Spaces:

vomebook
/

Hitokoto

Running

fix(tts): read list numbers as enumerations, not separate sentences

8c72c22 about 23 hours ago

9.39 kB

"use strict";
// Pure planning/bookmark helpers; positions are original JavaScript UTF-16 offsets.
globalThis.ReaderPlan = (() => {
  const maxText = 20000;
  const batchSize = 600;
  const bookmarkKey = "hitokoto-reader-v1";

  // "1." / "一." / "(3)." numbering its item at a line or clause start: the period
  // is enumeration, not a sentence end. Mirrors tts/text.py list_marker.
  function listMarker(chars, index) {
    let cursor = index;
    const closed = cursor > 0 && /[)）]/u.test(chars[cursor - 1]);
    if (closed) cursor -= 1;
    const digits = cursor;
    while (cursor > 0 && digits - cursor < 4 && /[0-9０-９一二三四五六七八九十百]/u.test(chars[cursor - 1])) cursor -= 1;
    if (cursor === digits) return false;
    if (cursor > 0 && /[(（]/u.test(chars[cursor - 1])) cursor -= 1;
    else if (closed) return false;
    return cursor === 0 || /[ \t\r\n。！？!?；;：:，,、]/u.test(chars[cursor - 1]);
  }
  function numberSpans(text) {
    // Same recognised atoms as tts.numbers; source positions remain UTF-16.
    const label = "(?:订单编号|訂單編號|订单号|訂單號|编号|編號|手机号|手機號|电话|電話|学号|學號|邮编|郵編|ISBN(?:-13|-10)?|IP)";
    const pattern = new RegExp(label + "[ \\t]*[:：]?[ \\t]*[A-Za-z0-9]+(?:[._/\\-][A-Za-z0-9]+)*" +
      "|(?<![A-Za-z0-9])(?:版本|[Vv]ersion|[Vv])[ \\t]*[:：]?[ \\t]*[vV]?[0-9]+(?:\\.[0-9]+)+" +
      "|(?<![A-Za-z0-9])[0-9]{4}[-/][0-9]{1,2}[-/][0-9]{1,2}(?![0-9])" +
      "|(?<![A-Za-z0-9])[0-9]{4}年[0-9]{1,2}月[0-9]{1,2}[日号]" +
      "|(?<![A-Za-z0-9])[0-9]{4}年|(?<![0-9])[0-9]+(?:\\.[0-9]+){2,}" +
      "|(?<![A-Za-z0-9])[+-]?[0-9]+/[0-9]+(?![0-9/])" +
      "|[+-]?[0-9]+(?:,[0-9]{3})*(?:\\.[0-9]+)?(?:[%％]|℃)?", "gu");
    return Array.from(text.matchAll(pattern), (match) => ({ start: match.index, end: match.index + match[0].length }));
  }

