tracker

16.0 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. // ONE PROCESS PER STORAGE (hybriel#21): the index lives in the app's process and is kept up to date by importTitle; the
15. // daily sync never changes a title. New rows get self-assigned 16-hex ids (hybriel#113), like tmdbsync.hl.
16. import { now } from 'hl:time'
17. import { randomBytes } from 'hl:crypto'
18. import { showsTable, personsTable, genresTable } from './shows.hl'
19. import { tmdbGet, tmdbImageBase, syncShow, persistSync, textOr } from './tmdbsync.hl'
21. // ---- text → words ---------------------------------------------------------------------------------
22. // lower case, apostrophes dropped ("Grey's" → "greys"), every other punctuation mark a word break, the common Latin
23. // accents folded ("Amélie" → "amelie") — the same for the titles and the query
24. static apostrophes = ["'", '’', '`', '´']
25. static breaks = ['.', ',', ':', ';', '!', '?', '(', ')', '[', ']', '{', '}', '"', '“', '”', '„', '-', '–', '—', '/', '&', '+', '*', '#', '_', '|', '…', '·', '@', '=', '~', '<', '>']
26. 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'],
27. ['í', 'i'], ['ì', 'i'], ['î', 'i'], ['ï', 'i'], ['ñ', 'n'], ['ó', 'o'], ['ò', 'o'], ['ô', 'o'], ['ö', 'o'], ['õ', 'o'], ['ø', 'o'], ['œ', 'oe'],
28. ['ú', 'u'], ['ù', 'u'], ['û', 'u'], ['ü', 'u'], ['ý', 'y'], ['ÿ', 'y'], ['ß', 'ss'], ['š', 's'], ['ž', 'z'], ['č', 'c'], ['ł', 'l']]
30. static normalize = (s) => {
31. if (s == null) { return '' }
32. let t = ('' + s).toLowerCase()
33. for (p of apostrophes) { if (t.includes(p)) { t = t.replaceAll(p, '') } }
34. for (p of breaks) { if (t.includes(p)) { t = t.replaceAll(p, ' ') } }
35. for (f of folds) { if (t.includes(f[0])) { t = t.replaceAll(f[0], f[1]) } }
36. return t
37. }
38. static wordsOf = (s) => {
39. let out = []
40. for (w of normalize(s).split(' ')) { if (w != '') { out.push(w) } }
41. return out
42. }
44. // ---- the index --------------------------------------------------------------------------------------
45. // entries[i] = { kind ('title' | 'person'), id, words, text (the words joined by one space, without the year) }
46. static entries = []
47. static buckets = {}
48. // what we have, by TMDB id: 'series:<id>' / 'movie:<id>' → our show id (the web list leaves these out)
49. static haveTmdb = {}
50. // every show slug in use (an imported title gets a free one)
51. static slugTaken = {}
52. static indexStats = { built = false titles = 0 people = 0 keys = 0 ms = 0 }
54. static bucketKeys = (words) => {
55. let seen = {}
56. let out = []
57. for (w of words) {
58. let n = 2
59. while (n <= 3) {
60. if (w.length >= n) {
61. let k = w.slice(0, n)
62. if (seen[k] != true) { seen[k] = true out.push(k) }
63. }
64. n = n + 1
65. }
66. }
67. return out
68. }
70. // `tie` orders equally good matches: lower first (a title: series before movies, then the newer one)
71. static addEntry = (kind, id, name, extra, tie) => {
72. let words = wordsOf(name)
73. if (words.length == 0) { return false }
74. let i = entries.length
75. let all = []
76. for (w of words) { all.push(w) }
77. if (extra != null && extra != '') { all.push(extra) }
78. entries.push({ kind = kind id = id words = all text = words.join(' ') tie = tie })
79. for (k of bucketKeys(all)) {
80. if (buckets[k] == null) { buckets[k] = [i] indexStats.keys = indexStats.keys + 1 } else { buckets[k].push(i) }
81. }
82. return true
83. }
85. static tmdbKeyOf = (type, tmdbId) => { return (type == 'movie' ? 'movie:' : 'series:') + tmdbId }
87. static yearWordOf = (show) => {
88. if (show.year != null) { return '' + show.year }
89. if (show.release != null && hlTypeName(show.release) == 'String' && show.release.length >= 4) { return show.release.slice(0, 4) }
90. return ''
91. }
93. static indexShow = (s) => {
94. if (s.urlSegment != null) { slugTaken[s.urlSegment] = true }
95. if (s.tmdbId != null) { haveTmdb[tmdbKeyOf(s.type, s.tmdbId)] = s.id }
