Removed unicode character classes from 'features not supported' list

Reviewed-on: #3

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.

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

@@ -12,16 +12,43 @@ 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 eturns the number of sub-expressions in the given [Reg]. This is equivalent
+// NumSubexp returns the number of sub-expressions in the given [Reg]. This is equivalent
 // to the number of capturing groups.
-func (r Reg) NumSubexp() int {
+func (re Reg) NumSubexp() int {
-	return r.numGroups
+	return re.numGroups
+}
+
+// String returns the string used to compile the expression.
+func (re Reg) String() string {
+	return re.str
+}
+
+// MarshalText implements [encoding.TextMarshaler]. The output is equivalent to that of [Reg.String].
+// Any flags passed as arguments (including calling [Reg.Longest]) are lost.
+func (re *Reg) MarshalText() ([]byte, error) {
+	return []byte(re.String()), nil
+}
+
+// UnmarshalText implements [encoding.TextUnmarshaler]. It calls [Reg.Compile] on the given byte-slice. If it returns successfully,
+// then the result of the compilation is stored in re. The result of [Reg.Compile] is returned.
+func (re *Reg) UnmarshalText(text []byte) error {
+	newReg, err := Compile(string(text))
+	if err == nil {
+		*re = newReg
+	}
+	return err
+}
+
+func (re *Reg) Longest() {
+	re.preferLongest = true
 }
 
 const concatRune rune = 0xF0001
@@ -81,6 +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
 
@@ -282,17 +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])
 			}
@@ -338,6 +434,8 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
 
 	// Actual algorithm
 	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
@@ -393,11 +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++
 				}
@@ -410,6 +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 {
@@ -561,11 +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++
 						}
@@ -769,6 +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.
@@ -785,6 +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--
 		}
 	}
@@ -799,6 +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
 }
 
@@ -1010,6 +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:
@@ -1128,7 +1328,8 @@ func thompson(re []postfixNode) (Reg, error) {
 
 	concatenate(nfa[0], &lastState)
 
-	return Reg{nfa[0], numGroups}, nil
+	// The string is empty here, because we add it in Compile()
+	return Reg{nfa[0], numGroups, "", false}, nil
 
 }
 
@@ -1146,10 +1347,11 @@ 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 panicks if Compile returns an error. They are identical in all other respects.
+// 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 {

regex/doc.go

@@ -4,6 +4,8 @@ Package regex implements regular expression search, using a custom non-bracktrac
 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
@@ -16,7 +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
@@ -31,7 +33,7 @@ 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_])
+	\W				Match any non-word character ([^a-zA-Z0-9_])
 	\s				Match any whitespace character ([ \t\n])
 	\S				Match any non-whitespace character ([^ \t\n])
 
@@ -55,8 +57,8 @@ POSIX classes (inside normal character classes):
 Composition:
 
 	def				Match d, followed by e, followed by f
-	x|y				Match x or y (prefer longer one)
+	x|y				Match x or y (prefer x)
-	xy|z			Match xy or z
+	xy|z			Match xy or z (prefer xy)
 
 Repitition (always greedy, preferring more):
 
@@ -91,34 +93,23 @@ 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)
+	\<x				Match a literal '<' followed by x
 
 # Key Differences with regexp
 
-The engine and the API differ from [regexp] in a number of ways, some of them very subtle.
+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
+This engine currently does not support non-greedy operators.
-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:
 
@@ -132,7 +123,7 @@ Rather than using primitives for return values, my engine defines two types that
 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:
+equivalent expression for this engine is shown below. Note that 'Index' is the default.
 
 	Find(All)?(String)?(Submatch)?
 
@@ -140,7 +131,7 @@ equivalent expression for this engine is:
 
 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 indices.
+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.
@@ -156,5 +147,20 @@ and the input string:
 
 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. Embedded flags (flags are instead passed as arguments to [Compile])
+ 4. 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.
 */
 package regex

regex/example_test.go

@@ -2,6 +2,7 @@ package regex_test
 
 import (
 	"fmt"
+	"strings"
 
 	"gitea.twomorecents.org/Rockingcool/kleingrep/regex"
 )
@@ -32,12 +33,12 @@ func ExampleReg_FindAll() {
 }
 
 func ExampleReg_FindString() {
-	regexStr := `\d+`
+	regexStr := `\w+\s+(?=sheep)`
 	regexComp := regex.MustCompile(regexStr)
 
