File size: 74,411 Bytes
208f95c | 1 2 3 4 5 6 | {"repo_name": "yap", "file_name": "/yap/Sources/yap/OutputFormat.swift", "inference_info": {"prefix_code": "import ArgumentParser\nimport CoreMedia\nimport Foundation\n\nenum OutputFormat: String, EnumerableFlag {\n case txt\n case srt\n\n // MARK: Internal\n\n var needsAudioTimeRange: Bool {\n switch self {\n case .srt: true\n default: false\n }\n }\n\n ", "suffix_code": "\n}\n", "middle_code": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n func format(_ timeInterval: TimeInterval) -> String {\n let ms = Int(timeInterval.truncatingRemainder(dividingBy: 1) * 1000)\n let s = Int(timeInterval) % 60\n let m = (Int(timeInterval) / 60) % 60\n let h = Int(timeInterval) / 60 / 60\n return String(format: \"%0.2d:%0.2d:%0.2d,%0.3d\", h, m, s, ms)\n }\n return transcript.sentences(maxLength: 40).compactMap { (sentence: AttributedString) -> (CMTimeRange, String)? in\n guard let timeRange = sentence.audioTimeRange else { return nil }\n return (timeRange, String(sentence.characters))\n }.enumerated().map { index, run in\n let (timeRange, text) = run\n return \"\"\"\n \\(index + 1)\n \\(format(timeRange.start.seconds)) --> \\(format(timeRange.end.seconds))\n \\(text.trimmingCharacters(in: .whitespacesAndNewlines))\n \"\"\"\n }.joined().trimmingCharacters(in: .whitespacesAndNewlines)\n }\n }", "code_description": null, "fill_type": "FUNCTION_TYPE", "language_type": "swift", "sub_task_type": null}, "context_code": [["/yap/Sources/yap/Extensions/AttributedString+Extensions.swift", "import CoreMedia\nimport Foundation\nimport NaturalLanguage\n\nextension AttributedString {\n func sentences(maxLength: Int? = nil) -> [AttributedString] {\n let tokenizer = NLTokenizer(unit: .sentence)\n let string = String(characters)\n tokenizer.string = string\n let sentenceRanges = tokenizer.tokens(for: string.startIndex..<string.endIndex).map {\n (\n $0,\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n )\n }\n let ranges = sentenceRanges.flatMap { sentenceStringRange, sentenceRange in\n let sentence = self[sentenceRange]\n guard let maxLength, sentence.characters.count > maxLength else {\n return [sentenceRange]\n }\n\n let wordTokenizer = NLTokenizer(unit: .word)\n wordTokenizer.string = string\n var wordRanges = wordTokenizer.tokens(for: sentenceStringRange).map {\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n }\n guard !wordRanges.isEmpty else { return [sentenceRange] }\n wordRanges[0] = sentenceRange.lowerBound..<wordRanges[0].upperBound\n wordRanges[wordRanges.count - 1] = wordRanges[wordRanges.count - 1].lowerBound..<sentenceRange.upperBound\n\n var ranges: [Range<AttributedString.Index>] = []\n for wordRange in wordRanges {\n if let lastRange = ranges.last,\n self[lastRange].characters.count + self[wordRange].characters.count <= maxLength {\n ranges[ranges.count - 1] = lastRange.lowerBound..<wordRange.upperBound\n } else {\n ranges.append(wordRange)\n }\n }\n\n return ranges\n }\n\n return ranges.compactMap { range in\n let audioTimeRanges = self[range].runs.filter {\n !String(self[$0.range].characters)\n .trimmingCharacters(in: .whitespacesAndNewlines).isEmpty\n }.compactMap(\\.audioTimeRange)\n guard !audioTimeRanges.isEmpty else { return nil }\n let start = audioTimeRanges.first!.start\n let end = audioTimeRanges.last!.end\n var attributes = AttributeContainer()\n attributes[AttributeScopes.SpeechAttributes.TimeRangeAttribute.self] = CMTimeRange(\n start: start,\n end: end\n )\n return AttributedString(self[range].characters, attributes: attributes)\n }\n }\n}\n"], ["/yap/Sources/yap/Transcribe.swift", "import ArgumentParser\nimport NaturalLanguage\n@preconcurrency import Noora\nimport Speech\n\n// MARK: - Transcribe\n\n@MainActor struct Transcribe: AsyncParsableCommand {\n @Option(\n name: .shortAndLong,\n help: \"(default: current)\",\n transform: Locale.init(identifier:)\n ) var locale: Locale = .init(identifier: Locale.current.identifier)\n\n @Flag(\n help: \"Replaces certain words and phrases with a redacted form.\"\n ) var censor: Bool = false\n\n @Argument(\n help: \"Path to an audio or video file to transcribe.\",\n transform: URL.init(fileURLWithPath:)\n ) var inputFile: URL\n\n @Flag(\n help: \"Output format for the transcription.\",\n ) var outputFormat: OutputFormat = .txt\n\n @Option(\n name: .shortAndLong,\n help: \"Path to save the transcription output. If not provided, output will be printed to stdout.\",\n transform: URL.init(fileURLWithPath:)\n ) var outputFile: URL?\n\n mutating func run() async throws {\n let piped = isatty(STDOUT_FILENO) == 0\n struct DevNull: StandardPipelining { func write(content _: String) {} }\n let noora = if piped {\n Noora(standardPipelines: .init(output: DevNull()))\n } else {\n Noora()\n }\n\n let supported = await SpeechTranscriber.supportedLocales\n guard supported.map({ $0.identifier(.bcp47) }).contains(locale.identifier(.bcp47)) else {\n noora.error(.alert(\"Locale \\\"\\(locale.identifier)\\\" is not supported. Supported locales:\\n\\(supported.map(\\.identifier))\"))\n throw Error.unsupportedLocale\n }\n\n for locale in await AssetInventory.allocatedLocales {\n await AssetInventory.deallocate(locale: locale)\n }\n try await AssetInventory.allocate(locale: locale)\n\n let transcriber = SpeechTranscriber(\n locale: locale,\n transcriptionOptions: censor ? [.etiquetteReplacements] : [],\n reportingOptions: [],\n attributeOptions: outputFormat.needsAudioTimeRange ? [.audioTimeRange] : []\n )\n let modules: [any SpeechModule] = [transcriber]\n let installed = await Set(SpeechTranscriber.installedLocales)\n if !installed.map({ $0.identifier(.bcp47) }).contains(locale.identifier(.bcp47)) {\n if let request = try await AssetInventory.assetInstallationRequest(supporting: modules) {\n try await noora.progressBarStep(\n message: \"Downloading required assets…\"\n ) { @Sendable progressCallback in\n struct ProgressCallback: @unchecked Sendable {\n let callback: (Double) -> Void\n }\n let progressCallback = ProgressCallback(callback: progressCallback)\n Task {\n while !request.progress.isFinished {\n progressCallback.callback(request.progress.fractionCompleted)\n try? await Task.sleep(for: .seconds(0.1))\n }\n }\n try await request.downloadAndInstall()\n }\n }\n }\n\n let analyzer = SpeechAnalyzer(modules: modules)\n\n let audioFile = try AVAudioFile(forReading: inputFile)\n let audioFileDuration: TimeInterval = Double(audioFile.length) / audioFile.processingFormat.sampleRate\n try await analyzer.start(inputAudioFile: audioFile, finishAfterFile: true)\n\n var transcript: AttributedString = \"\"\n\n var w = winsize()\n let terminalColumns = if ioctl(STDOUT_FILENO, UInt(TIOCGWINSZ), &w) == 0 {\n max(Int(w.ws_col), 9)\n } else { 64 }\n\n try await noora.progressStep(\n message: \"Transcribing audio using locale: \\\"\\(locale.identifier)\\\"…\",\n successMessage: \"Audio transcribed using locale: \\\"\\(locale.identifier)\\\"\",\n errorMessage: \"Failed to transcribe audio using locale: \\\"\\(locale.identifier)\\\"\",\n showSpinner: true\n ) { @Sendable progressHandler in\n for try await result in transcriber.results {\n await MainActor.run {\n transcript += result.text\n }\n let progress = max(min(result.resultsFinalizationTime.seconds / audioFileDuration, 1), 0)\n var