tracker

38.5 KB

1. // project.hl — tracker.worldapi.org: STEP 1 (tracker.worldapi.org#1), an empty shell. Login copied unchanged from
2. // calendar.worldapi.org (rejected once for a centered sign-in and no header selector — architect, 2026-09-27):
3. // ident only, no own passwords (README "How apps use ident" of ident.worldapi.org), the identity selector in the
4. // header (components/main.hl). No shows, no data yet — later steps come from the creator (CONCEPT.md).
5. //
6. // the header's ident-selector / "Log in with ident" -> <ident>/login?key=&return=<this app>/login/callback
7. // /login/callback?ident_code=<code> the app's server exchanges the code for the identity's short id
8. // (POST <ident>/api/exchange) and signs this app's OWN session in
9. // (session.user = { id = <users @id> }, users.hl)
10. // the header's "Log out" button signs this app's session out via the face trackerLogOut (components/main.hl)
11. // — ident's own session, and the browser's ident cookie, are untouched
12. //
13. // Config (env; no committed secret — the first deploy registers this app in ident and sets these, done by the
14. // architect, as with ident's own .env for SMTP):
15. // TRACKER_PORT (45008), TRACKER_URL (this app's own public address, for the return URL),
16. // IDENT_URL (https://ident.worldapi.org), TRACKER_KEY (pk_…), TRACKER_SECRET (sk_…)
17. import WebFramework from 'hl:web'
18. import { Response } from 'hl:http1'
19. import { randomBytes } from 'hl:crypto'
20. import { env } from 'hl:proc'
21. import { readBytes, readFile, exists } from 'hl:fs'
22. import { now, timestamp, every } from 'hl:time'
24. import Styles from './styles.hl'
25. import { exchangeCode, ensureUser } from './users.hl'
26. import Home from './components/home.hl'
27. import LoginFailed from './components/loginfailed.hl'
28. import Show from './components/show.hl'
29. import Unwatched from './components/unwatched.hl'
30. import Schedule from './components/schedule.hl'
31. import MyShows from './components/myshows.hl'
32. import MyMovies from './components/mymovies.hl'
33. import Movies from './components/movies.hl'
34. import AllShows from './components/allshows.hl'
35. import Search from './components/search.hl'
36. import Person from './components/person.hl'
37. import Genre from './components/genre.hl'
38. import { ensureIndex, searchTitleChanged, searchTitleRenamed } from './search.hl'
39. import { syncEnabled, syncShow, syncShowDelta, persistSync, emptyTotals, addTotals, totalsLine, pauseMsAfter, msPerRequest, adultBackfillQueue, backfillTitle } from './tmdbsync.hl'
40. import { detailsRepairQueue, completeStep, completeLine, persistDetails, persistCredits, applySynced, creditsQueue, creditsStep } from './details.hl'
41. import { planStart, planStep, planIds, planLine, markSynced, clearCopiedSummaries, isFollowedInRun } from './deltasync.hl'
42. import { allFollowedShowIds } from './follows.hl'
43. import { refreshCatalogShow } from './catalog.hl'
44. import { ensureShortIds, shortIdStep, shortIdPath, shortIdsLeft, shortIdStats, persistShortIds } from './shortids.hl'
45. import { guestPersonId } from './people.hl'
46. import { personById } from './shows.hl'
47. import { dark, darker } from './shared/tokens.hl'
49. static appTitle = "tracker"
51. // ---- THE INSTALLABLE APP (tracker.worldapi.org#10), the same way calendar.worldapi.org does it: hl:web's own web app
52. // manifest (/__hl/manifest.webmanifest, linked from every head with the apple-touch-icon and theme-color) and service
53. // worker (/__hl/sw.js) from these settings — no JavaScript of ours. Icons in icons/ (icon.svg is the source, the PNGs
54. // are rendered from it with rsvg-convert, README "Icons"). The theme colour is the header's background (darker), the
55. // splash background the page's (dark) — both off the tokens, not written out again.
56. appThemeColor = darker.value
57. appBackgroundColor = dark.value
58. appIcons = [
59. { src = '/icons/icon-192.png' sizes = '192x192' purpose = 'any' }
60. { src = '/icons/icon-512.png' sizes = '512x512' purpose = 'any' }
61. { src = '/icons/icon-192.png' sizes = '192x192' purpose = 'maskable' }
62. { src = '/icons/icon-512.png' sizes = '512x512' purpose = 'maskable' }
63. ]
64. appTouchIcon = '/icons/apple-touch-icon.png'
65. appFavicon = '/icons/favicon.svg'
66. // OFFLINE: only the shell — the header and `/` (components/main.hl says "You are offline" while the browser has no
67. // network). The data pages are not kept: without a network they get hl:web's "Unavailable offline" page.
68. offline = [ Home ]
70. styles = Styles
72. port = env('TRACKER_PORT') != null ? toNumber(env('TRACKER_PORT')) : 45008
73. watching = env('TRACKER_WATCH') != '0'
74. // read by hl:web's own manifest config (WebFramework.hl `cfg.sessionDir`), not the constructor call below
75. sessionDir = env('TRACKER_SESSIONS') != null ? env('TRACKER_SESSIONS') : null
77. // ---- THE LOGIN BUTTON'S RETURN (ident README "How apps use ident"), copied from calendar.worldapi.org's project.hl
78. // BACK TO THE PAGE: /login.js puts `?next=` into the button's return URL at the click. Only a same-origin PATH
79. // goes (one `/`, URL-safe characters, ≤ 500). Anything else → `/`.
