412 lines
9.7 KiB
TypeScript
412 lines
9.7 KiB
TypeScript
/**
|
||
* @fileoverview 元数据行清理器
|
||
*
|
||
* 用于清理歌词中开头和结尾的元数据行，例如：
|
||
*
|
||
* (歌曲名) - (歌手名)
|
||
* 词：...
|
||
* 曲：...
|
||
* 编曲：...
|
||
* 和声编唱：...
|
||
* 人声编辑：...
|
||
* 混音：...
|
||
* 母带：...
|
||
* 监制：...
|
||
* 出品：...
|
||
* 真正的歌词行 1
|
||
* 真正的歌词行 2
|
||
*
|
||
* @see https://github.com/apoint123/Unilyric/blob/afd351c54eca7137cf8ee4ea5652d9ee55c20e32/lyrics_helper_rs/src/converter/processors/metadata_stripper.rs
|
||
*/
|
||
|
||
import type { LyricLine } from '@applemusic-like-lyrics/lyric'
|
||
|
||
const STRICT_MATCH_SEPARATORS = [
|
||
':',
|
||
'：',
|
||
',',
|
||
'，',
|
||
'.',
|
||
'。',
|
||
'!',
|
||
'！',
|
||
'-',
|
||
'_',
|
||
'(',
|
||
'（',
|
||
'[',
|
||
'【',
|
||
'{',
|
||
'『',
|
||
'「'
|
||
]
|
||
|
||
/**
|
||
* 扫描限制配置
|
||
*/
|
||
export interface ScanLimitConfig {
|
||
/**
|
||
* 扫描比例 (0.0 - 1.0)
|
||
*/
|
||
ratio: number
|
||
/**
|
||
* 最小扫描行数
|
||
*/
|
||
minLines: number
|
||
/**
|
||
* 最大扫描行数
|
||
*/
|
||
maxLines: number
|
||
}
|
||
|
||
/**
|
||
* 歌词清理配置
|
||
*/
|
||
export interface StripOptions {
|
||
/**
|
||
* 要匹配的关键词列表
|
||
*/
|
||
keywords: string[]
|
||
/**
|
||
* 要匹配的正则字符串列表
|
||
*/
|
||
regexPatterns: string[]
|
||
/**
|
||
* 这里的正则表达式会被视为弱匹配
|
||
*/
|
||
softMatchRegexes?: string[]
|
||
/**
|
||
* 歌曲元数据，用于检查第一行是否为 "歌曲 - 歌手" 格式
|
||
*/
|
||
matchMetadata?: {
|
||
title?: string
|
||
artists?: string[]
|
||
}
|
||
}
|
||
|
||
const DEFAULT_HEADER_LIMIT: ScanLimitConfig = {
|
||
ratio: 0.2,
|
||
minLines: 20,
|
||
maxLines: 70
|
||
}
|
||
|
||
const DEFAULT_FOOTER_LIMIT: ScanLimitConfig = {
|
||
ratio: 0.2,
|
||
minLines: 20,
|
||
maxLines: 50
|
||
}
|
||
|
||
function calculateScanLimit(config: ScanLimitConfig, totalLines: number): number {
|
||
const proportional = Math.ceil(totalLines * config.ratio)
|
||
|
||
return Math.min(Math.min(Math.max(proportional, config.minLines), config.maxLines), totalLines)
|
||
}
|
||
|
||
function getLineText(line: LyricLine): string {
|
||
if (!line || !line.words) return ''
|
||
return line.words
|
||
.map((w) => w.word)
|
||
.join('')
|
||
.trim()
|
||
}
|
||
|
||
/**
|
||
* 清理文本以便检查
|
||
*
|
||
* 移除行首尾的括号，例如 (作曲: xxx) -> 作曲: xxx
|
||
*/
|
||
function cleanTextForCheck(text: string): string {
|
||
let processed = text.trim()
|
||
const brackets = [
|
||
['(', ')'],
|
||
['（', '）'],
|
||
['【', '】'],
|
||
['[', ']'],
|
||
['{', '}'],
|
||
['『', '』'],
|
||
['「', '」']
|
||
]
|
||
|
||
let changed = true
|
||
let loopCount = 0
|
||
|
||
while (changed && loopCount < 5) {
|
||
changed = false
|
||
loopCount++
|
||
|
||
for (const [open, close] of brackets) {
|
||
if (processed.startsWith(open)) {
|
||
// 类似于这样的格式：(作曲：周杰伦)
|
||
if (processed.endsWith(close)) {
|
||
processed = processed.slice(open.length, processed.length - close.length).trim()
|
||
changed = true
|
||
break
|
||
}
|
||
|
||
// 类似于这样的格式：(Live) 作曲：周杰伦
|
||
const closeIdx = processed.indexOf(close)
|
||
if (closeIdx > -1) {
|
||
const contentAfter = processed.slice(closeIdx + close.length).trim()
|
||
|
||
if (contentAfter.length > 0) {
|
||
processed = contentAfter
|
||
changed = true
|
||
break
|
||
}
|
||
}
|
||
}
|
||
}
|
||
}
|
||
return processed
|
||
}
|
||