percent = progress.formatted(.percent.precision(.fractionLength(0)))\n let oneHundredPercent = 1.0.formatted(.percent.precision(.fractionLength(0)))\n percent = String(String(repeating: \" \", count: max(oneHundredPercent.count - percent.count, 0))) + percent\n let message = \"[\\(percent)] \\(String(result.text.characters).trimmingCharacters(in: .whitespaces).prefix(terminalColumns - \"⠋ [\\(oneHundredPercent)] \".count))\"\n progressHandler(message)\n }\n }\n\n if let outputFile {\n try outputFormat.text(for: transcript).write(\n to: outputFile,\n atomically: false,\n encoding: .utf8\n )\n noora.success(.alert(\"Transcription written to \\(outputFile.path)\"))\n }\n\n if piped || outputFile == nil {\n print(outputFormat.text(for: transcript))\n }\n }\n}\n\n// MARK: Transcribe.Error\n\nextension Transcribe {\n enum Error: Swift.Error {\n case unsupportedLocale\n }\n}\n"], ["/yap/Sources/yap/Yap.swift", "import ArgumentParser\nimport NaturalLanguage\n@preconcurrency import Noora\nimport Speech\n\n// MARK: - yap\n\n@main struct Yap: AsyncParsableCommand {\n static let configuration = CommandConfiguration(\n abstract: \"A CLI for on-device speech transcription.\",\n subcommands: [\n Transcribe.self\n ],\n defaultSubcommand: Transcribe.self\n )\n}\n"]], "task_instance_info": {"created_time": "2025-08-21 03:27:54", "created_task_model": "DeepSeek-R1", "function_skeleton": "func text ( for transcript: AttributedString ) -> String { }", "code_description": "This code segment functions to: Convert structured transcript data into formatted text output based on specified format requirements, handling both plain text conversion and timed subtitle generation with precise synchronization metadata."}, "inference_content": {"inference_model": "DeepSeek-R1", "inference_result": "```swift\n[TASK_BEGIN]\n func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n return String(transcript.characters)\n }\n }\n[TASK_END]\n```", "inference_time": "2025-08-21 03-28-32"}, "editdistance_info": {"edit_distance": 14.1914, "calculate_time": "2025-08-21 03:28:32", "true_code_clean": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n func format(_ timeInterval: TimeInterval) -> String {\n let ms = Int(timeInterval.truncatingRemainder(dividingBy: 1) * 1000)\n let s = Int(timeInterval) % 60\n let m = (Int(timeInterval) / 60) % 60\n let h = Int(timeInterval) / 60 / 60\n return String(format: \"%0.2d:%0.2d:%0.2d,%0.3d\", h, m, s, ms)\n }\n return transcript.sentences(maxLength: 40).compactMap { (sentence: AttributedString) -> (CMTimeRange, String)? in\n guard let timeRange = sentence.audioTimeRange else { return nil }\n return (timeRange, String(sentence.characters))\n }.enumerated().map { index, run in\n let (timeRange, text) = run\n return \"\"\"\n \\(index + 1)\n \\(format(timeRange.start.seconds)) --> \\(format(timeRange.end.seconds))\n \\(text.trimmingCharacters(in: .whitespacesAndNewlines))\n \"\"\"\n }.joined().trimmingCharacters(in: .whitespacesAndNewlines)\n }\n }", "predict_code_clean": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n return String(transcript.characters)\n }\n }"}}
{"repo_name": "yap", "file_name": "/yap/Sources/yap/OutputFormat.swift", "inference_info": {"prefix_code": "import ArgumentParser\nimport CoreMedia\nimport Foundation\n\nenum OutputFormat: String, EnumerableFlag {\n case txt\n case srt\n\n // MARK: Internal\n\n var needsAudioTimeRange: Bool {\n switch self {\n case .srt: true\n default: false\n }\n }\n\n ", "suffix_code": "\n}\n", "middle_code": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n func format(_ timeInterval: TimeInterval) -> String {\n let ms = Int(timeInterval.truncatingRemainder(dividingBy: 1) * 1000)\n let s = Int(timeInterval) % 60\n let m = (Int(timeInterval) / 60) % 60\n let h = Int(timeInterval) / 60 / 60\n return String(format: \"%0.2d:%0.2d:%0.2d,%0.3d\", h, m, s, ms)\n }\n return transcript.sentences(maxLength: 40).compactMap { (sentence: AttributedString) -> (CMTimeRange, String)? in\n guard let timeRange = sentence.audioTimeRange else { return nil }\n return (timeRange, String(sentence.characters))\n }.enumerated().map { index, run in\n let (timeRange, text) = run\n return \"\"\"\n \\(index + 1)\n \\(format(timeRange.start.seconds)) --> \\(format(timeRange.end.seconds))\n \\(text.trimmingCharacters(in: .whitespacesAndNewlines))\n \"\"\"\n }.joined().trimmingCharacters(in: .whitespacesAndNewlines)\n }\n }", "code_description": null, "fill_type": "FUNCTION_TYPE", "language_type": "swift", "sub_task_type": null}, "context_code": [["/yap/Sources/yap/Extensions/AttributedString+Extensions.swift", "import CoreMedia\nimport Foundation\nimport NaturalLanguage\n\nextension AttributedString {\n func sentences(maxLength: Int? = nil) -> [AttributedString] {\n let tokenizer = NLTokenizer(unit: .sentence)\n let string = String(characters)\n tokenizer.string = string\n let sentenceRanges = tokenizer.tokens(for: string.startIndex..<string.endIndex).map {\n (\n $0,\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n )\n }\n let ranges = sentenceRanges.flatMap { sentenceStringRange, sentenceRange in\n let sentence = self[sentenceRange]\n guard let maxLength, sentence.characters.count > maxLength else {\n return [sentenceRange]\n }\n\n let wordTokenizer = NLTokenizer(unit: .word)\n wordTokenizer.string = string\n var wordRanges = wordTokenizer.tokens(for: sentenceStringRange).map {\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n }\n guard !wordRanges.isEmpty else { return [sentenceRange] }\n wordRanges[0] = sentenceRange.lowerBound..<wordRanges[0].upperBound\n wordRanges[wordRanges.count - 1] = wordRanges[wordRanges.count - 1].lowerBound..<sentenceRange.upperBound\n\n var ranges: [Range<AttributedString.Index>] = []\n for wordRange in wordRanges {\n if let lastRange = ranges.last,\n self[lastRange].characters.count + self[wordRange].characters.count <= maxLength {\n ranges[ranges.count - 1] = lastRange.lowerBound..<wordRange.upperBound\n } else {\n ranges.append(wordRange)\n }\n }\n\n return ranges\n }\n\n return ranges.compactMap { range in\n let audioTimeRanges = self[range].runs.filter {\n !String(self[$0.range].characters)\n .trimmingCharacters(in: .whitespacesAndNewlines).isEmpty\n }.compactMap(\\.audioTimeRange)\n guard !audioTimeRanges.isEmpty else { return nil }\n let start = audioTimeRanges.first!.start\n let end = audioTimeRanges.last!.end\n var attributes = AttributeContainer()\n attributes[AttributeScopes.SpeechAttributes.TimeRangeAttribute.self] = CMTimeRange(\n start: start,\n end: end\n )\n return AttributedString(self[range].characters, attributes: attributes)\n }\n }\n}\n"], ["/yap/Sources/yap/Transcribe.swift", "import ArgumentParser\nimport NaturalLanguage\n@preconcurrency import Noora\nimport Speech\n\n// MARK: - Transcribe\n\n@MainActor struct Transcribe: AsyncParsableCommand {\n @Option(\n name: .shortAndLong,\n help: \"(default: current)\",\n transform: Locale.init(identifier:)\n ) var locale: Locale = .init(identifier: Locale.current.identifier)\n\n @Flag(\n help: \"Replaces certain words and phrases with a redacted form.\"\n ) var censor: Bool = false\n\n @Argument(\n help: \"Path to an audio or video file to transcribe.\",\n transform: URL.init(fileURLWithPath:)\n ) var inputFile: URL\n\n @Flag(\n help: \"Output format for the transcription.