80. nextChars = 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789-._~/?&=%+,;@!$()*:'
81. safePath = (want) => {
82. if (want == null || hlTypeName(want) != 'String' || want == '' || want.length > 500) { return '/' }
83. if (want.slice(0, 1) != '/' || want.slice(0, 2) == '//' || want.slice(0, 7) == '/login/') { return '/' }
84. let i = 0
85. while (i < want.length) {
86. if (!nextChars.includes(want[i])) { return '/' }
87. i = i + 1
88. }
89. return want
90. }
92. // A FAILED LOGIN is a page (components/loginfailed.hl): the reason is parked in the session, then → /login/failed
93. failed = (req, why) => {
94. let s = req.session
95. let fresh = s == null
96. if (fresh) { s = server.sessions.mint() }
97. s.data.loginError = why
98. server.sessions.save(s)
99. let res = new Response('login failed: ' + why, { status = 302 headers = { 'Location' = '/login/failed' 'Cache-Control' = 'no-store' 'Content-Type' = 'text/plain; charset=utf-8' } })
100. if (fresh) { res.headers['Set-Cookie'] = server.sessions.cookieHeader(s.id) }
101. return res
102. }
104. // the function route gets the cookie's session as req.session (hybriel #11); none yet → minted here
105. loginCallback = (route, req) => {
106. if (req.method != 'GET') { return failed(req, 'GET only') }
107. let q = req.query != null ? req.query : {}
108. let code = q.ident_code
109. if (code == null || code == '') { return failed(req, 'ident sent no login code') }
110. let x = exchangeCode(code)
111. if (x.error != null) { return failed(req, x.error) }
112. let u = ensureUser(x.identity)
113. if (u == null) { return failed(req, 'could not store the user') }
114. let s = req.session
115. let fresh = s == null
116. if (fresh) { s = server.sessions.mint() }
117. s.user = { id = u.id }
118. s.data.tag = randomBytes(16)
119. s.data.loginError = null
120. server.sessions.save(s)
121. let res = new Response('logged in', { status = 302 headers = { 'Location' = safePath(q.next) 'Cache-Control' = 'no-store' 'Content-Type' = 'text/plain; charset=utf-8' } })
122. if (fresh) { res.headers['Set-Cookie'] = server.sessions.cookieHeader(s.id) }
123. return res
124. }
126. // ---- /posters/<name> (tracker.worldapi.org#4): a show's poster image, served dynamically from
127. // storage/mpackdb/posters/ (the poster's file name is the show's oldId + its migrated extension,
128. // shows.hl posterName) — falls back to a placeholder svg when the real file isn't there yet (the
129. // mongo export never included the actual poster bytes, see the report's `open`).
130. postersDir = (env('TRACKER_STORAGE') != null ? env('TRACKER_STORAGE') : './storage/mpackdb') + '/posters'
131. placeholderPoster = './assets/poster-placeholder.svg'
132. posterNameChars = 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789.-_'
133. isSafePosterName = (name) => {
134. if (name == null || hlTypeName(name) != 'String' || name == '' || name.length > 100) { return false }
135. let i = 0
136. while (i < name.length) {
137. if (!posterNameChars.includes(name[i])) { return false }
138. i = i + 1
139. }
140. return true
141. }
142. endsWithStr = (s, suffix) => { return s.length >= suffix.length && s.slice(s.length - suffix.length) == suffix }
143. mimeOfPoster = (name) => {
144. if (endsWithStr(name, '.png')) { return 'image/png' }
145. if (endsWithStr(name, '.webp')) { return 'image/webp' }
146. if (endsWithStr(name, '.svg')) { return 'image/svg+xml' }
147. return 'image/jpeg'
148. }
149. posterRoute = (route, req) => {
150. let name = route.params != null ? route.params.name : null
151. if (!isSafePosterName(name)) { return new Response('bad poster name', { status = 400 headers = { 'Content-Type' = 'text/plain; charset=utf-8' } }) }
152. let path = postersDir + '/' + name
153. if (exists(path)) { return new Response(readBytes(path), { headers = { 'Content-Type' = mimeOfPoster(name) 'Cache-Control' = 'public, max-age=86400' } }) }
154. return new Response(readBytes(placeholderPoster), { headers = { 'Content-Type' = 'image/svg+xml' 'Cache-Control' = 'public, max-age=3600' } })
155. }
157. // ---- /profiles/<name> (tracker.worldapi.org#16): a person's photo, from storage/mpackdb/profiles/ (people.hl syncProfile
158. // downloads it on the first visit of the person page) — the same checks and placeholder as /posters/<name>
159. profilesDir = (env('TRACKER_STORAGE') != null ? env('TRACKER_STORAGE') : './storage/mpackdb') + '/profiles'
160. profileRoute = (route, req) => {
161. let name = route.params != null ? route.params.name : null
162. if (!isSafePosterName(name)) { return new Response('bad photo name', { status = 400 headers = { 'Content-Type' = 'text/plain; charset=utf-8' } }) }
163. let path = profilesDir + '/' + name
164. if (exists(path)) { return new Response(readBytes(path), { headers = { 'Content-Type' = mimeOfPoster(name) 'Cache-Control' = 'public, max-age=86400' } }) }
165. return new Response(readBytes(placeholderPoster), { headers = { 'Content-Type' = 'image/svg+xml' 'Cache-Control' = 'public, max-age=3600' } })
166. }
168. // ---- the old addresses of the personal lists (tracker.worldapi.org#8): they all live under /my/ now —
169. // a permanent redirect so old links keep working (architect's decision)
170. movedTo = (path) => { return new Response('moved to ' + path, { status = 301 headers = { 'Location' = path 'Content-Type' = 'text/plain; charset=utf-8' } }) }
171. unwatchedMoved = (route, req) => { return movedTo('/my/unwatched') }
172. scheduleMoved = (route, req) => { return movedTo('/my/schedule') }
173. // mission 058: the show page's genre pills linked `/genre/<x>` (no page there) — the genre pages are `/genres/<x>`
