Continued implementing Thompson's algorithm

The git diff claims that a ton of code was changed, but most of it was just indentation changes.

regex/matching.go

@@ -151,6 +151,11 @@ func pruneIndices(indices []Match) []Match {
 	return toRet
 }
 
+func copyThread(to *nfaState, from nfaState) {
+	to.threadSP = from.threadSP
+	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) {
@@ -271,7 +276,8 @@ func findAllSubmatchHelper(start *nfaState, str []rune, offset int, numGroups in
 	//	foundPath := false
 	//startIdx := offset
 	//endIdx := offset
-	currentStates := make([]*nfaState, 0)
+	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
@@ -302,57 +308,139 @@ func findAllSubmatchHelper(start *nfaState, str []rune, offset int, numGroups in
 	//}
 
 	start.threadSP = i
-	currentStates = append(currentStates, start)
+	currentStates = append(currentStates, *start)
 	var foundMatch bool
+	var isEmptyAndNoAssertion bool
 	// Main loop
-	for len(currentStates) > 0 {
-		currentState, _ := pop(&currentStates)
-		idx := currentState.threadSP
-		foundMatch = false
+	for idx := i; idx <= len(str); idx++ {
+		for currentStateIdx := 0; currentStateIdx < len(currentStates); currentStateIdx++ {
+			currentState := currentStates[currentStateIdx]
+			foundMatch = false
+			isEmptyAndNoAssertion = false
 
-		if currentState.threadGroups == nil {
-			currentState.threadGroups = newMatch(numGroups + 1)
-			currentState.threadGroups[0].StartIdx = idx
-		}
-		if currentState.groupBegin {
-			currentState.threadGroups[currentState.groupNum].StartIdx = idx
-		} else if currentState.groupEnd {
-			currentState.threadGroups[currentState.groupNum].EndIdx = idx
-		} else if currentState.isKleene {
-			// Append the
-		} else if currentState.isAlternation {
-			rightState := currentState.rightState
-			rightState.threadGroups = currentState.threadGroups
-			rightState.threadSP = currentState.threadSP
-			currentStates = append(currentStates, currentState.rightState)
-			leftState := currentState.leftState
-			leftState.threadGroups = currentState.threadGroups
-			leftState.threadSP = currentState.threadSP
-			currentStates = append(currentStates, currentState.leftState)
-			continue
-		} else if currentState.contentContains(str, idx) {
-			foundMatch = true
-			allMatches := make([]*nfaState, 0)
-			for _, v := range currentState.transitions {
-				allMatches = append(allMatches, v...)
+			if currentState.threadGroups == nil {
+				currentState.threadGroups = newMatch(numGroups + 1)
+				currentState.threadGroups[0].StartIdx = idx
 			}
-			slices.Reverse(allMatches)
-			for _, m := range allMatches {
-				m.threadGroups = currentState.threadGroups
-				if currentState.assert == noneAssert {
-					m.threadSP = idx + 1
+
+			if currentState.groupBegin {
+				currentState.threadGroups[currentState.groupNum].StartIdx = idx
+				//		allMatches := make([]nfaState, 0)
+				//		for _, v := range currentState.transitions {
+				//			dereferenced := funcMap(v, func(s *nfaState) nfaState {
+				//				return *s
+				//			})
+				//			allMatches = append(allMatches, dereferenced...)
+				//		}
+				//		slices.Reverse(allMatches)
+				//		for i := range allMatches {
+				//			copyThread(&allMatches[i], currentState)
+				//		}
+				//		currentStates = append(currentStates, allMatches...)
+			}
+			if currentState.groupEnd {
+				currentState.threadGroups[currentState.groupNum].EndIdx = idx
+				//			allMatches := make([]nfaState, 0)
+				//			for _, v := range currentState.transitions {
+				//				dereferenced := funcMap(v, func(s *nfaState) nfaState {
+				//					return *s
+				//				})
+				//				allMatches = append(allMatches, dereferenced...)
+				//			}
+				//			slices.Reverse(allMatches)
+				//			for i := range allMatches {
+				//				copyThread(&allMatches[i], currentState)
+				//			}
+				//			currentStates = append(currentStates, allMatches...)
+			}
+
+			//		if currentState.isKleene {
+			//			// Append the next-state (after the kleene), then append the kleene state
+			//			allMatches := make([]*nfaState, 0)
+			//			for _, v := range currentState.transitions {
+			//				allMatches = append(allMatches, v...)
+			//			}
+			//			slices.Reverse(allMatches)
+			//			for _, m := range allMatches {
+			//				m.threadGroups = currentState.threadGroups
+			//				m.threadSP = idx
+			//			}
+			//			currentStates = append(currentStates, allMatches...)
+			//
+			//			//	kleeneState := currentState.kleeneState
+			//			//	kleeneState.threadGroups = currentState.threadGroups
+			//			//	kleeneState.threadSP = currentState.threadSP
+			//			//	currentStates = append(currentStates, kleeneState)
+			//			continue
+			//		}
+
+			// Alternation - enqueue left then right state, and continue
+			if currentState.isAlternation {
+				leftState := currentState.leftState
+				copyThread(leftState, currentState)
+				currentStates = append(currentStates, *currentState.leftState)
+				rightState := currentState.rightState
+				copyThread(rightState, currentState)
+				currentStates = append(currentStates, *currentState.rightState)
+				continue
+			}
+
+			// Empty state - enqueue next state, do _not_ increment the SP
+			if currentState.isEmpty && currentState.assert == noneAssert { //&& currentState.groupBegin == false && currentState.groupEnd == false {
+				isEmptyAndNoAssertion = true
+			}
+
+			if currentState.contentContains(str, idx) {
+				foundMatch = true
+			}
+
+			if isEmptyAndNoAssertion || foundMatch {
+				allMatches := make([]nfaState, 0)
+				for _, v := range currentState.transitions {
+					dereferenced := funcMap(v, func(s *nfaState) nfaState {
+						return *s
+					})
+					allMatches = append(allMatches, dereferenced...)
+				}
+				slices.Reverse(allMatches)
+				for i := range allMatches {
+					copyThread(&allMatches[i], currentState)
+					if foundMatch && currentState.assert == noneAssert {
+						allMatches[i].threadSP += 1
+					}
+				}
+				if currentState.groupBegin {
+					currentStates = slices.Insert(currentStates, currentStateIdx+1, allMatches...)
+				} else if currentState.groupEnd {
+					currentStates = append(currentStates, allMatches...)
 				} else {
-					m.threadSP = idx
+					nextStates = append(nextStates, allMatches...)
 				}
 			}
-			currentStates = append(currentStates, allMatches...)
-		}
 