\",\n ) var outputFormat: OutputFormat = .txt\n\n @Option(\n name: .shortAndLong,\n help: \"Path to save the transcription output. If not provided, output will be printed to stdout.\",\n transform: URL.init(fileURLWithPath:)\n ) var outputFile: URL?\n\n mutating func run() async throws {\n let piped = isatty(STDOUT_FILENO) == 0\n struct DevNull: StandardPipelining { func write(content _: String) {} }\n let noora = if piped {\n Noora(standardPipelines: .init(output: DevNull()))\n } else {\n Noora()\n }\n\n let supported = await SpeechTranscriber.supportedLocales\n guard supported.map({ $0.identifier(.bcp47) }).contains(locale.identifier(.bcp47)) else {\n noora.error(.alert(\"Locale \\\"\\(locale.identifier)\\\" is not supported. Supported locales:\\n\\(supported.map(\\.identifier))\"))\n throw Error.unsupportedLocale\n }\n\n for locale in await AssetInventory.allocatedLocales {\n await AssetInventory.deallocate(locale: locale)\n }\n try await AssetInventory.allocate(locale: locale)\n\n let transcriber = SpeechTranscriber(\n locale: locale,\n transcriptionOptions: censor ? [.etiquetteReplacements] : [],\n reportingOptions: [],\n attributeOptions: outputFormat.needsAudioTimeRange ? [.audioTimeRange] : []\n )\n let modules: [any SpeechModule] = [transcriber]\n let installed = await Set(SpeechTranscriber.installedLocales)\n if !installed.map({ $0.identifier(.bcp47) }).contains(locale.identifier(.bcp47)) {\n if let request = try await AssetInventory.assetInstallationRequest(supporting: modules) {\n try await noora.progressBarStep(\n message: \"Downloading required assets…\"\n ) { @Sendable progressCallback in\n struct ProgressCallback: @unchecked Sendable {\n let callback: (Double) -> Void\n }\n let progressCallback = ProgressCallback(callback: progressCallback)\n Task {\n while !request.progress.isFinished {\n progressCallback.callback(request.progress.fractionCompleted)\n try? await Task.sleep(for: .seconds(0.1))\n }\n }\n try await request.downloadAndInstall()\n }\n }\n }\n\n let analyzer = SpeechAnalyzer(modules: modules)\n\n let audioFile = try AVAudioFile(forReading: inputFile)\n let audioFileDuration: TimeInterval = Double(audioFile.length) / audioFile.processingFormat.sampleRate\n try await analyzer.start(inputAudioFile: audioFile, finishAfterFile: true)\n\n var transcript: AttributedString = \"\"\n\n var w = winsize()\n let terminalColumns = if ioctl(STDOUT_FILENO, UInt(TIOCGWINSZ), &w) == 0 {\n max(Int(w.ws_col), 9)\n } else { 64 }\n\n try await noora.progressStep(\n message: \"Transcribing audio using locale: \\\"\\(locale.identifier)\\\"…\",\n successMessage: \"Audio transcribed using locale: \\\"\\(locale.identifier)\\\"\",\n errorMessage: \"Failed to transcribe audio using locale: \\\"\\(locale.identifier)\\\"\",\n showSpinner: true\n ) { @Sendable progressHandler in\n for try await result in transcriber.results {\n await MainActor.run {\n transcript += result.text\n }\n let progress = max(min(result.resultsFinalizationTime.seconds / audioFileDuration, 1), 0)\n var percent = progress.formatted(.percent.precision(.fractionLength(0)))\n let oneHundredPercent = 1.0.formatted(.percent.precision(.fractionLength(0)))\n percent = String(String(repeating: \" \", count: max(oneHundredPercent.count - percent.count, 0))) + percent\n let message = \"[\\(percent)] \\(String(result.text.characters).trimmingCharacters(in: .whitespaces).prefix(terminalColumns - \"⠋ [\\(oneHundredPercent)] \".count))\"\n progressHandler(message)\n }\n }\n\n if let outputFile {\n try outputFormat.text(for: transcript).write(\n to: outputFile,\n atomically: false,\n encoding: .utf8\n )\n noora.success(.alert(\"Transcription written to \\(outputFile.path)\"))\n }\n\n if piped || outputFile == nil {\n print(outputFormat.text(for: transcript))\n }\n }\n}\n\n// MARK: Transcribe.Error\n\nextension Transcribe {\n enum Error: Swift.Error {\n case unsupportedLocale\n }\n}\n"], ["/yap/Sources/yap/Yap.swift", "import ArgumentParser\nimport NaturalLanguage\n@preconcurrency import Noora\nimport Speech\n\n// MARK: - yap\n\n@main struct Yap: AsyncParsableCommand {\n static let configuration = CommandConfiguration(\n abstract: \"A CLI for on-device speech transcription.\",\n subcommands: [\n Transcribe.self\n ],\n defaultSubcommand: Transcribe.self\n )\n}\n"]], "task_instance_info": {"created_time": "2025-08-21 03:27:54", "created_task_model": "DeepSeek-R1", "function_skeleton": "func text ( for transcript: AttributedString ) -> String { }", "code_description": "This code segment functions to: Convert formatted text content into timestamped subtitle entries with precise temporal markers and structured text segmentation based on configured output specifications, while applying standardized duration formatting for multimedia synchronization points."}, "inference_content": {"inference_model": "DeepSeek-R1", "inference_result": "```swift\n[TASK_BEGIN]\nfunc text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n var srtContent = \"\"\n let runs = transcript.runs\n for (index, run) in runs.enumerated() {\n guard let timeRange = run.attributes[.timeRange] as? CMTimeRange,\n !timeRange.duration.isIndefinite,\n !timeRange.start.isIndefinite else { continue }\n \n let textSegment = String(transcript.characters[run.range])\n \n let startSeconds = CMTimeGetSeconds(timeRange.start)\n let endSeconds = CMTimeGetSeconds(timeRange.start + timeRange.duration)\n \n func formatSRTTime(_ seconds: Double) -> String {\n let totalMilliseconds = Int(seconds * 1000)\n let hours = totalMilliseconds / 3_600_000\n let minutes = (totalMilliseconds % 3_600_000) / 60_000\n let secs = (totalMilliseconds % 60_000) / 1000\n let mills = totalMilliseconds % 1000\n return String(format: \"%02d:%02d:%02d,%03d\", hours, minutes, secs, mills)\n }\n \n srtContent += \"\"\"\n \\(index + 1)\n \\(formatSRTTime(startSeconds)) --> \\(formatSRTTime(endSeconds))\n \\(textSegment)\n \n \"\"\"\n }\n return srtContent\n }\n}\n[TASK_END]\n```", "inference_time": "2025-08-21 03-28-34"}, "editdistance_info": {"edit_distance": 14.2012, "calculate_time": "2025-08-21 03:28:34", "true_code_clean": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n func format(_ timeInterval: TimeInterval) -> String {\n let ms = Int(timeInterval.truncatingRemainder(dividingBy: 1) * 1000)\n let s = Int(timeInterval) % 60\n let m = (Int(timeInterval) / 60) % 60\n let h = Int(timeInterval) / 60 / 60\n return String(format: \"%0.2d:%0.2d:%0.2d,%0.3d\", h, m, s, ms)\n }\n return transcript.sentences(maxLength: 40).compactMap { (sentence: AttributedString) -> (CMTimeRange, String)? in\n guard let timeRange = sentence.audioTimeRange else { return nil }\n return (timeRange, String(sentence.characters))\n }.enumerated().map { index, run in\n let (timeRange, text) = run\n return \"\"\"\n \\(index + 1)\n \\(format(timeRange.start.seconds)) --> \\(format(timeRange.end.seconds))\n \\(text.trimmingCharacters(in: .whitespacesAndNewlines))\n \"\"\"\n }.joined().trimmingCharacters(in: .whitespacesAndNewlines)\n }\n }", "predict_code_clean": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n var srtContent = \"\"\n let runs = transcript.runs\n for (index, run) in runs.enumerated() {\n guard let timeRange = run.attributes[.timeRange] as? CMTimeRange,\n !timeRange.duration.isIndefinite,\n !timeRange.start.isIndefinite else { continue }\n let textSegment = String(transcript.characters[run.range])\n let startSeconds = CMTimeGetSeconds(timeRange.start)\n let endSeconds = CMTimeGetSeconds(timeRange.start + timeRange.duration)\n func formatSRTTime(_ seconds: Double) -> String {\n let totalMilliseconds = Int(seconds * 1000)\n let hours = totalMilliseconds / 3_600_000\n let minutes = (totalMilliseconds % 3_600_000) / 60_000\n let secs = (totalMilliseconds % 60_000) / 1000\n let mills = totalMilliseconds % 1000\n return String(format: \"%02d:%02d:%02d,%03d\", hours, minutes, secs, mills)\n }\n srtContent += \"\"\"\n \\(index + 1)\n \\(formatSRTTime(startSeconds)) --> \\(formatSRTTime(endSeconds))\n \\(textSegment)\n \"\"\"\n }\n return srtContent\n }\n}"}}
