Merge pull request 'Implement PCRE Matching (prefer left-branch)' (#2) from implementPCREMatchingRules into master

Reviewed-on: #2

LICENSE

@@ -0,0 +1,11 @@
+The MIT License (MIT)
+
+Copyright (c) 2025 Aadhavan Srinivasan
+
+All rights reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Makefile

@@ -1,9 +1,13 @@
-.DEFAULT_GOAL := build
-.PHONY: fmt vet build
+.DEFAULT_GOAL := buildCmd
+.PHONY: fmt vet buildLib buildCmd test
 
 fmt:
 	go fmt ./...
 vet: fmt
 	go vet ./...
-build: vet
-	go build -gcflags="-N -l" ./...
+buildLib: vet
+	go build -gcflags="all=-N -l" ./...
+buildCmd: buildLib
+	go build -C cmd/ -gcflags="all=-N -l" -o re ./...
+test: buildCmd
+	go test -v ./...

cmd/main.go

@@ -121,12 +121,12 @@ func main() {
 		}
 		matchIndices := make([]reg.Match, 0)
 		if matchNumFlagEnabled {
-			tmp, err := reg.FindNthMatch(regComp, test_str, *matchNum)
+			tmp, err := regComp.FindNthMatch(test_str, *matchNum)
 			if err == nil {
 				matchIndices = append(matchIndices, tmp)
 			}
 		} else {
-			matchIndices = reg.FindAllMatches(regComp, test_str)
+			matchIndices = regComp.FindAllSubmatch(test_str)
 		}
 
 		if *printMatchesFlag {
@@ -137,7 +137,7 @@ func main() {
 					fmt.Fprintf(out, "Line %d:\n", lineNum)
 				}
 				for _, m := range matchIndices {
-					fmt.Fprintf(out, "%s\n", m.ToString())
+					fmt.Fprintf(out, "%s\n", m.String())
 				}
 				err := out.Flush()
 				if err != nil {

cmd/unique_array.go

@@ -16,7 +16,6 @@ func (s *uniq_arr[T]) add(vals ...T) {
 			s.backingMap[item] = struct{}{}
 		}
 	}
-	return
 }
 
 func (s uniq_arr[T]) contains(val T) bool {

regex/compile.go

@@ -12,10 +12,27 @@ var notDotChars []rune
 
 // A Reg represents the result of compiling a regular expression. It contains
 // the startState of the NFA representation of the regex, and the number of capturing
-// groups in the regex.
+// groups in the regex. It also contains the expression string.
 type Reg struct {
 	start         *nfaState
 	numGroups     int
+	str           string
+	preferLongest bool
+}
+
+// NumSubexp returns the number of sub-expressions in the given [Reg]. This is equivalent
+// to the number of capturing groups.
+func (re Reg) NumSubexp() int {
+	return re.numGroups
+}
+
+// String returns the string used to compile the expression.
+func (re Reg) String() string {
+	return re.str
+}
+
+func (re *Reg) Longest() {
+	re.preferLongest = true
 }
 
 const concatRune rune = 0xF0001
@@ -171,7 +188,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 			}
 			re_runes = append(re_runes, []rune(regex)...)
 		} else if c == '(' && i < len(re_runes_orig)-2 && re_runes_orig[i+1] == '?' && re_runes_orig[i+2] == ':' {
-			re_runes = append(re_runes, NONCAPLPAREN_CHAR)
+			re_runes = append(re_runes, nonCapLparenRune)
 			i += 2
 		} else if c == '\\' && i < len(re_runes_orig)-1 && re_runes_orig[i+1] == '\\' { // Escaped backslash
 			re_runes = append(re_runes, escBackslashRune)
@@ -233,7 +250,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 					i++
 				}
 				if re_runes[i] == '-' && (i > 0 && re_runes[i-1] != '\\') && (i < len(re_runes)-1 && re_runes[i+1] != rbracketRune) { // Unescaped hyphen, that has some character (not a RBRACKET) after it - This represents a character range, so we replace with CHAR_RANGE. This metacharacter will be used later on to construct the range
-					re_runes[i] = CHAR_RANGE
+					re_runes[i] = charRangeRune
 				}
 				toAppend = append(toAppend, re_runes[i])
 				i++
@@ -254,7 +271,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 			re_postfix = append(re_postfix, re_runes[i]) // Append closing brace
 		}
 		if i < len(re_runes)-3 && string(re_runes[i+1:i+4]) == "(?:" { // Non-capturing lparen
-			re_postfix = append(re_postfix, NONCAPLPAREN_CHAR)
+			re_postfix = append(re_postfix, nonCapLparenRune)
 			i += 3
 		}
 		if i < len(re_runes) && re_runes[i] == '\\' { // Something is being escaped (I don't add the backslash to re_postfix, because it was already added earlier)
@@ -303,7 +320,7 @@ 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] == NONCAPLPAREN_CHAR {
+				if re_runes[i] == '(' || re_runes[i] == nonCapLparenRune {
 					numOpenParens++
 				}
 				if re_runes[i] == ')' {
@@ -317,7 +334,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 			}
 			continue
 		}
-		if i < len(re_runes) && (re_runes[i] != '(' && re_runes[i] != NONCAPLPAREN_CHAR && re_runes[i] != '|' && re_runes[i] != '\\') || (i > 0 && re_runes[i-1] == '\\') { // Every character should be concatenated if it is escaped
+		if i < len(re_runes) && (re_runes[i] != '(' && re_runes[i] != nonCapLparenRune && re_runes[i] != '|' && re_runes[i] != '\\') || (i > 0 && re_runes[i-1] == '\\') { // Every character should be concatenated if it is escaped
 			if i < len(re_runes)-1 {
 				if re_runes[i+1] != '|' && re_runes[i+1] != '*' && re_runes[i+1] != '+' && re_runes[i+1] != '?' && re_runes[i+1] != ')' && re_runes[i+1] != '{' {
 					re_postfix = append(re_postfix, concatRune)
@@ -407,7 +424,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 			} else {
 				escapedNode, err := newEscapedNode(re_postfix[i], false)
 				if err != nil {
-					return nil, fmt.Errorf("invalid escape character in expression")
+					return nil, err
 				}
 				outQueue = append(outQueue, escapedNode)
 			}
@@ -433,7 +450,7 @@ 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] == NONCAPLPAREN_CHAR {
+				if re_postfix[i] == '(' || re_postfix[i] == nonCapLparenRune {
 					numOpenParens++
 				}
 				if re_postfix[i] == ')' {
@@ -524,7 +541,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 				if firstCharAdded && re_postfix[i] == rbracketRune {
 					break
 				}
-				if re_postfix[i] == CHAR_RANGE {
+				if re_postfix[i] == charRangeRune {
 					endOfRange = true
 					i++
 					continue
@@ -575,7 +592,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 					} else {
 						escapedNode, err := newEscapedNode(re_postfix[i], true)
 						if err != nil {
-							return nil, fmt.Errorf("invalid escape character in character class")
+							return nil, err
 						}
 						chars = append(chars, escapedNode)
 						i++
@@ -757,7 +774,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 			outQueue[idx].startReps = startRangeNum
 			outQueue[idx].endReps = endRangeNum
 		}
-		if c == '(' || c == NONCAPLPAREN_CHAR {
+		if c == '(' || c == nonCapLparenRune {
 			opStack = append(opStack, c)
 			if c == '(' { // We only push _capturing_ group parentheses to outQueue
 				outQueue = append(outQueue, newPostfixNode(c))
@@ -768,7 +785,7 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 			// Keep popping from opStack until we encounter an opening parantheses or a NONCAPLPAREN_CHAR. Throw error if we reach the end of the stack.
 			var val rune
 			var err error
-			for val, err = peek(opStack); val != '(' && val != NONCAPLPAREN_CHAR; val, err = peek(opStack) {
+			for val, err = peek(opStack); val != '(' && val != nonCapLparenRune; val, err = peek(opStack) {
 				if err != nil {
 					return nil, fmt.Errorf("imbalanced parantheses")
 				}
@@ -810,13 +827,12 @@ func thompson(re []postfixNode) (Reg, error) {
 	// In these cases, we will return an NFA with 1 state, with an assertion that is always true.
 	if len(re) == 0 {
 		start := zeroLengthMatchState()
-		nfa = append(nfa, &start)
+		nfa = append(nfa, start)
 	}
 