-		if currentState.isLast && foundMatch { // Last state reached
-			currentState.threadGroups[0].EndIdx = idx + 1
-			return true, currentState.threadGroups, idx + 1
+			if currentState.isLast && len(nextStates) == 0 { // Last state reached
+				if foundMatch {
+					if currentState.assert != noneAssert {
+						currentState.threadGroups[0].EndIdx = idx
+					} else {
+						currentState.threadGroups[0].EndIdx = idx + 1
+					}
+					if idx == currentState.threadGroups[0].StartIdx {
+						idx += 1
+					}
+					return true, currentState.threadGroups, idx
+				} else if isEmptyAndNoAssertion {
+					currentState.threadGroups[0].EndIdx = idx
+					if idx == currentState.threadGroups[0].StartIdx {
+						idx++
+					}
+					return true, currentState.threadGroups, idx
+				}
 
+			}
 		}
+		currentStates = append([]nfaState{}, nextStates...)
+		nextStates = nil
 	}
 	return false, []Group{}, i + 1
 	//		zeroStates := make([]*nfaState, 0)

regex/nfa.go

@@ -116,7 +116,6 @@ func cloneStateHelper(stateToClone *nfaState, cloneMap map[*nfaState]*nfaState)
 		clone.rightState = clone
 	}
 	clone.rightState = cloneStateHelper(stateToClone.rightState, cloneMap)
-
 	return clone
 }
 
@@ -326,12 +325,16 @@ func kleene(s1 nfaState) (*nfaState, error) {
 		return nil, fmt.Errorf("previous token is not quantifiable")
 	}
 
-	toReturn := &nfaState{}
-	toReturn.transitions = make(map[int][]*nfaState)
-	toReturn.content = newContents(epsilon)
+	emptyState := zeroLengthMatchState()
+	emptyState.assert = noneAssert
+	toReturn := alternate(&s1, &emptyState)
+
+	//	toReturn := &nfaState{}
+	//	toReturn.transitions = make(map[int][]*nfaState)
+	//	toReturn.content = newContents(epsilon)
 	toReturn.isEmpty = true
 	toReturn.isKleene = true
-	toReturn.output = append(toReturn.output, toReturn)
+	toReturn.output = []*nfaState{&emptyState}
 	for i := range s1.output {
 		for _, c := range toReturn.content {
 			s1.output[i].transitions[c], _ = uniqueAppend(s1.output[i].transitions[c], toReturn)
@@ -340,6 +343,7 @@ func kleene(s1 nfaState) (*nfaState, error) {
 	for _, c := range s1.content {
 		toReturn.transitions[c], _ = uniqueAppend(toReturn.transitions[c], &s1)
 	}
+	//toReturn.kleeneState = &s1
 	return toReturn, nil
 }
 
@@ -374,7 +378,6 @@ func question(s1 *nfaState) *nfaState { // Use the fact that ab? == a(b|)
 	s2.content = newContents(epsilon)
 	s2.output = append(s2.output, s2)
 	s2.isEmpty = true
-	s2.isAlternation = true
 	s3 := alternate(s1, s2)
 	return s3
 }