gitoriaLog in with ident

tracker

All repositories: gitoria

ReadmeCodePull requestsReleasesTicketsSettings
Commit27c916fa27c916fatracker mission 028: README ("Code order", the new file map), STATUS (counts before/after, tests, how to repeat, open), LOGmre27c916fa/reports/016-tracker-short-ids.md

5.7 KB

  1. # w060: tracker#27, short ids (mission 016 (old 060))
  2. **Commit**: `d8b12d6` on main in loreana:/media/STORAGE/projects/tracker.worldapi.org. Not pushed and not deployed. I did not touch tracker-wt26.
  3. ## What was built
  4. - Every show (movie or series) and every person has a field `shortId`: 5 characters from `[a-z0-9]`.
  5. - Each short id is unique across both tables. One in-memory map in `shortids.hl` covers both tables and is built at start (0.12–0.22 s on real data).
  6. - I did not use an mpackdb index, because an mpackdb index covers only one table.
  7. - **Old ids**: `data/old-short-ids.json` maps the old tracker's 702 ids from oldId to id.
  8. - `tools/old-short-ids.hl` made the file once from `Show.jsonl`. All 702 old ids are already `[a-z0-9]{5}`.
  9. - The file is part of the code, so the deploy sends it along.
  10. - All 702 old ids are reserved: a new random id is never one of them.
  11. - **New records** get a short id when they are created (`claimShortId`): the search import and the titles the person fill adds.
  12. - Those are the only places where the app creates shows.
  13. - Persons are only created by `tools/migrate.hl`, and the sync only creates seasons and episodes.
  14. - **Backfill**: chosen to run in the background, not at start.
  15. - Doing all records at start would block the server for at least 16 s (25k records).
  16. - It runs on its own 0.1 s clock and works in batches of up to 40 ms.
  17. - It saves to disk every 20 batches and at the end.
  18. - After a restart it continues with the records that still have no short id.
  19. - If a stored short id is already held by another record, that record gets a new one (tested in the gate).
  20. - `TRACKER_SHORTIDS=0` switches it off. `TRACKER_SHORTID_BUDGET_MS` and `TRACKER_SHORTID_LOG_EVERY` are its other settings.
  21. - **Shown** as `#shortid` under the poster, and under the person's photo. A person without a photo has it under the name and facts. It is small (.8rem), muted and monospace.
  22. - **Redirect**: `/<shortId>` answers 301 to `/shows/<slug>` or `/person/<slug>`. Upper case works too. Anything else answers 404. This is the last route, so every named route wins over it.
  23. ## Gate
  24. `TRACKER_GATE_SHOTS=/tmp/w060/shots TRACKER_GATE_PORT=8720 TRACKER_GATE_IDENT_PORT=8721 TRACKER_GATE_CHROME=8722-8726 TRACKER_GATE_TMDB_PORT=8727 node tests/browser.mjs`
  25. → **`259 passed, 0 failed`** (was 238; 21 new checks). The gate takes about 1.5 min.
  26. The new checks prove:
  27. - The old id is kept: sh1 → 100jh.
  28. - A movie and a series get new ids.
  29. - A search import has its short id right away.
  30. - The person fill's titles get theirs when they are created.
  31. - The id is under the poster and the photo, and under the facts for a person without a photo. Checked: monospace, muted and smaller, at 390 and 1280 px.
  32. - The 301 redirects and the 404s work, and the browser lands on the right page.
