You cannot select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
664 lines
23 KiB
Go
664 lines
23 KiB
Go
package regex
|
|
|
|
import (
|
|
"fmt"
|
|
"slices"
|
|
"sort"
|
|
)
|
|
|
|
// A Match represents a match found by the regex in a given string.
|
|
// It is represented as a list of groups, where the nth element contains
|
|
// the contents of the nth capturing group. Note that the group may not be valid
|
|
// (see [Group.IsValid]). The element at index 0 is known
|
|
// as the 0-group, and represents the contents of the entire match.
|
|
//
|
|
// See [Reg.FindSubmatch] for an example.
|
|
type Match []Group
|
|
|
|
// a Group represents a group. It contains the start index and end index of the match
|
|
type Group struct {
|
|
StartIdx int
|
|
EndIdx int
|
|
}
|
|
|
|
func newMatch(size int) Match {
|
|
toRet := make([]Group, size)
|
|
for i := range toRet {
|
|
toRet[i].StartIdx = -1
|
|
toRet[i].EndIdx = -1
|
|
}
|
|
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
|
|
for i, g := range m {
|
|
if g.IsValid() {
|
|
toRet += fmt.Sprintf("Group %d\n", i)
|
|
toRet += g.String()
|
|
toRet += "\n"
|
|
}
|
|
}
|
|
return toRet
|
|
}
|
|
|
|
// String converts the Group into a string representation.
|
|
func (idx Group) String() string {
|
|
return fmt.Sprintf("%d\t%d", idx.StartIdx, idx.EndIdx)
|
|
}
|
|
|
|
// Returns whether a group 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
|
|
}
|
|
|
|
// Simple function, makes it easier to map this over a list of matches
|
|
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) {
|
|
match, err := regex.FindNthMatch(str, 1)
|
|
if err != nil {
|
|
return Group{}, fmt.Errorf("no matches found")
|
|
}
|
|
return getZeroGroup(match), 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 {
|
|
indices := regex.FindAllSubmatch(str)
|
|
zeroGroups := funcMap(indices, getZeroGroup)
|
|
return zeroGroups
|
|
}
|
|
|
|
// FindString returns the text of the leftmost match of the regex in the given string.
|
|
// The return value will be an empty string in two situations:
|
|
// 1. No match was found
|
|
// 2. The match was an empty string
|
|
func (regex Reg) FindString(str string) string {
|
|
match, err := regex.FindNthMatch(str, 1)
|
|
if err != nil {
|
|
return ""
|
|
}
|
|
zeroGroup := getZeroGroup(match)
|
|
return str[zeroGroup.StartIdx:zeroGroup.EndIdx]
|
|
}
|
|
|
|
// FindSubmatch returns the leftmost match of the regex in the given string, including
|
|
// the submatches matched by capturing groups. The returned [Match] will always contain the same
|
|
// number of groups. The validity of a group (whether or not it matched anything) can be determined with
|
|
// [Group.IsValid], or by checking that both indices of the group are >= 0.
|
|
// The second-return value is nil if no match was found.
|
|
func (regex Reg) FindSubmatch(str string) (Match, error) {
|
|
match, err := regex.FindNthMatch(str, 1)
|
|
if err != nil {
|
|
return Match{}, fmt.Errorf("no match found")
|
|
} else {
|
|
return match, nil
|
|
}
|
|
}
|
|
|
|
// 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 {
|
|
zerogroups := regex.FindAll(str)
|
|
matchStrs := funcMap(zerogroups, func(g Group) string {
|
|
return str[g.StartIdx:g.EndIdx]
|
|
})
|
|
return matchStrs
|
|
}
|
|
|
|
// FindNthMatch return the 'n'th match of the regex in the given string.
|
|
// It returns an error (!= nil) if there are fewer than 'n' matches in the string.
|
|
func (regex 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)
|
|
if matchFound {
|
|
matchNum++
|
|
}
|
|
if matchNum == n {
|
|
return matchIdx, nil
|
|
}
|
|
}
|
|
// We haven't found the nth match after scanning the string - Return an error
|
|
return nil, fmt.Errorf("invalid match index - too few matches found")
|
|
}
|
|
|
|
// FindAllSubmatch returns a slice of matches in the given string.