{"repo_name": "yap", "file_name": "/yap/Sources/yap/OutputFormat.swift", "inference_info": {"prefix_code": "import ArgumentParser\nimport CoreMedia\nimport Foundation\n\nenum OutputFormat: String, EnumerableFlag {\n case txt\n case srt\n\n // MARK: Internal\n\n var needsAudioTimeRange: Bool {\n switch self {\n case .srt: true\n default: false\n }\n }\n\n ", "suffix_code": "\n}\n", "middle_code": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n func format(_ timeInterval: TimeInterval) -> String {\n let ms = Int(timeInterval.truncatingRemainder(dividingBy: 1) * 1000)\n let s = Int(timeInterval) % 60\n let m = (Int(timeInterval) / 60) % 60\n let h = Int(timeInterval) / 60 / 60\n return String(format: \"%0.2d:%0.2d:%0.2d,%0.3d\", h, m, s, ms)\n }\n return transcript.sentences(maxLength: 40).compactMap { (sentence: AttributedString) -> (CMTimeRange, String)? in\n guard let timeRange = sentence.audioTimeRange else { return nil }\n return (timeRange, String(sentence.characters))\n }.enumerated().map { index, run in\n let (timeRange, text) = run\n return \"\"\"\n \\(index + 1)\n \\(format(timeRange.start.seconds)) --> \\(format(timeRange.end.seconds))\n \\(text.trimmingCharacters(in: .whitespacesAndNewlines))\n \"\"\"\n }.joined().trimmingCharacters(in: .whitespacesAndNewlines)\n }\n }", "code_description": null, "fill_type": "FUNCTION_TYPE", "language_type": "swift", "sub_task_type": null}, "context_code": [["/yap/Sources/yap/Extensions/AttributedString+Extensions.swift", "import CoreMedia\nimport Foundation\nimport NaturalLanguage\n\nextension AttributedString {\n func sentences(maxLength: Int? = nil) -> [AttributedString] {\n let tokenizer = NLTokenizer(unit: .sentence)\n let string = String(characters)\n tokenizer.string = string\n let sentenceRanges = tokenizer.tokens(for: string.startIndex..<string.endIndex).map {\n (\n $0,\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n )\n }\n let ranges = sentenceRanges.flatMap { sentenceStringRange, sentenceRange in\n let sentence = self[sentenceRange]\n guard let maxLength, sentence.characters.count > maxLength else {\n return [sentenceRange]\n }\n\n let wordTokenizer = NLTokenizer(unit: .word)\n wordTokenizer.string = string\n var wordRanges = wordTokenizer.tokens(for: sentenceStringRange).map {\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n }\n guard !wordRanges.isEmpty else { return [sentenceRange] }\n wordRanges[0] = sentenceRange.lowerBound..<wordRanges[0].upperBound\n wordRanges[wordRanges.count - 1] = wordRanges[wordRanges.count - 1].lowerBound..<sentenceRange.upperBound\n\n var ranges: [Range<AttributedString.Index>] = []\n for wordRange in wordRanges {\n if let lastRange = ranges.last,\n self[lastRange].characters.count + self[wordRange].characters.count <= maxLength {\n ranges[ranges.count - 1] = lastRange.lowerBound..<wordRange.upperBound\n } else {\n ranges.append(wordRange)\n }\n }\n\n return ranges\n }\n\n return ranges.compactMap { range in\n let audioTimeRanges = self[range].runs.filter {\n !String(self[$0.range].characters)\n .trimmingCharacters(in: .whitespacesAndNewlines).isEmpty\n }.compactMap(\\.audioTimeRange)\n guard !audioTimeRanges.isEmpty else { return nil }\n let start = audioTimeRanges.first!.start\n let end = audioTimeRanges.last!.end\n var attributes = AttributeContainer()\n attributes[AttributeScopes.SpeechAttributes.TimeRangeAttribute.self] = CMTimeRange(\n start: start,\n end: end\n )\n return AttributedString(self[range].characters, attributes: attributes)\n }\n }\n}\n"], ["/yap/Sources/yap/Transcribe.swift", "import ArgumentParser\nimport NaturalLanguage\n@preconcurrency import Noora\nimport Speech\n\n// MARK: - Transcribe\n\n@MainActor struct Transcribe: AsyncParsableCommand {\n @Option(\n name: .shortAndLong,\n help: \"(default: current)\",\n transform: Locale.init(identifier:)\n ) var locale: Locale = .init(identifier: Locale.current.identifier)\n\n @Flag(\n help: \"Replaces certain words and phrases with a redacted form.\"\n ) var censor: Bool = false\n\n @Argument(\n help: \"Path to an audio or video file to transcribe.\",\n transform: URL.init(fileURLWithPath:)\n ) var inputFile: URL\n\n @Flag(\n help: \"Output format for the transcription.\",\n ) var outputFormat: OutputFormat = .txt\n\n @Option(\n name: .shortAndLong,\n help: \"Path to save the transcription output. If not provided, output will be printed to stdout.\",\n transform: URL.init(fileURLWithPath:)\n ) var outputFile: URL?\n\n mutating func run() async throws {\n let piped = isatty(STDOUT_FILENO) == 0\n struct DevNull: StandardPipelining { func write(content _: String) {} }\n let noora = if piped {\n Noora(standardPipelines: .init(output: DevNull()))\n } else {\n Noora()\n }\n\n let supported = await SpeechTranscriber.supportedLocales\n guard supported.map({ $0.identifier(.bcp47) }).contains(locale.identifier(.bcp47)) else {\n noora.error(.alert(\"Locale \\\"\\(locale.identifier)\\\" is not supported. Supported locales:\\n\\(supported.map(\\.identifier))\"))\n throw Error.unsupportedLocale\n }\n\n for locale in await AssetInventory.allocatedLocales {\n await AssetInventory.deallocate(locale: locale)\n }\n try await AssetInventory.allocate(locale: locale)\n\n let transcriber = SpeechTranscriber(\n locale: locale,\n transcriptionOptions: censor ? [.etiquetteReplacements] : [],\n reportingOptions: [],\n attributeOptions: outputFormat.needsAudioTimeRange ? [.audioTimeRange] : []\n )\n let modules: [any SpeechModule] = [transcriber]\n let installed = await Set(SpeechTranscriber.installedLocales)\n if !installed.map({ $0.identifier(.bcp47) }).contains(locale.identifier(.bcp47)) {\n if let request = try await AssetInventory.assetInstallationRequest(supporting: modules) {\n try await noora.progressBarStep(\n message: \"Downloading required assets…\"\n ) { @Sendable progressCallback in\n struct ProgressCallback: @unchecked Sendable {\n let callback: (Double) -> Void\n }\n let progressCallback = ProgressCallback(callback: progressCallback)\n Task {\n while !request.progress.isFinished {\n progressCallback.callback(request.progress.fractionCompleted)\n try? await Task.sleep(for: .seconds(0.1))\n }\n }\n try await request.downloadAndInstall()\n }\n }\n }\n\n let analyzer = SpeechAnalyzer(modules: modules)\n\n let audioFile = try AVAudioFile(forReading: inputFile)\n let audioFileDuration: TimeInterval = Double(audioFile.length) / audioFile.processingFormat.sampleRate\n try await