deepvoice_client/local_plugins/azure_speech/ios/Classes/AzureTtsHelper.swift

import Foundation
import AVFoundation
import MicrosoftCognitiveServicesSpeech
// 自定义语音处理组件，提供音频流处理等功能
import speech
import os.log

/**
 * Azure TTS Helper
 * 
 * 基于微软Azure语音服务的TTS实现
 * 参考文档: https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-speech-synthesis
 * 
 * 特性:
 * - 对外接口保持同步，内部异步处理
 * - 使用专用队列保证合成顺序
 * - 避免阻塞主线程和其他后台线程
 * - 事件回调由调用方处理线程切换
 * 
 * 使用示例:
 * let success = ttsHelper.speakOnce("Hello World") // 立即返回，后台异步处理
 */
 public class AzureTtsHelper: NSObject, ITtsService {
    private let tag = "AzureTtsHelper"
    // 日志对象
    private let log = OSLog(subsystem: "com.azure.speech", category: "AzureTtsHelper")
    
    private static let DEFAULT_LANGUAGE = "zh-CN"
    private static let DEFAULT_VOICE = "zh-CN-XiaoxiaoNeural"
    
    // Azure语音服务配置
    private var speechConfig: SPXSpeechConfiguration?
    private var synthesizer: SPXSpeechSynthesizer?
    private var isInitialized = false
    private var speaking = false
    
    // 当前配置
    private var currentVoice = DEFAULT_VOICE
    private var currentLanguage = DEFAULT_LANGUAGE
    private var currentRate = "+0%"
    private var currentPitch = "+0%"
    private var currentVolume = "100%"
    
    // 事件监听器列表
    private var eventListeners = NSHashTable<AnyObject>.weakObjects()
    
    // 流式文本处理的缓冲区
    private var streamBuffer = ""
    private var lastSpeakTime: TimeInterval = 0
    
    // 内部异步处理队列，保证顺序执行
    private let synthesisQueue = DispatchQueue(label: "com.azure.tts.synthesis", qos: .userInitiated)
    private let synthesisGroup = DispatchGroup()
    private var pendingTasks: [() -> Void] = []
    private let taskLock = NSLock()
    
    /**
     * 初始化语音合成服务
     * 
     * @param appId 服务应用ID
     * @param token 服务访问令牌/订阅密钥
     * @param resource 服务资源ID/区域（可选）
     * @param language 语言代码，如"zh-CN"
     * @return 是否初始化成功
     */
    public func initialize(ttsAppId: String, ttsAppToken: String, ttsResource: String, language: String) -> Bool {
        do {
            // 创建语音配置
            if ttsResource.isEmpty {
                speechConfig = try SPXSpeechConfiguration(subscription: ttsAppToken, region: "eastasia")
            } else {
                speechConfig = try SPXSpeechConfiguration(subscription: ttsAppToken, region: ttsResource)
            }
            
            // 设置语言
            speechConfig?.speechSynthesisLanguage = language
            currentLanguage = language
            
            // 设置音频输出格式 - 使用16k、16位的PCM格式
            speechConfig?.setPropertyTo("Audio16Khz16BitMonoPcm", 
                                       byName: "SpeechServiceConnection_SynthOutputFormat")
            
            // 创建合成器，使用默认音频输出（扬声器）
            synthesizer = try SPXSpeechSynthesizer(speechConfig!)
            
            // 设置事件监听
            setupEventListeners()
            
            // 设置默认音色 - 中文默认使用晓晓，英文默认使用Jenny
            if language.lowercased().starts(with: "zh") {
                _ = setVoice("zh-CN-XiaoxiaoNeural")
            } else {
                _ = setVoice("en-US-JennyNeural")
            }
            
            // 设置初始化完成
            isInitialized = true
            
            return true
        } catch {
            os_log("语音合成服务初始化失败: %{public}@", log: log, type: .error, error.localizedDescription)
            return false
        }
    }
    
    /**
     * 设置语音角色
     * 
     * @param voiceName 语音角色名称（不同服务的语音角色命名可能不同）
     * @return 是否设置成功
     */
    public func setVoice(_ voiceName: String) -> Bool {
        if !isInitialized { return false }
        
        do {
            currentVoice = voiceName
            
            // 更新语音名称
            if let config = speechConfig {
                config.speechSynthesisVoiceName = voiceName
            }
            
            // 重新创建合成器
            recreateSynthesizer()
            return true
        } catch {
            os_log("设置语音失败: %{public}@", log: log, type: .error, error.localizedDescription)
            notifyEvent(eventType: .error, params: [
                "errorCode": "VOICE_SET_FAILED",
                "errorMessage": "设置语音失败: \(error.localizedDescription)"
            ])
            return false
        }
    }
    