 	for _, c := range re {
 		if c.nodetype == characterNode || c.nodetype == assertionNode {
 			stateToAdd := nfaState{}
-			stateToAdd.transitions = make(map[int][]*nfaState)
 			if c.allChars {
 				stateToAdd.allChars = true
 				if len(c.except) != 0 {
@@ -878,6 +894,10 @@ func thompson(re []postfixNode) (Reg, error) {
 						stateToAdd.assert = wboundAssert
 					case 'B':
 						stateToAdd.assert = nonwboundAssert
+					case 'A':
+						stateToAdd.assert = soiAssert
+					case 'z':
+						stateToAdd.assert = eoiAssert
 					}
 				} else { // Lookaround
 					stateToAdd.lookaroundRegex = string(c.contents)
@@ -924,7 +944,6 @@ func thompson(re []postfixNode) (Reg, error) {
 			s.isEmpty = true
 			s.output = make([]*nfaState, 0)
 			s.output = append(s.output, s)
-			s.transitions = make(map[int][]*nfaState)
 			// LPAREN nodes are just added normally
 			if c.nodetype == lparenNode {
 				numGroups++
@@ -939,7 +958,9 @@ func thompson(re []postfixNode) (Reg, error) {
 			// and added back in.
 			// If the middle node doesn't exist (ie. something like '()' ), that's fine, I just connect the LPAREN
 			// and RPAREN nodes.
-			// If neither node exists, that's a problem so I return an error.
+			// If the middle node exists but is itself the start of a group, then that _must_ be the opening paren for
+			// the closing paren that I'm on. I put the third node back (because it isn't involved in the capturing group), then
+			// I concatenate those two and add them. If neither node exists, that's a problem so I return an error.
 			if c.nodetype == rparenNode {
 				s.groupEnd = true
 				middleNode, err1 := pop(&nfa)
@@ -954,6 +975,11 @@ func thompson(re []postfixNode) (Reg, error) {
 					s.groupNum = lparenNode.groupNum
 					to_add := concatenate(lparenNode, s)
 					nfa = append(nfa, to_add)
+				} else if middleNode.groupBegin && middleNode.numTransitions() == 0 { // The middle node is a lone lparen - something like '(())', and I'm looking at the first rparen
+					nfa = append(nfa, lparenNode)    // I shouldn't have popped this out, because it is not involved in the current capturing group
+					s.groupNum = middleNode.groupNum // In this case, the 'middle' node is actually an lparen
+					to_add := concatenate(middleNode, s)
+					nfa = append(nfa, to_add)
 				} else {
 					// At this point, we assume all three nodes are valid ('lparenNode', 'middleNode' and 's')
 					if lparenNode.groupBegin {
@@ -972,7 +998,8 @@ func thompson(re []postfixNode) (Reg, error) {
 		if c.nodetype == charclassNode { // A Character class consists of all the nodes in it, alternated
 			// Map the list of nodes to a list of states, each state containing the contents of a specific node
 			states := funcMap(c.nodeContents, func(node postfixNode) *nfaState {
-				s := newState()
+				s := &nfaState{}
+				s.output = append(s.output, s)
 				nodeContents := node.contents
 				if caseInsensitive {
 					nodeContents = slices.Concat(funcMap(nodeContents, func(r rune) []rune {
@@ -986,7 +1013,7 @@ func thompson(re []postfixNode) (Reg, error) {
 						return n.contents
 					})...)
 				}
-				return &s
+				return s
 			})
 			// Reduce the list of states down to a single state by alternating them
 			toAdd := funcReduce(states, func(s1 *nfaState, s2 *nfaState) *nfaState {
@@ -1000,7 +1027,7 @@ func thompson(re []postfixNode) (Reg, error) {
 			s2 := mustPop(&nfa)
 			// Relax the requirements for concatenation a little bit - If
 			// the second element is not found ie. the postfixNodes look
-			// like 'a~', then that's fine, we just skip the concatenation.
+			// like 'a'+CONCAT, then that's fine, we just skip the concatenation.
 			s1, err := pop(&nfa)
 			if err != nil {
 				nfa = append(nfa, s2)
@@ -1013,14 +1040,14 @@ func thompson(re []postfixNode) (Reg, error) {
 			if err != nil {
 				return Reg{}, fmt.Errorf("error applying kleene star")
 			}
-			stateToAdd, err := kleene(*s1)
+			stateToAdd, err := kleene(s1)
 			if err != nil {
 				return Reg{}, err
 			}
 			nfa = append(nfa, stateToAdd)
 		case plusNode: // a+ is equivalent to aa*
 			s1 := mustPop(&nfa)
-			s2, err := kleene(*s1)
+			s2, err := kleene(s1)
 			if err != nil {
 				return Reg{}, err
 			}
@@ -1031,7 +1058,10 @@ func thompson(re []postfixNode) (Reg, error) {
 			if err != nil {
 				return Reg{}, fmt.Errorf("error applying question operator")
 			}
-			s2 := question(s1)
+			s2, err := question(s1)
+			if err != nil {
+				return Reg{}, err
+			}
 			nfa = append(nfa, s2)
 		case pipeNode:
 			// A pipe operator doesn't actually need either operand to be present. If an operand isn't present,
@@ -1042,21 +1072,21 @@ func thompson(re []postfixNode) (Reg, error) {
 			// 	'|a'
 			// 	'^a|'
 			// 	'^|a'
-			s1, err1 := pop(&nfa)
-			s2, err2 := pop(&nfa)
-			if err2 != nil || (s2.groupBegin && len(s2.transitions) == 0) { // Doesn't exist, or its just an LPAREN
+			s2, err1 := pop(&nfa)
+			s1, err2 := pop(&nfa)
+			if err2 != nil || (s2.groupBegin && s2.numTransitions() == 0) { // Doesn't exist, or its just an LPAREN
 				if err2 == nil { // Roundabout way of saying that this node existed, but it was an LPAREN, so we append it back
 					nfa = append(nfa, s2)
 				}
 				tmp := zeroLengthMatchState()
-				s2 = &tmp
+				s2 = tmp
 			}
-			if err1 != nil || (s1.groupBegin && len(s1.transitions) == 0) { // Doesn't exist, or its just an LPAREN
+			if err1 != nil || (s1.groupBegin && s1.numTransitions() == 0) { // Doesn't exist, or its just an LPAREN
 				if err1 == nil { // See above for explanation
 					nfa = append(nfa, s1)
 				}
 				tmp := zeroLengthMatchState()
-				s1 = &tmp
+				s1 = tmp
 			}
 			s3 := alternate(s1, s2)
 			nfa = append(nfa, s3)
@@ -1083,14 +1113,18 @@ func thompson(re []postfixNode) (Reg, error) {
 				stateToAdd = concatenate(stateToAdd, cloneState(poppedState))
 			}
 			if c.endReps == infinite_reps { // Case 3
-				s2, err := kleene(*poppedState)
+				s2, err := kleene(poppedState)
 				if err != nil {
 					return Reg{}, err
 				}
 				stateToAdd = concatenate(stateToAdd, s2)
 			} else { // Case 2
 				for i := c.startReps; i < c.endReps; i++ {
-					stateToAdd = concatenate(stateToAdd, question(cloneState(poppedState)))
+					tmp, err := question(cloneState(poppedState))
+					if err != nil {
+						return Reg{}, fmt.Errorf("error processing bounded repetition")
+					}
+					stateToAdd = concatenate(stateToAdd, tmp)
 				}
 			}
 			nfa = append(nfa, stateToAdd)
@@ -1100,16 +1134,21 @@ func thompson(re []postfixNode) (Reg, error) {
 		return Reg{}, fmt.Errorf("invalid regex")
 	}
 
-	verifyLastStates(nfa)
+	lastState := newState()
+	lastState.isLast = true
 
-	return Reg{nfa[0], numGroups}, nil
+	concatenate(nfa[0], &lastState)
+
+	// The string is empty here, because we add it in Compile()
+	return Reg{nfa[0], numGroups, "", false}, nil
 
 }
 
-// Compiles the given regular expression into a Reg type, suitable for use with the
-// matching functions. The second return value is non-nil if a compilation error has
-// occured. As such, the error value must be checked before using the Reg returned by this function.
-// The second parameter is an optional list of flags, passed to the parsing function shuntingYard.
+// Compile compiles the given regular expression into a [Reg].
+//
+// An error value != nil indicates that the regex was invalid; the error message should provide
+// detailed information on the nature of the error.
+// The second parameter is a sequence of zero or more [ReFlag] values, that modify the behavior of the regex.
 func Compile(re string, flags ...ReFlag) (Reg, error) {
 	nodes, err := shuntingYard(re, flags...)
 	if err != nil {
@@ -1119,5 +1158,15 @@ func Compile(re string, flags ...ReFlag) (Reg, error) {
 	if err != nil {
 		return Reg{}, fmt.Errorf("error compiling regex: %w", err)
 	}
+	reg.str = re
 	return reg, nil
 }
+
+// MustCompile panics if Compile returns an error. They are identical in all other respects.
+func MustCompile(re string, flags ...ReFlag) Reg {
+	reg, err := Compile(re, flags...)
+	if err != nil {
+		panic(err)
+	}
+	return reg
+}

regex/doc.go

@@ -0,0 +1,178 @@
+/*
+Package regex implements regular expression search, using a custom non-bracktracking engine with support for lookarounds and numeric ranges.
+
+The engine relies completely on UTF-8 codepoints. As such, it is capable of matching characters
+from other languages, emojis and symbols.
+
+The API and regex syntax are largely compatible with that of the stdlib's [regexp], with a few key differences (see 'Key Differences with regexp').
+
+The full syntax is specified below.
+
+# Syntax
+
+Single characters:
+
+	.				Match any character. Newline matching is dependent on the RE_SINGLE_LINE flag.
+	[abc]			Character class - match a, b or c
+	[a-z]			Character range - match any character from a to z
+	[^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)
+	\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
+	\a				Bell character
+	\f				Form-feed character
+	\r				Carriage return
+	\t				Horizontal tab
+	\v				Vertical tab
+
+Perl classes:
+
+	\d				Match any digit character ([0-9])
+	\D				Match any non-digit character ([^0-9])
+	\w				Match any word character ([a-zA-Z0-9_])
+	\W				Match any word character ([^a-zA-Z0-9_])
+	\s				Match any whitespace character ([ \t\n])
+	\S				Match any non-whitespace character ([^ \t\n])
+
+POSIX classes (inside normal character classes):
+
+	[:digit:]		All digit characters ([0-9])
+	[:upper:]		All upper-case letters ([A-Z])
+	[:lower:]		All lower-case letters ([a-z])
+	[:alpha:]		All letters ([a-zA-Z])
+	[:alnum:]		All alphanumeric characters ([a-zA-Z0-9])
+	[:xdigit:]		All hexadecimal characters ([a-fA-F0-9])
+	[:blank:]		All blank characters ([ \t])
+	[:space:]		All whitespace characters ([ \t\n\r\f\v])
+	[:cntrl:]		All control characters ([\x00-\x1F\x7F])
+	[:punct:]		All punctuation characters
+	[:graph:]		All graphical characters ([\x21-\x7E])
+	[:print:]		All graphical characters + space ([\x20-\x7E])
+	[:word:]		All word characters (\w)
+	[:ascii:]		All ASCII values ([\x00-\x7F])
+
+Composition:
+
+	def				Match d, followed by e, followed by f
+	x|y				Match x or y (prefer x)
+	xy|z			Match xy or z (prefer xy)
+
+Repitition (always greedy, preferring more):
+
+	x*				Match x zero or more times
+	x+				Match x one or more times
+	x?				Match x zero or one time
+	x{m,n}			Match x between m and n times (inclusive)
+	x{m,}			Match x atleast m times
+	x{,n}			Match x between 0 and n times (inclusive)
+	x{m}			Match x exactly m times
+
+Grouping:
+
+	(expr)			Create a capturing group. The contents of the group can be retrieved with [FindAllMatches]
+	x(y|z)			Match x followed by y or z. Given a successful match, the contents of group 1 will include either y or z
+	(?:expr)		Create a non-capturing group. The contents of the group aren't saved.
+	x(?:y|z)		Match x followed by y or z. No groups are created.
+
+Assertions:
+
+	^				Match at the start of the input string. If RE_MULTILINE is enabled, it also matches at the start of every line.
+	$				Match at the end of the input string. If RE_MULTILINE is enabled, it also matches at the end of every line.
+	\A				Always match at the start of the string, regardless of RE_MULTILINE
+	\z				Always match at the end of the string, regardless of RE_MULTILINE
+	\b				Match at a word boundary (a word character followed by a non-word character, or vice-versa)
+	\B				Match at a non-word boundary (a word character followed by a word character, or vice-versa)
+
+Lookarounds:
+
+	x(?=y)			Positive lookahead - Match x if followed by y
+	x(?!y)			Negative lookahead - Match x if NOT followed by y
+	(?<=x)y			Positive lookbehind - Match y if preceded by x
+	(?<!x)y			Negative lookbehind - Match y if NOT preceded by x
+
+Numeric ranges:
+
+	<x-y>			Match any number from x to y (inclusive) (x and y must be positive numbers)
+	\<x				Match a literal '<' followed by x
+
+# Key Differences with regexp
+
+The engine and the API differ from [regexp] in a few ways, some of them very subtle.
+The key differences are mentioned below.
+
+1. Greediness:
+
+This engine does not support non-greedy operators. All operators are always greedy in nature, and will try
+to match as much as they can, while still allowing for a successful match. For example, given the regex:
+
+	y*y
+
+The engine will match as many 'y's as it can, while still allowing the trailing 'y' to be matched.
+
+Another, more subtle example is the following regex:
+
+	x|xx
+
+While the stdlib implementation (and most other engines) will prefer matching the first item of the alternation,
+this engine will go for the longest possible match, regardless of the order of the alternation. Although this
+strays from the convention, it results in a nice rule-of-thumb - the engine is ALWAYS greedy.
+
+The stdlib implementation has a function [regexp.Regexp.Longest] which makes future searches prefer the longest match.
+That is the default (and unchangable) behavior in this engine.
+
+2. Byte-slices and runes:
+
+My engine does not support byte-slices. When a matching function receives a string, it converts it into a
+rune-slice to iterate through it. While this has some space overhead, the convenience of built-in unicode
+support made the tradeoff worth it.
+
+3. Return values
+
+Rather than using primitives for return values, my engine defines two types that are used as return
+values: a [Group] represents a capturing group, and a [Match] represents a list of groups.
+
+[regexp] specifies a regular expression that gives a list of all the matching functions that it supports. The
+equivalent expression for this engine is shown below. Note that 'Index' is the default.
+
+	Find(All)?(String)?(Submatch)?
+
+[Reg.Find] returns the index of the leftmost match in the string.
+
+If a function contains 'All' it returns all matches instead of just the leftmost one.
+
+If a function contains 'String' it returns the matched text, rather than the index in the string.
+
+If a function contains 'Submatch' it returns the match, including all submatches found by
+capturing groups.
+
+The term '0-group' is used to refer to the 0th capturing group of a match (which is the entire match).
+Given the following regex:
+
+	x(y)
+
+and the input string:
+
+	xyz
+
+The 0th group would contain 'xy' and the 1st group would contain 'y'. Any matching function without 'Submatch' in its name
+returns the 0-group.
+
+# Feature Differences
+
+The following features from [regexp] are (currently) NOT supported:
+ 1. Named capturing groups
+ 2. Non-greedy operators
+ 3. Unicode character classes
+ 4. Embedded flags (flags are 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
+
+I hope to shorten the first list, and expand the second.
+*/
+package regex

