AliAlavi
/
soniox-mac-voice-typing
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
Browse Source

fix(input-box): eliminate trailing word duplication via aligned index mapping & synchronous paste guard (v2.08)

main
Ali Alavi 1 week ago
parent
14f155732e
commit
eedd895c70
3 changed files with 80 additions and 93 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 2
      Info.plist
  2. 167
      src/InputBoxManager.swift
  3. 4
      src/TextProcessor.swift

2
Info.plist
View File

@ -19,7 +19,7 @@
<key>CFBundlePackageType</key>
<string>APPL</string>
<key>CFBundleShortVersionString</key>
<string>2.07</string>
<string>2.08</string>
<key>CFBundleVersion</key>
<string>1</string>
<key>LSMinimumSystemVersion</key>

167
src/InputBoxManager.swift
View File

@ -14,7 +14,7 @@ public final class InputBoxManager {
private let pasteQueue = DispatchQueue(label: "com.soniox.pasteQueue", qos: .userInteractive)
private static let punctRegex: NSRegularExpression = {
return try! NSRegularExpression(pattern: #"[.,!?:;،؛؟"'()\[\]{}«»–—\-]"#, options: [])
return try! NSRegularExpression(pattern: #"[.,!?:;،؛؟"'()\[\]{}«»–—\-…]"#, options: [])
}()
private init() {}
@ -101,11 +101,7 @@ public final class InputBoxManager {
self.lock.unlock()
guard valid else { return }
let pb = NSPasteboard.general
pb.clearContents()
pb.setString(textToPaste, forType: .string)
self.postCleanCmdV()
usleep(10000) // Ultra-fast 10ms pacing between streaming chunks
self.pasteSynchronously(textToPaste)
}
return
}
@ -138,11 +134,7 @@ public final class InputBoxManager {
self.lock.unlock()
guard valid else { return }
let pb = NSPasteboard.general
pb.clearContents()
pb.setString(toFlush, forType: .string)
self.postCleanCmdV()
usleep(10000)
self.pasteSynchronously(toFlush)
}
return
}
@ -169,11 +161,7 @@ public final class InputBoxManager {
self.lock.unlock()
guard valid else { return }
let pb = NSPasteboard.general
pb.clearContents()
pb.setString(textToPaste, forType: .string)
self.postCleanCmdV()
usleep(10000)
self.pasteSynchronously(textToPaste)
}
return
}
@ -201,11 +189,7 @@ public final class InputBoxManager {
self.lock.unlock()
guard valid else { return }
let pb = NSPasteboard.general
pb.clearContents()
pb.setString(toFlush, forType: .string)
self.postCleanCmdV()
usleep(10000)
self.pasteSynchronously(toFlush)
}
return
}
@ -213,18 +197,13 @@ public final class InputBoxManager {
lock.unlock()
}
/// Commits the final speech result: ensures 100% complete text in input box & clipboard
/// Commits the final speech result: ensures 100% complete text in input box & clipboard without ANY duplication
public func commitFinalText(fullText: String, addTrailingSpace: Bool) {
let trimmed = fullText.trimmingCharacters(in: .whitespacesAndNewlines)
guard !trimmed.isEmpty else { return }
let textWithTrailing = addTrailingSpace ? (trimmed + " ") : trimmed
// Always place the full processed text on the system clipboard
let pb = NSPasteboard.general
pb.clearContents()
pb.setString(textWithTrailing, forType: .string)
lock.lock()
let expectedGen = sessionGeneration
// Flush any pending buffer if recording ended before prefix resolution
@ -237,11 +216,14 @@ public final class InputBoxManager {
accumulatedStreamedText += pending
}
}
let streamed = accumulatedStreamedText
let rawStreamed = accumulatedStreamedText
accumulatedStreamedText = ""
lock.unlock()
let cleanStreamed = streamed.trimmingCharacters(in: .whitespacesAndNewlines)
// CRITICAL: Process the streamed text through the EXACT same TextProcessor pipeline as fullText!
// This guarantees that numbers (۱ -> 1), ellipses (... -> . . .), and punctuation tokens have identical word counts & tokens!
let processedStreamed = rawStreamed.isEmpty ? "" : TextProcessor.shared.process(text: rawStreamed)
let cleanStreamed = processedStreamed.trimmingCharacters(in: .whitespacesAndNewlines)
pasteQueue.async { [weak self] in
guard let self = self else { return }
@ -250,7 +232,6 @@ public final class InputBoxManager {
self.lock.unlock()
guard valid else { return }
// Safe 250ms clipboard restore with generation cancellation to prevent double-paste in heavy apps
