Skip to content

perf: retire duplicate nested docs HTML objects - #261

Merged
hannesrudolph merged 9 commits into
mainfrom
fix/docs-r2-single-html-object
Oct 3, 2026
Merged

hannesrudolph merged 9 commits into
mainfrom
fix/docs-r2-single-html-object

Conversation

@hannesrudolph

@hannesrudolph hannesrudolph commented Oct 3, 2026 •

Copy link
Copy Markdown
Member

Retire the redundant nested HTML objects from the docs R2 artifact while retaining both public URL forms. The compatible Worker from #260 serves /route/index.html from /route, with legacy fallback.

The preparer omits a nested HTML entry only when exactly one canonical alias refers to the same hashed source and every manifest field other than the storage key matches, including all HTTP/custom metadata. Missing, ambiguous or mismatched aliases retain both objects. Physical index files remain for previews. Full publication uses the existing manifest cleanup; partial uploads keep legacy keys.

Compatible reader verified: Worker 1e1d3eb1-e61e-4d97-9caf-c0e57de329c9 was deployed from d493cce3b3c4531aa7234303b1c7ffd4c2f923cc by Pages run 37099918404, successfully completed at 2026-10-03 05:29:01 UTC. The served 39-request matrix passed canonical/explicit GET/HEAD, locale, dotted/encoded, root/404, asset and Markdown checks. Proof SHA256: 1e766554296b2ebdc724f5ab2d94dc6190e9a747c7701312b88fc1672d89e769. Keep this PR draft until its required CI is green and fresh head/base guards pass. No manual bucket deletion or overlapping publication. The hosting runbook documents rollback: restore dual-key artifacts on current main before an older Worker, using the dual-key generator baseline 60a3699e2d39047044bd11d99f64b758efde8417 while preserving current docs.

Validation:

  • Full Node suite: 294 pass, 1 optional SDK skip; structured code review clean.
  • Full local build/smoke: 15,459 pages, 62,665 objects (external source-code index excluded).
  • Frozen corpus: 91,197 -> 62,665 objects; exactly 28,532 redundant nested keys omitted, zero retained object changes. Every omitted key has identical retained metadata/bytes, and every physical file was rehashed and preserved.
  • Real builder/Worker tests retain canonical and explicit HTML URLs, GET/HEAD, dotted/encoded/localized routes, Markdown redirects and asset contracts.
  • Dry-run full upload deletes only the expected duplicate key; partial upload retains it.

The local 28,532 count is a frozen-corpus measurement. The production receipt will report the actual count from its current docs snapshot.

@hannesrudolph
hannesrudolph marked this pull request as ready for review October 3, 2026 05:30
@hannesrudolph
hannesrudolph requested a review from a team as a code owner October 3, 2026 05:30
@hannesrudolph
hannesrudolph merged commit 6ad89b0 into main Oct 3, 2026
7 checks passed
@hannesrudolph
hannesrudolph deleted the fix/docs-r2-single-html-object branch October 3, 2026 05:30
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