  33. - `tools/count-short-ids.hl`: every record has a short id and there are 0 duplicates.
  34. - Resume: `tests/seed-shortids.hl` adds 60 persons without an id and 2 that share `aaaa1`. A start with a 1 ms budget is stopped part way, and the next start does only the rest. One of the two keeps `aaaa1` and the other gets a new id. A third start has nothing to do.
  35. - I looked at the screenshots in Loreana `/tmp/w060/shots/shortid-*`.
  36. ## Real data
  37. Copy of Byrodin `storage/` (10,010 shows, 15,157 persons), server on :8730.
  38. | | |
  39. |---|---|
  40. | **Backfill duration** | 25,167 records in **297 s** (another run: 307 s) |
  41. | Batches | 1,859; slowest 62 ms |
  42. | Slowest save to disk | 31 ms |
  43. | **Old kept** | **702 / 702** (Raised by Wolves `100jh`, Foundation `100sp`) |
  44. | Lost | 0 |
  45. | New ids that are an old one | 0 |
  46. | **Duplicates across shows + persons** | **0** |
  47. | Duplicate @ids (all tables, `count-duplicate-ids`) | 0 |
  48. **Pages during the backfill** (curl, back to back, about 520 requests each; avg / max / number over 1 s):
  49. | page | during the backfill | without it (baseline, 120 s) |
  50. |---|---|---|
  51. | `/shows/Foundation` | 0.234 / 1.56 / 9 | 0.173 / 1.34 / 5 |
  52. | `/person/Jared-Harris` | 0.170 / 1.37 / 9 | 0.046 / 1.10 / 5 |
  53. | `/` | 0.070 / 1.34 / 2 | 0.022 / 0.99 / 0 |
  54. | `/movies` | 0.073 / 1.18 / 2 | 0.015 / 1.01 / 1 |
  55. - Answers over 1 s happen without the backfill too. Loreana was busy with other workers.
  56. - The backfill's own steps never took more than 62 ms. On average it adds about 0.1 s per page.
  57. **Full page loads** (signed in as the creator, ms, 3 loads each):
  58. | page | 390 px | 1280 px |
  59. |---|---|---|
  60. | Raised by Wolves (opened via `/100jh`) | 1372 / 297 / 310 | 335 / 266 / 301 |
  61. | Foundation | 302 / 306 / 250 | 1246 / 205 / 195 |
  62. | Star Trek: First Contact (`aafif`) | 275 / 287 / 275 | 211 / 235 / 212 |
  63. | Jared Harris (`yg7lk`) | 164 / 138 / 122 | 104 / 119 / 121 |
  64. - Every short id answers 301 in 1–27 ms, and the browser lands on the page. `/zzzzz` answers 404. There were 0 console problems.
  65. - Screenshots on Loreana: `/tmp/w060/real/{old-id,series,movie,person}-{390,1280}.png`. I looked at them. That folder also has older `Grant-Gustin*`/`Jared-Harris*` files that are not mine.
  66. - The copy and its sessions are deleted. Nothing of mine is still running on 8720–8731.
  67. ## Open points
  68. 1. Deploy: the first start runs the backfill (about 5 minutes). Short ids appear page by page while it runs.
  69. 2. Inside the app, one record costs about 3 ms; in a plain script it costs about 0.6 ms. I did not find out why.
  70. - **Hybriel finding**: reading a list kept as a field of a `static` object (`state.queue[i]`) seems to copy the whole list on every access. Batches got about 5× faster after I made the list a plain `static`. I made no minimal repro.
  71. 3. Nothing in the app creates persons today. A later step that does must call `claimShortId('person', id, oldId)`.
  72. 4. `/<shortId>` of an adult title still redirects. That page then says "Not found" to anyone who does not follow it, as before.