|
|
func (regex 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)
|
|
if matchFound {
|
|
indices = append(indices, matchIdx)
|
|
}
|
|
}
|
|
if len(indices) > 0 {
|
|
return pruneIndices(indices)
|
|
}
|
|
|
|
return indices
|
|
}
|
|
|
|
func addStateToList(idx int, list []nfaState, state nfaState) []nfaState {
|
|
if stateExists(list, state) {
|
|
return list
|
|
}
|
|
if state.isAlternation {
|
|
copyThread(state.next, state)
|
|
list = append(list, addStateToList(idx, list, *state.next)...)
|
|
copyThread(state.splitState, state)
|
|
list = append(list, addStateToList(idx, list, *state.splitState)...)
|
|
return list
|
|
}
|
|
if state.isKleene {
|
|
copyThread(state.splitState, state)
|
|
list = append(list, addStateToList(idx, list, *state.splitState)...)
|
|
copyThread(state.next, state)
|
|
list = append(list, addStateToList(idx, list, *state.next)...)
|
|
return list
|
|
}
|
|
if state.groupBegin {
|
|
state.threadGroups[state.groupNum].StartIdx = idx
|
|
}
|
|
if state.groupEnd {
|
|
state.threadGroups[state.groupNum].StartIdx = idx
|
|
}
|
|
copyThread(state.next, state)
|
|
return append(list, *state.next)
|
|
|
|
}
|
|
|
|
// Helper for FindAllMatches. Returns whether it found a match, the
|
|
// first Match it finds, and how far it got into the string ie. where
|
|
// the next search should start from.
|
|
//
|
|
// Might return duplicates or overlapping indices, so care must be taken to prune the resulting array.
|
|
func 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)
|
|
return false, []Group{}, offset
|
|
}
|
|
resetThreads(start)
|
|
|
|
// Hold a list of match indices for the current run. When we
|
|
// can no longer find a match, the match with the largest range is
|
|
// chosen as the match for the entire string.
|
|
// This allows us to pick the longest possible match (which is how greedy matching works).
|
|
// COMMENT ABOVE IS CURRENTLY NOT UP-TO-DATE
|
|
// tempIndices := newMatch(numGroups + 1)
|
|
|
|
// foundPath := false
|
|
//startIdx := offset
|
|
//endIdx := offset
|
|
currentStates := make([]nfaState, 0)
|
|
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 {
|
|
i++
|
|
return false, []Group{}, i
|
|
}
|
|
}
|
|
// Increment until we hit a character matching the start state (assuming not 0-state)
|
|
// if start.isEmpty == false {
|
|
// for i < len(str) && !start.contentContains(str, i) {
|
|
// i++
|
|
// }
|
|
// startIdx = i
|
|
// startingFrom = i
|
|
// i++ // Advance to next character (if we aren't at a 0-state, which doesn't match anything), so that we can check for transitions. If we advance at a 0-state, we will never get a chance to match the first character
|
|
// }
|
|
|
|
// start.threadGroups = newMatch(numGroups + 1)
|
|
// Check if the start state begins a group - if so, add the start index to our list
|
|
//if start.groupBegin {
|
|
// start.threadGroups[start.groupNum].StartIdx = i
|
|
// tempIndices[start.groupNum].startIdx = i
|
|
//}
|
|
|
|
currentStates = append(currentStates, *start)
|
|
var foundMatch bool
|
|
var isEmptyAndNoAssertion bool
|
|
// Main loop
|
|
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
|
|
// 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 {
|
|
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
|
|
}
|
|
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.
|
|
// topStateItem := currentStates.peek()
|
|
// topState := topStateItem.(*priorQueueItem).state
|
|
// zeroStates, isZero := takeZeroState([]*nfaState{topState}, numGroups, i)
|
|
// tempStates = append(tempStates, zeroStates...)
|
|
// num_appended := 0
|
|
// 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
|
|
// }
|
|
// }
|
|
// if isZero == true {
|
|
// currentStates.Pop()
|
|
// }
|
|
//
|
|
// for _, state := range tempStates {
|
|
// heap.Push(currentStates, newPriorQueueItem(state))
|
|
// }
|
|
// tempStates = nil
|
|
//
|
|
// // Take any transitions corresponding to current character
|
|
// numStatesMatched := 0 // The number of states which had at least 1 match for this round
|
|
// assertionFailed := false // Whether or not an assertion failed for this round
|
|
// lastStateInList := false // Whether or not a last state was in our list of states
|
|
// var lastStatePtr *nfaState = nil // Pointer to the last-state, if it was found
|
|
// lastLookaroundInList := false // Whether or not a last state (that is a lookaround) was in our list of states
|
|
// for numStatesMatched == 0 && lastStateInList == false {
|
|
// if currentStates.Len() == 0 {
|
|
// break
|
|
// }
|
|
// stateItem := heap.Pop(currentStates)
|
|
// state := stateItem.(*priorQueueItem).state
|
|
// matches, numMatches := state.matchesFor(str, i)
|
|
// if numMatches > 0 {
|
|
// numStatesMatched++
|
|
// tempStates = append([]*nfaState(nil), matches...)
