571 lines
21 KiB
JavaScript
571 lines
21 KiB
JavaScript
/**
|
|
* SentenceQueueModule
|
|
* Manages the preparation pipeline for sentences, including TTS generation
|
|
*/
|
|
import { BaseModule } from './base-module.js';
|
|
|
|
class SentenceQueueModule extends BaseModule {
|
|
constructor() {
|
|
super('sentence-queue', 'Sentence Queue');
|
|
|
|
// Dependencies
|
|
this.dependencies = ['text-buffer', 'tts-factory', 'paragraph-layout', 'audio-manager'];
|
|
|
|
// Queue state
|
|
this.sentenceQueue = [];
|
|
this.isProcessing = false;
|
|
this.onSentenceReadyCallback = null;
|
|
|
|
// Cache for prefetched sentences
|
|
this.preparedCache = new Map();
|
|
|
|
// Bind methods
|
|
this.bindMethods([
|
|
'initialize',
|
|
'addSentence',
|
|
'processNextSentence',
|
|
'setOnSentenceReady',
|
|
'completeSentence',
|
|
'prepareSentence',
|
|
'prepareLayout',
|
|
'extractWords',
|
|
'getDropCapText',
|
|
'extractDropCapText',
|
|
'calculateAnimationTiming',
|
|
'clear'
|
|
]);
|
|
}
|
|
|
|
/**
|
|
* Initialize the module
|
|
* @returns {Promise<boolean>} - Resolves with success status
|
|
*/
|
|
async initialize() {
|
|
try {
|
|
// Get dependencies
|
|
const textBuffer = this.getModule('text-buffer');
|
|
|
|
if (!textBuffer) {
|
|
console.error("SentenceQueue: TextBuffer dependency not found");
|
|
return false;
|
|
}
|
|
|
|
// Set up the text buffer to send sentences to this queue
|
|
textBuffer.setOnSentenceReady((sentence, callback) => {
|
|
this.addSentence(sentence, callback);
|
|
});
|
|
|
|
this.reportProgress(100, "Sentence queue ready");
|
|
return true;
|
|
} catch (error) {
|
|
console.error("Error initializing Sentence Queue:", error);
|
|
return false;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Set callback for when a sentence is ready for display
|
|
* @param {Function} callback - Function to call with prepared sentence
|
|
*/
|
|
setOnSentenceReady(callback) {
|
|
if (typeof callback === 'function') {
|
|
this.onSentenceReadyCallback = callback;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Add a sentence to the queue
|
|
* @param {string} sentence - Sentence to add
|
|
* @param {Function} callback - Callback to call when sentence is processed
|
|
*/
|
|
addSentence(sentence, callback) {
|
|
const queueItem = typeof sentence === 'object' && sentence !== null
|
|
? { ...sentence, callback }
|
|
: { text: sentence, callback };
|
|
|
|
this.sentenceQueue.push({
|
|
...queueItem,
|
|
text: String(queueItem.text || '').trim()
|
|
});
|
|
|
|
// Process the queue if not already processing
|
|
if (!this.isProcessing) {
|
|
this.processNextSentence();
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Process the next sentence in the queue
|
|
*/
|
|
async processNextSentence() {
|
|
if (this.sentenceQueue.length === 0 || this.isProcessing) {
|
|
return;
|
|
}
|
|
|
|
this.isProcessing = true;
|
|
const item = this.sentenceQueue[0];
|
|
|
|
try {
|
|
// Check if sentence is already in cache
|
|
const cacheKey = `${item.id || ''}:${item.text}`;
|
|
let sentence = this.preparedCache.get(cacheKey);
|
|
|
|
if (!sentence) {
|
|
// Prepare complete sentence object (TTS + layout in parallel)
|
|
sentence = await this.prepareSentence(item);
|
|
} else {
|
|
console.log('SentenceQueue: Using cached sentence');
|
|
this.preparedCache.delete(cacheKey);
|
|
}
|
|
|
|
// Prefetch next sentence while current displays
|
|
if (this.sentenceQueue.length > 1) {
|
|
const nextItem = this.sentenceQueue[1];
|
|
const nextCacheKey = `${nextItem.id || ''}:${nextItem.text}`;
|
|
if (!this.preparedCache.has(nextCacheKey)) {
|
|
document.dispatchEvent(new CustomEvent('story:process-state', {
|
|
detail: { state: 'playing-generating', reason: 'prefetch-start', sentenceId: nextItem.id }
|
|
}));
|
|