174. genreMoved = (route, req) => {
175. let g = route.params != null ? route.params.genre : null
176. return movedTo(isSafePosterName(g) ? '/genres/' + g : '/')
177. }
178. // mission 060 (tracker.worldapi.org#27): `/<shortId>` → 301 to the title's or person's page (shortids.hl); upper case is
179. // read as lower case; anything else is a 404
180. shortIdRoute = (route, req) => {
181. let sid = route.params != null ? route.params.sid : null
182. let path = sid != null && hlTypeName(sid) == 'String' ? shortIdPath(sid.toLowerCase()) : null
183. if (path == null) { return new Response('not found', { status = 404 headers = { 'Content-Type' = 'text/plain; charset=utf-8' } }) }
184. return movedTo(path)
185. }
187. // mission 062: a GUEST STAR's link (components/show.hl castRowsOf) — guests are no person records until someone opens one:
188. // `/person/tmdb/<tmdbId>?show=<show id>` creates the person (people.hl guestPersonId: name + role from that title, a short id) or
189. // finds the one we have, then 302 → `/person/<slug>` (the page fills from TMDB as for any new person). A function route: the
190. // browser loads it itself (hl:web's client router hands every non-component link to the browser)
191. guestRoute = (route, req) => {
192. let t = route.params != null ? route.params.tmdb : null
193. let tmdbId = t != null && hlTypeName(t) == 'String' && t.length <= 10 ? toNumber(t) : null
194. if (tmdbId == null || tmdbId < 1) { return new Response('not found', { status = 404 headers = { 'Content-Type' = 'text/plain; charset=utf-8' } }) }
195. let q = req.query != null ? req.query : {}
196. let r = guestPersonId(tmdbId, q.show)
197. let p = r.id != null ? personById(r.id) : null
198. if (p == null || p.urlSegment == null) { return new Response('not found: ' + (r.error != null ? r.error : 'no such person'), { status = 404 headers = { 'Content-Type' = 'text/plain; charset=utf-8' } }) }
199. return new Response('see /person/' + p.urlSegment, { status = 302 headers = { 'Location' = '/person/' + p.urlSegment 'Cache-Control' = 'no-store' 'Content-Type' = 'text/plain; charset=utf-8' } })
200. }
202. routes = [
203. { pattern = "/favicon.ico" file = "./icons/favicon.ico" headers = { 'Cache-Control' = 'no-cache' } }
204. // the installable app (tracker.worldapi.org#10): the icons (manifest and service worker are hl:web's own, from appIcons / offline)
205. { pattern = "/icons/icon-192.png" file = "./icons/icon-192.png" headers = { 'Cache-Control' = 'no-cache' } }
206. { pattern = "/icons/icon-512.png" file = "./icons/icon-512.png" headers = { 'Cache-Control' = 'no-cache' } }
207. { pattern = "/icons/apple-touch-icon.png" file = "./icons/apple-touch-icon.png" headers = { 'Cache-Control' = 'no-cache' } }
208. { pattern = "/icons/favicon.svg" file = "./icons/favicon.svg" headers = { 'Cache-Control' = 'no-cache' } }
209. { pattern = "/login/callback" function = loginCallback }
210. { pattern = "/login/failed" component = LoginFailed }
211. { pattern = "/login.js" file = "./login.js" headers = { 'Cache-Control' = 'no-cache' } }
212. { pattern = "/posters/:name" function = posterRoute }
213. // tracker.worldapi.org#13: every movie / every series, 24 per page — the page is a PATH segment (`/movies/page/2`), a
214. // page component cannot read `?page=` (hybriel#11) — and the movies the user follows
215. { pattern = "/movies" component = Movies }
216. { pattern = "/movies/page/:page" component = Movies }
217. { pattern = "/shows" component = AllShows }
218. { pattern = "/shows/page/:page" component = AllShows }
219. { pattern = "/shows/:slug" component = Show }
220. { pattern = "/my/shows" component = MyShows }
221. { pattern = "/my/movies" component = MyMovies }
222. { pattern = "/my/unwatched" component = Unwatched }
223. { pattern = "/my/schedule" component = Schedule }
224. { pattern = "/unwatched" function = unwatchedMoved }
225. { pattern = "/schedule" function = scheduleMoved }
226. // tracker.worldapi.org#14: the search — `/search/<text>` (hl:web gives a page no query string, so not `?q=`)
227. { pattern = "/search/:q?" component = Search }
228. // tracker.worldapi.org#16: a person's page (the show page's cast and the search link here) and their photo
229. { pattern = "/person/tmdb/:tmdb" function = guestRoute }
230. { pattern = "/person/:slug" component = Person }
231. { pattern = "/profiles/:name" function = profileRoute }
232. // mission 058 (tracker.worldapi.org#25): a genre's movies and series, like /movies (the page in the path)
233. { pattern = "/genres/:genre" component = Genre }
234. { pattern = "/genres/:genre/page/:page" component = Genre }
235. { pattern = "/genre/:genre" function = genreMoved }
236. { pattern = "/" component = Home }
237. // mission 060 (tracker.worldapi.org#27): a short id → its page (LAST: every named one-segment route above wins)
238. { pattern = "/:sid" function = shortIdRoute }
239. ]
241. // WHO GETS THE PUSH: the login state reaches the tabs of one session (components/main.hl trackerSignedIn/Out).
242. tagOf = (session) => { return session != null && session.data != null ? session.data.tag : null }
243. audience = {
244. trackerSignedIn = (tag, session) => { return tag != null && tagOf(session) == tag }
245. trackerSignedOut = (tag, session) => { return tag != null && tagOf(session) == tag }
246. }
248. // cookies are per host, not per port: an own name keeps this app's session apart from
249. // ident's own (and from any other worldapi app sharing a dev host), see ident README "Design
250. // tokens" / hybriel#10 hybriel#17.
251. sessionCookie = 'trackersid'
252. // tracker.worldapi.org#14: the search index (search.hl) is built once here at boot (~0.6 s on the real data), not by the
253. // first search
254. searchIndexReady = ensureIndex()
255. // mission 060: every short id in use (one map over shows AND persons) and the records still without one, before the first request
256. shortIdsReady = ensureShortIds()
257. // tracker#18, ONCE (deltasync.hl clearCopiedSummaries): a `summary` that is only the migrated copy of tmdbSummary is cleared — the
258. // current TMDB text shows; the creator's own texts stay
259. summariesCleared = clearCopiedSummaries()
260. if (summariesCleared.ran) { console.log('summaries: cleared ' + summariesCleared.cleared + ' copies of tmdbSummary, kept ' + summariesCleared.kept + ' own texts, errors ' + summariesCleared.errors + ', ' + summariesCleared.ms + ' ms') }
261. server = new WebFramework(routes = routes, styles = styles, port = port, sessionCookie = sessionCookie, watchMode = watching)
263. // ---- THE DAILY TMDB SYNC (tracker.worldapi.org#9, tmdbsync.hl; tracker#18, deltasync.hl) -------------------------
264. // Once a day at TRACKER_SYNC_HOUR (UTC, default 4 — hl:time has no time zones; the container runs UTC). hl:web serves ONE
265. // request at a time, so the run never takes the server for itself: a 1-second clock does ONE step per tick and returns — the
266. // requests that queued meanwhile are served between two ticks, and a tick never starts inside a page render.
267. // tracker#18: the run first asks TMDB's and TVmaze's CHANGE LISTS since the last run (deltasync.hl — one list page per step)
268. // and then syncs only the titles of ours that changed, followed ones first — every title stays current, not only the followed
269. // ones. FULL WALK (every followed show, as before #18) on the first run, after a gap > 14 days, or when a change list fails.
270. // A title's step: tmdbsync.hl syncShow (1–3 TMDB requests + maybe its poster, TVmaze) + details.hl applySynced (the record's
271. // title, overview, tagline … and cast/crew from the same details answer). After a step that made n requests the next waits
272. // n × 260 ms (TMDB: ≤ 40 requests / 10 s) + 550 ms per TVmaze request (≤ 20 / 10 s, tracker.worldapi.org#12).
273. // The finished run stores its day (deltasync.hl markSynced). Log: "tmdb sync: start, …", the change lists ("tmdb sync:
274. // changes since …"), errors, slow steps, "tmdb sync done: mode=… changeRequests=… <totals>".
275. // TRACKER_SYNC=0 switches it off; without TMDB_READ_TOKEN it is off too (logged once).
276. syncHour = env('TRACKER_SYNC_HOUR') != null ? toNumber(env('TRACKER_SYNC_HOUR')) : 4
277. syncOn = env('TRACKER_SYNC') != '0' && syncEnabled
278. syncQueue = []
279. syncPlanning = false
280. syncMode = ''
281. syncMark = false
282. syncChangeRequests = 0
283. syncAt = 0
284. syncDay = ''
285. syncNextAt = 0
286. syncTotals = emptyTotals()
287. syncFinish = () => {
288. persistDetails()
289. if (syncMark) { markSynced(syncDay) }
290. console.log('tmdb sync done: mode=' + syncMode + ' changeRequests=' + syncChangeRequests + ' ' + totalsLine(syncTotals) + ' seconds=' + ((now() - syncAt) / 1000))
291. return null
292. }
293. syncTick = () => {
294. let t = now()
295. if (syncQueue.length == 0 && !syncPlanning) {
296. let stamp = timestamp(t)
297. let day = stamp.slice(0, 10)
298. if (toNumber(stamp.slice(11, 13)) != syncHour || day == syncDay) { return null }
299. syncDay = day
300. syncAt = t
301. syncNextAt = 0
302. syncTotals = emptyTotals()
303. syncChangeRequests = 0
304. let p = planStart(day, null)
305. syncMode = p.mode
306. syncMark = true
307. if (p.mode == 'delta') {
308. syncPlanning = true
309. console.log('tmdb sync: start, changes ' + p.why)
310. return null
311. }
312. syncQueue = allFollowedShowIds()
313. console.log('tmdb sync: start, ' + syncQueue.length + ' followed shows, full walk (' + p.why + ')')
314. if (syncQueue.length == 0) { syncFinish() }
315. return null
316. }
317. if (t < syncNextAt) { return null }
318. if (syncPlanning) {
319. let ps = planStep()
320. syncChangeRequests = syncChangeRequests + ps.requests + ps.tvmazeRequests
321. syncTotals.requests = syncTotals.requests + ps.requests
322. syncTotals.tvmazeRequests = syncTotals.tvmazeRequests + ps.tvmazeRequests
323. syncNextAt = now() + pauseMsAfter(ps)
324. if (!ps.done) { return null }
325. syncPlanning = false
326. console.log('tmdb sync: ' + planLine())
327. if (ps.failed != '') {
328. // a change list could not be read: the old full walk, and the day is not advanced (the next run asks again)
329. syncMode = 'full'
330. syncMark = false
331. syncQueue = allFollowedShowIds()
332. console.log('tmdb sync: change list failed, full walk of ' + syncQueue.length + ' followed shows instead')
333. } else {
334. syncQueue = planIds()
335. }
336. if (syncQueue.length == 0) { syncFinish() }
337. return null
338. }
339. let id = syncQueue[0]
340. let rest = []
341. let i = 1
342. while (i < syncQueue.length) { rest.push(syncQueue[i]) i = i + 1 }
343. syncQueue = rest
344. // tracker#18: in a delta run a title nobody follows gets the light step (changed seasons, no TVmaze)
345. let r = syncMode == 'delta' && !isFollowedInRun(id) ? syncShowDelta(id) : syncShow(id)
346. // tracker#18 + #28: the record's own fields (title, overview, tagline …) and the cast + crew from the same details answer
347. if (r.details != null) {
348. let c = applySynced(id, r.details)
349. if (c.error != '') { console.log('tmdb sync: ' + c.error) }
350. if (c.renamed) { searchTitleRenamed(id) }
351. }
352. // the public lists (catalog.hl, tracker.worldapi.org#13): this show's place by its newest released episode; the sync
353. // also brings TMDB's adult flag (mission 054) → the search's view of it too
354. refreshCatalogShow(id)
355. searchTitleChanged(id)
356. let took = now() - t