  function split(text, limit = batchSize, firstLimit = Math.min(limit, 180)) {
    const chars = Array.from(text);
    if (!text.trim() || chars.length > maxText || !Number.isInteger(limit) || limit < 1 || limit > 800 ||
        !Number.isInteger(firstLimit) || firstLimit < 1 || firstLimit > limit) {
      throw new Error(`请输入 1–${maxText.toLocaleString()} 个 Unicode 字符的正文`);
    }
    const offsets = [0];
    for (const char of chars) offsets.push(offsets[offsets.length - 1] + char.length);
    const positions = new Map(offsets.map((offset, index) => [offset, index]));
    const atoms = numberSpans(text).map((part) => ({ start: positions.get(part.start), end: positions.get(part.end) }));
    const result = [];
    for (let start = 0; start < chars.length;) {
      const width = result.length ? limit : firstLimit;
      let end = Math.min(start + width, chars.length);
      if (end < chars.length) {
        // Prefer a recent sentence boundary, without reducing every batch to a
        // tiny sentence or splitting decimal points, years and ASCII words.
        for (let index = end - 1; index >= start + Math.floor(width / 2); index -= 1) {
          if (/[。！？!?；;\n]/u.test(chars[index]) ||
              (chars[index] === "." && !(/[0-9]/u.test(chars[index - 1] || "") && /[0-9]/u.test(chars[index + 1] || "")) && !listMarker(chars, index))) {
            end = index + 1;
            break;
          }
        }
        const insideToken = (index) => index < chars.length &&
          ((/[a-zA-Z0-9]/u.test(chars[index - 1]) && /[a-zA-Z0-9]/u.test(chars[index])) ||
           (/[0-9.]/u.test(chars[index - 1]) && /[0-9.]/u.test(chars[index])));
        while (end > start && insideToken(end)) end -= 1;
        // A long but otherwise valid first English word need not fit the short
        // first-batch target. Fall back to the ordinary bounded batch size.
        if (end === start && width < limit) {
          end = Math.min(start + limit, chars.length);
          while (end > start && insideToken(end)) end -= 1;
        }
        if (end === start) throw new Error("单个英文词或数字过长，无法在保留原文的同时分批朗读");
        for (const atom of atoms) {
          if (atom.start < end && end < atom.end) {
            end = atom.start > start ? atom.start : atom.end;
            break;
          }
        }
        if (end - start > limit) throw new Error("单个日期、数字或编号超过批次上限，请整理后重试");
      }
      const body = chars.slice(start, end).join("");
      if (!/[\p{L}\p{N}]/u.test(body)) throw new Error("正文含有过长的连续空白或标点，请整理后重试");
      result.push({ text: body, start: offsets[start], end: offsets[end],
        context_before: chars.slice(Math.max(0, start - 32), start).join(""),
        context_after: chars.slice(end, end + 32).join("") });
      start = end;
    }
    return result;
  }

  function readBookmark(storage) {
    try {
      const value = JSON.parse(storage.getItem(bookmarkKey));
      if (value?.version !== 1 || typeof value.text !== "string" || Array.from(value.text).length > maxText ||
          !value.text.trim() || !["zf_001", "zf_017", "zm_010", "zm_029"].includes(value.voice) ||
          !Number.isFinite(value.speed) || value.speed < 0.7 || value.speed > 1.4 ||
          !Number.isInteger(value.offset) || value.offset < 0 || value.offset > value.text.length ||
          !Number.isFinite(value.seconds) || value.seconds < 0 || value.seconds > 240 ||
          (value.pitchSemitones !== undefined && (!Number.isInteger(value.pitchSemitones) || value.pitchSemitones < -3 || value.pitchSemitones > 3)) ||
          (value.playbackRate !== undefined && (!Number.isFinite(value.playbackRate) || value.playbackRate < .5 || value.playbackRate > 2))) return null;
      return value;
    } catch (_) { return null; }
  }

  function clipPlan(text) {
    // Mirror tts.text's lossless source segmentation for navigation before audio
    // is generated. The API's published spans remain authoritative for playback.
    const chars = Array.from(text), offsets = [0], raw = [], result = [];
    for (const char of chars) offsets.push(offsets[offsets.length - 1] + char.length);
    const positions = new Map(offsets.map((offset, index) => [offset, index]));
    const atoms = numberSpans(text).map((part) => ({ start: positions.get(part.start), end: positions.get(part.end) }));
    const digit = (char) => /\p{Nd}/u.test(char || "");
    const ascii = (char) => /^[a-zA-Z0-9]$/u.test(char || "");
    for (let start = 0; start < chars.length;) {
      let end = Math.min(chars.length, start + 60);
      for (let index = start; index < end; index += 1) {
        if (/[。！？!?；;\n]/u.test(chars[index]) || (chars[index] === "." && !(digit(chars[index - 1]) && digit(chars[index + 1])) && !listMarker(chars, index))) {
          end = index + 1;
          while (end < chars.length && end - start < 60 && /[”’"'」』）) \t\r]/u.test(chars[end])) end += 1;
          break;
        }
      }
      if (end < chars.length) {
        let boundary = end;
        while (boundary > start && ((ascii(chars[boundary - 1]) && ascii(chars[boundary])) ||
               (/^[0-9.]$/u.test(chars[boundary - 1]) && /^[0-9.]$/u.test(chars[boundary])))) boundary -= 1;
        if (boundary > start) end = boundary;
      }
      for (const atom of atoms) {
        if (atom.start < end && end < atom.end) { end = atom.start > start ? atom.start : atom.end; break; }
      }
      raw.push(end);
      start = end;
    }
    let start = 0;
    for (const end of raw) {
      const body = chars.slice(start, end).join("");
      const speech = /[\p{L}\p{N}]/u.test(body);
      if (end === chars.length && !speech) {
        if (!result.length) throw new Error("正文不能只有标点或空白");
        result[result.length - 1].end = offsets[end];
        start = end;
      } else if (speech && (end === chars.length || (result.length < 63 &&
                 (end - start >= (result.length ? 40 : 20) || /\n[ \t\r]*$/u.test(body))))) {
        result.push({ start: offsets[start], end: offsets[end] });
        start = end;
      }
    }
    return result;
  }