regex/example_test.go

@@ -0,0 +1,91 @@
+package regex_test
+
+import (
+	"fmt"
+
+	"gitea.twomorecents.org/Rockingcool/kleingrep/regex"
+)
+
+func ExampleReg_Find() {
+	regexStr := "b|a"
+	regexComp := regex.MustCompile(regexStr)
+
+	match, _ := regexComp.Find("banana")
+	fmt.Println(match.String())
+
+	// Output: 0	1
+}
+
+func ExampleReg_FindAll() {
+	regexStr := "b|a"
+	regexComp := regex.MustCompile(regexStr)
+
+	matches := regexComp.FindAll("banana")
+	for _, group := range matches {
+		fmt.Println(group.String())
+	}
+
+	// Output: 0	1
+	// 1	2
+	// 3	4
+	// 5	6
+}
+
+func ExampleReg_FindString() {
+	regexStr := `\d+`
+	regexComp := regex.MustCompile(regexStr)
+
+	matchStr := regexComp.FindString("The year of our lord, 2025")
+	fmt.Println(matchStr)
+	// Output: 2025
+}
+
+func ExampleReg_FindSubmatch() {
+	regexStr := `(\d)\.(\d)(\d)`
+	regexComp := regex.MustCompile(regexStr)
+
+	match, _ := regexComp.FindSubmatch("3.14")
+	fmt.Println(match[0])
+	fmt.Println(match[1])
+	fmt.Println(match[2])
+	// Output: 0	4
+	// 0	1
+	// 2	3
+}
+
+func ExampleReg_Expand() {
+	inputStr := `option1: value1
+	option2: value2`
+	regexStr := `(\w+): (\w+)`
+	templateStr := "$1 = $2\n"
+	regexComp := regex.MustCompile(regexStr, regex.RE_MULTILINE)
+	result := ""
+	for _, submatches := range regexComp.FindAllSubmatch(inputStr) {
+		result = regexComp.Expand(result, templateStr, inputStr, submatches)
+	}
+	fmt.Println(result)
+	// Output: option1 = value1
+	// option2 = value2
+
+}
+
+func ExampleReg_LiteralPrefix() {
+	regexStr := `a(b|c)d*`
+	regexComp := regex.MustCompile(regexStr)
+	prefix, complete := regexComp.LiteralPrefix()
+	fmt.Println(prefix)
+	fmt.Println(complete)
+	// Output: a
+	// false
+}
+
+func ExampleReg_Longest() {
+	regexStr := `x|xx`
+	inputStr := "xx"
+	regexComp := regex.MustCompile(regexStr)
+	fmt.Println(regexComp.FindString(inputStr))
+	regexComp.Longest()
+	fmt.Println(regexComp.FindString(inputStr))
+	// Output: x
+	// xx
+}

regex/matching.go

@@ -2,13 +2,20 @@ package regex
 
 import (
 	"fmt"
-	"sort"
+	"strconv"
+	"unicode"
 )
 
-// a Match stores a slice of all the capturing groups in a match.
+// A Match represents a match found by the regex in a given string.
+// It is represented as a list of groups, where the nth element contains
+// the contents of the nth capturing group. Note that the group may not be valid
+// (see [Group.IsValid]). The element at index 0 is known
+// as the 0-group, and represents the contents of the entire match.
+//
+// See [Reg.FindSubmatch] for an example.
 type Match []Group
 
-// a Group represents a group. It contains the start index and end index of the match
+// a Group represents a capturing group. It contains the start and index of the group.
 type Group struct {
 	StartIdx int
 	EndIdx   int
@@ -23,151 +30,153 @@ func newMatch(size int) Match {
 	return toRet
 }
 
-// Returns the number of valid groups in the match
-func (m Match) numValidGroups() int {
-	numValid := 0
-	for _, g := range m {
-		if g.StartIdx >= 0 && g.EndIdx >= 0 {
-			numValid++
-		}
-	}
-	return numValid
-}
-
 // Returns a string containing the indices of all (valid) groups in the match
-func (m Match) ToString() string {
+func (m Match) String() string {
 	var toRet string
 	for i, g := range m {
-		if g.isValid() {
+		if g.IsValid() {
 			toRet += fmt.Sprintf("Group %d\n", i)
-			toRet += g.toString()
+			toRet += g.String()
 			toRet += "\n"
 		}
 	}
 	return toRet
 }
 
-// Converts the Group into a string representation:
-func (idx Group) toString() string {
+// String converts the Group into a string representation.
+func (idx Group) String() string {
 	return fmt.Sprintf("%d\t%d", idx.StartIdx, idx.EndIdx)
 }
 
-// Returns whether a group contains valid indices
-func (g Group) isValid() bool {
+// IsValid returns whether a group is valid (ie. whether it matched any text). It
+// simply ensures that both indices of the group are >= 0.
+func (g Group) IsValid() bool {
 	return g.StartIdx >= 0 && g.EndIdx >= 0
 }
 
-// takeZeroState takes the 0-state (if such a transition exists) for all states in the
-// given slice. It returns the resulting states. If any of the resulting states is a 0-state,
-// the second ret val is true.
-// If a state begins or ends a capturing group, its 'thread' is updated to contain the correct index.
-func takeZeroState(states []*nfaState, numGroups int, idx int) (rtv []*nfaState, isZero bool) {
-	for _, state := range states {
-		if len(state.transitions[epsilon]) > 0 {
-			for _, s := range state.transitions[epsilon] {
-				if s.threadGroups == nil {
-					s.threadGroups = newMatch(numGroups + 1)
-				}
-				copy(s.threadGroups, state.threadGroups)
-				if s.groupBegin {
-					s.threadGroups[s.groupNum].StartIdx = idx
-					//					openParenGroups = append(openParenGroups, s.groupNum)
-				}
-				if s.groupEnd {
-					s.threadGroups[s.groupNum].EndIdx = idx
-					//					closeParenGroups = append(closeParenGroups, s.groupNum)
-				}
-			}
-			rtv = append(rtv, state.transitions[epsilon]...)
-		}
-	}
-	for _, state := range rtv {
-		if len(state.transitions[epsilon]) > 0 {
-			return rtv, true
-		}
-	}
-	return rtv, false
+// Simple function, makes it easier to map this over a list of matches
+func getZeroGroup(m Match) Group {
+	return m[0]
 }
 
-// zeroMatchPossible returns true if a zero-length match is possible
-// from any of the given states, given the string and our position in it.
-// It uses the same algorithm to find zero-states as the one inside the loop,
-// so I should probably put it in a function.
-func zeroMatchPossible(str []rune, idx int, numGroups int, states ...*nfaState) bool {
-	zeroStates, isZero := takeZeroState(states, numGroups, idx)
-	tempstates := make([]*nfaState, 0, len(zeroStates)+len(states))
-	tempstates = append(tempstates, states...)
-	tempstates = append(tempstates, zeroStates...)
-	num_appended := 0 // number of unique states addded to tempstates
-	for isZero == true {
-		zeroStates, isZero = takeZeroState(tempstates, numGroups, idx)
-		tempstates, num_appended = unique_append(tempstates, zeroStates...)
-		if num_appended == 0 { // break if we haven't appended any more unique values
-			break
-		}
-	}
-	for _, state := range tempstates {
-		if state.isEmpty && (state.assert == noneAssert || state.checkAssertion(str, idx)) && state.isLast {
-			return true
-		}
-	}
-	return false
+func copyThread(to *nfaState, from nfaState) {
+	to.threadGroups = append([]Group{}, from.threadGroups...)
 }
 
-// Prunes the slice by removing overlapping indices.
-func pruneIndices(indices []Match) []Match {
-	// First, sort the slice by the start indices
-	sort.Slice(indices, func(i, j int) bool {
-		return indices[i][0].StartIdx < indices[j][0].StartIdx
-	})
-	toRet := make([]Match, 0, len(indices))
-	current := indices[0]
-	for _, idx := range indices[1:] {
-		// idx doesn't overlap with current (starts after current ends), so add current to result
-		// and update the current.
-		if idx[0].StartIdx >= current[0].EndIdx {
-			toRet = append(toRet, current)
-			current = idx
-		} else if idx[0].EndIdx > current[0].EndIdx {
-			// idx overlaps, but it is longer, so update current
-			current = idx
+// Find returns the 0-group of the leftmost match of the regex in the given string.
+// An error value != nil indicates that no match was found.
+func (re Reg) Find(str string) (Group, error) {
+	match, err := re.FindNthMatch(str, 1)
+	if err != nil {
+		return Group{}, fmt.Errorf("no matches found")
 	}
-	}
-	// Add last state
-	toRet = append(toRet, current)
-	return toRet
+	return getZeroGroup(match), nil
 }
 
-// FindString returns a _string_ containing the _text_ of the _leftmost_ match of
-// the regex, in the given string. The return value will be an empty string in two situations:
+// Match returns a boolean value, indicating whether the regex found a match in the given string.
+func (re Reg) Match(str string) bool {
+	_, err := re.Find(str)
+	return err == nil
+}
+
+// CompileMatch compiles expr and returns true if str contains a match of the expression.
+// It is equivalent to [regexp.Match].
+// An optional list of flags may be provided (see [ReFlag]).
+// It returns an error (!= nil) if there was an error compiling the expression.
+func CompileMatch(expr string, str string, flags ...ReFlag) (bool, error) {
+	re, err := Compile(expr, flags...)
+	if err != nil {
+		return false, err
+	}
+	return re.Match(str), nil
+}
+
+// FindAll returns a slice containing all the 0-groups of the regex in the given string.
+// A 0-group represents the match without any submatches.
+func (re Reg) FindAll(str string) []Group {
+	indices := re.FindAllSubmatch(str)
+	zeroGroups := funcMap(indices, getZeroGroup)
+	return zeroGroups
+}
+
+// FindString returns the text of the leftmost match of the regex in the given string.
+// The return value will be an empty string in two situations:
 //  1. No match was found
 //  2. The match was an empty string
-func FindString(regex Reg, str string) string {
-	match, err := FindNthMatch(regex, str, 1)
+func (re Reg) FindString(str string) string {
+	match, err := re.FindNthMatch(str, 1)
 	if err != nil {
 		return ""
 	}
-	return str[match[0].StartIdx:match[0].EndIdx]
+	zeroGroup := getZeroGroup(match)
+	return str[zeroGroup.StartIdx:zeroGroup.EndIdx]
 }
 
