Generalize command word fuzzy matching to one edit
This commit is contained in:
Lawrence Chen
parent
a84e6304c9
commit
03e1fcf6c5
1 changed files with 182 additions and 72 deletions
|
|
@ -6035,11 +6035,33 @@ enum CommandPaletteSwitcherSearchIndexer {
|
|||
enum CommandPaletteFuzzyMatcher {
|
||||
private static let tokenBoundaryChars: Set<Character> = [" ", "-", "_", "/", ".", ":"]
|
||||
|
||||
private struct SingleOmittedCharacterWordPrefixMatch {
|
||||
private enum SingleEditWordPrefixEditKind {
|
||||
case candidateExtraCharacter
|
||||
case tokenExtraCharacter
|
||||
case substitutedCharacter
|
||||
case transposedCharacters
|
||||
|
||||
var basePenalty: Int {
|
||||
switch self {
|
||||
case .candidateExtraCharacter:
|
||||
return 0
|
||||
case .tokenExtraCharacter:
|
||||
return 10
|
||||
case .transposedCharacters:
|
||||
return 24
|
||||
case .substitutedCharacter:
|
||||
return 40
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
private struct SingleEditWordPrefixMatch {
|
||||
let matchedIndices: Set<Int>
|
||||
let segmentStart: Int
|
||||
let segmentLength: Int
|
||||
let omittedIndex: Int
|
||||
let prefixLength: Int
|
||||
let editPosition: Int
|
||||
let editKind: SingleEditWordPrefixEditKind
|
||||
}
|
||||
|
||||
struct PreparedQuery {
|
||||
|
|
@ -6127,8 +6149,8 @@ enum CommandPaletteFuzzyMatcher {
|
|||
continue
|
||||
}
|
||||
|
||||
if let omittedCharacterPrefix = singleOmittedCharacterWordPrefixMatch(token: token, candidate: loweredCandidate) {
|
||||
matched.formUnion(omittedCharacterPrefix.matchedIndices)
|
||||
if let singleEditPrefix = singleEditWordPrefixMatch(token: token, candidate: loweredCandidate) {
|
||||
matched.formUnion(singleEditPrefix.matchedIndices)
|
||||
continue
|
||||
}
|
||||
|
||||
|
|
@ -6172,11 +6194,11 @@ enum CommandPaletteFuzzyMatcher {
|
|||
if let wordPrefixScore = bestWordScore(tokenChars: tokenChars, candidateChars: candidateChars, requireExactWord: false) {
|
||||
bestScore = max(bestScore ?? wordPrefixScore, wordPrefixScore)
|
||||
}
|
||||
if let omittedCharacterPrefixScore = singleOmittedCharacterWordPrefixScore(
|
||||
if let singleEditPrefixScore = singleEditWordPrefixScore(
|
||||
tokenChars: tokenChars,
|
||||
candidateChars: candidateChars
|
||||
) {
|
||||
bestScore = max(bestScore ?? omittedCharacterPrefixScore, omittedCharacterPrefixScore)
|
||||
bestScore = max(bestScore ?? singleEditPrefixScore, singleEditPrefixScore)
|
||||
}
|
||||
|
||||
if let range = candidate.range(of: token) {
|
||||
|
|
@ -6238,31 +6260,33 @@ enum CommandPaletteFuzzyMatcher {
|
|||
return best
|
||||
}
|
||||
|
||||
private static func singleOmittedCharacterWordPrefixScore(
|
||||
private static func singleEditWordPrefixScore(
|
||||
tokenChars: [Character],
|
||||
candidateChars: [Character]
|
||||
) -> Int? {
|
||||
guard tokenChars.count >= 4 else { return nil }
|
||||
|
||||
var best: Int?