console.log('Process state: playing-generating', { reason: 'prefetch-start', sentenceId: nextItem.id });
|
|
this.prepareSentence(nextItem)
|
|
.then(prepared => {
|
|
this.preparedCache.set(nextCacheKey, prepared);
|
|
console.log('SentenceQueue: Prefetched next sentence');
|
|
document.dispatchEvent(new CustomEvent('story:process-state', {
|
|
detail: { state: 'playing-ready', reason: 'prefetch-complete', sentenceId: nextItem.id }
|
|
}));
|
|
console.log('Process state: playing-ready', { reason: 'prefetch-complete', sentenceId: nextItem.id });
|
|
})
|
|
.catch(err => console.warn('SentenceQueue: Prefetch failed:', err));
|
|
}
|
|
} else {
|
|
document.dispatchEvent(new CustomEvent('story:process-state', {
|
|
detail: { state: 'playing-ready', reason: 'no-prefetch-needed', sentenceId: item.id }
|
|
}));
|
|
console.log('Process state: playing-ready', { reason: 'no-prefetch-needed', sentenceId: item.id });
|
|
}
|
|
|
|
// Notify display handler with complete sentence
|
|
if (this.onSentenceReadyCallback) {
|
|
await new Promise(resolve => {
|
|
sentence.onComplete = resolve;
|
|
this.onSentenceReadyCallback(sentence, resolve);
|
|
});
|
|
}
|
|
|
|
// Remove from queue and continue
|
|
this.sentenceQueue.shift();
|
|
if (item.callback) item.callback({ success: true });
|
|
|
|
} catch (error) {
|
|
console.error("SentenceQueue: Error processing sentence:", error);
|
|
if (item.callback) item.callback({ success: false, error });
|
|
} finally {
|
|
this.isProcessing = false;
|
|
if (this.sentenceQueue.length > 0) {
|
|
this.processNextSentence();
|
|
} else {
|
|
document.dispatchEvent(new CustomEvent('story:process-state', {
|
|
detail: { state: 'ready', reason: 'queue-empty' }
|
|
}));
|
|
document.dispatchEvent(new CustomEvent('tts:queue-empty', {
|
|
detail: { reason: 'sentence-queue-empty' }
|
|
}));
|
|
console.log('Process state: ready', { reason: 'queue-empty' });
|
|
}
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Prepare speech metadata for a sentence
|
|
* @param {string} text - Text to prepare speech for
|
|
* @returns {Promise<Object>} - Speech metadata object
|
|
*/
|
|
async prepareSpeechMetadata(text) {
|
|
const ttsFactory = this.getModule('tts-factory');
|
|
|
|
if (!ttsFactory) {
|
|
throw new Error("TTS dependencies not found");
|
|
}
|
|
|
|
// Check if TTS is enabled via active handler
|
|
const activeHandler = ttsFactory.getActiveHandler();
|
|
const isTtsEnabled = activeHandler !== null;
|
|
|
|
// If TTS is disabled, estimate duration based on character count
|
|
if (!isTtsEnabled) {
|
|
return this.estimateSpeechDuration(text);
|
|
}
|
|
|
|
try {
|
|
// Preload the speech to get metadata
|
|
const result = await ttsFactory.preloadSpeech(text);
|
|
|
|
if (!result.success) {
|
|
console.warn("SentenceQueue: Speech preload failed, using estimated duration");
|
|
return this.estimateSpeechDuration(text);
|
|
}
|
|
|
|
// Create a speech metadata object
|
|
return {
|
|
text: text,
|
|
duration: result.duration || this.estimateSpeechDuration(text).duration,
|
|
handler: ttsFactory.getActiveHandler() ? ttsFactory.getActiveHandler().id : null,
|
|
play: async () => {
|
|
return ttsFactory.speak(text);
|
|
},
|
|
stop: () => {
|
|
return ttsFactory.stop();
|
|
},
|
|
isTtsEnabled: isTtsEnabled
|
|
};
|
|
} catch (error) {
|
|
console.error("Error preparing speech metadata:", error);
|
|
return this.estimateSpeechDuration(text);
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Estimate speech duration based on character count
|
|
* @param {string} text - Text to estimate duration for
|
|
* @returns {Object} - Speech metadata object with estimated duration
|
|
*/
|
|
estimateSpeechDuration(text) {
|
|
// Average aloud narration is around 12 characters per second at 1x.