  function locations(plan) {
    return plan.flatMap((batch, batchIndex) => clipPlan(batch.text).map((part, index) => (
      { batch: batchIndex, index, start: batch.start + part.start, end: batch.start + part.end }
    )));
  }

  class SleepTimer {
    constructor(clock = Date.now) { this.clock = clock; this.clear(); }
    clear() { this.mode = "off"; this.deadline = 0; this.anchor = null; }
    arm(mode, anchor = null) {
      if (mode === "off") { this.clear(); return; }
      if (mode === "clip") { this.mode = mode; this.deadline = 0; this.anchor = anchor; return; }
      if (!["5", "15", "30", "60"].includes(mode)) throw new Error("无效的睡眠定时");
      this.mode = mode; this.deadline = this.clock() + Number(mode) * 60000; this.anchor = null;
    }
    remaining() { return Math.max(0, Math.ceil((this.deadline - this.clock()) / 1000)); }
    due() { return this.deadline > 0 && this.clock() >= this.deadline; }
    bind(anchor) { if (this.mode === "clip" && this.anchor === null) this.anchor = anchor; }
    ended(anchor) { return this.mode === "clip" && this.anchor === anchor; }
  }

  return { maxText, batchSize, bookmarkKey, numberSpans, split, clipPlan, locations, readBookmark, SleepTimer };
})();

	"use strict";
	// Pure planning/bookmark helpers; positions are original JavaScript UTF-16 offsets.
	globalThis.ReaderPlan = (() => {
	const maxText = 20000;
	const batchSize = 600;
	const bookmarkKey = "hitokoto-reader-v1";

	// "1." / "一." / "(3)." numbering its item at a line or clause start: the period
	// is enumeration, not a sentence end. Mirrors tts/text.py list_marker.
	function listMarker(chars, index) {
	let cursor = index;
	const closed = cursor > 0 && /[)）]/u.test(chars[cursor - 1]);
	if (closed) cursor -= 1;
	const digits = cursor;
	while (cursor > 0 && digits - cursor < 4 && /[0-9０-９一二三四五六七八九十百]/u.test(chars[cursor - 1])) cursor -= 1;
	if (cursor === digits) return false;
	if (cursor > 0 && /[(（]/u.test(chars[cursor - 1])) cursor -= 1;
	else if (closed) return false;
	return cursor === 0 \|\| /[ \t\r\n。！？!?；;：:，,、]/u.test(chars[cursor - 1]);
	}
	function numberSpans(text) {
	// Same recognised atoms as tts.numbers; source positions remain UTF-16.
	const label = "(?:订单编号\|訂單編號\|订单号\|訂單號\|编号\|編號\|手机号\|手機號\|电话\|電話\|学号\|學號\|邮编\|郵編\|ISBN(?:-13\|-10)?\|IP)";
	const pattern = new RegExp(label + "[ \\t][:：]?[ \\t][A-Za-z0-9]+(?:[._/\\-][A-Za-z0-9]+)*" +
	"\|(?<![A-Za-z0-9])(?:版本\|[Vv]ersion\|[Vv])[ \\t][:：]?[ \\t][vV]?[0-9]+(?:\\.[0-9]+)+" +
	"\|(?<![A-Za-z0-9])[0-9]{4}[-/][0-9]{1,2}[-/][0-9]{1,2}(?![0-9])" +
	"\|(?<![A-Za-z0-9])[0-9]{4}年[0-9]{1,2}月[0-9]{1,2}[日号]" +
	"\|(?<![A-Za-z0-9])[0-9]{4}年\|(?<![0-9])[0-9]+(?:\\.[0-9]+){2,}" +
	"\|(?<![A-Za-z0-9])[+-]?[0-9]+/[0-9]+(?![0-9/])" +
	"\|[+-]?[0-9]+(?:,[0-9]{3})*(?:\\.[0-9]+)?(?:[%％]\|℃)?", "gu");
	return Array.from(text.matchAll(pattern), (match) => ({ start: match.index, end: match.index + match[0].length }));
	}

	function split(text, limit = batchSize, firstLimit = Math.min(limit, 180)) {
	const chars = Array.from(text);
	if (!text.trim() \|\| chars.length > maxText \|\| !Number.isInteger(limit) \|\| limit < 1 \|\| limit > 800 \|\|
	!Number.isInteger(firstLimit) \|\| firstLimit < 1 \|\| firstLimit > limit) {
	throw new Error(`请输入 1–${maxText.toLocaleString()} 个 Unicode 字符的正文`);
	}
	const offsets = [0];
	for (const char of chars) offsets.push(offsets[offsets.length - 1] + char.length);
	const positions = new Map(offsets.map((offset, index) => [offset, index]));
	const atoms = numberSpans(text).map((part) => ({ start: positions.get(part.start), end: positions.get(part.end) }));
	const result = [];
	for (let start = 0; start < chars.length;) {
	const width = result.length ? limit : firstLimit;
	let end = Math.min(start + width, chars.length);
	if (end < chars.length) {
	// Prefer a recent sentence boundary, without reducing every batch to a
	// tiny sentence or splitting decimal points, years and ASCII words.
	for (let index = end - 1; index >= start + Math.floor(width / 2); index -= 1) {
	if (/[。！？!?；;\n]/u.test(chars[index]) \|\|
	(chars[index] === "." && !(/[0-9]/u.test(chars[index - 1] \|\| "") && /[0-9]/u.test(chars[index + 1] \|\| "")) && !listMarker(chars, index))) {
	end = index + 1;
	break;
	}
	}
	const insideToken = (index) => index < chars.length &&
	((/[a-zA-Z0-9]/u.test(chars[index - 1]) && /[a-zA-Z0-9]/u.test(chars[index])) \|\|
	(/[0-9.]/u.test(chars[index - 1]) && /[0-9.]/u.test(chars[index])));
	while (end > start && insideToken(end)) end -= 1;
	// A long but otherwise valid first English word need not fit the short
	// first-batch target. Fall back to the ordinary bounded batch size.
	if (end === start && width < limit) {
	end = Math.min(start + limit, chars.length);
	while (end > start && insideToken(end)) end -= 1;
	}
	if (end === start) throw new Error("单个英文词或数字过长，无法在保留原文的同时分批朗读");
	for (const atom of atoms) {
	if (atom.start < end && end < atom.end) {
	end = atom.start > start ? atom.start : atom.end;
	break;
	}
	}
	if (end - start > limit) throw new Error("单个日期、数字或编号超过批次上限，请整理后重试");
	}
	const body = chars.slice(start, end).join("");
	if (!/[\p{L}\p{N}]/u.test(body)) throw new Error("正文含有过长的连续空白或标点，请整理后重试");
	result.push({ text: body, start: offsets[start], end: offsets[end],
	context_before: chars.slice(Math.max(0, start - 32), start).join(""),
	context_after: chars.slice(end, end + 32).join("") });
	start = end;
	}
	return result;
	}

	function readBookmark(storage) {
	try {
	const value = JSON.parse(storage.getItem(bookmarkKey));
	if (value?.version !== 1 \|\| typeof value.text !== "string" \|\| Array.from(value.text).length > maxText \|\|
	!value.text.trim() \|\| !["zf_001", "zf_017", "zm_010", "zm_029"].includes(value.voice) \|\|
	!Number.isFinite(value.speed) \|\| value.speed < 0.7 \|\| value.speed > 1.4 \|\|
	!Number.isInteger(value.offset) \|\| value.offset < 0 \|\| value.offset > value.text.length \|\|
	!Number.isFinite(value.seconds) \|\| value.seconds < 0 \|\| value.seconds > 240 \|\|
	(value.pitchSemitones !== undefined && (!Number.isInteger(value.pitchSemitones) \|\| value.pitchSemitones < -3 \|\| value.pitchSemitones > 3)) \|\|
	(value.playbackRate !== undefined && (!Number.isFinite(value.playbackRate) \|\| value.playbackRate < .5 \|\| value.playbackRate > 2))) return null;
	return value;
	} catch (_) { return null; }
	}

	function clipPlan(text) {
	// Mirror tts.text's lossless source segmentation for navigation before audio
	// is generated. The API's published spans remain authoritative for playback.
	const chars = Array.from(text), offsets = [0], raw = [], result = [];
	for (const char of chars) offsets.push(offsets[offsets.length - 1] + char.length);
	const positions = new Map(offsets.map((offset, index) => [offset, index]));
	const atoms = numberSpans(text).map((part) => ({ start: positions.get(part.start), end: positions.get(part.end) }));
	const digit = (char) => /\p{Nd}/u.test(char \|\| "");
	const ascii = (char) => /^[a-zA-Z0-9]$/u.test(char \|\| "");
	for (let start = 0; start < chars.length;) {
	let end = Math.min(chars.length, start + 60);
	for (let index = start; index < end; index += 1) {
	if (/[。！？!?；;\n]/u.test(chars[index]) \|\| (chars[index] === "." && !(digit(chars[index - 1]) && digit(chars[index + 1])) && !listMarker(chars, index))) {
	end = index + 1;
	while (end < chars.length && end - start < 60 && /[”’"'」』）) \t\r]/u.test(chars[end])) end += 1;
	break;
	}
	}
	if (end < chars.length) {
	let boundary = end;
	while (boundary > start && ((ascii(chars[boundary - 1]) && ascii(chars[boundary])) \|\|
	(/^[0-9.]$/u.test(chars[boundary - 1]) && /^[0-9.]$/u.test(chars[boundary])))) boundary -= 1;
	if (boundary > start) end = boundary;
	}
	for (const atom of atoms) {
	if (atom.start < end && end < atom.end) { end = atom.start > start ? atom.start : atom.end; break; }
	}
	raw.push(end);
	start = end;
	}
	let start = 0;
	for (const end of raw) {
	const body = chars.slice(start, end).join("");
	const speech = /[\p{L}\p{N}]/u.test(body);
	if (end === chars.length && !speech) {
	if (!result.length) throw new Error("正文不能只有标点或空白");
	result[result.length - 1].end = offsets[end];
	start = end;
	} else if (speech && (end === chars.length \|\| (result.length < 63 &&
	(end - start >= (result.length ? 40 : 20) \|\| /\n[ \t\r]*$/u.test(body))))) {
	result.push({ start: offsets[start], end: offsets[end] });
	start = end;
	}
	}
	return result;
	}

	function locations(plan) {
	return plan.flatMap((batch, batchIndex) => clipPlan(batch.text).map((part, index) => (
	{ batch: batchIndex, index, start: batch.start + part.start, end: batch.start + part.end }
	)));
	}

	class SleepTimer {
	constructor(clock = Date.now) { this.clock = clock; this.clear(); }
	clear() { this.mode = "off"; this.deadline = 0; this.anchor = null; }
	arm(mode, anchor = null) {
	if (mode === "off") { this.clear(); return; }
	if (mode === "clip") { this.mode = mode; this.deadline = 0; this.anchor = anchor; return; }
	if (!["5", "15", "30", "60"].includes(mode)) throw new Error("无效的睡眠定时");
	this.mode = mode; this.deadline = this.clock() + Number(mode) * 60000; this.anchor = null;
	}
	remaining() { return Math.max(0, Math.ceil((this.deadline - this.clock()) / 1000)); }
	due() { return this.deadline > 0 && this.clock() >= this.deadline; }
	bind(anchor) { if (this.mode === "clip" && this.anchor === null) this.anchor = anchor; }
	ended(anchor) { return this.mode === "clip" && this.anchor === anchor; }
	}

	return { maxText, batchSize, bookmarkKey, numberSpans, split, clipPlan, locations, readBookmark, SleepTimer };
	})();