-// FindAllString is the 'all' version of FindString.
-// It returns a _slice of strings_ containing the _text_ of _all_ matches of
-// the regex, in the given string.
-//func FindAllString(regex Reg, str []string) []string {
-//
-//}
+// FindSubmatch returns the leftmost match of the regex in the given string, including
+// the submatches matched by capturing groups. The returned [Match] will always contain the same
+// number of groups. The validity of a group (whether or not it matched anything) can be determined with
+// [Group.IsValid], or by checking that both indices of the group are >= 0.
+// The second-return value is nil if no match was found.
+func (re Reg) FindSubmatch(str string) (Match, error) {
+	match, err := re.FindNthMatch(str, 1)
+	if err != nil {
+		return Match{}, fmt.Errorf("no match found")
+	} else {
+		return match, nil
+	}
+}
 
-// FindNthMatch finds the 'n'th match of the regex represented by the given start-state, with
-// the given string.
+// FindStringSubmatch is the 'string' version of [FindSubmatch]. It returns a slice of strings,
+// where the string at index i contains the text matched by the i-th capturing group.
+// The 0-th index represents the entire match.
+// An empty string at index n could mean:
+// ,
+//  1. Group n did not find a match
+//  2. Group n found a zero-length match
+//
+// A return value of nil indicates no match.
+func (re Reg) FindStringSubmatch(str string) []string {
+	matchStr := make([]string, re.numGroups+1)
+	match, err := re.FindSubmatch(str)
+	if err != nil {
+		return nil
+	}
+	nonEmptyMatchFound := false
+	for i := range match {
+		if match[i].IsValid() {
+			matchStr[i] = str[match[i].StartIdx:match[i].EndIdx]
+			nonEmptyMatchFound = true
+		} else {
+			matchStr[i] = ""
+		}
+	}
+	if nonEmptyMatchFound == false {
+		return nil
+	}
+	return matchStr
+}
+
+// FindAllString is the 'all' version of [FindString].
+// It returns a slice of strings containing the text of all matches of
+// the regex in the given string.
+func (re Reg) FindAllString(str string) []string {
+	zerogroups := re.FindAll(str)
+	matchStrs := funcMap(zerogroups, func(g Group) string {
+		return str[g.StartIdx:g.EndIdx]
+	})
+	return matchStrs
+}
+
+// FindNthMatch return the 'n'th match of the regex in the given string.
 // It returns an error (!= nil) if there are fewer than 'n' matches in the string.
-func FindNthMatch(regex Reg, str string, n int) (Match, error) {
+func (re Reg) FindNthMatch(str string, n int) (Match, error) {
 	idx := 0
 	matchNum := 0
 	str_runes := []rune(str)
 	var matchFound bool
 	var matchIdx Match
 	for idx <= len(str_runes) {
-		matchFound, matchIdx, idx = findAllMatchesHelper(regex.start, str_runes, idx, regex.numGroups)
+		matchFound, matchIdx, idx = findAllSubmatchHelper(re.start, str_runes, idx, re.numGroups, re.preferLongest)
 		if matchFound {
 			matchNum++
 		}
@@ -179,237 +188,187 @@ func FindNthMatch(regex Reg, str string, n int) (Match, error) {
 	return nil, fmt.Errorf("invalid match index - too few matches found")
 }
 
-// FindAllMatches tries to find all matches of the regex represented by given start-state, with
-// the given string
-func FindAllMatches(regex Reg, str string) []Match {
+// FindAllSubmatch returns a slice of matches in the given string.
+func (re Reg) FindAllSubmatch(str string) []Match {
 	idx := 0
 	str_runes := []rune(str)
 	var matchFound bool
 	var matchIdx Match
 	indices := make([]Match, 0)
 	for idx <= len(str_runes) {
-		matchFound, matchIdx, idx = findAllMatchesHelper(regex.start, str_runes, idx, regex.numGroups)
+		matchFound, matchIdx, idx = findAllSubmatchHelper(re.start, str_runes, idx, re.numGroups, re.preferLongest)
 		if matchFound {
 			indices = append(indices, matchIdx)
 		}
 	}
-	if len(indices) > 0 {
-		return pruneIndices(indices)
-	}
+
 	return indices
 }
 
+func addStateToList(str []rune, idx int, list []nfaState, state nfaState, threadGroups []Group, visited []nfaState, preferLongest bool) []nfaState {
+	if stateExists(list, state) || stateExists(visited, state) {
+		return list
+	}
+	visited = append(visited, state)
+
+	if state.isKleene || state.isQuestion {
+		copyThread(state.splitState, state)
+		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)
+		copyThread(state.splitState, state)
+		list = addStateToList(str, idx, list, *state.splitState, threadGroups, visited, preferLongest)
+		return list
+	}
+	state.threadGroups = append([]Group{}, threadGroups...)
+	if state.assert != noneAssert {
+		if state.checkAssertion(str, idx, preferLongest) {
+			copyThread(state.next, state)
+			return addStateToList(str, idx, list, *state.next, state.threadGroups, visited, preferLongest)
+		}
+	}
+	if state.groupBegin {
+		state.threadGroups[state.groupNum].StartIdx = idx
+		return addStateToList(str, idx, list, *state.next, state.threadGroups, visited, preferLongest)
+	}
+	if state.groupEnd {
+		state.threadGroups[state.groupNum].EndIdx = idx
+		return addStateToList(str, idx, list, *state.next, state.threadGroups, visited, preferLongest)
+	}
+	return append(list, state)
+
+}
+
 // Helper for FindAllMatches. Returns whether it found a match, the
 // first Match it finds, and how far it got into the string ie. where
 // the next search should start from.
-//
-//	Might return duplicates or overlapping indices, so care must be taken to prune the resulting array.
-func findAllMatchesHelper(start *nfaState, str []rune, offset int, numGroups int) (bool, Match, int) {
+func findAllSubmatchHelper(start *nfaState, str []rune, offset int, numGroups int, preferLongest bool) (bool, Match, int) {
 	// Base case - exit if offset exceeds string's length
 	if offset > len(str) {
 		// The second value here shouldn't be used, because we should exit when the third return value is > than len(str)
 		return false, []Group{}, offset
 	}
+	resetThreads(start)
 
-	// Hold a list of match indices for the current run. When we
-	// can no longer find a match, the match with the largest range is
-	// chosen as the match for the entire string.
-	// This allows us to pick the longest possible match (which is how greedy matching works).
-	// COMMENT ABOVE IS CURRENTLY NOT UP-TO-DATE
-	tempIndices := newMatch(numGroups + 1)
-
-	foundPath := false
-	startIdx := offset
-	endIdx := offset
-	currentStates := make([]*nfaState, 0)
-	tempStates := make([]*nfaState, 0) // Used to store states that should be used in next loop iteration
+	currentStates := make([]nfaState, 0)
+	nextStates := make([]nfaState, 0)
 	i := offset // Index in string
-	startingFrom := i                  // Store starting index
 
 	// If the first state is an assertion, makes sure the assertion
 	// is true before we do _anything_ else.
 	if start.assert != noneAssert {
-		if start.checkAssertion(str, offset) == false {
+		if start.checkAssertion(str, offset, preferLongest) == false {
 			i++
 			return false, []Group{}, i
 		}
 	}
-	// Increment until we hit a character matching the start state (assuming not 0-state)
-	if start.isEmpty == false {
-		for i < len(str) && !start.contentContains(str, i) {
-			i++
-		}
-		startIdx = i
-		startingFrom = i
-		i++ // Advance to next character (if we aren't at a 0-state, which doesn't match anything), so that we can check for transitions. If we advance at a 0-state, we will never get a chance to match the first character
-	}
 
 	start.threadGroups = newMatch(numGroups + 1)
-	// Check if the start state begins a group - if so, add the start index to our list
-	if start.groupBegin {
-		start.threadGroups[start.groupNum].StartIdx = i
-		//		tempIndices[start.groupNum].startIdx = i
-	}
-
-	currentStates = append(currentStates, start)
-
-	// Main loop
-	for i < len(str) {
-		foundPath = false
-
-		zeroStates := make([]*nfaState, 0)
-		// Keep taking zero-states, until there are no more left to take
-		// Objective: If any of our current states have transitions to 0-states, replace them with the 0-state. Do this until there are no more transitions to 0-states, or there are no more unique 0-states to take.
-		zeroStates, isZero := takeZeroState(currentStates, numGroups, i)
-		tempStates = append(tempStates, zeroStates...)
-		num_appended := 0
-		for isZero == true {
-			zeroStates, isZero = takeZeroState(tempStates, numGroups, i)
-			tempStates, num_appended = unique_append(tempStates, zeroStates...)
-			if num_appended == 0 { // Break if we haven't appended any more unique values
+	start.threadGroups[0].StartIdx = i
+	currentStates = addStateToList(str, i, currentStates, *start, start.threadGroups, nil, preferLongest)
+	var match Match = nil
+	for idx := i; idx <= len(str); idx++ {
+		if len(currentStates) == 0 {
 			break
 		}
+		for currentStateIdx := 0; currentStateIdx < len(currentStates); currentStateIdx++ {
+			currentState := currentStates[currentStateIdx]
+
+			if currentState.threadGroups == nil {
+				currentState.threadGroups = newMatch(numGroups + 1)
+				currentState.threadGroups[0].StartIdx = idx
 			}
 
-		currentStates, _ = unique_append(currentStates, tempStates...)
-		tempStates = nil
+			if currentState.isLast {
+				currentState.threadGroups[0].EndIdx = idx
+				match = append([]Group{}, currentState.threadGroups...)
+				if !preferLongest {
+					break
+				}
+			} else if !currentState.isAlternation && !currentState.isKleene && !currentState.isQuestion && !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)
+				}
+			}
+		}
+		currentStates = append([]nfaState{}, nextStates...)
+		nextStates = nil
+	}
+	if match != nil {
+		if offset == match[0].EndIdx {
+			return true, match, match[0].EndIdx + 1
+		}
+		return true, match, match[0].EndIdx
+	}
+	return false, []Group{}, i + 1
+}
 