|
|
const charactersPerSecond = 12;
|
|
|
|
let speedMultiplier = 1.0;
|
|
const ttsFactory = this.getModule('tts-factory');
|
|
if (ttsFactory) {
|
|
speedMultiplier = Number.isFinite(ttsFactory.speed) ? Math.max(0.25, ttsFactory.speed) : 1.0;
|
|
}
|
|
|
|
// Calculate estimated duration in milliseconds
|
|
const charCount = text.length;
|
|
const durationSeconds = charCount / (charactersPerSecond * speedMultiplier);
|
|
const durationMs = Math.max(durationSeconds * 1000, 800);
|
|
|
|
return {
|
|
text: text,
|
|
duration: durationMs,
|
|
handler: null,
|
|
play: async () => ({ success: false, reason: 'tts_disabled' }),
|
|
stop: () => true,
|
|
isTtsEnabled: false,
|
|
isEstimated: true
|
|
};
|
|
}
|
|
|
|
/**
|
|
* Prepare a complete sentence object with TTS and layout
|
|
* @param {string} text - Text to prepare
|
|
* @returns {Promise<Object>} - Complete sentence object
|
|
*/
|
|
async prepareSentence(item) {
|
|
const text = typeof item === 'string' ? item : item.text;
|
|
const id = item.id || `paragraph-${Date.now()}-${Math.random().toString(36).substr(2, 9)}`;
|
|
const metadata = typeof item === 'object' && item !== null ? item : {};
|
|
|
|
try {
|
|
if (metadata.type && !['paragraph', 'heading'].includes(metadata.type)) {
|
|
if (metadata.type === 'music') {
|
|
const audioManager = this.getModule('audio-manager');
|
|
if (audioManager && typeof audioManager.playMusic === 'function') {
|
|
audioManager.getAssetUrl('music', metadata.filename);
|
|
}
|
|
}
|
|
|
|
return {
|
|
id,
|
|
kind: metadata.type,
|
|
text: text || '',
|
|
status: 'ready',
|
|
metadata,
|
|
tts: { duration: 0, provider: null, audioData: null, play: null, stop: null, enabled: false },
|
|
animation: { wordTimings: [], cueTimings: [], totalDuration: 0 },
|
|
element: null,
|
|
onComplete: null
|
|
};
|
|
}
|
|
|
|
const audioManager = this.getModule('audio-manager');
|
|
if (audioManager && typeof audioManager.preloadMediaCues === 'function') {
|
|
await audioManager.preloadMediaCues(metadata.cueMarkers || []);
|
|
}
|
|
|
|
// Prepare TTS and layout in parallel
|
|
const [ttsData, layoutData] = await Promise.all([
|
|
this.prepareSpeechMetadata(text),
|
|
this.prepareLayout(text, metadata)
|
|
]);
|
|
|
|
// Calculate animation timing based on TTS duration
|
|
const words = this.extractWords(layoutData.nodes);
|
|
const animation = this.calculateAnimationTiming(words, ttsData.duration, metadata.cueMarkers || []);
|
|
|
|
console.log(`SentenceQueue: Prepared sentence "${text.substring(0, 50)}..." - TTS duration: ${ttsData.duration}ms, Words: ${words.length}, Animation total: ${animation.totalDuration}ms, Layout breaks: ${layoutData.breaks.length}`);
|
|
|
|
return {
|
|
id,
|
|
kind: metadata.type === 'heading' ? 'heading' : 'paragraph',
|
|
text,
|
|
paragraphIndex: metadata.paragraphIndex ?? null,
|
|
isFirstParagraphInChapter: Boolean(metadata.isFirstParagraphInChapter),
|
|
role: metadata.role || (metadata.type === 'heading' ? 'chapter-heading' : 'body'),
|
|
dropCap: Boolean(metadata.dropCap),
|
|
addTopSpace: Boolean(metadata.addTopSpace),
|
|
cueMarkers: metadata.cueMarkers || [],
|
|
status: 'ready',
|
|
layout: layoutData,
|
|
tts: {
|
|
duration: ttsData.duration,
|
|
provider: ttsData.handler,
|
|
audioData: ttsData.audioData || null,
|
|
play: ttsData.play,
|
|
stop: ttsData.stop,
|
|
enabled: ttsData.isTtsEnabled
|
|
},
|
|
animation: animation,
|
|
element: null,
|
|
onComplete: null
|
|
};
|
|
} catch (error) {
|
|
console.error('SentenceQueue: Error preparing sentence:', error);
|
|
throw error;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Prepare layout for a sentence
|
|
* @param {string} text - Text to prepare layout for
|
|
* @returns {Promise<Object>} - Layout data
|
|
*/
|
|
async prepareLayout(text, metadata = {}) {
|
|
const paragraphLayout = this.getModule('paragraph-layout');
|
|
|
|
if (!paragraphLayout) {
|
|
throw new Error("ParagraphLayout module not found");
|
|
}
|
|
|
|
try {
|
|
if (document.fonts && document.fonts.ready) {
|
|
await document.fonts.ready;
|
|
}
|
|
|
|
// Calculate layout with Knuth-Plass
|
|
const storyElement = document.getElementById('story');
|
|
if (!storyElement) {
|
|
throw new Error("Story container not found");
|
|
}
|
|
|
|
// Get actual CSS values from the paragraph typography rule, not the
|
|
// container. The measured font and rendered font must be identical.
