tracker

18.5 KB

1. // search.hl — THE SEARCH (tracker.worldapi.org#14): our own database first, then "Fetch from web" (TMDB) for titles
2. // we don't have yet, and the import of the one picked. Used by components/search.hl (`/search/<text>`).
3. //
4. // dbSearch(text) our shows, movies and people — from an IN-MEMORY INDEX built once at boot (no table scan per
5. // keystroke): every title/name split into words; each word's first 2 and first 3 characters are
6. // a bucket key → the entries holding such a word. A query looks at ONE bucket (its longest word's
7. // first 3 characters, or 2 for a 2-letter word) and keeps the entries where EVERY query word is
8. // the start of one of the entry's words. A title's year is one of its words ("doctor who 2005").
9. // webSearch(text) GET <TMDB_BASE_URL>/search/multi?query=… (one request; TV + movies only, people dropped) —
10. // the titles we already have (same TMDB id and kind) are left out
11. // importTitle(kind, id) a new show record from GET /tv/<id>?append_to_response=external_ids,aggregate_credits (a movie:
12. // /movie/<id>?…,credits), then tmdbsync.hl `syncShowWith` fills it from that same answer exactly like
13. // the daily sync does (seasons, episodes, poster, external ids); added to the index. The answer goes
14. // back as `details`: the caller (components/search.hl) stores its cast + crew (details.hl applyCredits,
15. // tracker#28 — details.hl imports this file, so it cannot be called from here)
16. //
17. // ADULT TITLES (mission 054): our database lists only titles with `adult == false` (shows.hl isPublicTitle; unknown is
18. // hidden) — `titleOk` holds that per show id, set by indexShow and by `searchTitleChanged` (the adult backfill and the daily
19. // sync call it per title). The web list asks TMDB with include_adult=false AND keeps only results with `adult` false.
20. //
21. // ONE PROCESS PER STORAGE (hybriel#21): the index lives in the app's process and is kept up to date by importTitle; the
22. // daily sync never changes a title. New rows get self-assigned 16-hex ids (hybriel#113), like tmdbsync.hl.
23. import { now } from 'hl:time'
24. import { randomBytes } from 'hl:crypto'
25. import { showsTable, personsTable, genresTable, isPublicTitle } from './shows.hl'
26. import { tmdbGet, tmdbImageBase, syncShowWith, persistSync, textOr, adultOf } from './tmdbsync.hl'
27. import { refreshCatalogShow } from './catalog.hl'
28. import { claimShortId } from './shortids.hl'
30. // ---- text → words ---------------------------------------------------------------------------------
31. // lower case, apostrophes dropped ("Grey's" → "greys"), every other punctuation mark a word break, the common Latin
32. // accents folded ("Amélie" → "amelie") — the same for the titles and the query
33. static apostrophes = ["'", '’', '`', '´']
34. static breaks = ['.', ',', ':', ';', '!', '?', '(', ')', '[', ']', '{', '}', '"', '“', '”', '„', '-', '–', '—', '/', '&', '+', '*', '#', '_', '|', '…', '·', '@', '=', '~', '<', '>']
35. static folds = [['á', 'a'], ['à', 'a'], ['â', 'a'], ['ä', 'a'], ['ã', 'a'], ['å', 'a'], ['æ', 'ae'], ['ç', 'c'], ['é', 'e'], ['è', 'e'], ['ê', 'e'], ['ë', 'e'],
36. ['í', 'i'], ['ì', 'i'], ['î', 'i'], ['ï', 'i'], ['ñ', 'n'], ['ó', 'o'], ['ò', 'o'], ['ô', 'o'], ['ö', 'o'], ['õ', 'o'], ['ø', 'o'], ['œ', 'oe'],
37. ['ú', 'u'], ['ù', 'u'], ['û', 'u'], ['ü', 'u'], ['ý', 'y'], ['ÿ', 'y'], ['ß', 'ss'], ['š', 's'], ['ž', 'z'], ['č', 'c'], ['ł', 'l']]
39. static normalize = (s) => {
40. if (s == null) { return '' }
41. let t = ('' + s).toLowerCase()
42. for (p of apostrophes) { if (t.includes(p)) { t = t.replaceAll(p, '') } }
43. for (p of breaks) { if (t.includes(p)) { t = t.replaceAll(p, ' ') } }
44. for (f of folds) { if (t.includes(f[0])) { t = t.replaceAll(f[0], f[1]) } }
45. return t
46. }
47. static wordsOf = (s) => {
48. let out = []
49. for (w of normalize(s).split(' ')) { if (w != '') { out.push(w) } }
50. return out
51. }
53. // ---- the index --------------------------------------------------------------------------------------