    /**
     * 单次合成并播放
     * 
     * @param text 要合成的文本
     * @return 是否成功开始合成
     */
    public func speakOnce(_ text: String) -> Bool {
        if !isInitialized {
            os_log("语音合成未初始化", log: log, type: .error)
            return false
        }
        
        // 立即返回成功，内部异步处理
        enqueueSynthesisTask {
            self.performSynthesis(text: text)
        }
        
        return true
    }
    
    /**
     * 将合成任务加入队列，保证顺序执行
     */
    private func enqueueSynthesisTask(_ task: @escaping () -> Void) {
        taskLock.lock()
        defer { taskLock.unlock() }
        
        pendingTasks.append(task)
        
        // 如果当前没有任务在执行，开始处理队列
        if pendingTasks.count == 1 {
            processNextTask()
        }
    }
    
    /**
     * 处理队列中的下一个任务
     */
    private func processNextTask() {
        synthesisQueue.async {
            self.synthesisGroup.enter()
            
            self.taskLock.lock()
            guard !self.pendingTasks.isEmpty else {
                self.taskLock.unlock()
                self.synthesisGroup.leave()
                return
            }
            let task = self.pendingTasks.removeFirst()
            self.taskLock.unlock()
            
            // 执行任务
            task()
            
            self.synthesisGroup.leave()
            
            // 处理下一个任务
            self.taskLock.lock()
            if !self.pendingTasks.isEmpty {
                self.taskLock.unlock()
                self.processNextTask()
            } else {
                self.taskLock.unlock()
            }
        }
    }
    
    /**
     * 实际执行合成的方法
     */
    private func performSynthesis(text: String) {
        // 重置状态
        speaking = true
        
        // 生成SSML
        let ssml = generateSsml(text)
        
        do {
            os_log("开始合成: %{public}@", log: log, type: .debug, ssml)
            // 使用同步方法，在后台队列中执行
            let result = try synthesizer?.startSpeakingSsml(ssml)
            
            // 检查结果
            if let result = result {
                os_log("合成完成，结果: %{public}@", log: log, type: .debug, String(describing: result.reason))
            }
        } catch {
            os_log("语音合成失败: %{public}@", log: log, type: .error, error.localizedDescription)
            notifyEvent(eventType: .error, params: [
                "errorCode": "SYNTHESIS_FAILED",
                "errorMessage": error.localizedDescription
            ])
            speaking = false
        }
    }
    
    /**
     * 流式合成文本
     * 
     * @param text 要合成的文本片段
     * @return 是否成功处理
     */
    public func speakStream(_ text: String) -> Bool {
        if !isInitialized || text.isEmpty {
            if !isInitialized {
                notifyEvent(eventType: .error, params: [
                    "errorCode": "NOT_INITIALIZED",
                    "errorMessage": "TTS引擎未初始化"
                ])
            }
            return false
        }
        
        do {
            // 添加新文本到缓冲区
            streamBuffer.append(text)
            
            // 增加500ms防抖逻辑
            let currentTime = Date().timeIntervalSince1970
            if currentTime - lastSpeakTime < 0.6 {
                return true
            }
            lastSpeakTime = currentTime
            
            let currentText = streamBuffer
            
            // 定义标点符号列表
            let punctuationMarks: [Character] = [".", "。", "!", "！", "?", "？", ";", "；", ",", "，", ":", "：", "\n"]
            
            // 查找最后一个标点符号的位置
            var lastPunctuationIndex = -1
            for (i, char) in currentText.enumerated().reversed() {
                if punctuationMarks.contains(char) {
                    lastPunctuationIndex = i
                    break
                }
            }
            
            // 如果找到标点符号，则播放到该标点符号
            if lastPunctuationIndex >= 0 {
                // 提取要播放的文本（包含标点符号）
                let textToSpeak = String(currentText.prefix(lastPunctuationIndex + 1)).trimmingCharacters(in: .whitespacesAndNewlines)
                
                // 剩余的文本保存在缓冲区中
                let startIndex = currentText.index(currentText.startIndex, offsetBy: lastPunctuationIndex + 1)
                streamBuffer = String(currentText[startIndex...])
                