|
|
const containerWidth = storyElement.clientWidth;
|
|
const probe = document.createElement('p');
|
|
probe.style.visibility = 'hidden';
|
|
probe.style.position = 'absolute';
|
|
probe.style.left = '-8000px';
|
|
probe.style.top = '-8000px';
|
|
storyElement.appendChild(probe);
|
|
const computedStyle = window.getComputedStyle(probe);
|
|
const fontSize = parseFloat(computedStyle.fontSize);
|
|
const lineHeight = parseFloat(computedStyle.lineHeight);
|
|
const fontFamily = computedStyle.fontFamily;
|
|
probe.remove();
|
|
|
|
console.log(`SentenceQueue: Container metrics - width: ${containerWidth}px, fontSize: ${fontSize}px, lineHeight: ${lineHeight}px`);
|
|
|
|
// Standard book indentation: no indent on the first chapter paragraph,
|
|
// first-line indent on following paragraphs.
|
|
const isHeading = metadata.type === 'heading' || metadata.role === 'chapter-heading' || metadata.role === 'section-heading';
|
|
const dropCapLines = metadata.dropCap ? 2 : 0;
|
|
const dropCapWidth = metadata.dropCap ? lineHeight * 1.45 : 0;
|
|
const indentWidth = (isHeading || metadata.isFirstParagraphInChapter || metadata.addTopSpace) ? 0 : lineHeight * 1.5;
|
|
const layoutText = metadata.layoutText || text;
|
|
const layoutPlainText = metadata.dropCap ? this.extractDropCapText(layoutText) : layoutText;
|
|
|
|
// Measures are consumed in line order by the line breaker.
|
|
const measures = isHeading
|
|
? [containerWidth]
|
|
: metadata.dropCap
|
|
? [
|
|
Math.max(120, containerWidth - dropCapWidth),
|
|
Math.max(120, containerWidth - dropCapWidth),
|
|
containerWidth
|
|
]
|
|
: [
|
|
Math.max(120, containerWidth - indentWidth),
|
|
containerWidth,
|
|
containerWidth
|
|
];
|
|
|
|
console.log(`SentenceQueue: Layout calculation - indentWidth: ${indentWidth.toFixed(1)}px, measures: [${measures.map(m => m.toFixed(1)).join(', ')}]`);
|
|
|
|
const layout = paragraphLayout.calculateLayout(layoutPlainText, {
|
|
measures,
|
|
fontSize: `${fontSize}px`,
|
|
fontFamily,
|
|
lineHeight: lineHeight / fontSize,
|
|
lineHeightPx: lineHeight
|
|
});
|
|
|
|
if (!layout) {
|
|
throw new Error('Paragraph layout calculation failed');
|
|
}
|
|
|
|
return {
|
|
breaks: layout.breaks,
|
|
nodes: layout.nodes,
|
|
processedText: layout.processedText || text,
|
|
sourceLayoutText: layoutText,
|
|
measures,
|
|
indentWidth,
|
|
dropCap: Boolean(metadata.dropCap),
|
|
dropCapText: metadata.dropCap ? this.getDropCapText(layoutText) : '',
|
|
dropCapLines,
|
|
addTopSpace: Boolean(metadata.addTopSpace),
|
|
role: metadata.role || (isHeading ? 'chapter-heading' : 'body'),
|
|
align: isHeading ? 'center' : 'justify',
|
|
fontSize: layout.fontSize,
|
|
fontFamily: layout.fontFamily,
|
|
lineHeight: layout.lineHeight,
|
|
lineHeightPx: layout.lineHeightPx
|
|
};
|
|
} catch (error) {
|
|
console.error('SentenceQueue: Error preparing layout:', error);
|
|
throw error;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Extract words from layout nodes
|
|
* @param {Array} nodes - Layout nodes from Knuth-Plass algorithm
|
|
* @returns {Array<string>} - Array of words
|
|
*/
|
|
extractWords(nodes) {
|
|
if (!nodes || !Array.isArray(nodes)) {
|
|
return [];
|
|
}
|