54. // entries[i] = { kind ('title' | 'person'), id, words, text (the words joined by one space, without the year) }
55. static entries = []
56. static buckets = {}
57. // what we have, by TMDB id: 'series:<id>' / 'movie:<id>' → our show id (the web list leaves these out)
58. static haveTmdb = {}
59. // every show slug in use (an imported title gets a free one)
60. static slugTaken = {}
61. // show id → true when the title may be listed (not adult, mission 054)
62. static titleOk = {}
63. static indexStats = { built = false titles = 0 people = 0 keys = 0 ms = 0 }
65. static bucketKeys = (words) => {
66. let seen = {}
67. let out = []
68. for (w of words) {
69. let n = 2
70. while (n <= 3) {
71. if (w.length >= n) {
72. let k = w.slice(0, n)
73. if (seen[k] != true) { seen[k] = true out.push(k) }
74. }
75. n = n + 1
76. }
77. }
78. return out
79. }
81. // `tie` orders equally good matches: lower first (a title: series before movies, then the newer one)
82. static addEntry = (kind, id, name, extra, tie) => {
83. let words = wordsOf(name)
84. if (words.length == 0) { return false }
85. let i = entries.length
86. let all = []
87. for (w of words) { all.push(w) }
88. if (extra != null && extra != '') { all.push(extra) }
89. entries.push({ kind = kind id = id words = all text = words.join(' ') tie = tie })
90. for (k of bucketKeys(all)) {
91. if (buckets[k] == null) { buckets[k] = [i] indexStats.keys = indexStats.keys + 1 } else { buckets[k].push(i) }
92. }
93. return true
94. }
96. static tmdbKeyOf = (type, tmdbId) => { return (type == 'movie' ? 'movie:' : 'series:') + tmdbId }
98. static yearWordOf = (show) => {
99. if (show.year != null) { return '' + show.year }
100. if (show.release != null && hlTypeName(show.release) == 'String' && show.release.length >= 4) { return show.release.slice(0, 4) }
101. return ''
102. }
104. static indexShow = (s) => {
105. if (s.urlSegment != null) { slugTaken[s.urlSegment] = true }
106. if (s.tmdbId != null) { haveTmdb[tmdbKeyOf(s.type, s.tmdbId)] = s.id }
107. titleOk[s.id] = isPublicTitle(s)
108. let y = toNumber(yearWordOf(s))
109. if (addEntry('title', s.id, s.title, yearWordOf(s), (s.type == 'movie' ? 10000 : 0) + (y != null ? 3000 - y : 2999))) { indexStats.titles = indexStats.titles + 1 }
110. return null
111. }
113. // tracker#26 (details.hl): a person added from a title's cast — findable by the search at once
114. static indexPerson = (p) => {
115. ensureIndex()
116. if (addEntry('person', p.id, p.name, '', 0)) { indexStats.people = indexStats.people + 1 }
117. return null
118. }
120. // a title's adult flag changed (the backfill / the daily sync): listed from now on, or not
121. static searchTitleChanged = (showId) => {
122. let s = showsTable.fetch(showId)
123. if (s != null) { titleOk[s.id] = isPublicTitle(s) }
124. return null
125. }
127. // built once: at boot (project.hl calls it), else by the first search
128. static ensureIndex = () => {
129. if (indexStats.built) { return true }
130. indexStats.built = true
131. let t0 = now()
132. for (s of showsTable.find(null, null)) { indexShow(s) }
133. for (p of personsTable.find(null, null)) {
134. if (addEntry('person', p.id, p.name, '', 0)) { indexStats.people = indexStats.people + 1 }
135. }
136. indexStats.ms = now() - t0
137. console.log('search index: ' + indexStats.titles + ' titles, ' + indexStats.people + ' people, ' + indexStats.keys + ' keys, ' + indexStats.ms + ' ms')
138. return true
139. }
141. // tracker.worldapi.org#16 (people.hl): our show id for a TMDB title — `type` 'series' | 'movie' — or null
142. static titleIdByTmdb = (type, tmdbId) => {
143. ensureIndex()
144. return haveTmdb[tmdbKeyOf(type, tmdbId)]
145. }
147. // ---- the query as it comes in the address ------------------------------------------------------------
148. // `/search/<text>`: a full page load hands the page the path segment already DECODED by the server (measured: `%2F` splits
149. // the segment → 404, so the page never puts one in the address), a client-side navigation the RAW one — decoded here.