357. syncTotals = addTotals(syncTotals, r)
358. if (r.error != '') { console.log('tmdb sync: ' + r.error) }
359. // a step blocks every request meanwhile (Saturday Night Live, 53 seasons, 4 requests: ~1.7 s) — say so in the log
360. if (took > 1500) { console.log('tmdb sync: slow step, show ' + id + ', ' + r.requests + ' requests, ' + took + ' ms') }
361. syncNextAt = now() + pauseMsAfter(r)
362. if (syncQueue.length > 0 && syncTotals.shows % 10 == 0) { persistDetails() }
363. if (syncQueue.length == 0) { syncFinish() }
364. return null
365. }
367. // ---- THE ADULT-FLAG BACKFILL (mission 054, tmdbsync.hl backfillTitle) -----------------------------------
368. // The public lists and the search show only titles TMDB says are not adult (shows.hl isPublicTitle; unknown = hidden). At
369. // start the app collects every title whose flag is still unknown and asks TMDB for ONE title per step — like the daily
370. // sync, on the same clock, so requests queued meanwhile are served between two steps: one details request (adult,
371. // poster_path, external ids) + the poster file when missing, then a pause of 260 ms per TMDB API request (≤ 40 / 10 s;
372. // the poster comes from TMDB's image host, not the API, and is not counted). Each answered title goes into the lists
373. // (catalog.hl refreshCatalogShow) and the search at once. RESUMES after a restart: the queue is built from what is
374. // still unknown. Stops when the queue is empty; titles TMDB answers 404 for are marked (`adultCheck`) and stay hidden; a
375. // passing failure (no answer, 429, 5xx) goes to the back of the queue, at most 2 more tries per start.
376. // While the daily sync runs, the backfill waits. Log: start, every TRACKER_BACKFILL_LOG_EVERY (500) titles, done.
377. // TRACKER_BACKFILL=0 switches it off; without TMDB_READ_TOKEN it is off too.
378. backfillOn = env('TRACKER_BACKFILL') != '0' && syncEnabled
379. backfillLogEvery = env('TRACKER_BACKFILL_LOG_EVERY') != null ? toNumber(env('TRACKER_BACKFILL_LOG_EVERY')) : 500
380. backfillQueue = null
381. backfillPos = 0
382. backfillDone = false
383. backfillAt = 0
384. backfillNextAt = 0
385. backfillTries = {}
386. backfillTotals = { titles = 0 adult = 0 notAdult = 0 unknown = 0 posters = 0 ids = 0 requests = 0 errors = 0 retries = 0 }
387. backfillLine = () => {
388. let b = backfillTotals
389. return b.titles + '/' + backfillQueue.length + ' titles, adult=' + b.adult + ' notAdult=' + b.notAdult + ' unknown=' + b.unknown + ' posters=' + b.posters
390. + ' ids=' + b.ids + ' requests=' + b.requests + ' errors=' + b.errors + ' retries=' + b.retries + ' seconds=' + ((now() - backfillAt) / 1000)
391. }
392. backfillTick = () => {
393. if (backfillDone) { return null }
394. let t = now()
395. if (backfillQueue == null) {
396. backfillQueue = adultBackfillQueue()
397. backfillAt = t
398. console.log('adult backfill: start, ' + backfillQueue.length + ' titles without an adult flag (' + (now() - t) + ' ms to find them)')
399. if (backfillQueue.length == 0) { backfillDone = true console.log('adult backfill: nothing to do') }
400. return null
401. }
402. if (t < backfillNextAt) { return null }
403. let id = backfillQueue[backfillPos]
404. backfillPos = backfillPos + 1
405. let r = backfillTitle(id)
406. if (r.retry && (backfillTries[id] == null || backfillTries[id] < 2)) {
407. backfillTries[id] = backfillTries[id] == null ? 1 : backfillTries[id] + 1
408. let q = backfillQueue
409. q.push(id)
410. backfillQueue = q
411. backfillTotals.retries = backfillTotals.retries + 1
412. } else {
413. backfillTotals.titles = backfillTotals.titles + 1
414. if (r.adult == true) { backfillTotals.adult = backfillTotals.adult + 1 } else if (r.adult == false) { backfillTotals.notAdult = backfillTotals.notAdult + 1 } else { backfillTotals.unknown = backfillTotals.unknown + 1 }
415. }
416. backfillTotals.posters = backfillTotals.posters + r.posters
417. backfillTotals.ids = backfillTotals.ids + r.ids
418. backfillTotals.requests = backfillTotals.requests + r.requests
419. backfillTotals.errors = backfillTotals.errors + r.errors
420. if (r.error != '') { console.log('adult backfill: ' + r.error) }
421. if (r.adult != null) {
422. refreshCatalogShow(id)
423. searchTitleChanged(id)
424. }
425. let took = now() - t
426. if (took > 1500) { console.log('adult backfill: slow step, show ' + id + ', ' + took + ' ms') }
427. backfillNextAt = now() + r.requests * msPerRequest
428. let finished = backfillPos >= backfillQueue.length
429. if (finished || backfillPos % 25 == 0) { persistSync() }
430. if (!finished && backfillTotals.titles > 0 && backfillTotals.titles % backfillLogEvery == 0 && !r.retry) { console.log('adult backfill: ' + backfillLine()) }
431. if (finished) {
432. backfillDone = true
433. console.log('adult backfill done: ' + backfillLine())
434. }
435. return null
436. }
438. // ---- THE DETAILS REPAIR (tracker.worldapi.org#26, details.hl completeStep) ----------------------------------------
439. // Titles the person pages added from a filmography (#16) are MINIMAL (no poster, seasons, ids, cast …). Opening one completes
440. // it (components/show.hl); this job completes the rest in the background: at start it collects every incomplete, non-adult
441. // title and does ONE per step on the app's clock, like the adult backfill — a step is the daily sync's work for one title
442. // (1–3 TMDB requests + poster + TVmaze), requests queued meanwhile are served between two steps; then a pause of 260 ms per
443. // TMDB request + 550 ms per TVmaze request (pauseMsAfter: ≤ 40 / 10 s, ≤ 20 / 10 s). RESUMES after a restart (the queue is
444. // what is still incomplete); TMDB 404 marks the title (`detailsCheck`), a passing failure goes to the back of the queue (at
445. // most 2 more tries per start). Waits while the daily sync or the adult backfill runs. Persisted every 10 titles + at the end.