|
||
// 强匹配行：匹配关键词加冒号，或者匹配正则表达式的行
|
||
// 弱匹配行：带有冒号，但不匹配关键词或正则表达式的行。如果夹在强匹配行之间，多半是元数据行但是没有对应的规则。但也有可能是演唱者标识，“男：...”这样的，为了避免误删，如果后面全是弱匹配行，就不删它们
|
||
// 真正的歌词行：既不匹配规则，又没有冒号的行，作为防火墙来阻止对之后行的移除。避免元数据在歌词中间，把中间的歌词也移除了
|
||
|
||
function isStrictMatch(text: string, keywords: string[], regexes: RegExp[]): boolean {
|
||
const cleaned = cleanTextForCheck(text)
|
||
|
||
// 转小写并移除空格进行匹配
|
||
const normalizedText = cleaned.toLowerCase().replace(/\s+/g, '')
|
||
|
||
for (const kw of keywords) {
|
||
const normalizedKw = kw.toLowerCase().replace(/\s+/g, '')
|
||
|
||
if (normalizedText.startsWith(normalizedKw)) {
|
||
const remainder = normalizedText.slice(normalizedKw.length)
|
||
|
||
// 如果剩余部分为空，说明完全匹配
|
||
if (remainder.length === 0) {
|
||
return true
|
||
}
|
||
|
||
// 检查分隔符
|
||
// 允许的分隔符包括：冒号、逗号、句号、感叹号、连字符、括号等
|
||
// 注意：normalizedText 已经移除了空格
|
||
if (STRICT_MATCH_SEPARATORS.includes(remainder.charAt(0))) {
|
||
return true
|
||
}
|
||
}
|
||
}
|
||
|
||
for (const reg of regexes) {
|
||
if (reg.test(text)) {
|
||
return true
|
||
}
|
||
}
|
||
|
||
return false
|
||
}
|
||
|
||
function looksLikeMetadata(text: string, softRegexes: RegExp[]): boolean {
|
||
const cleaned = cleanTextForCheck(text)
|
||
if (
|
||
cleaned.includes(':') ||
|
||
cleaned.includes('：') ||
|
||
// 第一行的 歌曲名 - 歌手名 这样的格式
|
||
cleaned.includes('-')
|
||
) {
|
||
return true
|
||
}
|
||
|
||
for (const reg of softRegexes) {
|
||
if (reg.test(text)) {
|
||
return true
|
||
}
|
||
}
|
||
|
||
return false
|
||
}
|
||
|
||
/**
|
||
* 扫描头部，寻找正文开始的位置
|
||
*/
|
||
function findHeaderCutoff(
|
||
lines: readonly LyricLine[],
|
||
startIndex: number,
|
||
keywords: string[],
|
||
regexes: RegExp[],
|
||
softRegexes: RegExp[],
|
||
limit: number
|
||
): number {
|
||
let lastValidMetadataIndex = startIndex - 1
|
||
|
||
console.groupCollapsed(`[LyricStripper] ⬇️ 开始头部扫描 (Start: ${startIndex}, Limit: ${limit})`)
|
||
|
||
for (let i = startIndex; i < limit; i++) {
|
||
if (i >= lines.length) break
|
||
|
||
const text = getLineText(lines[i])
|
||
|
||
if (!text) {
|
||
continue
|
||
}
|
||
|
||
const strict = isStrictMatch(text, keywords, regexes)
|
||
const weak = looksLikeMetadata(text, softRegexes)
|
||
|
||
let status = '❌ NONE'
|
||
if (strict) status = '✅ STRICT'
|
||
else if (weak) status = '⚠️ WEAK'
|
||
|
||
console.log(`Line [${i}]: "${text}" | Result: ${status}`)
|
||
|
||
if (!strict && !weak) {
|
||
break
|
||
}
|
||
|
||
if (strict) {
|
||
lastValidMetadataIndex = i
|
||
}
|
||
}
|
||
console.groupEnd()
|
||
|
||
return lastValidMetadataIndex + 1
|
||
}
|
||
|
||
/**
|
||
* 扫描尾部，寻找正文结束的位置
|
||
*/
|
||
function findFooterCutoff(
|
||
lines: readonly LyricLine[],
|
||
startIndex: number,
|
||
keywords: string[],
|
||
regexes: RegExp[],
|
||
softRegexes: RegExp[],
|
||
limit: number
|
||
): number {
|
||
if (startIndex >= lines.length) return startIndex
|
||
|
||
const scanEnd = Math.max(startIndex, lines.length - limit)
|
||
let firstValidFooterIndex = lines.length
|
||
|
||
console.groupCollapsed(`[LyricStripper] ⬆️ 开始尾部扫描 (Limit: ${limit})`)
|
||
|
||
for (let i = lines.length - 1; i >= scanEnd; i--) {
|
||
const text = getLineText(lines[i])
|
||
|
||
if (!text) {
|
||
continue
|
||
}
|
||
|
||
const strict = isStrictMatch(text, keywords, regexes)
|
||
const weak = looksLikeMetadata(text, softRegexes)
|
||
|
||
let status = '❌ NONE'
|
||
if (strict) status = '✅ STRICT'
|
||
else if (weak) status = '⚠️ WEAK'
|
||
|
||
console.log(`Line [${i}]: "${text}" | Result: ${status}`)
|
||
|
||
if (!strict && !weak) {
|
||
break
|
||
}
|
||
|
||
if (strict) {
|
||
firstValidFooterIndex = i
|
||
}
|
||
}
|
||
console.groupEnd()
|
||
|
||
return firstValidFooterIndex
|
||
}
|
||
|
||
/**
|
||
* 剥离歌词中的元数据行
|
||
* @param lines 原始歌词行数组
|
||
* @param options 包含关键词和正则的配置
|
||
* @returns 清理后的新歌词行数组
|
||
*/
|
||
export function stripLyricMetadata(
|
||
lines: readonly LyricLine[],
|
||
options: StripOptions
|
||
): LyricLine[] {
|
||
if (!lines || lines.length === 0) return []
|
||
|
||
let scanStartIndex = 0
|
||
|
||
if (options.matchMetadata) {
|
||
const { title, artists } = options.matchMetadata
|
||
const firstLineText = getLineText(lines[0])
|
||
|
||
if (title && artists && artists.length > 0 && firstLineText) {
|
||
const lowerText = firstLineText.toLowerCase()
|
||
const lowerTitle = title.toLowerCase()
|
||
|
||
if (lowerText.includes(lowerTitle)) {
|
||
const hasAnyArtist = artists.some((artist) => lowerText.includes(artist.toLowerCase()))
|
||
|
||
if (hasAnyArtist) {
|
||
console.log(
|
||
`[LyricStripper] 在第一行匹配到歌曲元数据: "${firstLineText}" (Title: ${title}, Artists: ${artists.join(', ')})`
|
||
)
|
||
scanStartIndex = 1
|
||
}
|
||
}
|
||
}
|
||
}
|
||
|
||
if (
|
||
scanStartIndex === 0 &&
|
||
(!options.keywords || options.keywords.length === 0) &&
|
||
(!options.regexPatterns || options.regexPatterns.length === 0) &&
|
||
(!options.softMatchRegexes || options.softMatchRegexes.length === 0)
|
||
) {
|
||
return [...lines]
|
||
}
|
||
|
||
const regexes: RegExp[] = []
|
||
if (options.regexPatterns) {
|
||
options.regexPatterns.forEach((p) => {
|
||
try {
|
||
if (p.trim()) {
|
||
regexes.push(new RegExp(p, 'i')) // 忽略大小写
|
||
}
|
||
} catch (e) {
|
||
console.warn(`[LyricStripper] 无效的正则表达式: ${p}`, e)
|
||
}
|
||
})
|
||
}
|
||
|
||
const softRegexes: RegExp[] = []
|
||
if (options.softMatchRegexes) {
|
||
options.softMatchRegexes.forEach((p) => {
|
||
try {
|
||
if (p.trim()) softRegexes.push(new RegExp(p, 'i'))
|
||
} catch (e) {
|
||
console.warn(`[LyricStripper] 无效的正则表达式: ${p}`, e)
|
||
}
|
||
})
|
||
}
|
||
|
||
const keywords = options.keywords || []
|
||
const totalLines = lines.length
|
||
|
||
const headerConfig = DEFAULT_HEADER_LIMIT
|
||
const headerLimit = calculateScanLimit(headerConfig, totalLines)
|
||
|
||
const startIdx = findHeaderCutoff(
|
||
lines,
|
||
scanStartIndex,
|
||
keywords,
|
||
regexes,
|
||
softRegexes,
|
||
headerLimit
|
||
)
|
||
|
||
const footerConfig = DEFAULT_FOOTER_LIMIT
|
||
const footerLimit = calculateScanLimit(footerConfig, totalLines)
|
||
|
||
const endIdx = findFooterCutoff(lines, startIdx, keywords, regexes, softRegexes, footerLimit)
|
||
|
||
if (startIdx === 0 && endIdx === lines.length) {
|
||
return [...lines]
|
||
}
|
||
|
||
const newLength = endIdx - startIdx
|
||
console.log(`[LyricStripper] 清理完成，总行数从 ${totalLines} 变为 ${newLength}`)
|
||
|
||
return lines.slice(startIdx, endIdx)
|
||
}
|