-		// Take any transitions corresponding to current character
-		numStatesMatched := 0            // The number of states which had at least 1 match for this round
-		assertionFailed := false         // Whether or not an assertion failed for this round
-		lastStateInList := false         // Whether or not a last state was in our list of states
-		var lastStatePtr *nfaState = nil // Pointer to the last-state, if it was found
-		lastLookaroundInList := false    // Whether or not a last state (that is a lookaround) was in our list of states
-		for _, state := range currentStates {
-			matches, numMatches := state.matchesFor(str, i)
-			if numMatches > 0 {
-				numStatesMatched++
-				tempStates = append(tempStates, matches...)
-				foundPath = true
-				for _, m := range matches {
-					if m.threadGroups == nil {
-						m.threadGroups = newMatch(numGroups + 1)
+// Expand appends template to dst, expanding any variables in template to the relevant capturing group.
+//
+// A variable is of the form '$n', where 'n' is a number. It will be replaced by the contents of the n-th capturing group.
+// To insert a literal $, do not put a number after it. Alternatively, you can use $$.
+// src is the input string, and match must be the result of [Reg.FindSubmatch].
+func (re Reg) Expand(dst string, template string, src string, match Match) string {
+	templateRuneSlc := []rune(template)
+	srcRuneSlc := []rune(src)
+	i := 0
+	for i < len(templateRuneSlc) {
+		c := templateRuneSlc[i]
+		if c == '$' {
+			i += 1
+			// The dollar sign is the last character of the string, or it is proceeded by another dollar sign
+			if i >= len(templateRuneSlc) || templateRuneSlc[i] == '$' {
+				dst += "$"
+				i++
+			} else {
+				numStr := ""
+				for unicode.IsDigit(templateRuneSlc[i]) {
+					numStr += string(templateRuneSlc[i])
+					i++
 				}
-					copy(m.threadGroups, state.threadGroups)
+				if numStr == "" {
+					dst += "$"
+				} else {
+					num, _ := strconv.Atoi(numStr)
+					if num < len(match) {
+						dst += string(srcRuneSlc[match[num].StartIdx:match[num].EndIdx])
+					} else {
+						dst += "$" + numStr
 					}
 				}
-			if numMatches < 0 {
-				assertionFailed = true
+			}
+		} else {
+			dst += string(c)
+			i++
+		}
+	}
+	return dst
+}
+
+// LiteralPrefix returns a string that must begin any match of the given regular expression.
+// The second return value is true if the string comprises the entire expression.
+func (re Reg) LiteralPrefix() (prefix string, complete bool) {
+	state := re.start
+	if state.assert != noneAssert {
+		state = state.next
+	}
+	for !(state.isLast) && (!state.isAlternation) && len(state.content) == 1 && state.assert == noneAssert {
+		if state.groupBegin || state.groupEnd {
+			state = state.next
+			continue
+		}
+		prefix += string(rune(state.content[0]))
+		state = state.next
 	}
 	if state.isLast {
-				if state.isLookaround() {
-					lastLookaroundInList = true
-				}
-				lastStateInList = true
-				lastStatePtr = state
-			}
-		}
-
-		if assertionFailed && numStatesMatched == 0 { // Nothing has matched and an assertion has failed
-			// If I'm being completely honest, I'm not sure why I have to check specifically for a _lookaround_
-			// state. The explanation below is my attempt to explain this behavior.
-			// If you replace 'lastLookaroundInList' with 'lastStateInList', one of the test cases fails.
-			//
-			// One of the states in our list was a last state and a lookaround. In this case, we
-			// don't abort upon failure of the assertion, because we have found
-			// another path to a final state.
-			// Even if the last state _was_ an assertion, we can use the previously
-			// saved indices to find a match.
-			if lastLookaroundInList {
-				break
+		complete = true
 	} else {
-				if i == startingFrom {
-					i++
+		complete = false
 	}
-				return false, []Group{}, i
-			}
-		}
-		// Check if we can find a state in our list that is:
-		// 	a. A last-state
-		// 	b. Empty
-		// 	c. Doesn't assert anything
-		for _, s := range currentStates {
-			if s.isLast && s.isEmpty && s.assert == noneAssert {
-				lastStatePtr = s
-				lastStateInList = true
-			}
-		}
-		if lastStateInList { // A last-state was in the list of states. add the matchIndex to our MatchIndex list
-			for j := 1; j < numGroups+1; j++ {
-				tempIndices[j] = lastStatePtr.threadGroups[j]
-			}
-			endIdx = i
-			tempIndices[0] = Group{startIdx, endIdx}
-		}
-
-		// Check if we can find a zero-length match
-		if foundPath == false {
-			if ok := zeroMatchPossible(str, i, numGroups, currentStates...); ok {
-				if tempIndices[0].isValid() == false {
-					tempIndices[0] = Group{startIdx, startIdx}
-				}
-			}
-			// If we haven't moved in the string, increment the counter by 1
-			// to ensure we don't keep trying the same string over and over.
-			//			if i == startingFrom {
-			startIdx++
-			//	i++
-			//			}
-			if tempIndices.numValidGroups() > 0 && tempIndices[0].isValid() {
-				if tempIndices[0].StartIdx == tempIndices[0].EndIdx { // If we have a zero-length match, we have to shift the index at which we start. Otherwise we keep looking at the same paert of the string over and over.
-					return true, tempIndices, tempIndices[0].EndIdx + 1
-				} else {
-					return true, tempIndices, tempIndices[0].EndIdx
-				}
-			}
-			return false, []Group{}, startIdx
-		}
-		currentStates = make([]*nfaState, len(tempStates))
-		copy(currentStates, tempStates)
-		tempStates = nil
-
-		i++
-	}
-
-	// End-of-string reached. Go to any 0-states, until there are no more 0-states to go to. Then check if any of our states are in the end position.
-	// This is the exact same algorithm used inside the loop, so I should probably put it in a function.
-	zeroStates, isZero := takeZeroState(currentStates, numGroups, i)
-	tempStates = append(tempStates, zeroStates...)
-	num_appended := 0 // Number of unique states addded to tempStates
-	for isZero == true {
-		zeroStates, isZero = takeZeroState(tempStates, numGroups, i)
-		tempStates, num_appended = unique_append(tempStates, zeroStates...)
-		if num_appended == 0 { // Break if we haven't appended any more unique values
-			break
-		}
-	}
-
-	currentStates = append(currentStates, tempStates...)
-	tempStates = nil
-
-	for _, state := range currentStates {
-		// Only add the match if the start index is in bounds. If the state has an assertion,
-		// make sure the assertion checks out.
-		if state.isLast && i <= len(str) {
-			if state.assert == noneAssert || state.checkAssertion(str, i) {
-				for j := 1; j < numGroups+1; j++ {
-					tempIndices[j] = state.threadGroups[j]
-				}
-				endIdx = i
-				tempIndices[0] = Group{startIdx, endIdx}
-			}
-		}
-	}
-
-	if tempIndices.numValidGroups() > 0 {
-		if tempIndices[0].StartIdx == tempIndices[0].EndIdx { // If we have a zero-length match, we have to shift the index at which we start. Otherwise we keep looking at the same paert of the string over and over.
-			return true, tempIndices, tempIndices[0].EndIdx + 1
-		} else {
-			return true, tempIndices, tempIndices[0].EndIdx
-		}
-	}
-	if startIdx == startingFrom { // Increment starting index if we haven't moved in the string. Prevents us from matching the same part of the string over and over.
-		startIdx++
-	}
-	return false, []Group{}, startIdx
+	return prefix, complete
 }

regex/misc.go

@@ -15,7 +15,7 @@ var lparenRune rune = 0xF0005  // Parentheses in regex are concatenated with thi
 var rparenRune rune = 0xF0006
 var nonCapLparenRune rune = 0xF0007 // Represents a non-capturing group's LPAREN
 var escBackslashRune rune = 0xF0008 // Represents an escaped backslash
-var CHAR_RANGE rune = 0xF0009       // Represents a character range
+var charRangeRune rune = 0xF0009    // Represents a character range
 
 var specialChars = []rune{'?', '*', '\\', '^', '$', '{', '}', '(', ')', '[', ']', '+', '|', '.', concatRune, '<', '>', lbracketRune, rbracketRune, nonCapLparenRune}
 
@@ -48,30 +48,6 @@ func isNormalChar(c rune) bool {
 	return !slices.Contains(specialChars, c)
 }
 
-// Ensure that the given elements are only appended to the given slice if they
-// don't already exist. Returns the new slice, and the number of unique items appended.
-func unique_append[T comparable](slc []T, items ...T) ([]T, int) {
-	num_appended := 0
-	for _, item := range items {
-		if !slices.Contains(slc, item) {
-			slc = append(slc, item)
-			num_appended++
-		}
-	}
-	return slc, num_appended
-}
-
-// Returns true only if all the given elements are equal
-func allEqual[T comparable](items ...T) bool {
-	first := items[0]
-	for _, item := range items {
-		if item != first {
-			return false
-		}
-	}
-	return true
-}
-
 // Map function - convert a slice of T to a slice of V, based on a function
 // that maps a T to a V
 func funcMap[T, V any](slc []T, fn func(T) V) []V {

regex/nfa.go

@@ -11,8 +11,10 @@ type assertType int
 
 const (
 	noneAssert assertType = iota
-	sosAssert
-	eosAssert
+	sosAssert             // Start of string (^)
+	soiAssert             // Start of input (\A)
+	eosAssert             // End of string ($)
+	eoiAssert             // End of input (\Z)
 	wboundAssert
 	nonwboundAssert
 	plaAssert        // Positive lookahead
@@ -27,8 +29,12 @@ type nfaState struct {
 	isEmpty bool          // If it is empty - Union operator and Kleene star states will be empty
 	isLast  bool          // If it is the last state (acept state)
 	output  []*nfaState   // The outputs of the current state ie. the 'outward arrows'. A union operator state will have more than one of these.
-	transitions                map[int][]*nfaState // Transitions to different states (maps a character (int representation) to a _list of states. This is useful if one character can lead multiple states eg. ab|aa)
+	//	transitions                map[int][]*nfaState // Transitions to different states (maps a character (int representation) to a _list of states. This is useful if one character can lead multiple states eg. ab|aa)
+	next                       *nfaState  // The next state (not for alternation or kleene states)
 	isKleene                   bool       // Identifies whether current node is a 0-state representing Kleene star
+	isQuestion                 bool       // Identifies whether current node is a 0-state representing the question operator
+	isAlternation              bool       // Identifies whether current node is a 0-state representing an alternation
+	splitState                 *nfaState  // Only for alternation states - the 'other' branch of the alternation ('next' is the first)
 	assert                     assertType // Type of assertion of current node - NONE means that the node doesn't assert anything
 	allChars                   bool       // Whether or not the state represents all characters (eg. a 'dot' metacharacter). A 'dot' node doesn't store any contents directly, as it would take up too much space
 	except                     []rune     // Only valid if allChars is true - match all characters _except_ the ones in this block. Useful for inverting character classes.
@@ -66,8 +72,9 @@ func cloneStateHelper(stateToClone *nfaState, cloneMap map[*nfaState]*nfaState)
 		isEmpty:         stateToClone.isEmpty,
 		isLast:          stateToClone.isLast,
 		output:          make([]*nfaState, len(stateToClone.output)),
-		transitions:     make(map[int][]*nfaState),
 		isKleene:        stateToClone.isKleene,
+		isQuestion:      stateToClone.isQuestion,
+		isAlternation:   stateToClone.isAlternation,
 		assert:          stateToClone.assert,
 		zeroMatchFound:  stateToClone.zeroMatchFound,
 		allChars:        stateToClone.allChars,
@@ -85,26 +92,48 @@ func cloneStateHelper(stateToClone *nfaState, cloneMap map[*nfaState]*nfaState)
 			clone.output[i] = cloneStateHelper(s, cloneMap)
 		}
 	}
-	for k, v := range stateToClone.transitions {
-		clone.transitions[k] = make([]*nfaState, len(v))
-		for i, s := range v {
-			if s == stateToClone {
-				clone.transitions[k][i] = clone
-			} else {
-				clone.transitions[k][i] = cloneStateHelper(s, cloneMap)
-			}
-		}
-	}
 	if stateToClone.lookaroundNFA == stateToClone {
 		clone.lookaroundNFA = clone
 	}
 	clone.lookaroundNFA = cloneStateHelper(stateToClone.lookaroundNFA, cloneMap)
+	if stateToClone.splitState == stateToClone {
+		clone.splitState = clone
+	}
+	clone.splitState = cloneStateHelper(stateToClone.splitState, cloneMap)
+	if stateToClone.next == stateToClone {
+		clone.next = clone
+	}
+	clone.next = cloneStateHelper(stateToClone.next, cloneMap)
 	return clone
 }
 
+// Reset any thread-related fields of the NFA starting from the given state.
+func resetThreads(start *nfaState) {
+	visitedMap := make(map[*nfaState]bool) // The value type doesn't matter here
+	resetThreadsHelper(start, visitedMap)
+}
+
+func resetThreadsHelper(state *nfaState, visitedMap map[*nfaState]bool) {
+	if state == nil {
+		return
+	}
+	if _, ok := visitedMap[state]; ok {
+		return
+	}
+	// Assuming it hasn't been visited
+	state.threadGroups = nil
+	visitedMap[state] = true
+	if state.isAlternation {
+		resetThreadsHelper(state.next, visitedMap)
+		resetThreadsHelper(state.splitState, visitedMap)
+	} else {
+		resetThreadsHelper(state.next, visitedMap)
+	}
+}
+
 // Checks if the given state's assertion is true. Returns true if the given
 // state doesn't have an assertion.
-func (s nfaState) checkAssertion(str []rune, idx int) bool {
+func (s nfaState) checkAssertion(str []rune, idx int, preferLongest bool) bool {
 	if s.assert == alwaysTrueAssert {
 		return true
 	}
@@ -119,6 +148,15 @@ func (s nfaState) checkAssertion(str []rune, idx int) bool {
 		// Index is at the end of the string, or it points to the last character which is a newline
 		return idx == len(str) || (multilineMode && str[idx] == '\n')
 	}
+	if s.assert == soiAssert {
+		// Only true at the start of the input, regardless of mode
+		return idx == 0
+	}
+	if s.assert == eoiAssert {
+		// Only true at the end of the input, regardless of mode
+		return idx == len(str)
+	}
+
 	if s.assert == wboundAssert {
 		return isWordBoundary(str, idx)
 	}
@@ -145,17 +183,18 @@ func (s nfaState) checkAssertion(str []rune, idx int) bool {
 			strToMatch = string(runesToMatch)
 		}
 
-		matchIndices := FindAllMatches(Reg{startState, s.lookaroundNumCaptureGroups}, strToMatch)
+		regComp := Reg{startState, s.lookaroundNumCaptureGroups, s.lookaroundRegex, preferLongest}
+		matchIndices := regComp.FindAll(strToMatch)
 
 		numMatchesFound := 0
 		for _, matchIdx := range matchIndices {
 			if s.assert == plaAssert || s.assert == nlaAssert { // Lookahead - return true (or false) if at least one match starts at 0. Zero is used because the test-string _starts_ from idx.
-				if matchIdx[0].StartIdx == 0 {
+				if matchIdx.StartIdx == 0 {
 					numMatchesFound++
 				}
 			}
 			if s.assert == plbAssert || s.assert == nlbAssert { // Lookbehind - return true (or false) if at least one match _ends_ at the current index.
-				if matchIdx[0].EndIdx == idx {
+				if matchIdx.EndIdx == idx {
 					numMatchesFound++
 				}
 			}
@@ -171,9 +210,12 @@ func (s nfaState) checkAssertion(str []rune, idx int) bool {
 }
 
 // Returns true if the contents of 's' contain the value at the given index of the given string
-func (s nfaState) contentContains(str []rune, idx int) bool {
+func (s nfaState) contentContains(str []rune, idx int, preferLongest bool) bool {
 	if s.assert != noneAssert {
-		return s.checkAssertion(str, idx)
+		return s.checkAssertion(str, idx, preferLongest)
+	}
+	if idx >= len(str) {
+		return false
 	}
 	if s.allChars {
 		return !slices.Contains(slices.Concat(notDotChars, s.except), str[idx]) // Return true only if the index isn't a 'notDotChar', or isn't one of the exception characters for the current node.
@@ -186,74 +228,84 @@ func (s nfaState) isLookaround() bool {
 	return s.assert == plaAssert || s.assert == plbAssert || s.assert == nlaAssert || s.assert == nlbAssert
 }
 
+func (s nfaState) numTransitions() int {
+	if s.next == nil && s.splitState == nil {
+		return 0
+	}
+	if s.next == nil || s.splitState == nil {
+		return 1
+	}
+	return 2
+}
+
 // Returns the matches for the character at the given index of the given string.
 // Also returns the number of matches. Returns -1 if an assertion failed.
-func (s nfaState) matchesFor(str []rune, idx int) ([]*nfaState, int) {
-	// Assertions can be viewed as 'checks'. If the check fails, we return
-	// an empty array and 0.
-	// If it passes, we treat it like any other state, and return all the transitions.
-	if s.assert != noneAssert {
-		if s.checkAssertion(str, idx) == false {
-			return make([]*nfaState, 0), -1
-		}
-	}
-	listTransitions := s.transitions[int(str[idx])]
-	for _, dest := range s.transitions[int(anyCharRune)] {
-		if !slices.Contains(slices.Concat(notDotChars, dest.except), str[idx]) {
-			// Add an allChar state to the list of matches if:
-			// 		a. The current character isn't a 'notDotChars' character. In single line mode, this includes newline. In multiline mode, it doesn't.
-			// 		b. The current character isn't the state's exception list.
-			listTransitions = append(listTransitions, dest)
-		}
-	}
-	numTransitions := len(listTransitions)
-	return listTransitions, numTransitions
-}
+//func (s nfaState) matchesFor(str []rune, idx int) ([]*nfaState, int) {
+//	// Assertions can be viewed as 'checks'. If the check fails, we return
+//	// an empty array and 0.
+//	// If it passes, we treat it like any other state, and return all the transitions.
+//	if s.assert != noneAssert {
+//		if s.checkAssertion(str, idx) == false {
+//			return make([]*nfaState, 0), -1
+//		}
+//	}
+//	listTransitions := s.transitions[int(str[idx])]
+//	for _, dest := range s.transitions[int(anyCharRune)] {
+//		if !slices.Contains(slices.Concat(notDotChars, dest.except), str[idx]) {
+//			// Add an allChar state to the list of matches if:
+//			// 		a. The current character isn't a 'notDotChars' character. In single line mode, this includes newline. In multiline mode, it doesn't.
+//			// 		b. The current character isn't the state's exception list.
+//			listTransitions = append(listTransitions, dest)
+//		}
+//	}
+//	numTransitions := len(listTransitions)
+//	return listTransitions, numTransitions
+//}
 
 // verifyLastStatesHelper performs the depth-first recursion needed for verifyLastStates
-func verifyLastStatesHelper(st *nfaState, visited map[*nfaState]bool) {
-	if len(st.transitions) == 0 {
-		st.isLast = true
-		return
-	}
-	//	if len(state.transitions) == 1 && len(state.transitions[state.content]) == 1 && state.transitions[state.content][0] == state { // Eg. a*
-	if len(st.transitions) == 1 { // Eg. a*
-		var moreThanOneTrans bool // Dummy variable, check if all the transitions for the current's state's contents have a length of one
-		for _, c := range st.content {
-			if len(st.transitions[c]) != 1 || st.transitions[c][0] != st {
-				moreThanOneTrans = true
-			}
-		}
-		st.isLast = !moreThanOneTrans
-	}
-
-	if st.isKleene { // A State representing a Kleene Star has transitions going out, which loop back to it. If all those transitions point to the same (single) state, then it must be a last state
-		transitionDests := make([]*nfaState, 0)
-		for _, v := range st.transitions {
-			transitionDests = append(transitionDests, v...)
-		}
-		if allEqual(transitionDests...) {
-			st.isLast = true
-			return
-		}
-	}
-	if visited[st] == true {
-		return
-	}
-	visited[st] = true
-	for _, states := range st.transitions {
-		for i := range states {
-			if states[i] != st {
-				verifyLastStatesHelper(states[i], visited)
-			}
-		}
-	}
-}
+//func verifyLastStatesHelper(st *nfaState, visited map[*nfaState]bool) {
+//	if st.numTransitions() == 0 {
+//		st.isLast = true
+//		return
+//	}
+//	//	if len(state.transitions) == 1 && len(state.transitions[state.content]) == 1 && state.transitions[state.content][0] == state { // Eg. a*
+//	if st.numTransitions() == 1 { // Eg. a*
+//		var moreThanOneTrans bool // Dummy variable, check if all the transitions for the current's state's contents have a length of one
+//		for _, c := range st.content {
+//			if len(st.transitions[c]) != 1 || st.transitions[c][0] != st {
+//				moreThanOneTrans = true
+//			}
+//		}
+//		st.isLast = !moreThanOneTrans
+//	}
+//
+//	if st.isKleene { // A State representing a Kleene Star has transitions going out, which loop back to it. If all those transitions point to the same (single) state, then it must be a last state
+//		transitionDests := make([]*nfaState, 0)
+//		for _, v := range st.transitions {
+//			transitionDests = append(transitionDests, v...)
+//		}
+//		if allEqual(transitionDests...) {
+//			st.isLast = true
+//			return
+//		}
+//	}
+//	if visited[st] == true {
+//		return
+//	}
+//	visited[st] = true
+//	for _, states := range st.transitions {
+//		for i := range states {
+//			if states[i] != st {
+//				verifyLastStatesHelper(states[i], visited)
+//			}
+//		}
+//	}
+//}
 
 // verifyLastStates enables the 'isLast' flag for the leaf nodes (last states)
-func verifyLastStates(start []*nfaState) {
-	verifyLastStatesHelper(start[0], make(map[*nfaState]bool))
-}
+//func verifyLastStates(start []*nfaState) {
+//	verifyLastStatesHelper(start[0], make(map[*nfaState]bool))
+//}
 
 // Concatenates s1 and s2, returns the start of the concatenation.
 func concatenate(s1 *nfaState, s2 *nfaState) *nfaState {
@@ -261,73 +313,84 @@ func concatenate(s1 *nfaState, s2 *nfaState) *nfaState {
 		return s2
 	}
 	for i := range s1.output {
-		for _, c := range s2.content { // Create transitions for every element in s1's content to s2'
-			s1.output[i].transitions[c], _ = unique_append(s1.output[i].transitions[c], s2)
-		}
+		s1.output[i].next = s2
 	}
 	s1.output = s2.output
 	return s1
 }
 