-	matchStr := regexComp.FindString("The year of our lord, 2025")
+	matchStr := regexComp.FindString("pink cows and yellow sheep")
 	fmt.Println(matchStr)
-	// Output: 2025
+	// Output: yellow
 }
 
 func ExampleReg_FindSubmatch() {
@@ -52,3 +53,129 @@ func ExampleReg_FindSubmatch() {
 	// 0	1
 	// 2	3
 }
+
+func ExampleReg_FindStringSubmatch() {
+	regexStr := `(\d{4})-(\d{2})-(\d{2})`
+	regexComp := regex.MustCompile(regexStr)
+	inputStr := `The date is 2025-02-10`
+
+	match := regexComp.FindStringSubmatch(inputStr)
+	fmt.Println(match[1])
+	fmt.Println(match[3])
+	// Output: 2025
+	// 10
+}
+
+func ExampleReg_FindAllSubmatch() {
+	regexStr := `(\d)\.(\d)(\d)`
+	regexComp := regex.MustCompile(regexStr)
+
+	matches := regexComp.FindAllSubmatch("3.14+8.97")
+	fmt.Println(matches[0][0]) // 0-group (entire match) of 1st match (0-indexed)
+	fmt.Println(matches[0][1]) // 1st group of 1st match
+	fmt.Println(matches[1][0]) // 0-group of 2nd match
+	fmt.Println(matches[1][1]) // 1st group of 2nd math
+	// Output: 0	4
+	// 0	1
+	// 5	9
+	// 5	6
+}
+
+func ExampleReg_FindAllString() {
+	regexStr := `<0-255>\.<0-255>\.<0-255>\.<0-255>`
+	inputStr := `192.168.220.7 pings 9.9.9.9`
+	regexComp := regex.MustCompile(regexStr)
+
+	matchStrs := regexComp.FindAllString(inputStr)
+
+	fmt.Println(matchStrs[0])
+	fmt.Println(matchStrs[1])
+	// Output: 192.168.220.7
+	// 9.9.9.9
+}
+
+func ExampleReg_FindAllStringSubmatch() {
+	// 'https' ...
+	// followed by 1 or more alphanumeric characters (including period) ...
+	// then a forward slash ...
+	// followed by one more of :
+	// 		word character,
+	// 		question mark,
+	// 		period,
+	// 		equals sign
+	regexStr := `https://([a-z0-9\.]+)/([\w.?=]+)`
+	regexComp := regex.MustCompile(regexStr, regex.RE_CASE_INSENSITIVE)
+	inputStr := `You can find me at https://twomorecents.org/index.html and https://news.ycombinator.com/user?id=aadhavans`
+
+	matchIndices := regexComp.FindAllStringSubmatch(inputStr)
+	fmt.Println(matchIndices[0][1]) // 1st group of 1st match (0-indexed)
+	fmt.Println(matchIndices[0][2]) // 2nd group of 1st match
+	fmt.Println(matchIndices[1][1]) // 1st group of 2nd match
+	fmt.Println(matchIndices[1][2]) // 2nd group of 2nd match
+	// Output: twomorecents.org
+	// index.html
+	// news.ycombinator.com
+	// user?id=aadhavans
+
+}
+
+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
+}
+
+func ExampleReg_ReplaceAll() {
+	regexStr := `(\d)(\w)`
+	inputStr := "5d9t"
+	regexComp := regex.MustCompile(regexStr)
+	fmt.Println(regexComp.ReplaceAll(inputStr, `$2$1`))
+	// Output: d5t9
+}
+
+func ExampleReg_ReplaceAllLiteral() {
+	regexStr := `fox|dog`
+	inputStr := "the quick brown fox jumped over the lazy dog"
+	regexComp := regex.MustCompile(regexStr)
+	fmt.Println(regexComp.ReplaceAllLiteral(inputStr, `duck`))
+	// Output: the quick brown duck jumped over the lazy duck
+}
+
+func ExampleReg_ReplaceAllFunc() {
+	regexStr := `\w{5,}`
+	inputStr := `all five or more letter words in this string are capitalized`
+	regexComp := regex.MustCompile(regexStr)
+	fmt.Println(regexComp.ReplaceAllFunc(inputStr, strings.ToUpper))
+	// Output: all five or more LETTER WORDS in this STRING are CAPITALIZED
+}

regex/matching.go

@@ -2,7 +2,8 @@ package regex
 
 import (
 	"fmt"
-	"sort"
+	"strconv"
+	"unicode"
 )
 
