Keep timeline dates tied to note routes

[srid]

Timeline heatmap backlinks now get their dates from the daily note route, not from the rendered note title. That keeps daily-note detection in the same Haskell parser that already decides whether a backlink belongs in the timeline, so custom-titled daily notes still appear in the heatmap.

The data now flows as a route-derived Day through the backlink splice and into data-iso-date; the browser only groups already-serialized ISO dates.

Calendar.parseRouteDay
  -> daily backlink splice
  -> data-iso-date
  -> timeline heatmap grouping

The smoke fixture now retitles one daily note to cover the regression, and the backlinks guide/changelog describe the route-derived date contract.

Try it locally

nix run github:srid/emanote/fix/timeline-iso-date

Generated by /do on Codex (model gpt-5).

[srid]

Hickey/Lowy Analysis

#	Lens	Finding	Disposition
1	Hickey	Daily backlink date was recomputed instead of carried	Fixed in this PR
2	Lowy	No boundary findings	No-op

Hickey rationale

Hickey flagged that the first implementation classified daily backlinks with Calendar.isDailyNote, discarded the parsed Day, then parsed the route again while rendering backlink:note:iso-date. That split one domain value across classification and template data. The follow-up commit 560f44f0 carries the parsed Day through the daily backlink representation and renders data-iso-date from that carried value.

Lowy rationale

Lowy found the boundary aligned with the intended volatility: daily-note route semantics stay in Haskell behind Calendar.parseRouteDay, while timeline JS now consumes a stable ISO-date DOM attribute and only owns browser-side grouping/rendering. No boundary restructuring was needed beyond the carried date refinement above.

[srid]

Evidence

The custom-titled daily note still appears in the Timeline as a route-derived daily backlink.

[srid]

/do results

Step	Status	Duration	Verification
sync	✓	1s	git fetch ok; forge=github; noGit=false
research	✓	2m 7s	Timeline daily backlinks use backlinksSplice in Template.hs; timeline.tpl emits hidden li data-title/data-url; timeline-heatmap.js parses DATE_RE from title. Change: add backlink:note:iso-date from Calendar.parseRouteDay, emit data-iso-date, consume dataset.isoDate, and extend e2e assertion.
branch	✓	10s	Created feature branch fix/timeline-iso-date from origin/master
implement	✓	1m 3s	Added e2e assertion, Haskell backlink iso-date splice, timeline data-iso-date attribute, and JS parsing from dataset.isoDate instead of title regex
check	✓	1m 27s	cabal build all passed via nix develop (ambient shell lacked cabal)
docs	✓	44s	Updated backlinks guide and changelog; retitled a daily backlink fixture so e2e covers custom-titled daily notes
fmt	✓	11s	just fmt passed (pre-commit: cabal-fmt, fourmolu, hlint, nixpkgs-fmt)
commit	✓	16s	Committed 2bde2be fix(timeline): use route dates for heatmap entries and pushed origin/fix/timeline-iso-date
hickey+lowy	✓	5m 22s	Hickey found daily date not carried; fixed in commit 560f44f. Lowy found no actions.
police	✓	32s	All clear: rules, fact-check, and elegance passes found no actionable violations; cabal build remained up to date
test	✓	1m 33s	nix develop -c cabal test all passed; just e2e-static passed; just e2e-live passed
create-pr	✓	27s	Created draft PR #705 and posted Hickey/Lowy analysis comment
ci	✓	11m 31s	vira ci passed on b35fdcc with vira/aarch64-darwin and vira/x86_64-linux signoffs; just e2e-live, just e2e-static, and just e2e-morph passed
evidence	✓	1m 50s	Posted ## Evidence PR comment with screenshot of /dailyhost.html Timeline showing the custom-titled daily note still appears
done	✓	0s	Final /do results generated and posted to PR #705
Total		29m 30s

Slowest step: ci (11m 31s)

Optimization suggestions

• ci dominated this run. For follow-up-only changes on this PR, resume with $do --from ci-only after pushing the fix so the workflow does not replay the earlier implementation/review steps.
• The morph e2e timeout came from the live-test harness not draining stdout. That is fixed in b35fdccf; future JS/live-server PRs should run just e2e-morph before full CI because it is a fast detector for this class of issue.
• Hickey/Lowy took 5m 22s and caught a real duplicate-parse concern, so it was worthwhile here. For genuinely mechanical one-line fixes, $do --minimal would avoid that review cost.

Workflow completed at 2026-05-04T20:13:00Z.

[srid]

Merged current origin/master (716976d3) into this branch and removed the redundant local stdout-drain listener, so the PR diff no longer includes tests/support/hooks.ts.

Reverified on head 8d2b2009:

• vira ci passed, including vira/aarch64-darwin and vira/x86_64-linux signoffs
• just e2e-static passed: 42 scenarios, 3 skipped, 39 passed
• just e2e-live passed: 42 scenarios passed
• just e2e-morph passed: 42 scenarios passed