-func kleene(s1 nfaState) (*nfaState, error) {
+func kleene(s1 *nfaState) (*nfaState, error) {
 	if s1.isEmpty && s1.assert != noneAssert {
 		return nil, fmt.Errorf("previous token is not quantifiable")
 	}
 
 	toReturn := &nfaState{}
-	toReturn.transitions = make(map[int][]*nfaState)
-	toReturn.content = newContents(epsilon)
 	toReturn.isEmpty = true
+	toReturn.isAlternation = true
+	toReturn.content = newContents(epsilon)
+	toReturn.splitState = s1
+
+	//	toReturn := &nfaState{}
+	//	toReturn.transitions = make(map[int][]*nfaState)
+	//	toReturn.content = newContents(epsilon)
 	toReturn.isKleene = true
-	toReturn.output = append(toReturn.output, toReturn)
+	toReturn.output = append([]*nfaState{}, toReturn)
 	for i := range s1.output {
-		for _, c := range toReturn.content {
-			s1.output[i].transitions[c], _ = unique_append(s1.output[i].transitions[c], toReturn)
-		}
-	}
-	for _, c := range s1.content {
-		toReturn.transitions[c], _ = unique_append(toReturn.transitions[c], &s1)
+		s1.output[i].next = toReturn
 	}
+	//	for _, c := range s1.content {
+	//		toReturn.transitions[c], _ = uniqueAppend(toReturn.transitions[c], &s1)
+	//	}
+	//toReturn.kleeneState = &s1
 	return toReturn, nil
 }
 
 func alternate(s1 *nfaState, s2 *nfaState) *nfaState {
 	toReturn := &nfaState{}
-	toReturn.transitions = make(map[int][]*nfaState)
+	//	toReturn.transitions = make(map[int][]*nfaState)
 	toReturn.output = append(toReturn.output, s1.output...)
 	toReturn.output = append(toReturn.output, s2.output...)
-	// Unique append is used here (and elsewhere) to ensure that,
-	// for any given transition, a state can only be mentioned once.
-	// For example, given the transition 'a', the state 's1' can only be mentioned once.
-	// This would lead to multiple instances of the same set of match indices, since both
-	// 's1' states would be considered to match.
-	for _, c := range s1.content {
-		toReturn.transitions[c], _ = unique_append(toReturn.transitions[c], s1)
-	}
-	for _, c := range s2.content {
-		toReturn.transitions[c], _ = unique_append(toReturn.transitions[c], s2)
-	}
+	//	// Unique append is used here (and elsewhere) to ensure that,
+	//	// for any given transition, a state can only be mentioned once.
+	//	// For example, given the transition 'a', the state 's1' can only be mentioned once.
+	//	// This would lead to multiple instances of the same set of match indices, since both
+	//	// 's1' states would be considered to match.
+	//	for _, c := range s1.content {
+	//		toReturn.transitions[c], _ = uniqueAppend(toReturn.transitions[c], s1)
+	//	}
+	//	for _, c := range s2.content {
+	//		toReturn.transitions[c], _ = uniqueAppend(toReturn.transitions[c], s2)
+	//	}
 	toReturn.content = newContents(epsilon)
 	toReturn.isEmpty = true
+	toReturn.isAlternation = true
+	toReturn.next = s1
+	toReturn.splitState = s2
 
 	return toReturn
 }
 
-func question(s1 *nfaState) *nfaState { // Use the fact that ab? == a(b|)
-	s2 := &nfaState{}
-	s2.transitions = make(map[int][]*nfaState)
-	s2.content = newContents(epsilon)
-	s2.output = append(s2.output, s2)
-	s2.isEmpty = true
-	s3 := alternate(s1, s2)
-	return s3
+func question(s1 *nfaState) (*nfaState, error) { // Use the fact that ab? == a(b|)
+	if s1.isEmpty && s1.assert != noneAssert {
+		return nil, fmt.Errorf("previous token is not quantifiable")
+	}
+	toReturn := &nfaState{}
+	toReturn.isEmpty = true
+	toReturn.isAlternation = true
+	toReturn.isQuestion = true
+	toReturn.content = newContents(epsilon)
+	toReturn.splitState = s1
+	toReturn.output = append([]*nfaState{}, toReturn)
+	toReturn.output = append(toReturn.output, s1.output...)
+	//	s2.transitions = make(map[int][]*nfaState)
+	return toReturn, nil
 }
 
 // Creates and returns a new state with the 'default' values.
 func newState() nfaState {
 	ret := nfaState{
 		output: make([]*nfaState, 0),
-		transitions:     make(map[int][]*nfaState),
+		//		transitions:     make(map[int][]*nfaState),
 		assert:          noneAssert,
 		except:          append([]rune{}, 0),
 		lookaroundRegex: "",
@@ -339,10 +402,40 @@ func newState() nfaState {
 }
 
 // Creates and returns a state that _always_ has a zero-length match.
-func zeroLengthMatchState() nfaState {
-	start := newState()
+func zeroLengthMatchState() *nfaState {
+	start := &nfaState{}
 	start.content = newContents(epsilon)
 	start.isEmpty = true
 	start.assert = alwaysTrueAssert
+	start.output = append([]*nfaState{}, start)
 	return start
 }
+
+func (s nfaState) equals(other nfaState) bool {
+	return s.isEmpty == other.isEmpty &&
+		s.isLast == other.isLast &&
+		slices.Equal(s.output, other.output) &&
+		slices.Equal(s.content, other.content) &&
+		s.next == other.next &&
+		s.isKleene == other.isKleene &&
+		s.isQuestion == other.isQuestion &&
+		s.isAlternation == other.isAlternation &&
+		s.splitState == other.splitState &&
+		s.assert == other.assert &&
+		s.allChars == other.allChars &&
+		slices.Equal(s.except, other.except) &&
+		s.lookaroundNFA == other.lookaroundNFA &&
+		s.groupBegin == other.groupBegin &&
+		s.groupEnd == other.groupEnd &&
+		s.groupNum == other.groupNum &&
+		slices.Equal(s.threadGroups, other.threadGroups)
+}
+
+func stateExists(list []nfaState, s nfaState) bool {
+	for i := range list {
+		if list[i].equals(s) {
+			return true
+		}
+	}
+	return false
+}

regex/postfixNode.go

@@ -95,6 +95,16 @@ func newEscapedNode(c rune, inCharClass bool) (postfixNode, error) {
 			toReturn.nodetype = assertionNode
 			toReturn.contents = append(toReturn.contents, c)
 		}
+		if c == 'B' && inCharClass { // Invalid
+			return postfixNode{}, fmt.Errorf("word boundaries are not allowed in character class")
+		}
+	case 'A', 'z': // A is start of input, z is end of input (regardless of RE_MULTILINE)
+		if inCharClass {
+			return postfixNode{}, fmt.Errorf("input boundaries are not allowed in character class")
+		} else {
+			toReturn.nodetype = assertionNode
+			toReturn.contents = append(toReturn.contents, c)
+		}
 	case 'n': // Newline character
 		toReturn.nodetype = characterNode
 		toReturn.contents = append(toReturn.contents, '\n')

regex/range2regex.go

@@ -3,7 +3,9 @@ package regex
 import (
 	"fmt"
 	"math"
+	"slices"
 	"strconv"
+	"strings"
 )
 
 type numRange struct {
@@ -99,28 +101,39 @@ func range2regex(start int, end int) (string, error) {
 	// Last range - tmp to rangeEnd
 	ranges = append(ranges, numRange{tmp, rangeEnd})
 
-	regex := string(NONCAPLPAREN_CHAR)
+	regexSlice := make([]string, 0)
 	// Generate the regex
-	for i, rg := range ranges {
-		if i > 0 {
-			regex += "|"
-		}
-		regex += string(NONCAPLPAREN_CHAR)
+	for _, rg := range ranges {
+		tmpStr := ""
+		tmpStr += string(nonCapLparenRune)
 		startSlc := intToSlc(rg.start)
 		endSlc := intToSlc(rg.end)
 		if len(startSlc) != len(endSlc) {
-			return "", fmt.Errorf("Error parsing numeric range")
+			return "", fmt.Errorf("error parsing numeric range")
 		}
 		for i := range startSlc {
 			if startSlc[i] == endSlc[i] {
-				regex += string(rune(startSlc[i] + 48)) // '0' is ascii value 48, 1 is 49 etc. To convert the digit to its character form, we can just add 48.
+				tmpStr += string(rune(startSlc[i] + 48)) // '0' is ascii value 48, 1 is 49 etc. To convert the digit to its character form, we can just add 48.
 			} else {
-				regex += fmt.Sprintf("%c%c-%c%c", LBRACKET, rune(startSlc[i]+48), rune(endSlc[i]+48), RBRACKET)
+				tmpStr += fmt.Sprintf("%c%c-%c%c", lbracketRune, rune(startSlc[i]+48), rune(endSlc[i]+48), rbracketRune)
 			}
 		}
-		regex += ")"
+		tmpStr += ")"
+		regexSlice = append(regexSlice, tmpStr)
 	}
-	regex += ")"
+	// Each element of the slice represents one 'group'. Taking 0-255 as an example, the elements would be:
+	// 	1. 0-9
+	// 	2. 10-99
+	// 	3. 100-199
+	// 	4. 200-249
+	// 	5. 250-255
+	//
+	// The reason this is reversed before joining it, is because it is incompatible with the PCRE rule for matching.
+	// The PCRE rule specifies that the left-branch of an alternation is preferred. Even though this engine uses the POSIX
+	// rule at the moment (which prefers the longest match regardless of the order of the alternation), reversing the string
+	// has no downsides. It doesn't affect POSIX matching, and it will reduce my burden if I decide to switch to PCRE matching.
+	slices.Reverse(regexSlice)
+	regex := string(nonCapLparenRune) + strings.Join(regexSlice, "|") + ")"
 	return regex, nil
 
 }

regex/re_test.go

@@ -25,7 +25,9 @@ var reTests = []struct {
 	{"a*b", nil, "qwqw", []Group{}},
 	{"(abc)*", nil, "abcabcabc", []Group{{0, 9}, {9, 9}}},
 	{"((abc)|(def))*", nil, "abcdef", []Group{{0, 6}, {6, 6}}},
-	{"(abc)*|(def)*", nil, "abcdef", []Group{{0, 3}, {3, 6}, {6, 6}}},
+	// This match will only happen with Longest()
+	// {"(abc)*|(def)*", nil, "abcdef", []Group{{0, 3}, {3, 6}, {6, 6}}},
+	{"(abc)*|(def)*", nil, "abcdef", []Group{{0, 3}, {3, 3}, {4, 4}, {5, 5}, {6, 6}}},
 	{"b*a*a", nil, "bba", []Group{{0, 3}}},
 	{"(ab)+", nil, "abcabddd", []Group{{0, 2}, {3, 5}}},
 	{"a(b(c|d)*)*", nil, "abccbd", []Group{{0, 6}}},
@@ -105,6 +107,9 @@ var reTests = []struct {
 	{"(a|b){3,4}", nil, "ababaa", []Group{{0, 4}}},
 	{"(bc){5,}", nil, "bcbcbcbcbcbcbcbc", []Group{{0, 16}}},
 	{`\d{3,4}`, nil, "1209", []Group{{0, 4}}},
+	{`\d{3,4}`, nil, "120", []Group{{0, 3}}},
+	{`\d{3,4}`, nil, "12709", []Group{{0, 4}}},
+	{`\d{3,4}`, nil, "12", []Group{}},
 	{`\d{3,4}`, nil, "109", []Group{{0, 3}}},
 	{`\d{3,4}`, nil, "5", []Group{}},
 	{`\d{3,4}`, nil, "123135", []Group{{0, 4}}},
@@ -443,8 +448,11 @@ var reTests = []struct {
 	{`abc$`, []ReFlag{RE_MULTILINE}, "jkl\n123abc\nxyz", []Group{{7, 10}}},
 	{`^`, nil, "jkl\n123abc\nxyz", []Group{{0, 0}}},
 	{`^`, []ReFlag{RE_MULTILINE}, "jkl\n123abc\nxyz", []Group{{0, 0}, {4, 4}, {11, 11}}},
+	{`\A`, []ReFlag{RE_MULTILINE}, "jkl\n123abc\nxyz", []Group{{0, 0}}},
 	{`$`, nil, "jkl\n123abc\nxyz", []Group{{14, 14}}},
 	{`$`, []ReFlag{RE_MULTILINE}, "jkl\n123abc\nxyz", []Group{{3, 3}, {10, 10}, {14, 14}}},
+	{`\z`, []ReFlag{RE_MULTILINE}, "jkl\n123abc\nxyz", []Group{{14, 14}}},
+	{`^abc\z`, []ReFlag{RE_MULTILINE}, "abc\nabc\nabc", []Group{{8, 11}}},
 