150. // decodeURIComponent stops the request on a malformed one ('%zz', a broken UTF-8 run), so every %-run is checked first; a
151. // bad one is taken literally. (A text that itself looks like an escape, '%41', is decoded twice on a reload — accepted.)
152. static hexDigits = '0123456789abcdef'
153. static hexAt = (s, i) => {
154. if (i + 3 > s.length) { return -1 }
155. let a = hexDigits.indexOf(s.slice(i + 1, i + 2).toLowerCase())
156. let b = hexDigits.indexOf(s.slice(i + 2, i + 3).toLowerCase())
157. if (a < 0 || b < 0) { return -1 }
158. return a * 16 + b
159. }
160. // the bytes of one run of %XX are valid UTF-8 (the rules decodeURIComponent applies)
161. static utf8Ok = (bytes) => {
162. let i = 0
163. while (i < bytes.length) {
164. let b = bytes[i]
165. let need = 0
166. let lo = 128
167. let hi = 191
168. if (b < 128) { need = 0 } else if (b >= 194 && b <= 223) { need = 1 } else if (b >= 224 && b <= 239) {
169. need = 2
170. if (b == 224) { lo = 160 }
171. if (b == 237) { hi = 159 }
172. } else if (b >= 240 && b <= 244) {
173. need = 3
174. if (b == 240) { lo = 144 }
175. if (b == 244) { hi = 143 }
176. } else { return false }
177. if (i + need > bytes.length - 1) { return false }
178. let k = 1
179. while (k <= need) {
180. let c = bytes[i + k]
181. if (k == 1 && (c < lo || c > hi)) { return false }
182. if (k > 1 && (c < 128 || c > 191)) { return false }
183. k = k + 1
184. }
185. i = i + need + 1
186. }
187. return true
188. }
189. static decodableUri = (s) => {
190. let i = 0
191. while (i < s.length) {
192. if (s.slice(i, i + 1) == '%') {
193. let run = []
194. while (i < s.length && s.slice(i, i + 1) == '%') {
195. let v = hexAt(s, i)
196. if (v < 0) { return false }
197. run.push(v)
198. i = i + 3
199. }
200. if (!utf8Ok(run)) { return false }
201. } else {
202. i = i + 1
203. }
204. }
205. return true
206. }
207. static maxQuery = 100
208. static queryOfParam = (raw) => {
209. if (raw == null || hlTypeName(raw) != 'String' || raw == '') { return '' }
210. let t = raw.includes('%') && decodableUri(raw) ? decodeURIComponent(raw) : raw
211. return t.length > maxQuery ? t.slice(0, maxQuery) : t
212. }
214. // ---- our database ------------------------------------------------------------------------------------
215. static titleLimit = 30
216. static peopleLimit = 10
218. // lower is better: the whole name (0), the name starts with the query (1), its first word does (2), anything else (3);
219. // then the shorter name
220. static scoreOf = (e, qw, qtext) => {
221. if (e.text == qtext) { return 0 }
222. if (e.text.startsWith(qtext)) { return 1 }
223. if (e.words[0].startsWith(qw[0])) { return 2 }
224. return 3
225. }
226. static matchesAll = (e, qw) => {
227. for (q of qw) {
228. let hit = false
229. for (w of e.words) { if (!hit && w.startsWith(q)) { hit = true } }
230. if (!hit) { return false }
231. }
232. return true
233. }
234. // keeps the best `limit` of { score, len, tie, i } in order (hand-written: no list sort(), hybriel#1)
235. static insertTop = (top, item, limit) => {
236. let out = []
237. let inserted = false
238. for (x of top) {
239. if (!inserted && (item.score < x.score || (item.score == x.score && (item.len < x.len || (item.len == x.len && item.tie < x.tie))))) { out.push(item) inserted = true }
240. if (out.length < limit) { out.push(x) }
241. }
242. if (!inserted && out.length < limit) { out.push(item) }
243. return out
244. }
246. // { text, words, titles: [{ id }], people: [{ id }], titleCount, peopleCount, tooShort }
247. static dbSearch = (text) => {
248. ensureIndex()
249. let out = { text = text titles = [] people = [] titleCount = 0 peopleCount = 0 tooShort = false ms = 0 }
250. let t0 = now()
251. let qw = wordsOf(text)
252. let longest = ''
253. for (w of qw) { if (w.length > longest.length) { longest = w } }
254. if (longest.length < 2) { out.tooShort = true return out }
255. let key = longest.slice(0, longest.length >= 3 ? 3 : 2)
256. let cand = buckets[key]
257. if (cand == null) { return out }
258. let qtext = qw.join(' ')
259. let titles = []
260. let people = []
261. for (i of cand) {
262. let e = entries[i]
263. if ((e.kind != 'title' || titleOk[e.id] == true) && matchesAll(e, qw)) {
264. let item = { score = scoreOf(e, qw, qtext) len = e.text.length tie = e.tie i = i }
265. if (e.kind == 'title') {
266. out.titleCount = out.titleCount + 1
267. titles = insertTop(titles, item, titleLimit)
268. } else {
269. out.peopleCount = out.peopleCount + 1
270. people = insertTop(people, item, peopleLimit)
271. }
272. }
273. }
274. for (t of titles) { out.titles.push({ id = entries[t.i].id }) }
275. for (p of people) { out.people.push({ id = entries[p.i].id }) }
276. out.ms = now() - t0
277. return out
278. }
280. // ---- the web (TMDB) -----------------------------------------------------------------------------------
281. static yearOfDate = (d) => { return textOr(d) != null && d.length >= 4 ? d.slice(0, 4) : '' }
283. // TMDB's search: [{ kind ('tv' | 'movie'), tmdbId, title, year, overview, posterUrl }] of what we DON'T have, or
284. // { failed }. One request (search/multi, page 1 = up to 20 results); people and titles we have are left out.
285. static webSearch = (text) => {
286. ensureIndex()
287. let qw = wordsOf(text)
288. if (text == null || text.trim().length < 2 || qw.length == 0) { return { rows = [] left = 0 failed = null } }
289. let j = tmdbGet('/search/multi?query=' + encodeURIComponent(text.trim()) + '&include_adult=false&page=1')
290. if (j.failed != null) { return { rows = [] left = 0 failed = j.failed } }
291. let rows = []
292. let left = 0
293. if (j.results != null) {
294. for (r of j.results) {
295. let kind = r.media_type
296. // mission 054: only what TMDB says is not adult (include_adult=false already, and the result's own flag)
297. if ((kind == 'tv' || kind == 'movie') && r.id != null && hlTypeName(r.id) == 'Number' && adultOf(r) == false) {
298. if (haveTmdb[tmdbKeyOf(kind == 'tv' ? 'series' : 'movie', r.id)] != null) {
299. left = left + 1
300. } else {
301. let title = kind == 'tv' ? textOr(r.name) : textOr(r.title)
302. let year = yearOfDate(kind == 'tv' ? r.first_air_date : r.release_date)
303. if (title != null) {
304. rows.push({ kind = kind tmdbId = r.id title = title year = year
305. overview = textOr(r.overview) != null ? r.overview : ''
306. posterUrl = textOr(r.poster_path) != null ? tmdbImageBase + '/w92' + r.poster_path : '/posters/none' })
307. }
308. }
309. }
310. }
311. }
312. return { rows = rows left = left failed = null }
313. }
315. // ---- the import ------------------------------------------------------------------------------------
316. static slugChars = 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789'
317. // "Doctor Who" → "Doctor-Who" (the migrated urlSegment form: letters and digits, everything else one '-'); a slug in use
318. // gets "-2", "-3", … like the migrated ones ("Raised-by-Wolves-2"); a title with no Latin letter at all → "title-<tmdbId>"
319. static slugOf = (title, tmdbId) => {
320. let out = slugBaseOf(title, 'title-' + tmdbId)
321. if (slugTaken[out] != true) { return out }
322. let n = 2
323. while (slugTaken[out + '-' + n] == true) { n = n + 1 }
324. return out + '-' + n
325. }
326. // the slug of a text without the "in use" check — `fallback` when it has no Latin letter at all (people.hl personSlugOf
327. // uses it for new people, tracker#26)
328. static slugBaseOf = (title, fallback) => {
329. let t = normalizeSlugText(title)
330. let out = ''
331. let dash = false
332. let i = 0
333. while (i < t.length) {
334. let c = t.slice(i, i + 1)
335. if (slugChars.includes(c)) {
336. if (dash && out != '') { out = out + '-' }
337. out = out + c
338. dash = false
339. } else {
340. dash = true
341. }
342. i = i + 1
343. }
344. if (out == '') { out = fallback }
345. if (out.length > 80) { out = out.slice(0, 80) }
346. return out
347. }