                // 只有非空文本才播放
                if !textToSpeak.isEmpty {
                return speakOnce(textToSpeak)
                }
            }
            
            // 如果没有找到标点符号，则等待更多文本
            return true
        } catch {
            os_log("流式语音合成失败: %{public}@", log: log, type: .error, error.localizedDescription)
            notifyEvent(eventType: .error, params: [
                "errorCode": "STREAM_FAILED",
                "errorMessage": "流式语音合成失败: \(error.localizedDescription)"
            ])
            return false
        }
    }
    
    /**
     * 刷新并播放流式文本缓冲区中的剩余内容
     * 
     * @return 是否成功处理
     */
    public func flushStream() -> Bool {
        if !isInitialized {
            notifyEvent(eventType: .error, params: [
                "errorCode": "NOT_INITIALIZED",
                "errorMessage": "TTS引擎未初始化"
            ])
            return false
        }
        
        do {
            // 获取缓冲区中剩余的文本
            let remainingText = streamBuffer.trimmingCharacters(in: .whitespacesAndNewlines)
            
            // 清空缓冲区
            streamBuffer = ""
            
            // 如果缓冲区为空，直接返回成功
            if remainingText.isEmpty {
                return true
            }
            
            // 播放剩余文本
            return speakOnce(remainingText)
        } catch {
            os_log("刷新流式文本失败: %{public}@", log: log, type: .error, error.localizedDescription)
            notifyEvent(eventType: .error, params: [
                "errorCode": "FLUSH_FAILED",
                "errorMessage": "刷新流式文本失败: \(error.localizedDescription)"
            ])
            return false
        }
    }
    
    /**
     * 停止语音合成和播放
     * 
     * @return 是否成功停止
     */
    public func stop() -> Bool {
        // 立即更新状态
        speaking = false
        
        // 清除流式缓冲区中的待播放内容
        streamBuffer = ""
        
        // 清空待处理的任务队列
        taskLock.lock()
        pendingTasks.removeAll()
        taskLock.unlock()
        
        // 在后台队列停止合成器，避免阻塞主线程
        synthesisQueue.async {
            if let synthesizer = self.synthesizer {
                do {
                    try synthesizer.stopSpeaking()
                    os_log("语音合成已停止", log: self.log, type: .info)
                    
                    // 直接通知停止完成
                    self.notifyEvent(eventType: .synthesisCanceled)
                } catch {
                    os_log("停止语音合成失败: %{public}@", log: self.log, type: .error, error.localizedDescription)
                    self.notifyEvent(eventType: .error, params: [
                        "errorCode": "STOP_FAILED",
                        "errorMessage": error.localizedDescription
                    ])
                }
            }
        }
        
        return true
    }
    
    /**
     * 释放资源
     * 在不再需要服务时调用，释放底层资源
     */
    public func dispose() {
        // 停止播放
        _ = stop()
        
        // 等待所有任务完成
        synthesisGroup.wait()
        
        // 清理资源
        synthesisQueue.async {
            // 释放合成器
            self.synthesizer = nil
            
            // 释放配置
            self.speechConfig = nil
            
            // 清空流缓冲区
            self.streamBuffer = ""
            
            // 重置状态
            self.isInitialized = false
            self.speaking = false
            
            // 清空任务队列
            self.taskLock.lock()
            self.pendingTasks.removeAll()
            self.taskLock.unlock()
        }
    }
    
    /**
     * 添加TTS事件监听器
     * 
     * @param listener 事件监听器
     */
    public func addListener(_ listener: TtsEventListener) {
        eventListeners.add(listener as AnyObject)
    }
    
    /**
     * 移除TTS事件监听器
     * 
     * @param listener 要移除的事件监听器
     */
    public func removeListener(_ listener: TtsEventListener) {
        eventListeners.remove(listener as AnyObject)
    }
    
    /**
     * 添加音频数据监听器
     * 由于不再支持自定义音频流，此方法实际上不再有效
     * 
     * @param listener 音频数据监听器
     */
    public func addAudioDataListener(_ listener: AudioDataListener) {
        os_log("警告：不支持音频数据监听器功能", log: log, type: .info)
    }
    
    /**
     * 移除音频数据监听器
     * 由于不再支持自定义音频流，此方法实际上不再有效
     * 
     * @param listener 要移除的音频数据监听器
     */
    public func removeAudioDataListener(_ listener: AudioDataListener) {
        // 不做任何操作
    }
    
    /**
     * 当前是否正在播放/合成
     */
    public func isSpeaking() -> Bool {
        return speaking
    }
    
    // MARK: - 辅助方法
    
    /**
     * 设置事件监听器
     */
    private func setupEventListeners() {
        guard let synthesizer = synthesizer else { return }
        
        // 添加合成开始事件处理器
        synthesizer.addSynthesisStartedEventHandler { [weak self] _, _ in
            guard let self = self else { return }
            self.notifyEvent(eventType: .synthesisStarted)
        }
        