446. // Log: start, every TRACKER_REPAIR_LOG_EVERY (100) titles, each failure, done. TRACKER_REPAIR=0 switches it off (off too
447. // without TMDB_READ_TOKEN). TRACKER_REPAIR_FAST=1 (the gate, fake TMDB only): 20 ms pause instead.
448. repairOn = env('TRACKER_REPAIR') != '0' && syncEnabled
449. repairFast = env('TRACKER_REPAIR_FAST') == '1'
450. repairLogEvery = env('TRACKER_REPAIR_LOG_EVERY') != null ? toNumber(env('TRACKER_REPAIR_LOG_EVERY')) : 100
451. repairQueue = null
452. repairPos = 0
453. repairDone = false
454. repairAt = 0
455. repairNextAt = 0
456. repairTries = {}
457. repairTotals = { titles = 0 complete = 0 failed = 0 adult = 0 newSeasons = 0 newEpisodes = 0 posters = 0 cast = 0 requests = 0 tvmazeRequests = 0 retries = 0 }
458. repairLine = () => {
459. let b = repairTotals
460. return b.titles + '/' + repairQueue.length + ' titles, complete=' + b.complete + ' failed=' + b.failed + ' adult=' + b.adult + ' newSeasons=' + b.newSeasons + ' newEpisodes=' + b.newEpisodes
461. + ' posters=' + b.posters + ' cast=' + b.cast + ' requests=' + b.requests + ' tvmazeRequests=' + b.tvmazeRequests + ' retries=' + b.retries + ' seconds=' + ((now() - repairAt) / 1000)
462. }
463. repairTick = () => {
464. if (repairDone) { return null }
465. let t = now()
466. if (repairQueue == null) {
467. repairQueue = detailsRepairQueue()
468. repairAt = t
469. console.log('details repair: start, ' + repairQueue.length + ' incomplete titles (' + (now() - t) + ' ms to find them)')
470. if (repairQueue.length == 0) { repairDone = true console.log('details repair: nothing to do') }
471. return null
472. }
473. if (t < repairNextAt) { return null }
474. let id = repairQueue[repairPos]
475. let r = completeStep(id)
476. let took = now() - t
477. if (took > 1500) { console.log('details repair: slow step, show ' + id + ', ' + took + ' ms') }
478. if (r.error != '') { console.log('details repair: ' + r.error) }
479. repairNextAt = now() + (repairFast ? 20 : pauseMsAfter(r))
480. // a series takes several steps: the same title again next time
481. if (!r.done && !r.retry) { return null }
482. repairPos = repairPos + 1
483. if (r.retry && (repairTries[id] == null || repairTries[id] < 2)) {
484. repairTries[id] = repairTries[id] == null ? 1 : repairTries[id] + 1
485. let q = repairQueue
486. q.push(id)
487. repairQueue = q
488. repairTotals.retries = repairTotals.retries + 1
489. } else if (!r.skipped) {
490. // done: its counts (a failed first step: the request alone)
491. let sum = r.sum
492. repairTotals.titles = repairTotals.titles + 1
493. if (r.adult == true) { repairTotals.adult = repairTotals.adult + 1 } else if (sum.steps == 0) { repairTotals.failed = repairTotals.failed + 1 } else { repairTotals.complete = repairTotals.complete + 1 }
494. repairTotals.newSeasons = repairTotals.newSeasons + sum.newSeasons
495. repairTotals.newEpisodes = repairTotals.newEpisodes + sum.newEpisodes
496. repairTotals.posters = repairTotals.posters + sum.posters
497. repairTotals.cast = repairTotals.cast + sum.cast
498. repairTotals.requests = repairTotals.requests + (sum.steps == 0 ? r.requests : sum.requests)
499. repairTotals.tvmazeRequests = repairTotals.tvmazeRequests + sum.tvmazeRequests
500. }
501. let finished = repairPos >= repairQueue.length
502. if (finished || repairPos % 10 == 0) { persistDetails() }
503. if (!finished && !r.skipped && !r.retry && repairTotals.titles > 0 && repairTotals.titles % repairLogEvery == 0) { console.log('details repair: ' + repairLine()) }
504. if (finished) {
505. repairDone = true
506. console.log('details repair done: ' + repairLine())
507. }
508. return null
509. }
511. // ---- THE CREDITS JOB (tracker.worldapi.org#28, details.hl creditsStep) ------------------------------------------------------
512. // Every PUBLIC title (adult == false; adult/unknown are skipped) with a TMDB id gets its whole cast + crew once: at start (after
513. // the adult backfill and the details repair are done) the job collects every such title without `creditsAt` and asks ONE per
514. // step on the app's clock — one request (details + credits / aggregate_credits), then 260 ms (≤ 40 / 10 s); requests queued
515. // meanwhile are served between two steps. Titles the details repair, the daily sync or an import already did carry `creditsAt`
516. // and are not asked. RESUMES after a restart (the queue = what still has no credits); TMDB 404 / adult → `creditsCheck`, never
517. // asked again; a passing failure goes to the back of the queue (at most 2 more tries per start). Waits while the daily sync
518. // runs. Persisted (shows + people only) every 25 titles + at the end. Log: start, every TRACKER_CREDITS_LOG_EVERY (500) titles, failures, slow steps
519. // (> 1.5 s), done. TRACKER_CREDITS=0 switches it off (off too without TMDB_READ_TOKEN). TRACKER_CREDITS_FAST=1 (the gate, fake
520. // TMDB only): 20 ms pause instead.