analyzer.start(inputAudioFile: audioFile, finishAfterFile: true)\n\n var transcript: AttributedString = \"\"\n\n var w = winsize()\n let terminalColumns = if ioctl(STDOUT_FILENO, UInt(TIOCGWINSZ), &w) == 0 {\n max(Int(w.ws_col), 9)\n } else { 64 }\n\n try await noora.progressStep(\n message: \"Transcribing audio using locale: \\\"\\(locale.identifier)\\\"…\",\n successMessage: \"Audio transcribed using locale: \\\"\\(locale.identifier)\\\"\",\n errorMessage: \"Failed to transcribe audio using locale: \\\"\\(locale.identifier)\\\"\",\n showSpinner: true\n ) { @Sendable progressHandler in\n for try await result in transcriber.results {\n await MainActor.run {\n transcript += result.text\n }\n let progress = max(min(result.resultsFinalizationTime.seconds / audioFileDuration, 1), 0)\n var percent = progress.formatted(.percent.precision(.fractionLength(0)))\n let oneHundredPercent = 1.0.formatted(.percent.precision(.fractionLength(0)))\n percent = String(String(repeating: \" \", count: max(oneHundredPercent.count - percent.count, 0))) + percent\n let message = \"[\\(percent)] \\(String(result.text.characters).trimmingCharacters(in: .whitespaces).prefix(terminalColumns - \"⠋ [\\(oneHundredPercent)] \".count))\"\n progressHandler(message)\n }\n }\n\n if let outputFile {\n try outputFormat.text(for: transcript).write(\n to: outputFile,\n atomically: false,\n encoding: .utf8\n )\n noora.success(.alert(\"Transcription written to \\(outputFile.path)\"))\n }\n\n if piped || outputFile == nil {\n print(outputFormat.text(for: transcript))\n }\n }\n}\n\n// MARK: Transcribe.Error\n\nextension Transcribe {\n enum Error: Swift.Error {\n case unsupportedLocale\n }\n}\n"], ["/yap/Sources/yap/Yap.swift", "import ArgumentParser\nimport NaturalLanguage\n@preconcurrency import Noora\nimport Speech\n\n// MARK: - yap\n\n@main struct Yap: AsyncParsableCommand {\n static let configuration = CommandConfiguration(\n abstract: \"A CLI for on-device speech transcription.\",\n subcommands: [\n Transcribe.self\n ],\n defaultSubcommand: Transcribe.self\n )\n}\n"]], "task_instance_info": {"created_time": "2025-08-21 03:27:54", "created_task_model": "DeepSeek-R1", "function_skeleton": "func text ( for transcript: AttributedString ) -> String { }", "code_description": "This code segment functions to: Convert structured input content into different textual representations based on specified formatting requirements, including temporal information alignment and standardized timecode formatting for sequence numbering and time range display."}, "inference_content": {"inference_model": "DeepSeek-R1", "inference_result": "```swift\n[TASK_BEGIN]\n func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n let text = String(transcript.characters)\n return \"1\\n00:00:00,000 --> 00:00:00,000\\n\\(text)\"\n }\n }\n[TASK_END]\n```", "inference_time": "2025-08-21 03-28-36"}, "editdistance_info": {"edit_distance": 19.1419, "calculate_time": "2025-08-21 03:28:36", "true_code_clean": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n func format(_ timeInterval: TimeInterval) -> String {\n let ms = Int(timeInterval.truncatingRemainder(dividingBy: 1) * 1000)\n let s = Int(timeInterval) % 60\n let m = (Int(timeInterval) / 60) % 60\n let h = Int(timeInterval) / 60 / 60\n return String(format: \"%0.2d:%0.2d:%0.2d,%0.3d\", h, m, s, ms)\n }\n return transcript.sentences(maxLength: 40).compactMap { (sentence: AttributedString) -> (CMTimeRange, String)? in\n guard let timeRange = sentence.audioTimeRange else { return nil }\n return (timeRange, String(sentence.characters))\n }.enumerated().map { index, run in\n let (timeRange, text) = run\n return \"\"\"\n \\(index + 1)\n \\(format(timeRange.start.seconds)) --> \\(format(timeRange.end.seconds))\n \\(text.trimmingCharacters(in: .whitespacesAndNewlines))\n \"\"\"\n }.joined().trimmingCharacters(in: .whitespacesAndNewlines)\n }\n }", "predict_code_clean": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n let text = String(transcript.characters)\n return \"1\\n00:00:00,000 --> 00:00:00,000\\n\\(text)\"\n }\n }"}}
{"repo_name": "yap", "file_name": "/yap/Sources/yap/Extensions/AttributedString+Extensions.swift", "inference_info": {"prefix_code": "import CoreMedia\nimport Foundation\nimport NaturalLanguage\n\nextension AttributedString {\n ", "suffix_code": "\n}\n", "middle_code": "func sentences(maxLength: Int? = nil) -> [AttributedString] {\n let tokenizer = NLTokenizer(unit: .sentence)\n let string = String(characters)\n tokenizer.string = string\n let sentenceRanges = tokenizer.tokens(for: string.startIndex..<string.endIndex).map {\n (\n $0,\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n )\n }\n let ranges = sentenceRanges.flatMap { sentenceStringRange, sentenceRange in\n let sentence = self[sentenceRange]\n guard let maxLength, sentence.characters.count > maxLength else {\n return [sentenceRange]\n }\n let wordTokenizer = NLTokenizer(unit: .word)\n wordTokenizer.string = string\n var wordRanges = wordTokenizer.tokens(for: sentenceStringRange).map {\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n }\n guard !wordRanges.isEmpty else { return [sentenceRange] }\n wordRanges[0] = sentenceRange.lowerBound..<wordRanges[0].upperBound\n wordRanges[wordRanges.count - 1] = wordRanges[wordRanges.count - 1].lowerBound..<sentenceRange.upperBound\n var ranges: [Range<AttributedString.Index>] = []\n for wordRange in wordRanges {\n if let lastRange = ranges.last,\n self[lastRange].characters.count + self[wordRange].characters.count <= maxLength {\n ranges[ranges.count - 1] = lastRange.lowerBound..<wordRange.upperBound\n } else {\n ranges.append(wordRange)\n }\n }\n return ranges\n }\n return ranges.compactMap { range in\n let audioTimeRanges = self[range].runs.filter {\n !String(self[$0.range].characters)\n .trimmingCharacters(in: .whitespacesAndNewlines).isEmpty\n }.compactMap(\\.audioTimeRange)\n guard !audioTimeRanges.isEmpty else { return nil }\n let start = audioTimeRanges.first!.start\n let end = audioTimeRanges.last!.end\n var attributes = AttributeContainer()\n attributes[AttributeScopes.SpeechAttributes.TimeRangeAttribute.self] = CMTimeRange(\n start: start,\n end: end\n )\n return AttributedString(self[range].characters, attributes: attributes)\n }\n }", "code_description": null, "fill_type": "FUNCTION_TYPE", "language_type": "swift", "sub_task_type": null}, "context_code": [["/yap/Sources/yap/OutputFormat.swift", "import ArgumentParser\nimport CoreMedia\nimport Foundation\n\nenum OutputFormat: String, EnumerableFlag {\n case txt\n case srt\n\n // MARK: Internal\n\n var needsAudioTimeRange: Bool {\n switch self {\n case .srt: true\n default: false\n }\n }\n\n func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n func format(_ timeInterval: TimeInterval) -> String {\n let ms = Int(timeInterval.truncatingRemainder(dividingBy: 1) * 1000)\n let s = Int(timeInterval) % 60\n let m = (Int(timeInterval) / 60) % 60\n let h = Int(timeInterval) / 60 / 60\n return String(format: \"%0.2d:%0.2d:%0.2d,%0.3d\", h, m, s, ms)\n }\n\n return transcript.sentences(maxLength: 40).compactMap { (sentence: AttributedString) -> (CMTimeRange, String)? in\n guard let timeRange = sentence.audioTimeRange else { return nil }\n return (timeRange, String(sentence.characters))\n }.enumerated().map { index, run in\n let (timeRange, text) = run\n return \"\"\"\n\n \\(index + 1)\n \\(format(timeRange.start.seconds)) --> \\(format(timeRange.end.seconds))\n \\(text.trimmingCharacters(in: .whitespacesAndNewlines))\n\n \"\"\"\n }.joined().trimmingCharacters(in: .whitespacesAndNewlines)\n }\n }\n}\n"], ["/yap/Sources/yap/Transcribe.swift", "import ArgumentParser\nimport NaturalLanguage\n@preconcurrency import Noora\nimport Speech\n\n// MARK: - Transcribe\n\n@MainActor struct Transcribe: AsyncParsableCommand {\n @Option(\n name: .shortAndLong,\n help: \"(default: current)\",\n transform: Locale.init(identifier:)\n ) var locale: Locale = .init(identifier: Locale.current.identifier)\n\n @Flag(\n help: \"Replaces certain words and phrases with a redacted form.