let restoreClipboardFullText = { [weak self] in
guard let self = self else { return }
let work = DispatchWorkItem { [weak self] in
@ -267,7 +248,7 @@ public final class InputBoxManager {
self.clipboardRestoreWorkItem?.cancel()
self.clipboardRestoreWorkItem = work
self.lock.unlock()
DispatchQueue.main.asyncAfter(deadline: .now() + 0.25, execute: work)
DispatchQueue.main.asyncAfter(deadline: .now() + 0.30, execute: work)
}
// Case 1: Nothing was streamed live -> Paste entire text
@ -283,10 +264,7 @@ public final class InputBoxManager {
// Case 2: Exactly matches streamed text
if cleanStreamed == trimmed {
if addTrailingSpace {
let pb = NSPasteboard.general
pb.clearContents()
pb.setString(" ", forType: .string)
self.postCleanCmdV()
self.pasteSynchronously(" ")
}
restoreClipboardFullText()
return
@ -296,29 +274,34 @@ public final class InputBoxManager {
if trimmed.count > cleanStreamed.count && trimmed.hasPrefix(cleanStreamed) {
let suffix = String(trimmed.dropFirst(cleanStreamed.count))
let suffixToPaste = addTrailingSpace ? (suffix + " ") : suffix
let pb = NSPasteboard.general
pb.clearContents()
pb.setString(suffixToPaste, forType: .string)
self.postCleanCmdV()
self.pasteSynchronously(suffixToPaste)
restoreClipboardFullText()
return
}
// Case 4: Intelligent Word-level Tail-Anchor Matching
let streamedWords = cleanStreamed.components(separatedBy: .whitespacesAndNewlines).filter { !$0.isEmpty }
let finalWords = trimmed.components(separatedBy: .whitespacesAndNewlines).filter { !$0.isEmpty }
guard !finalWords.isEmpty else {
restoreClipboardFullText()
return
}
// Case 4: Word-level Tail-Anchor Matching with 1-to-1 Aligned Index Mapping
let rawStreamedWords = cleanStreamed.components(separatedBy: .whitespacesAndNewlines).filter { !$0.isEmpty }
let rawFinalWords = trimmed.components(separatedBy: .whitespacesAndNewlines).filter { !$0.isEmpty }
let cleanWord: (String) -> String = { w in
let noZwnj = w.replacingOccurrences(of: "\u{200C}", with: "").replacingOccurrences(of: "\u{200D}", with: "")
return InputBoxManager.stripPunctuation(noZwnj).lowercased()
}
let normStreamed = streamedWords.map(cleanWord).filter { !$0.isEmpty }
let normFinal = finalWords.map(cleanWord).filter { !$0.isEmpty }
// Keep ONLY words that contain actual alphanumeric/Persian letters so standalone punctuation (".", "...") NEVER shifts indices!
let streamedPairs: [(raw: String, norm: String)] = rawStreamedWords.compactMap { w in
let nw = cleanWord(w)
return nw.isEmpty ? nil : (w, nw)
}
let finalPairs: [(raw: String, norm: String)] = rawFinalWords.compactMap { w in
let nw = cleanWord(w)
return nw.isEmpty ? nil : (w, nw)
}
guard !finalPairs.isEmpty else {
restoreClipboardFullText()
return
}
let wordsMatch: (String, String) -> Bool = { w1, w2 in
if w1 == w2 { return true }
@ -328,83 +311,76 @@ public final class InputBoxManager {
return false
}
var unstreamedStartIndex: Int? = nil
var unstreamedPairIndex: Int? = nil
// Strategy 4A: Search for 3-word tail anchor from streamed text in final words
if normStreamed.count >= 3 {
let a1 = normStreamed[normStreamed.count - 3]
let a2 = normStreamed[normStreamed.count - 2]
let a3 = normStreamed[normStreamed.count - 1]
for j in stride(from: normFinal.count - 1, through: 2, by: -1) {
if wordsMatch(normFinal[j], a3) && wordsMatch(normFinal[j - 1], a2) && wordsMatch(normFinal[j - 2], a1) {
unstreamedStartIndex = j + 1
// Strategy 4A: 3-word tail anchor
if streamedPairs.count >= 3 {
let a1 = streamedPairs[streamedPairs.count - 3].norm
let a2 = streamedPairs[streamedPairs.count - 2].norm
let a3 = streamedPairs[streamedPairs.count - 1].norm
for j in stride(from: finalPairs.count - 1, through: 2, by: -1) {
if wordsMatch(finalPairs[j].norm, a3) && wordsMatch(finalPairs[j - 1].norm, a2) && wordsMatch(finalPairs[j - 2].norm, a1) {
unstreamedPairIndex = j + 1
break
}
}
}
// Strategy 4B: Search for 2-word tail anchor
if unstreamedStartIndex == nil && normStreamed.count >= 2 {
let a1 = normStreamed[normStreamed.count - 2]
let a2 = normStreamed[normStreamed.count - 1]
for j in stride(from: normFinal.count - 1, through: 1, by: -1) {
