tracker

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