521. // MEMORY LIMIT: the runtime never frees what a fetch + JSON parse took (mission 061 measured ~6 MB more RSS per title on the real
522. // data, 300 fetches of a 440 KB JSON = +2.7 GB in a 10-line script) — the whole job in one process would add ~35 GB. Before every
523. // step the job reads its own RSS (/proc/self/status); above TRACKER_CREDITS_MAX_MB (16000) it PAUSES for this process (logged) and
524. // goes on with what is left after the next restart (deploy, `docker restart`).
525. creditsOn = env('TRACKER_CREDITS') != '0' && syncEnabled
526. creditsFast = env('TRACKER_CREDITS_FAST') == '1'
527. creditsLogEvery = env('TRACKER_CREDITS_LOG_EVERY') != null ? toNumber(env('TRACKER_CREDITS_LOG_EVERY')) : 500
528. creditsQueueIds = null
529. creditsPos = 0
530. creditsDone = false
531. creditsAt = 0
532. creditsNextAt = 0
533. creditsTries = {}
534. creditsMaxMb = env('TRACKER_CREDITS_MAX_MB') != null ? toNumber(env('TRACKER_CREDITS_MAX_MB')) : 16000
535. creditsTotals = { titles = 0 done = 0 failed = 0 cast = 0 crew = 0 people = 0 requests = 0 retries = 0 slowest = 0 }
536. // this process' resident memory in MB (VmRSS), 0 when unknown
537. rssMb = () => {
538. if (!exists('/proc/self/status')) { return 0 }
539. let t = readFile('/proc/self/status')
540. let i = t != null ? t.indexOf('VmRSS:') : -1
541. if (i < 0) { return 0 }
542. let kb = toNumber(t.slice(i + 6, t.indexOf('kB', i)).trim())
543. return kb != null ? (kb - kb % 1024) / 1024 : 0
544. }
545. creditsLine = () => {
546. let b = creditsTotals
547. return b.titles + '/' + creditsQueueIds.length + ' titles, done=' + b.done + ' failed=' + b.failed + ' cast=' + b.cast + ' crew=' + b.crew + ' people=' + b.people
548. + ' requests=' + b.requests + ' retries=' + b.retries + ' slowestStepMs=' + b.slowest + ' seconds=' + ((now() - creditsAt) / 1000)
549. }
550. creditsTick = () => {
551. if (creditsDone) { return null }
552. let t = now()
553. if (creditsQueueIds == null) {
554. creditsQueueIds = creditsQueue()
555. creditsAt = t
556. console.log('credits job: start, ' + creditsQueueIds.length + ' titles without credits (' + (now() - t) + ' ms to find them)')
557. if (creditsQueueIds.length == 0) { creditsDone = true console.log('credits job: nothing to do') }
558. return null
559. }
560. if (t < creditsNextAt) { return null }
561. let mb = rssMb()
562. if (mb > creditsMaxMb) {
563. creditsDone = true
564. persistCredits()
565. console.log('credits job: paused, RSS ' + mb + ' MB > TRACKER_CREDITS_MAX_MB ' + creditsMaxMb + ' — ' + (creditsQueueIds.length - creditsPos) + ' titles left for the next start; ' + creditsLine())
566. return null
567. }
568. let id = creditsQueueIds[creditsPos]
569. creditsPos = creditsPos + 1
570. let r = creditsStep(id)
571. let took = now() - t
572. if (took > creditsTotals.slowest) { creditsTotals.slowest = took }
573. if (took > 1500) { console.log('credits job: slow step, show ' + id + ', ' + took + ' ms, cast=' + r.cast + ' crew=' + r.crew + ' people=' + r.people) }
574. if (r.error != '') { console.log('credits job: ' + r.error) }
575. creditsNextAt = now() + (creditsFast ? 20 : r.requests * msPerRequest)
576. if (r.retry && (creditsTries[id] == null || creditsTries[id] < 2)) {
577. creditsTries[id] = creditsTries[id] == null ? 1 : creditsTries[id] + 1
578. let q = creditsQueueIds
579. q.push(id)
580. creditsQueueIds = q
581. creditsTotals.retries = creditsTotals.retries + 1
582. } else if (!r.skipped) {
583. creditsTotals.titles = creditsTotals.titles + 1
584. if (r.error != '') { creditsTotals.failed = creditsTotals.failed + 1 } else { creditsTotals.done = creditsTotals.done + 1 }
585. creditsTotals.cast = creditsTotals.cast + r.cast
586. creditsTotals.crew = creditsTotals.crew + r.crew
587. creditsTotals.people = creditsTotals.people + r.people
588. }
589. creditsTotals.requests = creditsTotals.requests + r.requests
590. let finished = creditsPos >= creditsQueueIds.length
591. if (finished || creditsPos % 25 == 0) { persistCredits() }
592. if (!finished && !r.skipped && !r.retry && creditsTotals.titles > 0 && creditsTotals.titles % creditsLogEvery == 0) { console.log('credits job: ' + creditsLine()) }
593. if (finished) {
594. creditsDone = true
595. console.log('credits job done: ' + creditsLine())
596. }
597. return null
598. }