 // A Match represents a match found by the regex in a given string.
@@ -14,7 +15,7 @@ import (
 // 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
@@ -29,17 +30,6 @@ 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) String() string {
 	var toRet string
@@ -58,7 +48,7 @@ func (idx Group) String() string {
 	return fmt.Sprintf("%d\t%d", idx.StartIdx, idx.EndIdx)
 }
 
-// Returns whether a group is valid (ie. whether it matched any text). It
+// 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
@@ -69,105 +59,42 @@ func getZeroGroup(m Match) Group {
 	return m[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
-//}
-
-// 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 = uniqueAppend(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
-//}
-
-// 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
-		}
-	}
-	// Add last state
-	toRet = append(toRet, current)
-	return toRet
-}
-
 func copyThread(to *nfaState, from nfaState) {
 	to.threadGroups = append([]Group{}, from.threadGroups...)
 }
 
 // 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 (regex Reg) Find(str string) (Group, error) {
+func (re Reg) Find(str string) (Group, error) {
-	match, err := regex.FindNthMatch(str, 1)
+	match, err := re.FindNthMatch(str, 1)
 	if err != nil {
 		return Group{}, fmt.Errorf("no matches found")
 	}
 	return getZeroGroup(match), nil
 }
 
+// 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 (regex Reg) FindAll(str string) []Group {
+func (re Reg) FindAll(str string) []Group {
-	indices := regex.FindAllSubmatch(str)
+	indices := re.FindAllSubmatch(str)
 	zeroGroups := funcMap(indices, getZeroGroup)
 	return zeroGroups
 }
@@ -176,8 +103,8 @@ func (regex Reg) FindAll(str string) []Group {
 // The return value will be an empty string in two situations:
 //  1. No match was found
 //  2. The match was an empty string
-func (regex Reg) FindString(str string) string {
+func (re Reg) FindString(str string) string {
-	match, err := regex.FindNthMatch(str, 1)
+	match, err := re.FindNthMatch(str, 1)
 	if err != nil {
 		return ""
 	}
@@ -190,8 +117,8 @@ func (regex Reg) FindString(str string) string {
 // 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 (regex Reg) FindSubmatch(str string) (Match, error) {
+func (re Reg) FindSubmatch(str string) (Match, error) {
-	match, err := regex.FindNthMatch(str, 1)
+	match, err := re.FindNthMatch(str, 1)
 	if err != nil {
 		return Match{}, fmt.Errorf("no match found")
 	} else {
@@ -199,11 +126,41 @@ func (regex Reg) FindSubmatch(str string) (Match, error) {
 	}
 }
 
-// FindAllString is the 'all' version of FindString.
+// 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 (regex Reg) FindAllString(str string) []string {
+func (re Reg) FindAllString(str string) []string {
-	zerogroups := regex.FindAll(str)
+	zerogroups := re.FindAll(str)
 	matchStrs := funcMap(zerogroups, func(g Group) string {
 		return str[g.StartIdx:g.EndIdx]
 	})
@@ -212,14 +169,14 @@ func (regex Reg) FindAllString(str string) []string {
 
 // 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 (regex Reg) FindNthMatch(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 = findAllSubmatchHelper(regex.start, str_runes, idx, regex.numGroups)
+		matchFound, matchIdx, idx = findAllSubmatchHelper(re.start, str_runes, idx, re.numGroups, re.preferLongest)
 		if matchFound {
 			matchNum++
 		}
@@ -232,26 +189,46 @@ func (regex Reg) FindNthMatch(str string, n int) (Match, error) {
 }
 
 // FindAllSubmatch returns a slice of matches in the given string.
-func (regex Reg) FindAllSubmatch(str string) []Match {
+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 = findAllSubmatchHelper(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) []nfaState {
+// FindAllSubmatch returns a double-slice of strings. Each slice contains the text of a match, including all submatches.
+// A return value of nil indicates no match.
+func (re Reg) FindAllStringSubmatch(str string) [][]string {
+	match := re.FindAllSubmatch(str)
+	if len(match) == 0 {
+		return nil
+	}
+	rtv := make([][]string, len(match))
+	for i := range rtv {
+		rtv[i] = make([]string, re.numGroups+1)
+	}
+	rtv = funcMap(match, func(m Match) []string {
+		return funcMap(m, func(g Group) string {
+			if g.IsValid() {
+				return str[g.StartIdx:g.EndIdx]
+			} else {
+				return ""
+			}
+		})
+	})
+	return rtv
+}
+
+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
 	}
@@ -259,32 +236,34 @@ 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)
+		list := addStateToList(str, idx, list, *state.splitState, threadGroups, visited, preferLongest)
 		copyThread(state.next, state)
-		list = addStateToList(str, idx, list, *state.next, threadGroups, visited)
+		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)
+		list := addStateToList(str, idx, list, *state.next, threadGroups, visited, preferLongest)
 		copyThread(state.splitState, state)
-		list = addStateToList(str, idx, list, *state.splitState, threadGroups, visited)
+		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) {
+		if state.checkAssertion(str, idx, preferLongest) {
 			copyThread(state.next, state)
-			return addStateToList(str, idx, list, *state.next, state.threadGroups, visited)
+			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)
+		copyThread(state.next, state)
+		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)
+		copyThread(state.next, state)
+		return addStateToList(str, idx, list, *state.next, state.threadGroups, visited, preferLongest)
 	}
 	return append(list, state)
 