|
||||
for segment in wordSegments(candidateChars) {
|
||||
guard let match = singleOmittedCharacterWordPrefixMatch(
|
||||
tokenChars: tokenChars,
|
||||
candidateChars: candidateChars,
|
||||
segment: segment
|
||||
) else {
|
||||
continue
|
||||
}
|
||||
|
||||
let lengthPenalty = max(0, match.segmentLength - tokenChars.count - 1) * 6
|
||||
let distancePenalty = match.segmentStart * 8
|
||||
let trailingPenalty = max(0, candidateChars.count - match.segmentLength)
|
||||
let omissionPenalty = max(0, match.omittedIndex - match.segmentStart) * 10
|
||||
let score = 5000 - distancePenalty - lengthPenalty - trailingPenalty - omissionPenalty
|
||||
best = max(best ?? score, score)
|
||||
guard let match = singleEditWordPrefixMatch(
|
||||
tokenChars: tokenChars,
|
||||
candidateChars: candidateChars
|
||||
) else {
|
||||
return nil
|
||||
}
|
||||
return singleEditWordPrefixScore(match: match, candidateLength: candidateChars.count)
|
||||
}
|
||||
|
||||
return best
|
||||
private static func singleEditWordPrefixScore(
|
||||
match: SingleEditWordPrefixMatch,
|
||||
candidateLength: Int
|
||||
) -> Int {
|
||||
let lengthPenalty = max(0, match.segmentLength - match.prefixLength) * 6
|
||||
let distancePenalty = match.segmentStart * 8
|
||||
let trailingPenalty = max(0, candidateLength - match.segmentLength)
|
||||
let editPositionPenalty = max(0, match.editPosition - match.segmentStart) * 10
|
||||
return 5000
|
||||
- match.editKind.basePenalty
|
||||
- distancePenalty
|
||||
- lengthPenalty
|
||||
- trailingPenalty
|
||||
- editPositionPenalty
|
||||
}
|
||||
|
||||
private static func initialismScore(tokenChars: [Character], candidateChars: [Character]) -> Int? {
|
||||
|
|
@ -6299,9 +6323,10 @@ enum CommandPaletteFuzzyMatcher {
|
|||
candidateChars: [Character],
|
||||
candidateStart: Int
|
||||
) -> Bool {
|
||||
guard length > 0 else { return false }
|
||||
guard length >= 0 else { return false }
|
||||
guard tokenStart + length <= tokenChars.count else { return false }
|
||||
guard candidateStart + length <= candidateChars.count else { return false }
|
||||
guard length > 0 else { return true }
|
||||
|
||||
for offset in 0..<length where tokenChars[tokenStart + offset] != candidateChars[candidateStart + offset] {
|
||||
return false
|
||||
|
|
@ -6432,27 +6457,27 @@ enum CommandPaletteFuzzyMatcher {
|
|||
return matchedIndices
|
||||
}
|
||||
|
||||
private static func singleOmittedCharacterWordPrefixMatch(
|
||||
private static func singleEditWordPrefixMatch(
|
||||
token: String,
|
||||
candidate: String
|
||||
) -> SingleOmittedCharacterWordPrefixMatch? {
|
||||
singleOmittedCharacterWordPrefixMatch(
|
||||
) -> SingleEditWordPrefixMatch? {
|
||||
singleEditWordPrefixMatch(
|
||||
tokenChars: Array(token),
|
||||
candidateChars: Array(candidate)
|
||||
)
|
||||
}
|
||||
|
||||
private static func singleOmittedCharacterWordPrefixMatch(
|
||||
private static func singleEditWordPrefixMatch(
|
||||
tokenChars: [Character],
|
||||
candidateChars: [Character]
|
||||
) -> SingleOmittedCharacterWordPrefixMatch? {
|
||||
) -> SingleEditWordPrefixMatch? {
|
||||
guard tokenChars.count >= 4 else { return nil }
|
||||
|
||||
var bestMatch: SingleOmittedCharacterWordPrefixMatch?
|
||||
var bestMatch: SingleEditWordPrefixMatch?
|
||||
var bestScore: Int?