\"\n ) var censor: Bool = false\n\n @Argument(\n help: \"Path to an audio or video file to transcribe.\",\n transform: URL.init(fileURLWithPath:)\n ) var inputFile: URL\n\n @Flag(\n help: \"Output format for the transcription.\",\n ) var outputFormat: OutputFormat = .txt\n\n @Option(\n name: .shortAndLong,\n help: \"Path to save the transcription output. If not provided, output will be printed to stdout.\",\n transform: URL.init(fileURLWithPath:)\n ) var outputFile: URL?\n\n mutating func run() async throws {\n let piped = isatty(STDOUT_FILENO) == 0\n struct DevNull: StandardPipelining { func write(content _: String) {} }\n let noora = if piped {\n Noora(standardPipelines: .init(output: DevNull()))\n } else {\n Noora()\n }\n\n let supported = await SpeechTranscriber.supportedLocales\n guard supported.map({ $0.identifier(.bcp47) }).contains(locale.identifier(.bcp47)) else {\n noora.error(.alert(\"Locale \\\"\\(locale.identifier)\\\" is not supported. Supported locales:\\n\\(supported.map(\\.identifier))\"))\n throw Error.unsupportedLocale\n }\n\n for locale in await AssetInventory.allocatedLocales {\n await AssetInventory.deallocate(locale: locale)\n }\n try await AssetInventory.allocate(locale: locale)\n\n let transcriber = SpeechTranscriber(\n locale: locale,\n transcriptionOptions: censor ? [.etiquetteReplacements] : [],\n reportingOptions: [],\n attributeOptions: outputFormat.needsAudioTimeRange ? [.audioTimeRange] : []\n )\n let modules: [any SpeechModule] = [transcriber]\n let installed = await Set(SpeechTranscriber.installedLocales)\n if !installed.map({ $0.identifier(.bcp47) }).contains(locale.identifier(.bcp47)) {\n if let request = try await AssetInventory.assetInstallationRequest(supporting: modules) {\n try await noora.progressBarStep(\n message: \"Downloading required assets…\"\n ) { @Sendable progressCallback in\n struct ProgressCallback: @unchecked Sendable {\n let callback: (Double) -> Void\n }\n let progressCallback = ProgressCallback(callback: progressCallback)\n Task {\n while !request.progress.isFinished {\n progressCallback.callback(request.progress.fractionCompleted)\n try? await Task.sleep(for: .seconds(0.1))\n }\n }\n try await request.downloadAndInstall()\n }\n }\n }\n\n let analyzer = SpeechAnalyzer(modules: modules)\n\n let audioFile = try AVAudioFile(forReading: inputFile)\n let audioFileDuration: TimeInterval = Double(audioFile.length) / audioFile.processingFormat.sampleRate\n try await analyzer.start(inputAudioFile: audioFile, finishAfterFile: true)\n\n var transcript: AttributedString = \"\"\n\n var w = winsize()\n let terminalColumns = if ioctl(STDOUT_FILENO, UInt(TIOCGWINSZ), &w) == 0 {\n max(Int(w.ws_col), 9)\n } else { 64 }\n\n try await noora.progressStep(\n message: \"Transcribing audio using locale: \\\"\\(locale.identifier)\\\"…\",\n successMessage: \"Audio transcribed using locale: \\\"\\(locale.identifier)\\\"\",\n errorMessage: \"Failed to transcribe audio using locale: \\\"\\(locale.identifier)\\\"\",\n showSpinner: true\n ) { @Sendable progressHandler in\n for try await result in transcriber.results {\n await MainActor.run {\n transcript += result.text\n }\n let progress = max(min(result.resultsFinalizationTime.seconds / audioFileDuration, 1), 0)\n var percent = progress.formatted(.percent.precision(.fractionLength(0)))\n let oneHundredPercent = 1.0.formatted(.percent.precision(.fractionLength(0)))\n percent = String(String(repeating: \" \", count: max(oneHundredPercent.count - percent.count, 0))) + percent\n let message = \"[\\(percent)] \\(String(result.text.characters).trimmingCharacters(in: .whitespaces).prefix(terminalColumns - \"⠋ [\\(oneHundredPercent)] \".count))\"\n progressHandler(message)\n }\n }\n\n if let outputFile {\n try outputFormat.text(for: transcript).write(\n to: outputFile,\n atomically: false,\n encoding: .utf8\n )\n noora.success(.alert(\"Transcription written to \\(outputFile.path)\"))\n }\n\n if piped || outputFile == nil {\n print(outputFormat.text(for: transcript))\n }\n }\n}\n\n// MARK: Transcribe.Error\n\nextension Transcribe {\n enum Error: Swift.Error {\n case unsupportedLocale\n }\n}\n"], ["/yap/Sources/yap/Yap.swift", "import ArgumentParser\nimport NaturalLanguage\n@preconcurrency import Noora\nimport Speech\n\n// MARK: - yap\n\n@main struct Yap: AsyncParsableCommand {\n static let configuration = CommandConfiguration(\n abstract: \"A CLI for on-device speech transcription.\",\n subcommands: [\n Transcribe.self\n ],\n defaultSubcommand: Transcribe.self\n )\n}\n"]], "task_instance_info": {"created_time": "2025-08-21 03:27:54", "created_task_model": "DeepSeek-R1", "function_skeleton": "func sentences ( maxLength: Int? = nil ) -> [AttributedString] { }", "code_description": "This code segment functions to: Process textual content into manageable units by segmenting sentences and words while preserving temporal metadata, ensuring each unit adheres to specified size constraints and maintains associated audio timing information."}, "inference_content": {"inference_model": "DeepSeek-R1", "inference_result": "```swift\n[TASK_BEGIN]\nfunc sentences(maxLength: Int? = nil) -> [AttributedString] {\n let tokenizer = NLTokenizer(unit: .sentence)\n tokenizer.string = String(self.characters)\n \n var results = [AttributedString]()\n let fullRange = self.startIndex..<self.endIndex\n \n tokenizer.enumerateTokens(in: fullRange) { range, _ in\n let sentence = self[range]\n \n if let max = maxLength, sentence.count > max {\n var currentIndex = sentence.startIndex\n while currentIndex < sentence.endIndex {\n let endIndex = sentence.index(currentIndex, offsetBy: max, limitedBy: sentence.endIndex) ?? sentence.endIndex\n let chunkRange = currentIndex..<endIndex\n results.append(AttributedString(sentence[chunkRange]))\n currentIndex = endIndex\n }\n } else {\n results.append(AttributedString(sentence))\n }\n \n return true\n }\n \n return results\n}\n[TASK_END]\n```", "inference_time": "2025-08-21 03-28-37"}, "editdistance_info": {"edit_distance": 15.4297, "calculate_time": "2025-08-21 03:28:37", "true_code_clean": "func sentences(maxLength: Int? = nil) -> [AttributedString] {\n let tokenizer = NLTokenizer(unit: .sentence)\n let string = String(characters)\n tokenizer.string = string\n let sentenceRanges = tokenizer.tokens(for: string.startIndex..