Branches

Latest commits

  • 27c916fatracker mission 028: README ("Code order", the new file map), STATUS (counts before/after, tests, how to repeat, open), LOGmre
  • d924f398tracker mission 028: comments name the new files (sync.hl, dailysync.hl, backfill.hl, credits.hl, jobs.hl, images.hl …); tools/ref-params.py + tools/lambda-audit.py also scan lib/ (they globbed the root only), lambda-audit counts a plain `x = p` alias like `let x = p`mre
  • 2e89b968tracker mission 028 (code order) 5/5 let: `let` only where a variable is reassigned — 667 never-reassigned lets became plain declarations (project.hl, lib/, components/, tools/, tests/); kept: 264 in loop bodies (a plain declaration there is 'Cannot reassign' on the 2nd pass), 234 reassigned, 27 whose name is also a member/outer/free name (a plain write would rebind it); tools/let-audit.py decides and fixes (README 'Code order'); tests/realdata-m028.{sh,mjs} = the page-output diff on a real copy; gates 342/0, 32/0, 52/0, real-copy pages identicalmre
  • 54796ff2tracker mission 028 (code order) 4/5 thin faces + last copies: the show page's check/follow faces call lib/watches.hl toggleWatched / toggleSeasonWatched (seasonAllWatched moved there) and lib/follows.hl toggleFollowed; both logins (header selector face, /login/callback) share lib/users.hl userOfCode; todayStr/listOf copies in components and the export readers copied into tools/migrate.hl + tools/old-short-ids.hl now once (lib/util.hl, lib/export.hl); gates 342/0, 32/0, 52/0; old-short-ids output byte-identical, migrate output identicalmre
  • 06b078e3tracker mission 028 (code order) 3/5 project.hl is the map: config, routes, wiring and a feature → file index (914 → 258 lines); the background jobs (daily sync run, backfills, details repair, credits job, merge, short ids, collection seed) moved unchanged into lib/jobs.hl (a class: their state is reassigned every step, a static cannot be; one instance made after the server), the login callback into lib/users.hl, poster/photo serving into lib/images.hl, the /shows/<slug> rule into lib/shows.hl showsMovedPath; route handlers are thin wrappers; gates 342/0, 32/0, 52/0, real-copy pages identicalmre
  • 94716fd2tracker mission 028 (code order) 2/5 util + topics: lib/util.hl holds envOr, storageDir, postersDir, profilesDir, newId, hexDigits, todayStr, dateOr, textOr, hasId, listOr, firstOf, sortDesc once (were copied into up to 5 files); tmdbsync.hl split into tmdb.hl (TMDB/TVmaze requests), sync.hl (one title's sync), sync-helpers.hl, backfill.hl; details.hl split into details.hl, credits.hl, credits-helpers.hl (isIncomplete to shows.hl); search-helpers.hl (words, query, ranking, slugs); collections.hl (the TMDB collection seed, out of franchises.hl); deltasync.hl renamed dailysync.hl; no behaviour change: gates 342/0, 32/0, 52/0, real-copy pages identicalmre
  • 186079b0tracker mission 028 (code order) 1/5 move: every root .hl except project.hl into lib/ (styles.hl into components/), import paths only; gates 342/0, 32/0, 52/0; real-copy pages identicalmre
  • 4f47f181tracker: report 027mre
  • dc1d4be4tracker mission 027: Hybriel master 06617221 vendored (plugin allocator fixes 3a781359 + 413f60e4); real copy RSS through first-start jobs + 400 loads flat ~2.55 GB (190aa11d 2.3 -> 5.6 GB), page times <= 1.1x; gates 342/0, 32/0, 52/0mre
  • 84e1b3e1tracker: reports 025 + 026mre
  • dc40d859tracker#31 (mission 026): duplicate titles merged — the 68 type+tmdbId pairs held by 157 records were the old tracker's (all migrated); merge.hl repair job (own clock, before the TMDB jobs) keeps one keeper per title (follows/watches > old short id > oldest), moves follows, watches, seasons, cast, credits, timelines, tombstones the rest (mergedInto, never deleted), slugs + short ids 301 to the keeper; stray seasons merged into their listed twin (Reacher S3 watches) or linked when watched; search import re-checks before its put; deploy.sh waits up to 90 s for 200; real copy 68 -> 0 dup ids, az5b2 follows/watches equal; gates 342/0, 32/0, 52/0mre