|
|
|
return nodes
|
|
.filter(node => node.type === 'box')
|
|
.map(node => node.value || '');
|
|
}
|
|
|
|
getDropCapText(text) {
|
|
const plain = String(text || '').replace(/<[^>]+>/g, '');
|
|
const match = plain.match(/^([“"']?[A-Za-zÀ-ÖØ-öø-ÿ])/u);
|
|
return match ? match[1] : '';
|
|
}
|
|
|
|
extractDropCapText(text) {
|
|
const dropCap = this.getDropCapText(text);
|
|
if (!dropCap) return text;
|
|
return String(text).replace(dropCap, '').trimStart();
|
|
}
|
|
|
|
/**
|
|
* Calculate animation timing based on TTS duration
|
|
* @param {Array<string>} words - Array of words to animate
|
|
* @param {number} totalDuration - Total duration in milliseconds
|
|
* @returns {Object} - Animation timing data
|
|
*/
|
|
calculateAnimationTiming(words, totalDuration, cueMarkers = []) {
|
|
if (!words || words.length === 0) {
|
|
return {
|
|
wordTimings: [],
|
|
cueTimings: [],
|
|
totalDuration: 0
|
|
};
|
|
}
|
|
|
|
const totalChars = words.reduce((sum, word) => sum + word.length, 0);
|
|
|
|
if (totalChars === 0) {
|
|
return {
|
|
wordTimings: words.map(word => ({ word, delay: 0, duration: 0 })),
|
|
cueTimings: [],
|
|
totalDuration: 0
|
|
};
|
|
}
|
|
|
|
const msPerChar = totalDuration / totalChars;
|
|
|
|
let currentDelay = 0;
|
|
const wordTimings = words.map(word => {
|
|
const duration = word.length * msPerChar;
|
|
const timing = {
|
|
word: word,
|
|
delay: currentDelay,
|
|
duration: duration
|
|
};
|
|
currentDelay += duration;
|
|
return timing;
|
|
});
|
|
|
|
const cueTimings = (cueMarkers || []).map(cue => {
|
|
const wordIndex = Math.max(0, Math.min(cue.wordIndex || 0, wordTimings.length - 1));
|
|
const timing = wordTimings[wordIndex] || { delay: currentDelay };
|
|
return {
|
|
...cue,
|
|
delay: timing.delay
|
|
};
|
|
});
|
|
|
|
return {
|
|
wordTimings,
|
|
cueTimings,
|
|
totalDuration: Math.round(currentDelay)
|
|
};
|
|
}
|
|
|
|
/**
|
|
* Complete processing of a sentence
|
|
* @param {Object} item - Queue item
|
|
* @param {Object} result - Processing result
|
|
*/
|
|
completeSentence(item, result) {
|
|
// Remove from queue
|
|
this.sentenceQueue.shift();
|
|
|
|
// Call the original callback
|
|
if (item.callback) {
|
|
item.callback(result);
|
|
}
|
|
|
|
// Reset processing flag
|
|
this.isProcessing = false;
|
|
|
|
// Process next sentence if any
|
|
if (this.sentenceQueue.length > 0) {
|
|
this.processNextSentence();
|
|
}
|
|
}
|
|
|
|
clear() {
|
|
this.sentenceQueue = [];
|
|
this.isProcessing = false;
|
|
this.preparedCache.clear();
|
|
document.dispatchEvent(new CustomEvent('tts:queue-empty', {
|
|
detail: { reason: 'sentence-queue-cleared' }
|
|
}));
|
|
document.dispatchEvent(new CustomEvent('story:process-state', {
|
|
detail: { state: 'ready', reason: 'sentence-queue-cleared' }
|
|
}));
|
|
}
|
|
}
|
|
|
|
// Create the singleton instance
|
|
const SentenceQueue = new SentenceQueueModule();
|
|
|
|
// Export the module
|
|
export { SentenceQueue };
|
|
|
|
// Register with the module registry
|
|
if (window.moduleRegistry) {
|
|
window.moduleRegistry.register(SentenceQueue);
|
|
}
|
|
|
|
// Keep a reference in window for loader system
|
|
window.SentenceQueue = SentenceQueue;
|