|
||||
|
||||
for segment in wordSegments(candidateChars) {
|
||||
guard let match = singleOmittedCharacterWordPrefixMatch(
|
||||
guard let match = singleEditWordPrefixMatch(
|
||||
tokenChars: tokenChars,
|
||||
candidateChars: candidateChars,
|
||||
segment: segment
|
||||
|
|
@ -6460,12 +6485,7 @@ enum CommandPaletteFuzzyMatcher {
|
|||
continue
|
||||
}
|
||||
|
||||
let score = 5000
|
||||
- (match.segmentStart * 8)
|
||||
- (max(0, match.segmentLength - tokenChars.count - 1) * 6)
|
||||
- (max(0, candidateChars.count - match.segmentLength))
|
||||
- (max(0, match.omittedIndex - match.segmentStart) * 10)
|
||||
|
||||
let score = singleEditWordPrefixScore(match: match, candidateLength: candidateChars.count)
|
||||
if let bestScore, score <= bestScore {
|
||||
continue
|
||||
}
|
||||
|
|
@ -6476,49 +6496,139 @@ enum CommandPaletteFuzzyMatcher {
|
|||
return bestMatch
|
||||
}
|
||||
|
||||
private static func singleOmittedCharacterWordPrefixMatch(
|
||||
private static func singleEditWordPrefixMatch(
|
||||
tokenChars: [Character],
|
||||
candidateChars: [Character],
|
||||
segment: (start: Int, end: Int)
|
||||
) -> SingleOmittedCharacterWordPrefixMatch? {
|
||||
) -> SingleEditWordPrefixMatch? {
|
||||
guard tokenChars.count >= 4 else { return nil }
|
||||
|
||||
let segmentLength = segment.end - segment.start
|
||||
guard segmentLength >= tokenChars.count + 1 else { return nil }
|
||||
guard segmentLength + 1 >= tokenChars.count else { return nil }
|
||||
|
||||
var tokenIndex = 0
|
||||
var candidateIndex = segment.start
|
||||
var omittedIndex: Int?
|
||||
var matchedIndices: Set<Int> = []
|
||||
|
||||
while tokenIndex < tokenChars.count, candidateIndex < segment.end {
|
||||
if candidateChars[candidateIndex] == tokenChars[tokenIndex] {
|
||||
matchedIndices.insert(candidateIndex)
|
||||
tokenIndex += 1
|
||||
candidateIndex += 1
|
||||
continue
|
||||
}
|
||||
|
||||
guard omittedIndex == nil else { return nil }
|
||||
omittedIndex = candidateIndex
|
||||
candidateIndex += 1
|
||||
let exactPrefixLength = min(tokenChars.count, segmentLength)
|
||||
var mismatchOffset = 0
|
||||
while mismatchOffset < exactPrefixLength,
|
||||
candidateChars[segment.start + mismatchOffset] == tokenChars[mismatchOffset]
|
||||
{
|
||||
mismatchOffset += 1
|
||||
}
|
||||
|
||||
guard tokenIndex == tokenChars.count else { return nil }
|
||||
let resolvedOmittedIndex: Int
|
||||
if let omittedIndex {
|
||||
resolvedOmittedIndex = omittedIndex
|
||||
} else {
|
||||
guard candidateIndex < segment.end else { return nil }
|
||||
resolvedOmittedIndex = candidateIndex
|
||||
if mismatchOffset == tokenChars.count {
|
||||
let prefixLength = tokenChars.count + 1
|
||||
guard segmentLength >= prefixLength else { return nil }
|
||||
return SingleEditWordPrefixMatch(
|
||||
matchedIndices: Set(segment.start..<(segment.start + tokenChars.count)),
|
||||
segmentStart: segment.start,
|
||||
segmentLength: segmentLength,
|
||||
prefixLength: prefixLength,
|
||||
editPosition: segment.start + tokenChars.count,
|
||||
editKind: .candidateExtraCharacter
|
||||
)
|
||||
}
|
||||
|
||||
return SingleOmittedCharacterWordPrefixMatch(
|
||||
matchedIndices: matchedIndices,
|
||||
segmentStart: segment.start,
|
||||
segmentLength: segmentLength,
|
||||
omittedIndex: resolvedOmittedIndex
|
||||
)
|
||||
if mismatchOffset == segmentLength {
|
||||
let prefixLength = tokenChars.count - 1
|
||||
guard prefixLength > 0 else { return nil }
|
||||
guard tokenChars.count == segmentLength + 1 else { return nil }
|
||||
return SingleEditWordPrefixMatch(
|
||||