@@ -293,9 +272,7 @@ func addStateToList(str []rune, idx int, list []nfaState, state nfaState, thread
 // 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.
-//
+func findAllSubmatchHelper(start *nfaState, str []rune, offset int, numGroups int, preferLongest bool) (bool, Match, int) {
-//	Might return duplicates or overlapping indices, so care must be taken to prune the resulting array.
-func findAllSubmatchHelper(start *nfaState, str []rune, offset int, numGroups int) (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)
@@ -303,53 +280,23 @@ func findAllSubmatchHelper(start *nfaState, str []rune, offset int, numGroups in
 	}
 	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)
 	nextStates := make([]nfaState, 0)
-	//	tempStates := make([]*nfaState, 0) // Used to store states that should be used in next loop iteration
 	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
-	//}
 
 	start.threadGroups = newMatch(numGroups + 1)
 	start.threadGroups[0].StartIdx = i
-	currentStates = addStateToList(str, i, currentStates, *start, start.threadGroups, nil)
+	currentStates = addStateToList(str, i, currentStates, *start, start.threadGroups, nil, preferLongest)
 	var match Match = nil
-	//	var isEmptyAndNoAssertion bool
-	// Main loop
 	for idx := i; idx <= len(str); idx++ {
 		if len(currentStates) == 0 {
 			break
@@ -365,82 +312,28 @@ func findAllSubmatchHelper(start *nfaState, str []rune, offset int, numGroups in
 			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 { // Normal character or assertion
+				}
-				if currentState.contentContains(str, idx) {
+			} else if !currentState.isAlternation && !currentState.isKleene && !currentState.isQuestion && !currentState.isBackreference && !currentState.groupBegin && !currentState.groupEnd && currentState.assert == noneAssert { // Normal character
-					nextStates = addStateToList(str, idx+1, nextStates, *currentState.next, currentState.threadGroups, nil)
+				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)
+					}
 				}
 			}
 