 	{`a.b`, nil, "a\nb", []Group{}},
 	{`a.b`, []ReFlag{RE_SINGLE_LINE}, "a\nb", []Group{{0, 3}}},
@@ -522,7 +530,7 @@ var groupTests = []struct {
 }{
 	{"(a)(b)", nil, "ab", []Match{[]Group{{0, 2}, {0, 1}, {1, 2}}}},
 	{"((a))(b)", nil, "ab", []Match{[]Group{{0, 2}, {0, 1}, {0, 1}, {1, 2}}}},
-	{"(0)", nil, "ab", []Match{[]Group{}}},
+	{"(0)", nil, "ab", []Match{}},
 	{"(a)b", nil, "ab", []Match{[]Group{{0, 2}, {0, 1}}}},
 	{"a(b)", nil, "ab", []Match{[]Group{{0, 2}, {1, 2}}}},
 	{"(a|b)", nil, "ab", []Match{[]Group{{0, 1}, {0, 1}}, []Group{{1, 2}, {1, 2}}}},
@@ -531,10 +539,11 @@ var groupTests = []struct {
 	{"(a+)|(a)", nil, "aaaa", []Match{[]Group{{0, 4}, {0, 4}, {-1, -1}}}},
 	{"(a+)(aa)", nil, "aaaa", []Match{[]Group{{0, 4}, {0, 2}, {2, 4}}}},
 	{"(aaaa)|(aaaa)", nil, "aaaa", []Match{[]Group{{0, 4}, {0, 4}, {-1, -1}}}},
-	{"(aaa)|(aaaa)", nil, "aaaa", []Match{[]Group{{0, 4}, {-1, -1}, {0, 4}}}},
-	{"(aaa)|(aaaa)", nil, "aaaa", []Match{[]Group{{0, 4}, {-1, -1}, {0, 4}}}},
+	// This match will only happen with Longest()
+	//	{"(aaa)|(aaaa)", nil, "aaaa", []Match{[]Group{{0, 4}, {-1, -1}, {0, 4}}}},
+	{"(aaa)|(aaaa)", nil, "aaaa", []Match{[]Group{{0, 3}, {0, 3}, {-1, -1}}}},
 	{"(aaaa)|(aaa)", nil, "aaaa", []Match{[]Group{{0, 4}, {0, 4}, {-1, -1}}}},
-	{"(a)|(aa)", nil, "aa", []Match{[]Group{{0, 2}, {-1, -1}, {0, 2}}}},
+	{"(a)|(aa)", nil, "aa", []Match{[]Group{{0, 1}, {0, 1}}, []Group{{1, 2}, {1, 2}}}},
 	{"(a?)a?", nil, "b", []Match{[]Group{{0, 0}, {0, 0}}, []Group{{1, 1}, {1, 1}}}},
 	{"(a?)a?", nil, "ab", []Match{[]Group{{0, 1}, {0, 1}}, []Group{{1, 1}, {1, 1}}, []Group{{2, 2}, {2, 2}}}},
 	{"(a?)a?", nil, "aa", []Match{[]Group{{0, 2}, {0, 1}}, []Group{{2, 2}, {2, 2}}}},
@@ -572,7 +581,7 @@ 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)))))))))\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}}}},
 	{`(.*)c(.*)`, nil, `abcde`, []Match{[]Group{{0, 5}, {0, 2}, {3, 5}}}},
 	{`\((.*), (.*)\)`, nil, `(a, b)`, []Match{[]Group{{0, 6}, {1, 2}, {4, 5}}}},
 
@@ -627,7 +636,7 @@ 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)))))))))\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}}}},
 	{`(.*)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}}}},
@@ -668,9 +677,20 @@ var groupTests = []struct {
 	{`^([ab]*)(?<!(a))c`, nil, `abc`, []Match{[]Group{{0, 3}, {0, 2}}}},
 
 	{`(<389-400>)`, nil, `391`, []Match{[]Group{{0, 3}, {0, 3}}}},
+
+	// // Tests from https://wiki.haskell.org/Regex_Posix
+	// {`(()|.)(b)`, nil, `ab`, []Match{[]Group{{0, 2}, {0, 1}, {-1, -1}, {1, 2}}}},
+	// {`(()|[ab])(b)`, nil, `ab`, []Match{[]Group{{0, 2}, {0, 1}, {-1, -1}, {1, 2}}}},
+	// {`(()|[ab])+b`, nil, `aaab`, []Match{[]Group{{0, 4}, {2, 3}, {-1, -1}}}},
+	// {`([ab]|())+b`, nil, `aaab`, []Match{[]Group{{0, 4}, {2, 3}, {-1, -1}}}},
+	// // Bug - this should give {0,6},{3,6},{-1,-1} but it gives {0,6},{3,6},{3,3}
+	// //	{`yyyyyy`, nil, `(yyy|(x?)){2,4}`, []Match{[]Group{{0, 6}, {3, 6}, {-1, -1}}, []Group{{6, 6}, {6, 6}, {6, 6}}}},
+	// {`(a|ab|c|bcd)*(d*)`, nil, `ababcd`, []Match{[]Group{{0, 6}, {3, 6}, {6, 6}}, []Group{{6, 6}, {6, 6}, {6, 6}}}},
+	// // Bug - this should give {0,3},{0,3},{0,0},{0,3},{3,3} but it gives {0,3},{0,2},{0,1},{1,2},{2,3}
+	// //	{`((a*)(b|abc))(c*)`, nil, `abc`, []Match{[]Group{{0, 3}, {0, 3}, {0, 0}, {0, 3}, {3, 3}}}},
 }
 
-func TestFindAllMatches(t *testing.T) {
+func TestFind(t *testing.T) {
 	for _, test := range reTests {
 		t.Run(test.re+"	"+test.str, func(t *testing.T) {
 			regComp, err := Compile(test.re, test.flags...)
@@ -679,13 +699,35 @@ func TestFindAllMatches(t *testing.T) {
 					panic(fmt.Errorf("Test Error: %v", err))
 				}
 			} else {
-				matchIndices := FindAllMatches(regComp, test.str)
-				zeroGroups := make([]Group, len(matchIndices))
-				for i, m := range matchIndices {
-					zeroGroups[i] = m[0]
+				groupIndex, err := regComp.Find(test.str)
+				if err != nil { // No matches found
+					if len(test.result) == 0 {
+						return // Manually pass the test, because this is the expected behavior
+					} else {
+						t.Errorf("Wanted %v Got no matches\n", test.result)
 					}
-				if !slices.Equal(test.result, zeroGroups) {
-					t.Errorf("Wanted %v	Got %v\n", test.result, zeroGroups)
+				} else {
+					if groupIndex != test.result[0] {
+						t.Errorf("Wanted %v	Got %v\n", test.result, groupIndex)
+					}
+				}
+			}
+		})
+	}
+}
+
+func TestFindAll(t *testing.T) {
+	for _, test := range reTests {
+		t.Run(test.re+"	"+test.str, func(t *testing.T) {
+			regComp, err := Compile(test.re, test.flags...)
+			if err != nil {
+				if test.result != nil {
+					panic(fmt.Errorf("Test Error: %v", err))
+				}
+			} else {
+				matchIndices := regComp.FindAll(test.str)
+				if !slices.Equal(test.result, matchIndices) {
+					t.Errorf("Wanted %v	Got %v\n", test.result, matchIndices)
 				}
 			}
 		})
@@ -701,10 +743,10 @@ func TestFindString(t *testing.T) {
 					panic(err)
 				}
 			} else {
-				foundString := FindString(regComp, test.str)
+				foundString := regComp.FindString(test.str)
 				if len(test.result) == 0 {
 					if foundString != "" {
-						t.Errorf("Expected no match got %v\n", foundString)
+						t.Errorf("Wanted no match got %v\n", foundString)
 					}
 				} else {
 					expectedString := test.str[test.result[0].StartIdx:test.result[0].EndIdx]
@@ -717,7 +759,32 @@ func TestFindString(t *testing.T) {
 	}
 }
 
-func TestFindAllGroups(t *testing.T) {
+func TestFindAllString(t *testing.T) {
+	for _, test := range reTests {
+		t.Run(test.re+"	"+test.str, func(t *testing.T) {
+			regComp, err := Compile(test.re, test.flags...)
+			if err != nil {
+				if test.result != nil {
+					panic(err)
+				}
+			} else {
+				foundStrings := regComp.FindAllString(test.str)
+				if len(test.result) != len(foundStrings) {
+					t.Errorf("Differing number of matches: Wanted %v matches Got %v matches\n", len(test.result), len(foundStrings))
+				} else {
+					for idx, group := range test.result {
+						groupStr := test.str[group.StartIdx:group.EndIdx]
+						if groupStr != foundStrings[idx] {
+							t.Errorf("Wanted %v	Got %v\n", groupStr, foundStrings[idx])
+						}
+					}
+				}
+			}
+		})
+	}
+}
+
+func TestFindSubmatch(t *testing.T) {
 	for _, test := range groupTests {
 		t.Run(test.re+"	"+test.str, func(t *testing.T) {
 			regComp, err := Compile(test.re, test.flags...)
@@ -726,13 +793,94 @@ func TestFindAllGroups(t *testing.T) {
 					panic(err)
 				}
 			}
-			matchIndices := FindAllMatches(regComp, test.str)
+			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 {
+					if i < len(test.result) && test.result[0][i].IsValid() {
+						t.Errorf("Wanted %v	Got %v\n", test.result[0], match)
+					}
+				}
+			}
+		})
+	}
+}
+func TestFindStringSubmatch(t *testing.T) {
+	for _, test := range groupTests {
+		t.Run(test.re+"	"+test.str, func(t *testing.T) {
+			regComp, err := Compile(test.re, test.flags...)
+			if err != nil {
+				if test.result != nil {
+					panic(err)
+				}
+			}
+			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]
+					} 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)
+						}
+					}
+				}
+			}
+		})
+	}
+}
+func TestFindAllSubmatch(t *testing.T) {
+	for _, test := range groupTests {
+		t.Run(test.re+"	"+test.str, func(t *testing.T) {
+			regComp, err := Compile(test.re, test.flags...)
+			if err != nil {
+				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 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)
+						}
 					}
 				}
 			}