Added lookaround-related fields to State struct, added lookaround support to checkAssertion()
This commit is contained in:
80
nfa.go
80
nfa.go
@@ -1,6 +1,8 @@
|
||||
package main
|
||||
|
||||
import "slices"
|
||||
import (
|
||||
"slices"
|
||||
)
|
||||
|
||||
const EPSILON int = 0
|
||||
|
||||
@@ -12,19 +14,24 @@ const (
|
||||
EOS
|
||||
WBOUND
|
||||
NONWBOUND
|
||||
PLA // Positive lookahead
|
||||
NLA // Negative lookahead
|
||||
PLB // Positive lookbehind
|
||||
NLB // Negative lookbehind
|
||||
)
|
||||
|
||||
type State struct {
|
||||
content stateContents // Contents of current state
|
||||
isEmpty bool // If it is empty - Union operator and Kleene star states will be empty
|
||||
isLast bool // If it is the last state (acept state)
|
||||
output []*State // The outputs of the current state ie. the 'outward arrows'. A union operator state will have more than one of these.
|
||||
transitions map[int][]*State // Transitions to different states (maps a character (int representation) to a _list of states. This is useful if one character can lead multiple states eg. ab|aa)
|
||||
isKleene bool // Identifies whether current node is a 0-state representing Kleene star
|
||||
assert assertType // Type of assertion of current node - NONE means that the node doesn't assert anything
|
||||
zeroMatchFound bool // Whether or not the state has been used for a zero-length match - only relevant for zero states
|
||||
allChars bool // Whether or not the state represents all characters (eg. a 'dot' metacharacter). A 'dot' node doesn't store any contents directly, as it would take up too much space
|
||||
except []rune // Only valid if allChars is true - match all characters _except_ the ones in this block. Useful for inverting character classes.
|
||||
content stateContents // Contents of current state
|
||||
isEmpty bool // If it is empty - Union operator and Kleene star states will be empty
|
||||
isLast bool // If it is the last state (acept state)
|
||||
output []*State // The outputs of the current state ie. the 'outward arrows'. A union operator state will have more than one of these.
|
||||
transitions map[int][]*State // Transitions to different states (maps a character (int representation) to a _list of states. This is useful if one character can lead multiple states eg. ab|aa)
|
||||
isKleene bool // Identifies whether current node is a 0-state representing Kleene star
|
||||
assert assertType // Type of assertion of current node - NONE means that the node doesn't assert anything
|
||||
zeroMatchFound bool // Whether or not the state has been used for a zero-length match - only relevant for zero states
|
||||
allChars bool // Whether or not the state represents all characters (eg. a 'dot' metacharacter). A 'dot' node doesn't store any contents directly, as it would take up too much space
|
||||
except []rune // Only valid if allChars is true - match all characters _except_ the ones in this block. Useful for inverting character classes.
|
||||
lookaroundRegex string // Only for lookaround states - Contents of the regex that the lookaround state holds
|
||||
}
|
||||
|
||||
// Clones the NFA starting from the given state.
|
||||
@@ -43,16 +50,17 @@ func cloneStateHelper(state *State, cloneMap map[*State]*State) *State {
|
||||
// Recursive case - if the clone doesn't exist, create it, add it to the map,
|
||||
// and recursively call for each of the transition states.
|
||||
clone := &State{
|
||||
content: append([]int{}, state.content...),
|
||||
isEmpty: state.isEmpty,
|
||||
isLast: state.isLast,
|
||||
output: make([]*State, len(state.output)),
|
||||
transitions: make(map[int][]*State),
|
||||
isKleene: state.isKleene,
|
||||
assert: state.assert,
|
||||
zeroMatchFound: state.zeroMatchFound,
|
||||
allChars: state.allChars,
|
||||
except: append([]rune{}, state.except...),
|
||||
content: append([]int{}, state.content...),
|
||||
isEmpty: state.isEmpty,
|
||||
isLast: state.isLast,
|
||||
output: make([]*State, len(state.output)),
|
||||
transitions: make(map[int][]*State),
|
||||
isKleene: state.isKleene,
|
||||
assert: state.assert,
|
||||
zeroMatchFound: state.zeroMatchFound,
|
||||
allChars: state.allChars,
|
||||
except: append([]rune{}, state.except...),
|
||||
lookaroundRegex: state.lookaroundRegex,
|
||||
}
|
||||
cloneMap[state] = clone
|
||||
for i, s := range state.output {
|
||||
@@ -90,6 +98,36 @@ func (s State) checkAssertion(str []rune, idx int) bool {
|
||||
if s.assert == NONWBOUND {
|
||||
return !isWordBoundary(str, idx)
|
||||
}
|
||||
if s.assert == PLA || s.assert == PLB || s.assert == NLA || s.assert == NLB { // Lookaround
|
||||
// The process here is simple:
|
||||
// 1. Compile the regex stored in the state's contents.
|
||||
// 2. Run it on the test string.
|
||||
// 3. Based on the kind of lookaround (and the indices we get), determine what action to take.
|
||||
regex := s.lookaroundRegex
|
||||
re_postfix := shuntingYard(regex)
|
||||
startState := thompson(re_postfix)
|
||||
matchIndices := findAllMatches(startState, str)
|
||||
|
||||
numMatchesFound := 0
|
||||
for _, matchIdx := range matchIndices {
|
||||
if s.assert == PLA || s.assert == NLA { // Lookahead - return true (or false) if at least one match starts at the current index
|
||||
if matchIdx.startIdx == idx {
|
||||
numMatchesFound++
|
||||
}
|
||||
}
|
||||
if s.assert == PLB || s.assert == NLB { // Lookbehind - return true (or false) if at least one match _ends_ at the current index.
|
||||
if matchIdx.endIdx == idx {
|
||||
numMatchesFound++
|
||||
}
|
||||
}
|
||||
}
|
||||
if s.assert == PLA || s.assert == PLB { // Positive assertions want at least one match
|
||||
return numMatchesFound > 0
|
||||
}
|
||||
if s.assert == NLA || s.assert == NLB { // Negative assertions only want zero matches
|
||||
return numMatchesFound == 0
|
||||
}
|
||||
}
|
||||
return true
|
||||
}
|
||||
|
||||
|
Reference in New Issue
Block a user