348. // the accents folded (keeping the case) and apostrophes dropped, so "Grey's Anatomy" → "Greys-Anatomy"
349. static normalizeSlugText = (s) => {
350. let t = '' + s
351. for (p of apostrophes) { if (t.includes(p)) { t = t.replaceAll(p, '') } }
352. for (f of folds) {
353. if (t.includes(f[0])) { t = t.replaceAll(f[0], f[1]) }
354. let up = f[0].toUpperCase()
355. if (up != f[0] && t.includes(up)) { t = t.replaceAll(up, f[1].toUpperCase()) }
356. }
357. return t
358. }
360. // our genres by lower-case name (TMDB's "Drama" → ours); TMDB-only names ("Sci-Fi & Fantasy") are left out
361. static genreIdsOf = (tmdbGenres) => {
362. let out = []
363. if (tmdbGenres == null) { return out }
364. let byName = {}
365. for (g of genresTable.find(null, null)) { if (g.name != null) { byName[('' + g.name).toLowerCase()] = g.id } }
366. for (tg of tmdbGenres) {
367. let id = tg.name != null ? byName[('' + tg.name).toLowerCase()] : null
368. if (id != null && !out.includes(id)) { out.push(id) }
369. }
370. return out
371. }
373. // `kind` 'tv' | 'movie', `tmdbId` a number → { slug, show, created, sync, details } or { error }. A title we already have (same
374. // TMDB id and kind) is not added again: its slug is answered.
375. static importTitle = (kind, tmdbId) => {
376. if (kind != 'tv' && kind != 'movie') { return { error = 'unknown kind' } }
377. // `tid`, not `id`: inside the record literal below an entry named `id` would shadow it for every later entry
378. let tid = toNumber('' + tmdbId)
379. if (tid == null || tid <= 0 || tid % 1 != 0) { return { error = 'bad TMDB id' } }
380. ensureIndex()
381. let type = kind == 'tv' ? 'series' : 'movie'
382. let have = haveTmdb[tmdbKeyOf(type, tid)]
383. if (have != null) {
384. let s = showsTable.fetch(have)
385. if (s != null) { return { slug = s.urlSegment show = s.id created = false } }
386. }
387. let d = tmdbGet('/' + kind + '/' + tid + '?append_to_response=external_ids,' + (kind == 'tv' ? 'aggregate_credits' : 'credits'))
388. if (d.failed != null) { return { error = 'TMDB: ' + d.failed } }
389. let title = kind == 'tv' ? textOr(d.name) : textOr(d.title)
390. if (title == null) { return { error = 'TMDB has no title for ' + kind + ' ' + tid } }
391. let release = textOr(kind == 'tv' ? d.first_air_date : d.release_date)
392. let year = release != null && release.length >= 4 ? toNumber(release.slice(0, 4)) : null
393. let overview = textOr(d.overview) != null ? d.overview : ''
394. let record = {
395. id = randomBytes(8, 'hex') title = title summary = overview tmdbSummary = textOr(d.overview) tmdbId = tid
396. language = textOr(d.original_language) type = type release = release year = year image = null
397. urlSegment = slugOf(title, tid) episodesCount = null homepage = textOr(d.homepage) imdbId = null seasonsCount = null
398. status = textOr(d.status) tagline = textOr(d.tagline) tvdbId = null tvmzId = null genres = genreIdsOf(d.genres)
399. cast = [] seasons = [] oldId = 'tmdb-' + kind + '-' + tid imported = now() adult = adultOf(d)
400. }
401. // mission 060: its short id (shortids.hl — unique across shows and persons)
402. record.shortId = claimShortId('show', record.id, record.oldId)
403. if (showsTable.put(record) == null) { return { error = 'could not store the title: ' + showsTable.lastError() } }
404. indexShow(record)
405. let r = syncShowWith(record.id, d)
406. persistSync()
407. // the public lists (catalog.hl, #13) get it at once, like a show the daily sync touched (merge of #13 + #14)
408. refreshCatalogShow(record.id)
409. console.log('search import: ' + kind + ' ' + tid + ' "' + title + '" → /shows/' + record.urlSegment + ' (newSeasons=' + r.newSeasons + ' newEpisodes=' + r.newEpisodes + ' posters=' + r.posters + ' ids=' + r.ids + ' requests=' + r.requests + ' errors=' + r.errors + (r.error != '' ? ' ' + r.error : '') + ')')
410. return { slug = record.urlSegment show = record.id created = true sync = r details = d }
411. }