} else {
-				expectedStrs := funcMap(test.result, func(m Match) []string {
-					return funcMap(m, func(g Group) string {
-						if g.IsValid() {
-							return test.str[g.StartIdx:g.EndIdx]
-						} else {
-							return ""
-						}
-					})
-				})
-				for i, matchStr := range matchStrs {
-					for j, groupStr := range matchStr {
-						if groupStr == "" {
-							if j < len(expectedStrs[i]) && expectedStrs[i][j] != "" {
-								t.Errorf("Wanted %v	Got %v\n", expectedStrs, matchStrs)
-							}
-						} else {
-							if expectedStrs[i][j] != groupStr {
-								t.Errorf("Wanted %v	Got %v\n", expectedStrs, matchStrs)
+					for i, matchStr := range matchStrs {
+						for j, groupStr := range matchStr {
+							if groupStr == "" {
+								if j < len(expectedStrs[i]) && expectedStrs[i][j] != "" {
+									t.Errorf("Wanted %v	Got %v\n", expectedStrs, matchStrs)
+								}
+							} else {
+								if expectedStrs[i][j] != groupStr {
+									t.Errorf("Wanted %v	Got %v\n", expectedStrs, matchStrs)
+								}
 							}
 						}
 					}
@@ -923,17 +957,18 @@ func TestFindAllSubmatch(t *testing.T) {
 				if test.result != nil {
 					panic(err)
 				}
-			}
-			matchIndices := regComp.FindAllSubmatch(test.str)
-			for i := range matchIndices {
-				for j := range matchIndices[i] {
-					if matchIndices[i][j].IsValid() {
-						if test.result[i][j] != matchIndices[i][j] {
-							t.Errorf("Wanted %v	Got %v\n", test.result, matchIndices)
-						}
-					} else {
-						if i < len(test.result) && j < len(test.result[i]) && test.result[i][j].IsValid() {
-							t.Errorf("Wanted %v	Got %v\n", test.result, matchIndices)
+			} else {
+				matchIndices := regComp.FindAllSubmatch(test.str)
+				for i := range matchIndices {
+					for j := range matchIndices[i] {
+						if matchIndices[i][j].IsValid() {
+							if test.result[i][j] != matchIndices[i][j] {
+								t.Errorf("Wanted %v	Got %v\n", test.result, matchIndices)
+							}
+						} else {
+							if i < len(test.result) && j < len(test.result[i]) && test.result[i][j].IsValid() {
+								t.Errorf("Wanted %v	Got %v\n", test.result, matchIndices)
+							}
 						}
 					}
 				}