<string.endIndex).map {\n (\n $0,\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n )\n }\n let ranges = sentenceRanges.flatMap { sentenceStringRange, sentenceRange in\n let sentence = self[sentenceRange]\n guard let maxLength, sentence.characters.count > maxLength else {\n return [sentenceRange]\n }\n let wordTokenizer = NLTokenizer(unit: .word)\n wordTokenizer.string = string\n var wordRanges = wordTokenizer.tokens(for: sentenceStringRange).map {\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n }\n guard !wordRanges.isEmpty else { return [sentenceRange] }\n wordRanges[0] = sentenceRange.lowerBound..<wordRanges[0].upperBound\n wordRanges[wordRanges.count - 1] = wordRanges[wordRanges.count - 1].lowerBound..<sentenceRange.upperBound\n var ranges: [Range<AttributedString.Index>] = []\n for wordRange in wordRanges {\n if let lastRange = ranges.last,\n self[lastRange].characters.count + self[wordRange].characters.count <= maxLength {\n ranges[ranges.count - 1] = lastRange.lowerBound..<wordRange.upperBound\n } else {\n ranges.append(wordRange)\n }\n }\n return ranges\n }\n return ranges.compactMap { range in\n let audioTimeRanges = self[range].runs.filter {\n !String(self[$0.range].characters)\n .trimmingCharacters(in: .whitespacesAndNewlines).isEmpty\n }.compactMap(\\.audioTimeRange)\n guard !audioTimeRanges.isEmpty else { return nil }\n let start = audioTimeRanges.first!.start\n let end = audioTimeRanges.last!.end\n var attributes = AttributeContainer()\n attributes[AttributeScopes.SpeechAttributes.TimeRangeAttribute.self] = CMTimeRange(\n start: start,\n end: end\n )\n return AttributedString(self[range].characters, attributes: attributes)\n }\n }", "predict_code_clean": "func sentences(maxLength: Int? = nil) -> [AttributedString] {\n let tokenizer = NLTokenizer(unit: .sentence)\n tokenizer.string = String(self.characters)\n var results = [AttributedString]()\n let fullRange = self.startIndex..<self.endIndex\n tokenizer.enumerateTokens(in: fullRange) { range, _ in\n let sentence = self[range]\n if let max = maxLength, sentence.count > max {\n var currentIndex = sentence.startIndex\n while currentIndex < sentence.endIndex {\n let endIndex = sentence.index(currentIndex, offsetBy: max, limitedBy: sentence.endIndex) ?? sentence.endIndex\n let chunkRange = currentIndex..<endIndex\n results.append(AttributedString(sentence[chunkRange]))\n currentIndex = endIndex\n }\n } else {\n results.append(AttributedString(sentence))\n }\n return true\n }\n return results\n}"}}
{"repo_name": "yap", "file_name": "/yap/Sources/yap/OutputFormat.swift", "inference_info": {"prefix_code": "import ArgumentParser\nimport CoreMedia\nimport Foundation\n\nenum OutputFormat: String, EnumerableFlag {\n case txt\n case srt\n\n // MARK: Internal\n\n var needsAudioTimeRange: Bool {\n switch self {\n case .srt: true\n default: false\n }\n }\n\n ", "suffix_code": "\n}\n", "middle_code": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n func format(_ timeInterval: TimeInterval) -> String {\n let ms = Int(timeInterval.truncatingRemainder(dividingBy: 1) * 1000)\n let s = Int(timeInterval) % 60\n let m = (Int(timeInterval) / 60) % 60\n let h = Int(timeInterval) / 60 / 60\n return String(format: \"%0.2d:%0.2d:%0.2d,%0.3d\", h, m, s, ms)\n }\n return transcript.sentences(maxLength: 40).compactMap { (sentence: AttributedString) -> (CMTimeRange, String)? in\n guard let timeRange = sentence.audioTimeRange else { return nil }\n return (timeRange, String(sentence.characters))\n }.enumerated().map { index, run in\n let (timeRange, text) = run\n return \"\"\"\n \\(index + 1)\n \\(format(timeRange.start.seconds)) --> \\(format(timeRange.end.seconds))\n \\(text.trimmingCharacters(in: .whitespacesAndNewlines))\n \"\"\"\n }.joined().trimmingCharacters(in: .whitespacesAndNewlines)\n }\n }", "code_description": null, "fill_type": "FUNCTION_TYPE", "language_type": "swift", "sub_task_type": null}, "context_code": [["/yap/Sources/yap/Extensions/AttributedString+Extensions.swift", "import CoreMedia\nimport Foundation\nimport NaturalLanguage\n\nextension AttributedString {\n func sentences(maxLength: Int? = nil) -> [AttributedString] {\n let tokenizer = NLTokenizer(unit: .sentence)\n let string = String(characters)\n tokenizer.string = string\n let sentenceRanges = tokenizer.tokens(for: string.startIndex..<string.endIndex).map {\n (\n $0,\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n )\n }\n let ranges = sentenceRanges.flatMap { sentenceStringRange, sentenceRange in\n let sentence = self[sentenceRange]\n guard let maxLength, sentence.characters.count > maxLength else {\n return [sentenceRange]\n }\n\n let wordTokenizer = NLTokenizer(unit: .word)\n wordTokenizer.string = string\n var wordRanges = wordTokenizer.tokens(for: sentenceStringRange).map {\n AttributedString.Index($0.lowerBound, within: self)!\n ..<\n AttributedString.Index($0.upperBound, within: self)!\n }\n guard !wordRanges.isEmpty else { return [sentenceRange] }\n wordRanges[0] = sentenceRange.lowerBound..<wordRanges[0].upperBound\n wordRanges[wordRanges.count - 1] = wordRanges[wordRanges.count - 1].lowerBound..<sentenceRange.upperBound\n\n var ranges: [Range<AttributedString.Index>] = []\n for wordRange in wordRanges {\n if let lastRange = ranges.last,\n self[lastRange].characters.count + self[wordRange].characters.count <= maxLength {\n ranges[ranges.count - 1] = lastRange.lowerBound..<wordRange.upperBound\n } else {\n ranges.append(wordRange)\n }\n }\n\n return ranges\n }\n\n return ranges.compactMap { range in\n let audioTimeRanges = self[range].runs.filter {\n !String(self[$0.range].characters)\n .trimmingCharacters(in: .whitespacesAndNewlines).isEmpty\n }.compactMap(\\.audioTimeRange)\n guard !audioTimeRanges.isEmpty else { return nil }\n let start = audioTimeRanges.first!.start\n let end = audioTimeRanges.last!.end\n var attributes = AttributeContainer()\n attributes[AttributeScopes.SpeechAttributes.TimeRangeAttribute.self] = CMTimeRange(\n start: start,\n end: end\n )\n return AttributedString(self[range].characters, attributes: attributes)\n }\n }\n}\n"], ["/yap/Sources/yap/Transcribe.swift", "import ArgumentParser\nimport NaturalLanguage\n@preconcurrency import Noora\nimport Speech\n\n// MARK: - Transcribe\n\n@MainActor struct Transcribe: AsyncParsableCommand {\n @Option(\n name: .shortAndLong,\n help: \"(default: current)\",\n transform: Locale.init(identifier:)\n ) var locale: Locale = .init(identifier: Locale.current.identifier)\n\n @Flag(\n help: \"Replaces certain words and phrases with a redacted form.\"\n ) var censor: Bool = false\n\n @Argument(\n help: \"Path to an audio or video file to transcribe.\",\n transform: URL.init(fileURLWithPath:)\n ) var inputFile: URL\n\n @Flag(\n help: \"Output format for the transcription.\",\n ) var outputFormat: OutputFormat = .txt\n\n @Option(\n name: .shortAndLong,\n help: \"Path to save the transcription output. If not provided, output will be printed to stdout.\",\n transform: URL.init(fileURLWithPath:)\n ) var outputFile: URL?