96. let y = toNumber(yearWordOf(s))
97. if (addEntry('title', s.id, s.title, yearWordOf(s), (s.type == 'movie' ? 10000 : 0) + (y != null ? 3000 - y : 2999))) { indexStats.titles = indexStats.titles + 1 }
98. return null
99. }
101. // built once: at boot (project.hl calls it), else by the first search
102. static ensureIndex = () => {
103. if (indexStats.built) { return true }
104. indexStats.built = true
105. let t0 = now()
106. for (s of showsTable.find(null, null)) { indexShow(s) }
107. for (p of personsTable.find(null, null)) {
108. if (addEntry('person', p.id, p.name, '', 0)) { indexStats.people = indexStats.people + 1 }
109. }
110. indexStats.ms = now() - t0
111. console.log('search index: ' + indexStats.titles + ' titles, ' + indexStats.people + ' people, ' + indexStats.keys + ' keys, ' + indexStats.ms + ' ms')
112. return true
113. }
115. // ---- the query as it comes in the address ------------------------------------------------------------
116. // `/search/<text>`: a full page load hands the page the path segment already DECODED by the server (measured: `%2F` splits
117. // the segment → 404, so the page never puts one in the address), a client-side navigation the RAW one — decoded here.
118. // decodeURIComponent stops the request on a malformed one ('%zz', a broken UTF-8 run), so every %-run is checked first; a
119. // bad one is taken literally. (A text that itself looks like an escape, '%41', is decoded twice on a reload — accepted.)
120. static hexDigits = '0123456789abcdef'
121. static hexAt = (s, i) => {
122. if (i + 3 > s.length) { return -1 }
123. let a = hexDigits.indexOf(s.slice(i + 1, i + 2).toLowerCase())
124. let b = hexDigits.indexOf(s.slice(i + 2, i + 3).toLowerCase())
125. if (a < 0 || b < 0) { return -1 }
126. return a * 16 + b
127. }
128. // the bytes of one run of %XX are valid UTF-8 (the rules decodeURIComponent applies)
129. static utf8Ok = (bytes) => {
130. let i = 0
131. while (i < bytes.length) {
132. let b = bytes[i]
133. let need = 0
134. let lo = 128
135. let hi = 191
136. if (b < 128) { need = 0 } else if (b >= 194 && b <= 223) { need = 1 } else if (b >= 224 && b <= 239) {
137. need = 2
138. if (b == 224) { lo = 160 }
139. if (b == 237) { hi = 159 }
140. } else if (b >= 240 && b <= 244) {
141. need = 3
142. if (b == 240) { lo = 144 }
143. if (b == 244) { hi = 143 }
144. } else { return false }
145. if (i + need > bytes.length - 1) { return false }
146. let k = 1
147. while (k <= need) {
148. let c = bytes[i + k]
149. if (k == 1 && (c < lo || c > hi)) { return false }
150. if (k > 1 && (c < 128 || c > 191)) { return false }
151. k = k + 1
152. }
153. i = i + need + 1
154. }
155. return true
156. }
157. static decodableUri = (s) => {
158. let i = 0
159. while (i < s.length) {
160. if (s.slice(i, i + 1) == '%') {
161. let run = []
162. while (i < s.length && s.slice(i, i + 1) == '%') {
163. let v = hexAt(s, i)
164. if (v < 0) { return false }
165. run.push(v)
166. i = i + 3
167. }
168. if (!utf8Ok(run)) { return false }
169. } else {
170. i = i + 1
171. }
172. }
173. return true
174. }
175. static maxQuery = 100
176. static queryOfParam = (raw) => {
177. if (raw == null || hlTypeName(raw) != 'String' || raw == '') { return '' }
178. let t = raw.includes('%') && decodableUri(raw) ? decodeURIComponent(raw) : raw
179. return t.length > maxQuery ? t.slice(0, maxQuery) : t
180. }
182. // ---- our database ------------------------------------------------------------------------------------
183. static titleLimit = 30
184. static peopleLimit = 10
186. // lower is better: the whole name (0), the name starts with the query (1), its first word does (2), anything else (3);
187. // then the shorter name
188. static scoreOf = (e, qw, qtext) => {
189. if (e.text == qtext) { return 0 }
190. if (e.text.startsWith(qtext)) { return 1 }
191. if (e.words[0].startsWith(qw[0])) { return 2 }
192. return 3
193. }
194. static matchesAll = (e, qw) => {
195. for (q of qw) {
196. let hit = false
197. for (w of e.words) { if (!hit && w.startsWith(q)) { hit = true } }
198. if (!hit) { return false }
199. }
200. return true
201. }
202. // keeps the best `limit` of { score, len, tie, i } in order (hand-written: no list sort(), hybriel#1)
203. static insertTop = (top, item, limit) => {
204. let out = []
205. let inserted = false
206. for (x of top) {
207. 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 }
208. if (out.length < limit) { out.push(x) }
209. }
210. if (!inserted && out.length < limit) { out.push(item) }
211. return out
212. }
214. // { text, words, titles: [{ id }], people: [{ id }], titleCount, peopleCount, tooShort }
215. static dbSearch = (text) => {
216. ensureIndex()
217. let out = { text = text titles = [] people = [] titleCount = 0 peopleCount = 0 tooShort = false ms = 0 }
218. let t0 = now()
219. let qw = wordsOf(text)
220. let longest = ''
221. for (w of qw) { if (w.length > longest.length) { longest = w } }
222. if (longest.length < 2) { out.tooShort = true return out }
223. let key = longest.slice(0, longest.length >= 3 ? 3 : 2)
224. let cand = buckets[key]
225. if (cand == null) { return out }
226. let qtext = qw.join(' ')
227. let titles = []
228. let people = []
229. for (i of cand) {
230. let e = entries[i]
231. if (matchesAll(e, qw)) {
232. let item = { score = scoreOf(e, qw, qtext) len = e.text.length tie = e.tie i = i }
233. if (e.kind == 'title') {
234. out.titleCount = out.titleCount + 1
235. titles = insertTop(titles, item, titleLimit)
236. } else {
237. out.peopleCount = out.peopleCount + 1
238. people = insertTop(people, item, peopleLimit)
239. }
240. }
241. }
242. for (t of titles) { out.titles.push({ id = entries[t.i].id }) }
243. for (p of people) { out.people.push({ id = entries[p.i].id }) }
244. out.ms = now() - t0
245. return out
246. }
248. // ---- the web (TMDB) -----------------------------------------------------------------------------------
249. static yearOfDate = (d) => { return textOr(d) != null && d.length >= 4 ? d.slice(0, 4) : '' }
251. // TMDB's search: [{ kind ('tv' | 'movie'), tmdbId, title, year, overview, posterUrl }] of what we DON'T have, or
252. // { failed }. One request (search/multi, page 1 = up to 20 results); people and titles we have are left out.
253. static webSearch = (text) => {
254. ensureIndex()
255. let qw = wordsOf(text)
256. if (text == null || text.trim().length < 2 || qw.length == 0) { return { rows = [] left = 0 failed = null } }
257. let j = tmdbGet('/search/multi?query=' + encodeURIComponent(text.trim()) + '&include_adult=false&page=1')
258. if (j.failed != null) { return { rows = [] left = 0 failed = j.failed } }
259. let rows = []
260. let left = 0
261. if (j.results != null) {
262. for (r of j.results) {
263. let kind = r.media_type
264. if ((kind == 'tv' || kind == 'movie') && r.id != null && hlTypeName(r.id) == 'Number') {
265. if (haveTmdb[tmdbKeyOf(kind == 'tv' ? 'series' : 'movie', r.id)] != null) {
266. left = left + 1
267. } else {
268. let title = kind == 'tv' ? textOr(r.name) : textOr(r.title)
269. let year = yearOfDate(kind == 'tv' ? r.first_air_date : r.release_date)
270. if (title != null) {
271. rows.push({ kind = kind tmdbId = r.id title = title year = year
272. overview = textOr(r.overview) != null ? r.overview : ''
273. posterUrl = textOr(r.poster_path) != null ? tmdbImageBase + '/w92' + r.poster_path : '/posters/none' })
274. }
275. }
276. }
277. }
278. }
279. return { rows = rows left = left failed = null }
280. }
282. // ---- the import ------------------------------------------------------------------------------------
283. static slugChars = 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789'
284. // "Doctor Who" → "Doctor-Who" (the migrated urlSegment form: letters and digits, everything else one '-'); a slug in use
285. // gets "-2", "-3", … like the migrated ones ("Raised-by-Wolves-2"); a title with no Latin letter at all → "title-<tmdbId>"
286. static slugOf = (title, tmdbId) => {
287. let t = normalizeSlugText(title)
288. let out = ''
289. let dash = false
290. let i = 0
291. while (i < t.length) {
292. let c = t.slice(i, i + 1)
293. if (slugChars.includes(c)) {
294. if (dash && out != '') { out = out + '-' }
295. out = out + c
296. dash = false
297. } else {
298. dash = true
299. }
300. i = i + 1
301. }
302. if (out == '') { out = 'title-' + tmdbId }
303. if (out.length > 80) { out = out.slice(0, 80) }
304. if (slugTaken[out] != true) { return out }
305. let n = 2
306. while (slugTaken[out + '-' + n] == true) { n = n + 1 }
307. return out + '-' + n
308. }
309. // the accents folded (keeping the case) and apostrophes dropped, so "Grey's Anatomy" → "Greys-Anatomy"
310. static normalizeSlugText = (s) => {
311. let t = '' + s
312. for (p of apostrophes) { if (t.includes(p)) { t = t.replaceAll(p, '') } }
313. for (f of folds) {
314. if (t.includes(f[0])) { t = t.replaceAll(f[0], f[1]) }
315. let up = f[0].toUpperCase()
316. if (up != f[0] && t.includes(up)) { t = t.replaceAll(up, f[1].toUpperCase()) }
317. }
318. return t
319. }
321. // our genres by lower-case name (TMDB's "Drama" → ours); TMDB-only names ("Sci-Fi & Fantasy") are left out
322. static genreIdsOf = (tmdbGenres) => {
323. let out = []
324. if (tmdbGenres == null) { return out }
325. let byName = {}
326. for (g of genresTable.find(null, null)) { if (g.name != null) { byName[('' + g.name).toLowerCase()] = g.id } }
327. for (tg of tmdbGenres) {
328. let id = tg.name != null ? byName[('' + tg.name).toLowerCase()] : null
329. if (id != null && !out.includes(id)) { out.push(id) }
330. }
331. return out
332. }
334. // `kind` 'tv' | 'movie', `tmdbId` a number → { slug, show, created, sync } or { error }. A title we already have (same
335. // TMDB id and kind) is not added again: its slug is answered.
336. static importTitle = (kind, tmdbId) => {
337. if (kind != 'tv' && kind != 'movie') { return { error = 'unknown kind' } }
338. // `tid`, not `id`: inside the record literal below an entry named `id` would shadow it for every later entry
339. let tid = toNumber('' + tmdbId)
340. if (tid == null || tid <= 0 || tid % 1 != 0) { return { error = 'bad TMDB id' } }
341. ensureIndex()
342. let type = kind == 'tv' ? 'series' : 'movie'
343. let have = haveTmdb[tmdbKeyOf(type, tid)]
344. if (have != null) {
345. let s = showsTable.fetch(have)
346. if (s != null) { return { slug = s.urlSegment show = s.id created = false } }
347. }
348. let d = tmdbGet('/' + kind + '/' + tid)
349. if (d.failed != null) { return { error = 'TMDB: ' + d.failed } }
350. let title = kind == 'tv' ? textOr(d.name) : textOr(d.title)
351. if (title == null) { return { error = 'TMDB has no title for ' + kind + ' ' + tid } }
352. let release = textOr(kind == 'tv' ? d.first_air_date : d.release_date)
353. let year = release != null && release.length >= 4 ? toNumber(release.slice(0, 4)) : null
354. let overview = textOr(d.overview) != null ? d.overview : ''
355. let record = {
356. id = randomBytes(8, 'hex') title = title summary = overview tmdbSummary = textOr(d.overview) tmdbId = tid
357. language = textOr(d.original_language) type = type release = release year = year image = null
358. urlSegment = slugOf(title, tid) episodesCount = null homepage = textOr(d.homepage) imdbId = null seasonsCount = null
359. status = textOr(d.status) tagline = textOr(d.tagline) tvdbId = null tvmzId = null genres = genreIdsOf(d.genres)
360. cast = [] seasons = [] oldId = 'tmdb-' + kind + '-' + tid imported = now()
361. }
362. if (showsTable.put(record) == null) { return { error = 'could not store the title: ' + showsTable.lastError() } }
363. indexShow(record)
364. let r = syncShow(record.id)
365. persistSync()
366. 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 : '') + ')')
367. return { slug = record.urlSegment show = record.id created = true sync = r }
368. }