600. // ONE clock for all four (a tick every 0.1 s): the daily sync's step once a second as before; the backfill's steps in
601. // between, only while no daily run is going on; the details repair only when neither has work; the credits job (#28) last
602. clockOn = syncOn || backfillOn || repairOn || creditsOn
603. clockTicks = 0
604. if (syncOn) { console.log('tmdb sync: daily at ' + syncHour + ':00 UTC') } else { console.log('tmdb sync: off (' + (syncEnabled ? 'TRACKER_SYNC=0' : 'no TMDB_READ_TOKEN') + ')') }
605. if (!backfillOn) { console.log('adult backfill: off (' + (syncEnabled ? 'TRACKER_BACKFILL=0' : 'no TMDB_READ_TOKEN') + ')') }
606. if (!repairOn) { console.log('details repair: off (' + (syncEnabled ? 'TRACKER_REPAIR=0' : 'no TMDB_READ_TOKEN') + ')') }
607. if (!creditsOn) { console.log('credits job: off (' + (syncEnabled ? 'TRACKER_CREDITS=0' : 'no TMDB_READ_TOKEN') + ')') }
608. if (clockOn) {
609. syncClock = every(0.1)
610. on syncClock.tick(x) {
611. clockTicks = clockTicks + 1
612. let busy = syncQueue.length > 0 || syncPlanning
613. if (syncOn && clockTicks % 10 == 0) { syncTick() }
614. let free = !busy && syncQueue.length == 0 && !syncPlanning
615. if (backfillOn && free && !backfillDone) { backfillTick() }
616. else if (repairOn && free && (!backfillOn || backfillDone) && !repairDone) { repairTick() }
617. else if (creditsOn && free && (!backfillOn || backfillDone) && (!repairOn || repairDone)) { creditsTick() }
618. return null
619. }
620. }
622. // ---- THE SHORT-ID BACKFILL (mission 060, tracker.worldapi.org#27, shortids.hl shortIdStep) ----------------------------
623. // Every show and person without a short id gets one in the background: on its own clock (every 0.1 s) ONE batch per tick,
624. // as many records as fit into TRACKER_SHORTID_BUDGET_MS (default 40 ms) — no TMDB, only the tables — so a page waits at
625. // most one batch. The queue is built at start (`ensureShortIds`) from what is still missing = it resumes after a restart.
626. // Persisted every 20 batches and at the end. Log: start, every TRACKER_SHORTID_LOG_EVERY (5000) records, done; with nothing
627. // to do one line. TRACKER_SHORTIDS=0 switches it off (the gate's resume check).
628. shortIdOn = env('TRACKER_SHORTIDS') != '0'
629. shortIdBudget = env('TRACKER_SHORTID_BUDGET_MS') != null ? toNumber(env('TRACKER_SHORTID_BUDGET_MS')) : 40
630. shortIdLogEvery = env('TRACKER_SHORTID_LOG_EVERY') != null ? toNumber(env('TRACKER_SHORTID_LOG_EVERY')) : 5000
631. shortIdRun = { at = 0 batches = 0 assigned = 0 nextLog = 0 slowest = 0 slowestPersist = 0 done = false }
632. shortIdTick = () => {
633. if (shortIdRun.done) { return null }
634. let t = now()
635. let r = shortIdStep(shortIdBudget)
636. let took = now() - t
637. shortIdRun.batches = shortIdRun.batches + 1
638. shortIdRun.assigned = shortIdRun.assigned + r.assigned
639. if (took > shortIdRun.slowest) { shortIdRun.slowest = took }
640. if (r.done || shortIdRun.batches % 20 == 0) {
641. let tp = now()
642. persistShortIds()
643. if (now() - tp > shortIdRun.slowestPersist) { shortIdRun.slowestPersist = now() - tp }
644. }
645. if (!r.done && shortIdRun.assigned >= shortIdRun.nextLog) {
646. console.log('short ids: ' + shortIdRun.assigned + ' assigned, ' + r.left + ' left, ' + ((now() - shortIdRun.at) / 1000) + ' s')
647. shortIdRun.nextLog = shortIdRun.nextLog + shortIdLogEvery
648. }
649. if (r.done) {
650. shortIdRun.done = true
651. let s = shortIdStats()
652. console.log('short ids done: assigned=' + shortIdRun.assigned + ' oldKept=' + s.oldKept + ' batches=' + shortIdRun.batches + ' slowestBatchMs=' + shortIdRun.slowest + ' slowestPersistMs=' + shortIdRun.slowestPersist + ' seconds=' + ((now() - shortIdRun.at) / 1000))
653. }
654. return null
655. }
656. if (!shortIdOn) { console.log('short ids: backfill off (TRACKER_SHORTIDS=0), ' + shortIdsLeft() + ' records without one') }
657. else if (shortIdsLeft() == 0) { console.log('short ids: nothing to do') }
658. else {
659. shortIdRun.at = now()
660. shortIdRun.nextLog = shortIdLogEvery
661. console.log('short ids: backfill start, ' + shortIdsLeft() + ' records, ' + shortIdBudget + ' ms per batch')
662. shortIdClock = every(0.1)
663. on shortIdClock.tick(x) {
664. shortIdTick()
665. return null
666. }
667. }
669. on Error(e) { console.log('error absorbed: ' + e.message) }