-			//			if currentState.groupBegin {
-			//				currentState.threadGroups[currentState.groupNum].StartIdx = idx
-			//			}
-			//			if currentState.groupEnd {
-			//				currentState.threadGroups[currentState.groupNum].EndIdx = idx
-			//			}
-
-			// Alternation - enqueue left then right state, and continue
-			//			if currentState.isAlternation {
-			//				if currentState.isKleene { // Reverse order of adding things
-			//					rightState := currentState.splitState
-			//					copyThread(rightState, currentState)
-			//					currentStates = slices.Insert(currentStates, currentStateIdx+1, *rightState)
-			//					leftState := currentState.next
-			//					copyThread(leftState, currentState)
-			//					currentStates = slices.Insert(currentStates, currentStateIdx+2, *leftState)
-			//				} else {
-			//					leftState := currentState.next
-			//					copyThread(leftState, currentState)
-			//					currentStates = slices.Insert(currentStates, currentStateIdx+1, *leftState)
-			//					rightState := currentState.splitState
-			//					copyThread(rightState, currentState)
-			//					currentStates = slices.Insert(currentStates, currentStateIdx+2, *rightState)
-			//				}
-			//				continue
-			//			}
-
-			// Empty state - enqueue next state, do _not_ increment the SP
-			//			if !currentState.isAlternation && currentState.isEmpty && currentState.assert == noneAssert { //&& currentState.groupBegin == false && currentState.groupEnd == false {
-			//				isEmptyAndNoAssertion = true
-			//			}
-			//
-			//			if currentState.contentContains(str, idx) {
-			//				foundMatch = true
-			//			}
-			//
-			//			if isEmptyAndNoAssertion || foundMatch {
-			//				nextMatch := *(currentState.next)
-			//				copyThread(&nextMatch, currentState)
-			//				if currentState.groupBegin {
-			//					//	if !stateExists(currentStates, nextMatch) {
-			//					currentStates = slices.Insert(currentStates, currentStateIdx+1, nextMatch)
-			//					//}
-			//				} else if currentState.groupEnd {
-			//					if !stateExists(currentStates, nextMatch) {
-			//						currentStates = slices.Insert(currentStates, currentStateIdx+1, nextMatch) // append(currentStates, nextMatch)
-			//					}
-			//				} else if currentState.assert != noneAssert {
-			//					if !stateExists(currentStates, nextMatch) {
-			//						currentStates = append(currentStates, nextMatch)
-			//					}
-			//				} else if currentState.isEmpty && !currentState.groupBegin && !currentState.groupEnd {
-			//					if !stateExists(currentStates, nextMatch) {
-			//						currentStates = append(currentStates, nextMatch)
-			//					}
-			//				} else {
-			//					if !stateExists(nextStates, nextMatch) {
-			//						nextStates = append(nextStates, nextMatch)
-			//					}
-			//				}
-			//			}
-			//
-			//			if currentState.isLast && len(nextStates) == 0 { // Last state reached
-			//				currentState.threadGroups[0].EndIdx = idx
-			//				if idx == currentState.threadGroups[0].StartIdx {
-			//					idx += 1
-			//				}
-			//				return true, currentState.threadGroups, idx
-			//			}
 		}
 		currentStates = append([]nfaState{}, nextStates...)
 		nextStates = nil
@@ -452,196 +345,132 @@ func findAllSubmatchHelper(start *nfaState, str []rune, offset int, numGroups in
 		return true, match, match[0].EndIdx
 	}
 	return false, []Group{}, i + 1
-	//		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.
+// Expand appends template to dst, expanding any variables in template to the relevant capturing group.
-	//		topStateItem := currentStates.peek()
+//
-	//		topState := topStateItem.(*priorQueueItem).state
+// 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.
-	//		zeroStates, isZero := takeZeroState([]*nfaState{topState}, numGroups, i)
+// To insert a literal $, do not put a number after it. Alternatively, you can use $$.
-	//		tempStates = append(tempStates, zeroStates...)
+// src is the input string, and match must be the result of [Reg.FindSubmatch].
-	//		num_appended := 0
+func (re Reg) Expand(dst string, template string, src string, match Match) string {
-	//		for isZero == true {
+	templateRuneSlc := []rune(template)
-	//			zeroStates, isZero = takeZeroState(tempStates, numGroups, i)
+	srcRuneSlc := []rune(src)
-	//			tempStates, num_appended = uniqueAppend(tempStates, zeroStates...)
+	i := 0
-	//			if num_appended == 0 { // Break if we haven't appended any more unique values
+	for i < len(templateRuneSlc) {
-	//				break
+		c := templateRuneSlc[i]
-	//			}
+		if c == '$' {
-	//		}
+			i += 1
-	//		if isZero == true {
+			// The dollar sign is the last character of the string, or it is proceeded by another dollar sign
-	//			currentStates.Pop()
+			if i >= len(templateRuneSlc) || templateRuneSlc[i] == '$' {
-	//		}
+				dst += "$"
-	//
+				i++
-	//		for _, state := range tempStates {
+			} else {
-	//			heap.Push(currentStates, newPriorQueueItem(state))
+				numStr := ""
-	//		}
+				for i < len(templateRuneSlc) && unicode.IsDigit(templateRuneSlc[i]) {
-	//		tempStates = nil
+					numStr += string(templateRuneSlc[i])
-	//
+					i++
-	//		// Take any transitions corresponding to current character
+				}
-	//		numStatesMatched := 0            // The number of states which had at least 1 match for this round
+				if numStr == "" {
-	//		assertionFailed := false         // Whether or not an assertion failed for this round
+					dst += "$"
-	//		lastStateInList := false         // Whether or not a last state was in our list of states
+				} else {
-	//		var lastStatePtr *nfaState = nil // Pointer to the last-state, if it was found
+					num, _ := strconv.Atoi(numStr)
-	//		lastLookaroundInList := false    // Whether or not a last state (that is a lookaround) was in our list of states
+					if num < len(match) {
-	//		for numStatesMatched == 0 && lastStateInList == false {
+						dst += string(srcRuneSlc[match[num].StartIdx:match[num].EndIdx])
-	//			if currentStates.Len() == 0 {
+					} else {
-	//				break
+						dst += "$" + numStr
-	//			}
+					}
-	//			stateItem := heap.Pop(currentStates)
+				}
-	//			state := stateItem.(*priorQueueItem).state
+			}
-	//			matches, numMatches := state.matchesFor(str, i)
+		} else {
-	//			if numMatches > 0 {
+			dst += string(c)
-	//				numStatesMatched++
+			i++
-	//				tempStates = append([]*nfaState(nil), matches...)
+		}
-	//				foundPath = true
+	}
-	//				for _, m := range matches {
+	return dst
-	//					if m.threadGroups == nil {
+}
-	//						m.threadGroups = newMatch(numGroups + 1)
+
-	//					}