        // 添加合成中事件处理器
        synthesizer.addSynthesizingEventHandler { [weak self] _, _ in
            // 可以在这里处理合成中的事件，目前没有特别操作
        }
        
        // 添加合成完成事件处理器
        synthesizer.addSynthesisCompletedEventHandler { [weak self] _, e in
            guard let self = self else { return }
            self.speaking = false
            self.notifyEvent(eventType: .synthesisCompleted)
        }
        
        // 添加合成取消事件处理器
        synthesizer.addSynthesisCanceledEventHandler { [weak self] _, e in
            guard let self = self else { return }
            
            self.speaking = false
            
            var params: [String: Any] = [:]
            do {
                let cancellationDetails = try SPXSpeechSynthesisCancellationDetails(fromCanceledSynthesisResult: e.result)
                if cancellationDetails.reason == SPXCancellationReason.error {
                    params["reason"] = String(describing: cancellationDetails.reason.rawValue)
                    params["errorDetails"] = cancellationDetails.errorDetails ?? "未知错误"
                }
            } catch {
                params["errorDetails"] = "获取取消详情失败: \(error.localizedDescription)"
            }
            os_log("语音合成取消, %{public}@", log: self.log, type: .info, String(describing: params["errorDetails"] ?? "未知错误"))

            self.notifyEvent(eventType: .synthesisCanceled, params: params)
        }
    }
    
    /**
     * 触发事件通知
     */
    private func notifyEvent(eventType: TtsEventType, params: [String: Any] = [:]) {
        let event = TtsEvent(type: eventType, params: params)
        
        // 直接通知事件，由调用方处理线程切换
        for case let listener as TtsEventListener in self.eventListeners.allObjects {
            listener.onEvent(event)
        }
    }
    
    /**
     * 重新创建合成器
     */
    private func recreateSynthesizer() {
        do {
            // 使用默认音频输出配置创建合成器
            synthesizer = try SPXSpeechSynthesizer(speechConfig!)
            
            // 设置事件监听
            setupEventListeners()
        } catch {
            os_log("重新创建合成器失败: %{public}@", log: log, type: .error, error.localizedDescription)
            notifyEvent(eventType: .error, params: [
                "errorCode": "RECREATE_FAILED",
                "errorMessage": "重新创建合成器失败: \(error.localizedDescription)"
            ])
        }
    }
    
    /**
     * 设置语音参数
     */
    private func setSpeechParams(rate: Int = 0, pitch: Int = 0, volume: Int = 100) -> Bool {
        if !isInitialized { return false }
        
        do {
            currentRate = formatPercentage(rate)
            currentPitch = formatPercentage(pitch)
            currentVolume = "\(min(max(volume, 0), 100))%"
            return true
        } catch {
            os_log("设置语音参数失败: %{public}@", log: log, type: .error, error.localizedDescription)
            notifyEvent(eventType: .error, params: [
                "errorCode": "PARAMS_SET_FAILED",
                "errorMessage": "设置语音参数失败: \(error.localizedDescription)"
            ])
            return false
        }
    }
    
    /**
     * 格式化百分比值
     */
    private func formatPercentage(_ value: Int) -> String {
        return value >= 0 ? "+\(value)%" : "\(value)%"
    }
    
    /**
     * 生成SSML
     */
    private func generateSsml(_ rawText: String) -> String {
        // 1. 定义要静音的符号和表情符号列表
        let symbolsToMute = [
            "#", "*", 
            "😀", "😂", "😊", "😍", "😢", "😎", "😉", "👍", "🙌", "🎉"
        ]
        
        // 2. 转义 XML 保留字符
        var escapedText = rawText
            .replacingOccurrences(of: "&", with: "&amp;")
            .replacingOccurrences(of: "<", with: "&lt;")
            .replacingOccurrences(of: ">", with: "&gt;")
        
        // 3. 静音处理特殊符号和表情符号
        // 使用空白替换法，直接将符号替换为空字符串
        var processedText = escapedText
        for symbol in symbolsToMute {
            processedText = processedText.replacingOccurrences(of: symbol, with: "")
        }
        
        // 4. 构造简化的SSML文档，减少嵌套层级
        let ssml = """
        <speak version="1.0"
               xmlns="http://www.w3.org/2001/10/synthesis"
               xmlns:mstts="https://www.w3.org/2001/mstts"
               xml:lang="zh-CN">
          <voice name="\(currentVoice)">
            <mstts:express-as style="cheerful">
              <prosody rate="\(currentRate)" pitch="\(currentPitch)" volume="\(currentVolume)">
                <say-as interpret-as="text">\(processedText)</say-as>
              </prosody>
            </mstts:express-as>
          </voice>
        </speak>
        """
        
        return ssml
    }
}