|
|
// foundPath = true
|
|
// for _, m := range matches {
|
|
// if m.threadGroups == nil {
|
|
// m.threadGroups = newMatch(numGroups + 1)
|
|
// }
|
|
// m.threadSP = state.threadSP + 1
|
|
// copy(m.threadGroups, state.threadGroups)
|
|
// }
|
|
// }
|
|
// if numMatches < 0 {
|
|
// assertionFailed = true
|
|
// }
|
|
// if state.isLast {
|
|
// if state.isLookaround() {
|
|
// lastLookaroundInList = true
|
|
// }
|
|
// lastStateInList = true
|
|
// lastStatePtr = state
|
|
// }
|
|
// }
|
|
//
|
|
// if assertionFailed && numStatesMatched == 0 { // Nothing has matched and an assertion has failed
|
|
// // If I'm being completely honest, I'm not sure why I have to check specifically for a _lookaround_
|
|
// // state. The explanation below is my attempt to explain this behavior.
|
|
// // If you replace 'lastLookaroundInList' with 'lastStateInList', one of the test cases fails.
|
|
// //
|
|
// // One of the states in our list was a last state and a lookaround. In this case, we
|
|
// // don't abort upon failure of the assertion, because we have found
|
|
// // another path to a final state.
|
|
// // Even if the last state _was_ an assertion, we can use the previously
|
|
// // saved indices to find a match.
|
|
// if lastLookaroundInList {
|
|
// break
|
|
// } else {
|
|
// if i == startingFrom {
|
|
// i++
|
|
// }
|
|
// return false, []Group{}, i
|
|
// }
|
|
// }
|
|
// // Check if we can find a state in our list that is:
|
|
// // a. A last-state
|
|
// // b. Empty
|
|
// // c. Doesn't assert anything
|
|
// for _, stateItem := range *currentStates {
|
|
// s := stateItem.state
|
|
// if s.isLast && s.isEmpty && s.assert == noneAssert {
|
|
// lastStatePtr = s
|
|
// lastStateInList = true
|
|
// }
|
|
// }
|
|
// if lastStateInList && numStatesMatched == 0 { // A last-state was in the list of states. add the matchIndex to our MatchIndex list
|
|
// for j := 1; j < numGroups+1; j++ {
|
|
// tempIndices[j] = lastStatePtr.threadGroups[j]
|
|
// }
|
|
// endIdx = i
|
|
// tempIndices[0] = Group{startIdx, endIdx}
|
|
// if tempIndices[0].StartIdx == tempIndices[0].EndIdx {
|
|
// return true, tempIndices, tempIndices[0].EndIdx + 1
|
|
// } else {
|
|
// return true, tempIndices, tempIndices[0].EndIdx
|
|
// }
|
|
// }
|
|
//
|
|
// // Check if we can find a zero-length match
|
|
// if foundPath == false {
|
|
// currentStatesList := funcMap(*currentStates, func(item *priorQueueItem) *nfaState {
|
|
// return item.state
|
|
// })
|
|
// if ok := zeroMatchPossible(str, i, numGroups, currentStatesList...); ok {
|
|
// if tempIndices[0].IsValid() == false {
|
|
// tempIndices[0] = Group{startIdx, startIdx}
|
|
// }
|
|
// }
|
|
// // If we haven't moved in the string, increment the counter by 1
|
|
// // to ensure we don't keep trying the same string over and over.
|
|
// // if i == startingFrom {
|
|
// startIdx++
|
|
// // i++
|
|
// // }
|
|
// if tempIndices.numValidGroups() > 0 && tempIndices[0].IsValid() {
|
|
// if tempIndices[0].StartIdx == tempIndices[0].EndIdx { // If we have a zero-length match, we have to shift the index at which we start. Otherwise we keep looking at the same paert of the string over and over.
|
|
// return true, tempIndices, tempIndices[0].EndIdx + 1
|
|
// } else {
|
|
// return true, tempIndices, tempIndices[0].EndIdx
|
|
// }
|
|
// }
|
|
// return false, []Group{}, startIdx
|
|
// }
|
|
// currentStates = &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
|
|
}
|