+// LiteralPrefix returns a string that must begin any match of the given regular expression.
-	//					m.threadSP = state.threadSP + 1
+// The second return value is true if the string comprises the entire expression.
-	//					copy(m.threadGroups, state.threadGroups)
+func (re Reg) LiteralPrefix() (prefix string, complete bool) {
-	//				}
+	state := re.start
-	//			}
+	if state.assert != noneAssert {
-	//			if numMatches < 0 {
+		state = state.next
-	//				assertionFailed = true
+	}
-	//			}
+	for !(state.isLast) && (!state.isAlternation) && len(state.content) == 1 && state.assert == noneAssert {
-	//			if state.isLast {
+		if state.groupBegin || state.groupEnd {
-	//				if state.isLookaround() {
+			state = state.next
-	//					lastLookaroundInList = true
+			continue
-	//				}
+		}
-	//				lastStateInList = true
+		prefix += string(rune(state.content[0]))
-	//				lastStatePtr = state
+		state = state.next
-	//			}
+	}
-	//		}
+	if state.isLast {
-	//
+		complete = true
-	//		if assertionFailed && numStatesMatched == 0 { // Nothing has matched and an assertion has failed
+	} else {
-	//			// If I'm being completely honest, I'm not sure why I have to check specifically for a _lookaround_
+		complete = false
-	//			// state. The explanation below is my attempt to explain this behavior.
+	}
-	//			// If you replace 'lastLookaroundInList' with 'lastStateInList', one of the test cases fails.
+	return prefix, complete
-	//			//
+}
-	//			// 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
+// ReplaceAll replaces all matches of the expression in src, with the text in repl. In repl, variables are interpreted
-	//			// another path to a final state.
+// as they are in [Reg.Expand]. The resulting string is returned.
-	//			// Even if the last state _was_ an assertion, we can use the previously
+func (re Reg) ReplaceAll(src string, repl string) string {
-	//			// saved indices to find a match.
+	matches := re.FindAllSubmatch(src)
-	//			if lastLookaroundInList {
+	i := 0
-	//				break
+	currentMatch := 0
-	//			} else {
+	dst := ""
-	//				if i == startingFrom {
+	for i < len(src) {
-	//					i++
+		if currentMatch < len(matches) && matches[currentMatch][0].IsValid() && i == matches[currentMatch][0].StartIdx {
-	//				}
+			dst += re.Expand("", repl, src, matches[currentMatch])
-	//				return false, []Group{}, i
+			i = matches[currentMatch][0].EndIdx
-	//			}
+			currentMatch++
-	//		}
+		} else {
-	//		// Check if we can find a state in our list that is:
+			dst += string(src[i])
-	//		// 	a. A last-state
+			i++
-	//		// 	b. Empty
+		}
-	//		// 	c. Doesn't assert anything
+	}
-	//		for _, stateItem := range *currentStates {
+	return dst
-	//			s := stateItem.state
+}
-	//			if s.isLast && s.isEmpty && s.assert == noneAssert {
+
-	//				lastStatePtr = s
+// ReplaceAllLiteral replaces all matches of the expression in src, with the text in repl. The text is replaced directly,
-	//				lastStateInList = true
+// without any expansion.
-	//			}
+func (re Reg) ReplaceAllLiteral(src string, repl string) string {
-	//		}
+	zerogroups := re.FindAll(src)
-	//		if lastStateInList && numStatesMatched == 0 { // A last-state was in the list of states. add the matchIndex to our MatchIndex list
+	currentMatch := 0
-	//			for j := 1; j < numGroups+1; j++ {
+	i := 0
-	//				tempIndices[j] = lastStatePtr.threadGroups[j]
+	dst := ""
-	//			}
+
-	//			endIdx = i
+	for i < len(src) {
-	//			tempIndices[0] = Group{startIdx, endIdx}
+		if currentMatch < len(zerogroups) && i == zerogroups[currentMatch].StartIdx {
-	//			if tempIndices[0].StartIdx == tempIndices[0].EndIdx {
+			dst += repl
-	//				return true, tempIndices, tempIndices[0].EndIdx + 1
+			i = zerogroups[currentMatch].EndIdx
-	//			} else {
+			currentMatch += 1
-	//				return true, tempIndices, tempIndices[0].EndIdx
+		} else {
-	//			}
+			dst += string(src[i])
-	//		}
+			i++
-	//
+		}
-	//		// Check if we can find a zero-length match
+	}
-	//		if foundPath == false {
+	return dst
-	//			currentStatesList := funcMap(*currentStates, func(item *priorQueueItem) *nfaState {
+}
-	//				return item.state
+
-	//			})
+// ReplaceAllFunc replaces every match of the expression in src, with the return value of the function replFunc.
-	//			if ok := zeroMatchPossible(str, i, numGroups, currentStatesList...); ok {
+// replFunc takes in the matched string. The return value is substituted in directly without expasion.
-	//				if tempIndices[0].IsValid() == false {
+func (re Reg) ReplaceAllFunc(src string, replFunc func(string) string) string {
-	//					tempIndices[0] = Group{startIdx, startIdx}
+	zerogroups := re.FindAll(src)
-	//				}
+	currentMatch := 0
-	//			}
+	i := 0
-	//			// If we haven't moved in the string, increment the counter by 1
+	dst := ""
-	//			// to ensure we don't keep trying the same string over and over.
+
-	//			//			if i == startingFrom {
+	for i < len(src) {
-	//			startIdx++
+		if currentMatch < len(zerogroups) && i == zerogroups[currentMatch].StartIdx {
-	//			//	i++
+			dst += replFunc(src[zerogroups[currentMatch].StartIdx:zerogroups[currentMatch].EndIdx])
-	//			//			}
+			i = zerogroups[currentMatch].EndIdx
-	//			if tempIndices.numValidGroups() > 0 && tempIndices[0].IsValid() {
+			currentMatch += 1
-	//				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.
+		} else {
-	//					return true, tempIndices, tempIndices[0].EndIdx + 1
+			dst += string(src[i])
-	//				} else {
+			i++
-	//					return true, tempIndices, tempIndices[0].EndIdx
+		}
-	//				}
+	}
-	//			}
+	return dst
-	//			return false, []Group{}, startIdx
+
-	//		}
-	//		currentStates = &priorityQueue{}
-	//		slices.Reverse(tempStates)
-	//		for _, state := range tempStates {
-	//			heap.Push(currentStates, newPriorQueueItem(state))
-	//		}
-	//		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.
-	//
-	//	if currentStates.Len() > 0 {
-	//		topStateItem := currentStates.peek()
-	//		topState := topStateItem.(*priorQueueItem).state
-	//		zeroStates, isZero := takeZeroState([]*nfaState{topState}, 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 = uniqueAppend(tempStates, zeroStates...)
-	//			if num_appended == 0 { // Break if we haven't appended any more unique values
-	//				break
-	//			}
-	//		}
-	//	}
-	//
-	//	for _, state := range tempStates {
-	//		heap.Push(currentStates, newPriorQueueItem(state))
-	//	}
-	//
-	// tempStates = nil
-	//
-	//	for _, stateItem := range *currentStates {
-	//		state := stateItem.state
-	//		// 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
 }