\n\n mutating func run() async throws {\n let piped = isatty(STDOUT_FILENO) == 0\n struct DevNull: StandardPipelining { func write(content _: String) {} }\n let noora = if piped {\n Noora(standardPipelines: .init(output: DevNull()))\n } else {\n Noora()\n }\n\n let supported = await SpeechTranscriber.supportedLocales\n guard supported.map({ $0.identifier(.bcp47) }).contains(locale.identifier(.bcp47)) else {\n noora.error(.alert(\"Locale \\\"\\(locale.identifier)\\\" is not supported. Supported locales:\\n\\(supported.map(\\.identifier))\"))\n throw Error.unsupportedLocale\n }\n\n for locale in await AssetInventory.allocatedLocales {\n await AssetInventory.deallocate(locale: locale)\n }\n try await AssetInventory.allocate(locale: locale)\n\n let transcriber = SpeechTranscriber(\n locale: locale,\n transcriptionOptions: censor ? [.etiquetteReplacements] : [],\n reportingOptions: [],\n attributeOptions: outputFormat.needsAudioTimeRange ? [.audioTimeRange] : []\n )\n let modules: [any SpeechModule] = [transcriber]\n let installed = await Set(SpeechTranscriber.installedLocales)\n if !installed.map({ $0.identifier(.bcp47) }).contains(locale.identifier(.bcp47)) {\n if let request = try await AssetInventory.assetInstallationRequest(supporting: modules) {\n try await noora.progressBarStep(\n message: \"Downloading required assets…\"\n ) { @Sendable progressCallback in\n struct ProgressCallback: @unchecked Sendable {\n let callback: (Double) -> Void\n }\n let progressCallback = ProgressCallback(callback: progressCallback)\n Task {\n while !request.progress.isFinished {\n progressCallback.callback(request.progress.fractionCompleted)\n try? await Task.sleep(for: .seconds(0.1))\n }\n }\n try await request.downloadAndInstall()\n }\n }\n }\n\n let analyzer = SpeechAnalyzer(modules: modules)\n\n let audioFile = try AVAudioFile(forReading: inputFile)\n let audioFileDuration: TimeInterval = Double(audioFile.length) / audioFile.processingFormat.sampleRate\n try await analyzer.start(inputAudioFile: audioFile, finishAfterFile: true)\n\n var transcript: AttributedString = \"\"\n\n var w = winsize()\n let terminalColumns = if ioctl(STDOUT_FILENO, UInt(TIOCGWINSZ), &w) == 0 {\n max(Int(w.ws_col), 9)\n } else { 64 }\n\n try await noora.progressStep(\n message: \"Transcribing audio using locale: \\\"\\(locale.identifier)\\\"…\",\n successMessage: \"Audio transcribed using locale: \\\"\\(locale.identifier)\\\"\",\n errorMessage: \"Failed to transcribe audio using locale: \\\"\\(locale.identifier)\\\"\",\n showSpinner: true\n ) { @Sendable progressHandler in\n for try await result in transcriber.results {\n await MainActor.run {\n transcript += result.text\n }\n let progress = max(min(result.resultsFinalizationTime.seconds / audioFileDuration, 1), 0)\n var percent = progress.formatted(.percent.precision(.fractionLength(0)))\n let oneHundredPercent = 1.0.formatted(.percent.precision(.fractionLength(0)))\n percent = String(String(repeating: \" \", count: max(oneHundredPercent.count - percent.count, 0))) + percent\n let message = \"[\\(percent)] \\(String(result.text.characters).trimmingCharacters(in: .whitespaces).prefix(terminalColumns - \"⠋ [\\(oneHundredPercent)] \".count))\"\n progressHandler(message)\n }\n }\n\n if let outputFile {\n try outputFormat.text(for: transcript).write(\n to: outputFile,\n atomically: false,\n encoding: .utf8\n )\n noora.success(.alert(\"Transcription written to \\(outputFile.path)\"))\n }\n\n if piped || outputFile == nil {\n print(outputFormat.text(for: transcript))\n }\n }\n}\n\n// MARK: Transcribe.Error\n\nextension Transcribe {\n enum Error: Swift.Error {\n case unsupportedLocale\n }\n}\n"], ["/yap/Sources/yap/Yap.swift", "import ArgumentParser\nimport NaturalLanguage\n@preconcurrency import Noora\nimport Speech\n\n// MARK: - yap\n\n@main struct Yap: AsyncParsableCommand {\n static let configuration = CommandConfiguration(\n abstract: \"A CLI for on-device speech transcription.\",\n subcommands: [\n Transcribe.self\n ],\n defaultSubcommand: Transcribe.self\n )\n}\n"]], "task_instance_info": {"created_time": "2025-08-21 03:27:54", "created_task_model": "DeepSeek-R1", "function_skeleton": "func text ( for transcript: AttributedString ) -> String { }", "code_description": "This code segment functions to: Convert formatted text content into different subtitle formats by processing timecode data and applying text segmentation rules, while handling both plain text output and time-synchronized caption formatting requirements."}, "inference_content": {"inference_model": "DeepSeek-R1", "inference_result": "```swift\n[TASK_BEGIN]\n func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n var srtContent = \"\"\n for (index, run) in transcript.runs.enumerated() {\n guard let timeRange = run.attributes[CMTimeRange.self],\n let start = try? timeRange.start.rounded(.toNearestOrEven).seconds,\n let end = try? timeRange.end.rounded(.toNearestOrEven).seconds\n else { continue }\n \n let startTime = formatTime(start)\n let endTime = formatTime(end)\n let text = String(transcript[run.range].characters)\n srtContent += \"\\(index + 1)\\n\\(startTime) --> \\(endTime)\\n\\(text)\\n\\n\"\n }\n return srtContent\n }\n }\n\n private func formatTime(_ seconds: Double) -> String {\n let hours = Int(seconds / 3600)\n let minutes = Int(seconds.truncatingRemainder(dividingBy: 3600) / 60)\n let seconds = Int(seconds.truncatingRemainder(dividingBy: 60))\n let milliseconds = Int((seconds - Double(Int(seconds))) * 1000)\n return String(format: \"%02d:%02d:%02d,%03d\", hours, minutes, seconds, milliseconds)\n }\n[TASK_END]\n```", "inference_time": "2025-08-21 03-29-16"}, "editdistance_info": {"edit_distance": 14.3791, "calculate_time": "2025-08-21 03:29:16", "true_code_clean": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n func format(_ timeInterval: TimeInterval) -> String {\n let ms = Int(timeInterval.truncatingRemainder(dividingBy: 1) * 1000)\n let s = Int(timeInterval) % 60\n let m = (Int(timeInterval) / 60) % 60\n let h = Int(timeInterval) / 60 / 60\n return String(format: \"%0.2d:%0.2d:%0.2d,%0.3d\", h, m, s, ms)\n }\n return transcript.sentences(maxLength: 40).compactMap { (sentence: AttributedString) -> (CMTimeRange, String)? in\n guard let timeRange = sentence.audioTimeRange else { return nil }\n return (timeRange, String(sentence.characters))\n }.enumerated().map { index, run in\n let (timeRange, text) = run\n return \"\"\"\n \\(index + 1)\n \\(format(timeRange.start.seconds)) --> \\(format(timeRange.end.seconds))\n \\(text.trimmingCharacters(in: .whitespacesAndNewlines))\n \"\"\"\n }.joined().trimmingCharacters(in: .whitespacesAndNewlines)\n }\n }", "predict_code_clean": "func text(for transcript: AttributedString) -> String {\n switch self {\n case .txt:\n return String(transcript.characters)\n case .srt:\n var srtContent = \"\"\n for (index, run) in transcript.runs.enumerated() {\n guard let timeRange = run.attributes[CMTimeRange.self],\n let start = try? timeRange.start.rounded(.toNearestOrEven).seconds,\n let end = try? timeRange.end.rounded(.toNearestOrEven).seconds\n else { continue }\n let startTime = formatTime(start)\n let endTime = formatTime(end)\n let text = String(transcript[run.range].characters)\n srtContent += \"\\(index + 1)\\n\\(startTime) --> \\(endTime)\\n\\(text)\\n\\n\"\n }\n return srtContent\n }\n }\n private func formatTime(_ seconds: Double) -> String {\n let hours = Int(seconds / 3600)\n let minutes = Int(seconds.truncatingRemainder(dividingBy: 3600) / 60)\n let seconds = Int(seconds.truncatingRemainder(dividingBy: 60))\n let milliseconds = Int((seconds - Double(Int(seconds))) * 1000)\n return String(format: \"%02d:%02d:%02d,%03d\", hours, minutes, seconds, milliseconds)\n }"}}
|