matchedIndices: Set(segment.start..<(segment.start + prefixLength)),
|
||||
segmentStart: segment.start,
|
||||
segmentLength: segmentLength,
|
||||
prefixLength: prefixLength,
|
||||
editPosition: segment.start + prefixLength,
|
||||
editKind: .tokenExtraCharacter
|
||||
)
|
||||
}
|
||||
|
||||
let mismatchCandidateIndex = segment.start + mismatchOffset
|
||||
|
||||
if segmentLength >= tokenChars.count + 1,
|
||||
tokenPrefixMatches(
|
||||
tokenChars: tokenChars,
|
||||
tokenStart: mismatchOffset,
|
||||
length: tokenChars.count - mismatchOffset,
|
||||
candidateChars: candidateChars,
|
||||
candidateStart: mismatchCandidateIndex + 1
|
||||
)
|
||||
{
|
||||
var matchedIndices = Set(segment.start..<(segment.start + tokenChars.count + 1))
|
||||
matchedIndices.remove(mismatchCandidateIndex)
|
||||
return SingleEditWordPrefixMatch(
|
||||
matchedIndices: matchedIndices,
|
||||
segmentStart: segment.start,
|
||||
segmentLength: segmentLength,
|
||||
prefixLength: tokenChars.count + 1,
|
||||
editPosition: mismatchCandidateIndex,
|
||||
editKind: .candidateExtraCharacter
|
||||
)
|
||||
}
|
||||
|
||||
if tokenChars.count >= 2,
|
||||
segmentLength >= tokenChars.count - 1,
|
||||
tokenPrefixMatches(
|
||||
tokenChars: tokenChars,
|
||||
tokenStart: mismatchOffset + 1,
|
||||
length: tokenChars.count - mismatchOffset - 1,
|
||||
candidateChars: candidateChars,
|
||||
candidateStart: mismatchCandidateIndex
|
||||
)
|
||||
{
|
||||
return SingleEditWordPrefixMatch(
|
||||
matchedIndices: Set(segment.start..<(segment.start + tokenChars.count - 1)),
|
||||
segmentStart: segment.start,
|
||||
segmentLength: segmentLength,
|
||||
prefixLength: tokenChars.count - 1,
|
||||
editPosition: mismatchCandidateIndex,
|
||||
editKind: .tokenExtraCharacter
|
||||
)
|
||||
}
|
||||
|
||||
if segmentLength >= tokenChars.count,
|
||||
tokenPrefixMatches(
|
||||
tokenChars: tokenChars,
|
||||
tokenStart: mismatchOffset + 1,
|
||||
length: tokenChars.count - mismatchOffset - 1,
|
||||
candidateChars: candidateChars,
|
||||
candidateStart: mismatchCandidateIndex + 1
|
||||
)
|
||||
{
|
||||
var matchedIndices = Set(segment.start..<(segment.start + tokenChars.count))
|
||||
matchedIndices.remove(mismatchCandidateIndex)
|
||||
return SingleEditWordPrefixMatch(
|
||||
matchedIndices: matchedIndices,
|
||||
segmentStart: segment.start,
|
||||
segmentLength: segmentLength,
|
||||
prefixLength: tokenChars.count,
|
||||
editPosition: mismatchCandidateIndex,
|
||||
editKind: .substitutedCharacter
|
||||
)
|
||||
}
|
||||
|
||||
if segmentLength >= tokenChars.count,
|
||||
mismatchOffset + 1 < tokenChars.count,
|
||||
mismatchCandidateIndex + 1 < segment.end,
|
||||
tokenChars[mismatchOffset] == candidateChars[mismatchCandidateIndex + 1],
|
||||
tokenChars[mismatchOffset + 1] == candidateChars[mismatchCandidateIndex],
|
||||
tokenPrefixMatches(
|
||||
tokenChars: tokenChars,
|
||||
tokenStart: mismatchOffset + 2,
|
||||
length: tokenChars.count - mismatchOffset - 2,
|
||||
candidateChars: candidateChars,
|
||||
candidateStart: mismatchCandidateIndex + 2
|
||||
)
|
||||
{
|
||||
return SingleEditWordPrefixMatch(
|
||||
matchedIndices: Set(segment.start..<(segment.start + tokenChars.count)),
|
||||
segmentStart: segment.start,
|
||||
segmentLength: segmentLength,
|
||||
prefixLength: tokenChars.count,
|
||||
editPosition: mismatchCandidateIndex,
|
||||
editKind: .transposedCharacters
|
||||
)
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
private static func wordSegments(_ candidateChars: [Character]) -> [(start: Int, end: Int)] {
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue