import { get } from 'svelte/store'; import type { Media } from '../../../../Data/AniList/media'; import settings from '$stores/settings'; import type { SubsPlease } from '$lib/Media/Anime/Airing/Subtitled/subsPlease'; import stringSimilarity from 'string-similarity'; import excludeMatch from '$lib/Data/Static/matchExclude.json'; import { season } from '../../season'; export interface Time { title: string; time: string; day: string; } interface IndexedTime { time: Time; normalizedTitle: string; tokens: string[]; } interface DayScheduleIndex { entries: IndexedTime[]; exactTitleIndex: Map; tokenIndex: Map; } interface ScheduleIndex { byDay: Map; version: string; } const secondsUntil = (targetTime: string, targetDay: string) => { const now = new Date(); const [targetHour, targetMinute] = targetTime.split(':').map(Number); let dayDifference = ['Sunday', 'Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday'].indexOf( targetDay ) - now.getDay(); if (dayDifference < 0) dayDifference += 7; const targetDate = new Date(now); targetDate.setDate(now.getDate() + dayDifference); targetDate.setHours(targetHour, targetMinute, 0, 0); const secondsDifference = (Number(targetDate) - Number(now)) / 1000; return secondsDifference > 0 ? secondsDifference : secondsDifference + 7 * 24 * 60 * 60; }; const preprocessTitle = (title: string): string => { return title .toLowerCase() .replace(/\b(season|s|part|cour)\b/g, ' ') .replace(/[^a-z0-9\s]/gi, '') .trim() .split(/\s+/) .join(' '); }; const NON_DISTINCTIVE_TOKENS = new Set([ 'a', 'and', 'de', 'e', 'for', 'ga', 'in', 'na', 'ni', 'no', 'o', 'of', 'on', 'the', 'to', 'wa', 'wo' ]); const isMeaningfulToken = (token: string): boolean => /^\d+$/.test(token) || (token.length >= 3 && !NON_DISTINCTIVE_TOKENS.has(token)); const MIN_MATCH_SCORE = 0.3; const MIN_TOKEN_OVERLAP = 2; const MIN_MATCH_MARGIN = 0.08; const FALLBACK_MIN_SCORE = 0.82; const FALLBACK_MIN_MARGIN = 0.08; const MAX_MATCH_CACHE_ENTRIES = 10_000; const MAX_INJECT_CACHE_ENTRIES = 10_000; const STALE_AIRING_GRACE_SECONDS = 5 * 60; const MAX_EPISODE_SHIFT_WINDOW_SECONDS = 36 * 60 * 60; interface SimilarityAnalysis { score: number; tokenOverlap: number; numericTokenOverlap: number; } const calculateWeightedSimilarity = (title1: string, title2: string): SimilarityAnalysis => { const tokens1 = title1.split(' ').filter(isMeaningfulToken); const tokens2 = title2.split(' ').filter(isMeaningfulToken); if (tokens1.length === 0 || tokens2.length === 0) return { score: 0, tokenOverlap: 0, numericTokenOverlap: 0 }; const set2 = new Set(tokens2); let score = 0; let tokenOverlap = 0; let numericTokenOverlap = 0; const numericTokens1 = tokens1.filter((token) => /^\d+$/.test(token)); const numericTokens2 = tokens2.filter((token) => /^\d+$/.test(token)); tokens1.forEach((token) => { if (set2.has(token)) { tokenOverlap += 1; if (/^\d+$/.test(token)) numericTokenOverlap += 1; score += /^\d+$/.test(token) ? 2 : 1; } }); let finalScore = (score / ((Math.max(tokens1.length, tokens2.length) || 1) * 2)) * 0.7 + stringSimilarity.compareTwoStrings(title1, title2) * 0.3; if (numericTokens1.length > 0 && numericTokens2.length > 0 && numericTokenOverlap === 0) finalScore *= 0.5; return { score: finalScore, tokenOverlap, numericTokenOverlap }; }; const indexPush = (index: Map, key: string, entryIndex: number) => { const existing = index.get(key); if (existing) existing.push(entryIndex); else index.set(key, [entryIndex]); }; const scheduleIndexCache = new WeakMap(); const closestMatchCache = new Map(); const injectAiringTimeCache = new Map(); const setBoundedCacheValue = ( cache: Map, key: string, value: T, maxEntries: number ) => { if (cache.size >= maxEntries) cache.clear(); cache.set(key, value); }; const animeTitleFingerprint = (anime: Media) => [anime.title.romaji, anime.title.english, ...anime.synonyms] .filter(Boolean) .map(preprocessTitle) .join('|'); const fallbackClosestMatch = (dayIndex: DayScheduleIndex, searchTitles: string[]): Time | null => { let bestMatch: Time | null = null; let bestScore = 0; let secondBestScore = 0; for (const searchTitle of searchTitles) { if (searchTitle.includes('OVA') || searchTitle.includes('Special')) continue; const normalizedSearchTitle = preprocessTitle(searchTitle); for (const candidateEntry of dayIndex.entries) { const score = stringSimilarity.compareTwoStrings( normalizedSearchTitle, candidateEntry.normalizedTitle ); if (score > bestScore) { secondBestScore = bestScore; bestScore = score; bestMatch = candidateEntry.time; } else if (score > secondBestScore) { secondBestScore = score; } } } if (bestScore < FALLBACK_MIN_SCORE) return null; if (bestScore - secondBestScore < FALLBACK_MIN_MARGIN) return null; return bestMatch; }; const buildScheduleIndex = (subsPlease: SubsPlease): ScheduleIndex => { const byDay = new Map(); const versionParts: string[] = []; for (const [day, value] of Object.entries(subsPlease.schedule)) { const flattenedValue = Array.isArray(value) ? value.flat() : []; const firstEntry = flattenedValue[0]; const lastEntry = flattenedValue[flattenedValue.length - 1]; versionParts.push( `${day}:${flattenedValue.length}:${firstEntry?.title || ''}:${firstEntry?.time || ''}:${ lastEntry?.title || '' }:${lastEntry?.time || ''}` ); const dayIndex: DayScheduleIndex = { entries: [], exactTitleIndex: new Map(), tokenIndex: new Map() }; for (const scheduleTime of flattenedValue) { const time = { title: scheduleTime.title, time: scheduleTime.time, day }; const normalizedTitle = preprocessTitle(time.title); const tokens = normalizedTitle.split(' ').filter(isMeaningfulToken); const entryIndex = dayIndex.entries.length; dayIndex.entries.push({ time, normalizedTitle, tokens }); indexPush(dayIndex.exactTitleIndex, normalizedTitle, entryIndex); for (const token of tokens) indexPush(dayIndex.tokenIndex, token, entryIndex); } byDay.set(day, dayIndex); } return { byDay, version: `${subsPlease.tz}:${versionParts.join('|')}` }; }; export const findClosestMatch = (scheduleIndex: ScheduleIndex, anime: Media): Time | null => { if (excludeMatch.includes(anime.id)) { setBoundedCacheValue(closestMatchCache, `${anime.id}:excluded`, null, MAX_MATCH_CACHE_ENTRIES); return null; } const cacheKey = `${anime.id}:${anime.nextAiringEpisode?.airingAt || 0}:${animeTitleFingerprint( anime )}:${scheduleIndex.version}`; const cached = closestMatchCache.get(cacheKey); if (cached !== undefined) return cached; const airingDay = new Date((anime.nextAiringEpisode?.airingAt || 0) * 1000).toLocaleString( 'en-US', { weekday: 'long' } ); const dayIndex = scheduleIndex.byDay.get(airingDay); if (!dayIndex || dayIndex.entries.length === 0) { setBoundedCacheValue(closestMatchCache, cacheKey, null, MAX_MATCH_CACHE_ENTRIES); return null; } let bestMatch: Time | null = null; let bestScore = 0; let secondBestScore = 0; let bestTokenOverlap = 0; let bestNumericTokenOverlap = 0; const searchTitles = [anime.title.romaji, anime.title.english, ...anime.synonyms].filter(Boolean); for (const searchTitle of searchTitles) { if (searchTitle.includes('OVA') || searchTitle.includes('Special')) continue; const normalizedSearchTitle = preprocessTitle(searchTitle); const exactMatchIndexes = dayIndex.exactTitleIndex.get(normalizedSearchTitle); if (exactMatchIndexes && exactMatchIndexes.length > 0) { const exactMatch = dayIndex.entries[exactMatchIndexes[0]]; if (exactMatch) { setBoundedCacheValue(closestMatchCache, cacheKey, exactMatch.time, MAX_MATCH_CACHE_ENTRIES); return exactMatch.time; } } const searchTokens = normalizedSearchTitle.split(' ').filter(isMeaningfulToken); const candidateIndexSet = new Set(); for (const token of searchTokens) { for (const candidateIndex of dayIndex.tokenIndex.get(token) || []) candidateIndexSet.add(candidateIndex); } const candidateIndexes = candidateIndexSet.size > 0 ? [...candidateIndexSet] : dayIndex.entries.map((_, entryIndex) => entryIndex); for (const candidateIndex of candidateIndexes) { const candidateEntry = dayIndex.entries[candidateIndex]; const similarity = calculateWeightedSimilarity( normalizedSearchTitle, candidateEntry.normalizedTitle ); if (similarity.score > bestScore) { secondBestScore = bestScore; bestScore = similarity.score; bestTokenOverlap = similarity.tokenOverlap; bestNumericTokenOverlap = similarity.numericTokenOverlap; bestMatch = candidateEntry.time; } else if (similarity.score > secondBestScore) { secondBestScore = similarity.score; } } } if (bestScore < MIN_MATCH_SCORE) { const fallbackMatch = fallbackClosestMatch(dayIndex, searchTitles); setBoundedCacheValue(closestMatchCache, cacheKey, fallbackMatch, MAX_MATCH_CACHE_ENTRIES); return fallbackMatch; } if (bestScore - secondBestScore < MIN_MATCH_MARGIN) { const fallbackMatch = fallbackClosestMatch(dayIndex, searchTitles); setBoundedCacheValue(closestMatchCache, cacheKey, fallbackMatch, MAX_MATCH_CACHE_ENTRIES); return fallbackMatch; } if (bestNumericTokenOverlap === 0 && bestTokenOverlap < MIN_TOKEN_OVERLAP) { const fallbackMatch = fallbackClosestMatch(dayIndex, searchTitles); setBoundedCacheValue(closestMatchCache, cacheKey, fallbackMatch, MAX_MATCH_CACHE_ENTRIES); return fallbackMatch; } setBoundedCacheValue(closestMatchCache, cacheKey, bestMatch, MAX_MATCH_CACHE_ENTRIES); return bestMatch; }; const normalizeTitle = (title: string | null) => (title || '') .toLowerCase() .replace(/\b(s|season|part|cour)\s*\d+/g, '') .replace(/[\W_]+/g, ' ') .trim(); const findClosestMediaCache = new Map(); export const findClosestMedia = (media: Media[], matchFor: string) => { if (!matchFor) return null; const cached = findClosestMediaCache.get(matchFor); if (cached !== undefined) return cached; const normalisedMatchFor = normalizeTitle(matchFor); const matchForWords = normalisedMatchFor.split(' '); let bestFitMedia: Media | null = null; let bestDistance = -Infinity; for (const m of media) { const titles = [m.title.romaji, m.title.english, ...m.synonyms].filter(Boolean); if ( titles.some( (title) => title.toLowerCase().includes('special') || title.toLowerCase().includes('ova') ) ) continue; const normalisedTitles = titles.map(normalizeTitle); for (const normalisedTitle of normalisedTitles) { const distance = stringSimilarity.compareTwoStrings(normalisedMatchFor, normalisedTitle); if (distance <= bestDistance) continue; const wordMatch = matchForWords.every((word) => normalisedTitles.some((t) => t.includes(word))) || normalisedTitles.some((t) => t.includes(normalisedMatchFor)); if (wordMatch) { bestDistance = distance; bestFitMedia = m; if (distance === 1) break; } } if (bestDistance === 1) break; } findClosestMediaCache.set(matchFor, bestFitMedia); return bestFitMedia as Media | null; }; export const clearClosestMediaCache = () => findClosestMediaCache.clear(); const getScheduleIndex = (subsPlease: SubsPlease): ScheduleIndex => { const cached = scheduleIndexCache.get(subsPlease); if (cached) return cached; const built = buildScheduleIndex(subsPlease); scheduleIndexCache.set(subsPlease, built); return built; }; const buildInjectAiringTimeCacheKey = ( anime: Media, scheduleVersion: string, displayNativeCountdown: boolean ) => [ anime.id, anime.status, anime.nextAiringEpisode?.episode || 0, anime.nextAiringEpisode?.airingAt || 0, displayNativeCountdown ? 1 : 0, scheduleVersion, animeTitleFingerprint(anime) ].join(':'); export const injectAiringTime = (anime: Media, subsPlease: SubsPlease | null) => { if (season() !== anime.season) return anime; const displayNativeCountdown = get(settings).displayNativeCountdown; const scheduleVersion = subsPlease ? getScheduleIndex(subsPlease).version : 'native-only'; const cacheKey = buildInjectAiringTimeCacheKey(anime, scheduleVersion, displayNativeCountdown); const cached = injectAiringTimeCache.get(cacheKey); if (cached) return cached; const airingAt = anime.nextAiringEpisode?.airingAt; const now = new Date(); // const nativeUntilAiring = airingAt // ? Math.round((airingAt - Date.now() / 1000) * 100) / 100 // : undefined; const nativeTime = new Date(airingAt ? airingAt * 1000 : 0); let untilAiring; let time = new Date(airingAt ? airingAt * 1000 : 0); let nextEpisode = anime.nextAiringEpisode?.episode || 0; if (!(displayNativeCountdown || !subsPlease)) { const scheduleIndex = getScheduleIndex(subsPlease); if ((anime.nextAiringEpisode?.episode || 0) > 1) { const foundTime: Time | null = findClosestMatch(scheduleIndex, anime); if (foundTime) { untilAiring = secondsUntil((foundTime as Time).time, (foundTime as Time).day); time = new Date(Date.now() + untilAiring * 1000); } } } const SEVEN_DAYS = 7 * 24 * 60 * 60 * 1000; const nowEpochSeconds = Date.now() / 1000; const nativeAheadSeconds = nativeTime.getTime() / 1000 - time.getTime() / 1000; if ( nativeAheadSeconds > 0 && nativeAheadSeconds <= MAX_EPISODE_SHIFT_WINDOW_SECONDS && nativeTime.getTime() / 1000 > nowEpochSeconds + STALE_AIRING_GRACE_SECONDS ) { nextEpisode -= 1; } if (nativeTime.getTime() - now.getTime() > SEVEN_DAYS) { const beforeTime = time; time = nativeTime; time.setHours(beforeTime.getHours()); time.setMinutes(beforeTime.getMinutes()); } const injected = { ...anime, nextAiringEpisode: { episode: nextEpisode, airingAt: time.getTime() / 1000, nativeAiringAt: nativeTime.getTime() / 1000, nativeEpisode: anime.nextAiringEpisode?.episode || 0 } } as Media; setBoundedCacheValue(injectAiringTimeCache, cacheKey, injected, MAX_INJECT_CACHE_ENTRIES); return injected; };