if wordsMatch(normFinal[j], a2) && wordsMatch(normFinal[j - 1], a1) {
unstreamedStartIndex = j + 1
// Strategy 4B: 2-word tail anchor
if unstreamedPairIndex == nil && streamedPairs.count >= 2 {
let a1 = streamedPairs[streamedPairs.count - 2].norm
let a2 = streamedPairs[streamedPairs.count - 1].norm
for j in stride(from: finalPairs.count - 1, through: 1, by: -1) {
if wordsMatch(finalPairs[j].norm, a2) && wordsMatch(finalPairs[j - 1].norm, a1) {
unstreamedPairIndex = j + 1
break
}
}
}
// Strategy 4C: Search for 1-word tail anchor in recent final words
if unstreamedStartIndex == nil && !normStreamed.isEmpty {
let a1 = normStreamed[normStreamed.count - 1]
let searchStart = max(0, normFinal.count - max(15, normStreamed.count))
for j in stride(from: normFinal.count - 1, through: searchStart, by: -1) {
if wordsMatch(normFinal[j], a1) {
unstreamedStartIndex = j + 1
// Strategy 4C: 1-word tail anchor
if unstreamedPairIndex == nil && !streamedPairs.isEmpty {
let a1 = streamedPairs[streamedPairs.count - 1].norm
let searchStart = max(0, finalPairs.count - max(15, streamedPairs.count))
for j in stride(from: finalPairs.count - 1, through: searchStart, by: -1) {
if wordsMatch(finalPairs[j].norm, a1) {
unstreamedPairIndex = j + 1
break
}
}
}
// Strategy 4D: Forward prefix match fallback
if unstreamedStartIndex == nil {
if unstreamedPairIndex == nil {
var prefixMatchCount = 0
for i in 0..<min(normStreamed.count, normFinal.count) {
if wordsMatch(normStreamed[i], normFinal[i]) {
for i in 0..<min(streamedPairs.count, finalPairs.count) {
if wordsMatch(streamedPairs[i].norm, finalPairs[i].norm) {
prefixMatchCount = i + 1
} else {
break
}
}
if prefixMatchCount > 0 {
unstreamedStartIndex = prefixMatchCount
unstreamedPairIndex = prefixMatchCount
}
}
// Guaranteed Safety Fallback: If no anchor matched, point to streamedWords.count
if unstreamedStartIndex == nil {
unstreamedStartIndex = min(streamedWords.count, finalWords.count)
// Guaranteed Safety Fallback: aligned 1-to-1 on content-bearing pairs
if unstreamedPairIndex == nil {
unstreamedPairIndex = min(streamedPairs.count, finalPairs.count)
}
// Suffix Evaluation: Always paste all remaining unstreamed words
if let startIndex = unstreamedStartIndex {
if startIndex < finalWords.count {
let remainingWords = Array(finalWords.dropFirst(startIndex))
if let startIndex = unstreamedPairIndex {
if startIndex < finalPairs.count {
let remainingWords = finalPairs.dropFirst(startIndex).map { $0.raw }
let suffix = remainingWords.joined(separator: " ")
let suffixToPaste = addTrailingSpace ? (" " + suffix + " ") : (" " + suffix)
let pb = NSPasteboard.general
pb.clearContents()
pb.setString(suffixToPaste, forType: .string)
self.postCleanCmdV()
self.pasteSynchronously(suffixToPaste)
restoreClipboardFullText()
return
} else {
if addTrailingSpace {
let pb = NSPasteboard.general
pb.clearContents()
pb.setString(" ", forType: .string)
self.postCleanCmdV()
self.pasteSynchronously(" ")
}
restoreClipboardFullText()
return
@ -414,6 +390,15 @@ public final class InputBoxManager {
}
}
/// Places text on NSPasteboard, posts Cmd+V, and waits 45ms so target app consumes this chunk BEFORE next clipboard write!
private func pasteSynchronously(_ text: String) {
let pb = NSPasteboard.general
pb.clearContents()
pb.setString(text, forType: .string)
postCleanCmdV()
usleep(45000) // 45ms guard ensures browser/app reads NSPasteboard before next chunk overwrites it
}
private func postCleanCmdV() {
let src = CGEventSource(stateID: .privateState)
let vKeyCode: CGKeyCode = 0x09 // 'v'

4
src/TextProcessor.swift
View File

@ -261,9 +261,11 @@ public final class TextProcessor {
for p in punctuationList {
t = t.replacingOccurrences(of: " " + p, with: p)
}
for p in [".", "،", "؛", "؟", "!"] {
t = t.replacingOccurrences(of: "...", with: "…")
for p in [".", "،", "؛", "؟", "!", "…"] {
t = t.replacingOccurrences(of: p, with: p + " ")
}
t = t.replacingOccurrences(of: "…", with: "...")
let speechFixes: [(String, String)] = [
(" ماا ", " ما "),
(" اینپیگیری ", " این پیگیری "),

Write Preview
Loading…
Cancel
Save