  • 2667da05tracker#30 (mission 025): /my/ pages from slim cached title cards, episode rows and watch sets (after the jobs /my/series 1.8 s -> 0.06 s, /my/unwatched 4.1 -> 0.18 s); timeline page shows its name once; franchise widget under the poster/title; movies with TV leftovers (First Contact) go through the details repair; tools/count-tmdb-ids.hl; gates 327/0, 52/0, 32/0mre
  • 3909810dantcolony#40: mission references in README/STATUS/docs point to the moved missionsmre
  • e63b1d28antcolony#40: history (LOG.md), worker briefs (missions/) and reports moved here from antcolony, numbered per project; old numbers in antcolony docs/mission-map.mdmre
  • 1cda451dtracker: Hybriel master 190aa11d (#127 both shapes, GC correctness fc838894) — conductor adopts despite /my/series 2.4x after jobs (memory 8.0 → 1.7 GB boot); see reports/071mre
  • c1fa2f2etracker (mission 071): Hybriel master 190aa11d measured on the real copy vs the live binary 8590df63 — NOT adopted (after the first-start jobs /my/series 2.4x slower, /series 1.6x, RSS swings 7.4-12.2 GB; fresh it is flat at 1.7-2.2 GB and /my/unwatched faster), vendor stays 8efba065, candidate kept in .scratch/w071/vendor-190aa11d; tests/kinds.mjs: collection seed off (its TMDB request broke check 1 in 1 of 4 runs); tests/realdata-071.sh + realdata-071-bench.mjs + tools/realdata-071-table.py; README + STATUS (numbers, how to repeat); gates 325/0, 32/0, 50/0mre
  • 8081350atracker docs (mission 070): README (summary, Config HL_GC_BYTES — kept at Hybriel's default, the 256 MiB setting is taken out of docker-compose.yml again: the jobs grew to 12+ GB with it too, see STATUS), Test (three gates), Deploy (first start ~50 min: kinds then seed, restart once after collections done, memory numbers), Vendored Hybriel 8efba065 + #48 audit, Files; STATUS mission 070 entry (merges, migrated counts, lambda audit, gates, RSS old vs new, how to repeat, open points); docs/kinds.md + docs/franchises.md job order; tests/realdata-070-*.sh, tools/count-migrated.hl, tools/ref-params.py, tools/lambda-audit.pymre
  • 1ad19c8ctracker: re-vendor Hybriel master 8efba065 (#126 GC by bytes, #48 lambda parameters copy) (mission 070): bin/hybriel sha256 50361e95…, plugins core crypto data fetch fs http http1 mpackdb proc smtp time web; lambda audit: 13 lambdas change a passed record/list (11 through a local alias), no caller relies on it — unchanged; 319 read-only lambda parameters get & (no copy per call: /my/schedule 2.35 → 0.46 s, /my/unwatched 13.7 → 5.7 s on the real copy); /my/unwatched one merge sort instead of n² inserts; docker-compose HL_GC_BYTES=268435456; deploy.sh runs kinds.mjs + franchises.mjs too (default ports 8700–8710); tests/realdata-070.mjs; gates browser 325/0, kinds 32/0, franchises 50/0mre
  • 46b21389tracker (mission 070, conductor): the sync never destroys migrated data — the one-time summary step MOVES a copied summary to migratedSummary (marker summaries-moved.txt) instead of clearing it; a migrated record's first title/genres/homepage/tagline TMDB replaces → migratedTitle/migratedGenres/migratedHomepage/migratedTagline, a migrated season's/episode's title/summary → migratedTitle/migratedSummary (set once); tests/peek-shows.hl prints them; gate 325/0mre
  • 749019b0Merge t19 (tracker#19 franchises + timelines) into main (mission 070): conflicts README/STATUS/show.hl/project.hl/browser.mjs/faketmdb.mjs, both sides kept; franchise/timeline pages get #20's typed heading (Franchise | …, Timeline | …), their title links via titlePath (/movies|/series|/shows); the collection seed waits for repair, kinds and credits too; franchises.mjs URLs + 2 new checks; gates browser 323/0, kinds 32/0, franchises 50/0mre