regex/misc.go

@@ -48,49 +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 uniqueAppend[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
-}
-
-func uniqueAppendFunc[T any](slc []T, fn func(T, T) bool, items ...T) ([]T, int) {
-	toRet := make([]T, len(slc))
-	num_appended := 0
-	copy(toRet, slc)
-	for _, item := range items {
-		itemExists := false
-		for _, val := range slc {
-			if fn(item, val) {
-				itemExists = true
-			}
-		}
-		if !itemExists {
-			toRet = append(toRet, item)
-			num_appended++
-		}
-	}
-	return toRet, 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

@@ -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.
+	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)
@@ -133,7 +135,7 @@ func resetThreadsHelper(state *nfaState, visitedMap map[*nfaState]bool) {
 
 // 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
 	}
@@ -183,7 +185,7 @@ func (s nfaState) checkAssertion(str []rune, idx int) bool {
 			strToMatch = string(runesToMatch)
 		}
 
-		regComp := Reg{startState, s.lookaroundNumCaptureGroups}
+		regComp := Reg{startState, s.lookaroundNumCaptureGroups, s.lookaroundRegex, preferLongest}
 		matchIndices := regComp.FindAll(strToMatch)
 
 		numMatchesFound := 0
@@ -210,9 +212,9 @@ 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
@@ -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
@@ -40,6 +43,7 @@ type postfixNode struct {
 	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/priorityQueue.go

@@ -1,89 +0,0 @@
-package regex
-
-import "container/heap"
-
-// Implement a priority queue using container/heap
-
-const (
-	min_priority int = iota
-	zerostate_priority
-	alternation_priority
-	kleene_priority
-	char_priority
-	max_priority
-)
-
-func getPriority(state *nfaState) int {
-	if state.isKleene {
-		return zerostate_priority
-	} else if state.isAlternation {
-		return alternation_priority
-	} else {
-		if state.isEmpty {
-			return zerostate_priority
-		} else {
-			return char_priority
-		}
-	}
-}
-
-type priorQueueItem struct {
-	state    *nfaState
-	priority int
-	index    int
-}
-
-func newPriorQueueItem(state *nfaState) *priorQueueItem {
-	return &priorQueueItem{
-		state:    state,
-		index:    -1,
-		priority: getPriority(state),
-	}
-}
-
-type priorityQueue []*priorQueueItem
-
-func (pq priorityQueue) Len() int {
-	return len(pq)
-}
-
-func (pq priorityQueue) Less(i, j int) bool {
-	if pq[i].priority == pq[j].priority {
-		return pq[i].index < pq[j].index
-	}
-	return pq[i].priority > pq[j].priority // We want max-heap, so we use greater-than
-}
-
-func (pq priorityQueue) Swap(i, j int) {
-	pq[i], pq[j] = pq[j], pq[i]
-	pq[i].index = i
-	pq[j].index = j
-}
-
-func (pq *priorityQueue) Push(x any) {
-	length := len(*pq)
-	item := x.(*priorQueueItem)
-	item.index = length
-	*pq = append(*pq, item)
-}
-
-func (pq *priorityQueue) Pop() any {
-	old := *pq
-	n := len(old)
-	item := old[n-1]
-	old[n-1] = nil
-	item.index = -1
-	*pq = old[0 : n-1]
-	return item
-}
-func (pq *priorityQueue) peek() any {
-	queue := *pq
-	n := len(queue)
-	return queue[n-1]
-}
-
-func (pq *priorityQueue) update(item *priorQueueItem, value *nfaState, priority int) {
-	item.state = value
-	item.priority = priority
-	heap.Fix(pq, item.index)
-}

regex/range2regex.go

@@ -109,7 +109,7 @@ func range2regex(start int, end int) (string, error) {
 		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/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}}},
@@ -177,7 +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}}},
@@ -192,7 +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{}},
@@ -310,11 +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}}},
+	{`\0141`, 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.
 
 	{`*a`, nil, ``, nil},
 	{`(*)b`, nil, ``, nil},
@@ -431,7 +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}}},
 
@@ -462,8 +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}}},
@@ -471,7 +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}}},
@@ -518,6 +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 {
@@ -528,7 +537,7 @@ 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}}}},
@@ -537,10 +546,11 @@ 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}}}},
+	// 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, 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}}}},
@@ -578,13 +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
+	// Backreference tests
-	// of backreference tests. Since my engine doesn't
+	{`(abc)\1`, nil, `abcabc`, []Match{[]Group{{0, 6}, {0, 3}}}},
-	// implement backreferences, I've skipped those tests.
+	{`([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}}}},
@@ -633,7 +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}}}},
@@ -743,7 +777,7 @@ func TestFindString(t *testing.T) {
 				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]
@@ -789,18 +823,132 @@ func TestFindSubmatch(t *testing.T) {
 				if test.result != nil {
 					panic(err)
 				}
-			}
+			} 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 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)
+				}
+			} 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]
+						} 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 TestFindAllStringSubmatch(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)
+				}
+			} 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() {
+								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)
+								}
+							}
+						}
+					}
+				}
+			}
+		})
+	}
+}
+
 func TestFindAllSubmatch(t *testing.T) {
 	for _, test := range groupTests {
 		t.Run(test.re+"	"+test.str, func(t *testing.T) {
@@ -809,7 +957,7 @@ func TestFindAllSubmatch(t *testing.T) {
 				if test.result != nil {
 					panic(err)
 				}
-			}
+			} else {
 				matchIndices := regComp.FindAllSubmatch(test.str)
 				for i := range matchIndices {
 					for j := range matchIndices[i] {
@@ -817,6 +965,11 @@ func TestFindAllSubmatch(t *testing.T) {
 							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)
+							}
+						}
 					}
 				}
 			}