Skip to content

Restore missing podcasts and resolve duplicate S03E04 identity - #4

Merged
alexeygrigorev merged 1 commit into
mainfrom
fix/podcast-legacy-identities
Nov 15, 2026
Merged

alexeygrigorev merged 1 commit into
mainfrom
fix/podcast-legacy-identities

Conversation

@alexeygrigorev

Copy link
Copy Markdown
Member

Restore the five podcast source identities that still prevented legacy .html URLs from reaching a canonical page after website PR507.

  • Resolve the duplicate S03E04 from publisher RSS: Oleg Novikov is explicitly S03E05 (2021-05-07); Lior Barak stays S03E04 (2021-04-30, between E03 and E05). Move metadata and transcripts together; preserve slugs and platform IDs.
  • Restore S25E07/E08/E09 metadata, resources and unchanged transcript segments from legacy commit 6265f8fe8d9e68575281c7dcb8b8f1f01a6a7697. RSS confirms episode numbers; E08 was published after E09, so do not renumber by date. Copy E07 artwork; use the corresponding original YouTube thumbnails for E08/E09, whose legacy records lack images.
  • Repair existing red CI: validate the current catalogue on this exact head, while running frozen migration verification/attestations against their last green historical release 9ad8f50. Historical attestation counts/census cannot describe later editorial additions. No immutable contract is loosened.

Full evidence, dates, guests, GUIDs, YouTube/Spotify/Apple IDs, inference for Lior's omitted RSS episode number, CI scope and reversal are in migration/podcast-identity-resolution.md. Mapping is trivially reversible by reverting this source commit; no runtime special cases or fallback reads.

Validation: full UV_FROZEN=1 make check (168 tests); original migration verified against pinned legacy source and both historical attestations reproduced. Independent tester verifies field/segment fidelity, unique IDs, artwork and CI scope; PM accepts documented identities/restoration before commit. After merge, sync this repository into production's database-owned catalogue and rerun all 215 legacy URLs, including raw query preservation and exactly one 301 hop to 200. Website code/CI/deploy proof reuses PR507.

Refs DataTalksClub/website#162; Hub intake 9a6ec0b2. No courses.datatalks.club or DNS/CloudFront changes.

Refs DataTalksClub/website#162; Hub intake 9a6ec0b2. Publisher RSS confirms Oleg Novikov S03E05. Restore S25E07-E09 from legacy, retain exact transcripts, and scope immutable migration CI to its frozen release.
@alexeygrigorev

Copy link
Copy Markdown
Member Author

Independent tester gate passed: restored E07/E08/E09 metadata and every transcript segment match legacy, E07 image bytes match, S03 source/transcript moves preserve unrelated data, all 215 IDs are unique, focused tests pass, and fetch-depth: 0 correction reviewed. Independent PM acceptance passed: RSS evidence and reversible inference are documented, source restoration meets the five-failure scope, and no runtime exception is introduced. Full local validation: 168 tests passed, lint/format/diff clean; pinned historical verify-source and both attestations reproduced successfully. Production sync and all-215 live curl acceptance follow merge.

@alexeygrigorev
alexeygrigorev merged commit 032082b into main Nov 15, 2026
1 check passed
@alexeygrigorev

Copy link
Copy Markdown
Member Author

Merged after exact-head CI success: head54e10ac9eb4cea186f38679bb9dcb40f707a4c7a, validate run43733552228, merge032082b4852c27839150f85f9643da31023dc3d8. Production import first hit the existing 100MB archive limit (source media ~149MiB); retry uses the same management importer with a bounded256MiB setting for that process only. No persistent app configuration or runtime read path changes. Final live215 proof follows.

@alexeygrigorev

alexeygrigorev commented Nov 15, 2026 •

Copy link
Copy Markdown
Member Author

Intake 9a6ec0b2 LIVE FIX VERIFIED 2026-11-15: all215 legacy podcast .html URLs now pass on prod.datatalks.club.

Website PR507 DataTalksClub/website#507 head e2eaa7bf9f5234c7591a41a9ea01b6af88d00b24:2/2 checks green, merged e6a816e85d2e0b3626199184de05267e57aa4669, deployed descendant d3aecc9baf0c6fbfc037394498deeae2840c5a23 (ahead4,behind0). Hub website merge-check ok. Final /health/ready ready, homepage200.

Resolved five source failures in #4: exact head54e10ac9eb4cea186f38679bb9dcb40f707a4c7a CI green https://github.com/DataTalksClub/content/actions/runs/43733552228, merged032082b4852c27839150f85f9643da31023dc3d8; merge CI green https://github.com/DataTalksClub/content/actions/runs/43733853626. Independent tester+PM passed;168 tests/lint/format passed. Publisher RSS https://anchor.fm/s/41286f68/podcast/rss explicitly marks Oleg Novikov S03E05 (2021-05-07, GUID05b6b633-aae8-4bec-872b-beef313a9922, YouTubeAYi7b-8GPm4). Lior Barak stays S03E04 (2021-04-30 between RSS E03 and E05; GUIDf3b3e084-7cba-4132-90c3-9c4ef73ec2c7, YouTubegqroEsTyLD0). Source/slugs/platform IDs/transcripts preserved. Reversible reasoning and RSS/platform/date evidence documented in PR and migration/podcast-identity-resolution.md. Restored S25E07/E08/E09 from legacy6265f8fe8d9e68575281c7dcb8b8f1f01a6a7697 with exact transcript segments and matching source/video artwork; RSS confirms numbers despite E08/E09 date order.

Production sync_content dtc-content succeeded at exact merged032082b4: created6 updated2 unchanged1195 deleted0. Production has215 published episodes and215 unique season/episode IDs; five repaired canonical rows confirmed. Default100MB archive cap was exceeded; successful import used same management importer with process-only256MiB bound (media149MiB). Persistent config unchanged. Temporary operational binary/source workspaces removed from production and host.

Live verification 2026-11-15T16:18:36.972750+00:00: uv run --frozen python .tmp/podcast-resolution/verify_podcast_live.py live-all215.json;215/215 pass,0 failures,0 retries.860 curl invocations, sequential with0.2s spacing, normal Mozilla/5.0 User-Agent. For every original static-sitemap .html URL: plain and raw query utm_source=legacy%2Btest&x=a%2Fb&x=&q=A+B&q=A%20B return301 with canonical sXXeYY Location and exact query preservation, then200 after exactly one hop. Original sitemap SHA256266f7a7b91935f500fbe6bef130c228768e636bd861dd44ae5e5e2087fd46db0. Full report SHA25695a8d1b91175bc8ff7c611a99b1d3839d318d6e60967801fa63b95fa44dd52f6.

Evidence directory: /home/alexey/git/drain-81995d56/.local/hub-worktrees/ee2075555fa5d293de65e0a2e2fd02412f7015f313424bb07278889c1574a307/.tmp/podcast-resolution/ (live-all215.json/log, production-sync-final.log, production-source-proof.log, content-pr-ci.json, content-main-ci.json, content-merged-pr.json, content-merged-ancestry.json, hub-website-merge-gate.json, deployed-ancestry.json, readiness-final.json, courses-live.txt).

No courses.datatalks.club changes: live root200, no Location,0 hops. No DNS/CloudFront changes. Technical/live acceptance passed. Parent Hub append remains pending because the worker still receives HTTP403 missing intake:append; no session_finished record has been submitted.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant