mapDocument re-ran the alphabetical min() scan over the letter's tag set to
look up its already-resolved root, duplicating the work resolveLetterRootTags
had just done and leaving two independent definitions of "primary tag" that
could silently diverge. Key the resolved-root map by document id and compute
the primary tag exactly once per letter; drop the redundant resolvePrimaryRoot
helper.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add one row per requirement for the zeitstrahl-tag-chips feature, each mapped
to its implementation file(s) and the test(s) that prove it, Status=Done.
Refs #835
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The doc comment described escaping by naming the raw-HTML directive literally,
which trips the lib/timeline grep gate that forbids that token. Reword it the
way LetterCard already does — behavior unchanged.
Refs #835
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
LetterCard now renders a TagChip beneath the sender→receiver/date line
whenever the entry carries a rootTagName, mapping rootTagColor to the chip
(neutral when null). Because the chip lives on LetterCard it shows up wherever
a LetterCard does — the global timeline and the expanded YearLetterStrip — with
no per-surface special-casing; a tagless letter shows no chip. A long name
truncates inline so the card never overflows at 320px.
Refs #835
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
TagChip renders a letter's primary root tag as a small rounded pill — a
decorative aria-hidden colored square (var(--c-tag-{token}), neutral when the
color is null) plus the escaped tag name, prefixed by the sr-only theme label
so color is never the only cue. Truncation is set inline so a long name
ellipsizes without forcing the card into horizontal scroll, and the full name
stays reachable via the chip title. Timeline-local by design — lib/timeline may
not import lib/tag (eslint boundary).
Refs #835
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
timeline_tag_chip_label (de "Thema" / en "Topic" / es "Tema") is the sr-only
prefix the /zeitstrahl letter tag chip reads out so color is never the only
cue. Pinned per locale in messages.spec.ts; the tag name itself is rendered as
data, never translated.
Refs #835
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
openapi-typescript pickup of TimelineEntryDTO.rootTagId/rootTagName/
rootTagColor (all optional), so the SvelteKit timeline can read the new
letter chip fields. Regenerated from the live dev spec; only the additive
fields differ from the committed baseline.
Refs #835
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
TimelineEntryDTO gains three nullable letter-only fields — rootTagId,
rootTagName, rootTagColor (token) — assembled in-transaction in TimelineService
(ADR-036): id + name + token only, never a serialized Tag entity. A letter's
primary tag is the root ancestor of its alphabetically-first assigned tag
(#827 Resolved Decision 3); roots are resolved through TagService in one
batched pass over the distinct primary tags (no per-letter N+1). The fields are
null for non-letter entries, untagged letters, and (color only) a colorless
root, so they are deliberately not @Schema(requiredMode = REQUIRED).
Refs #835
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
TagService.resolveRootTags(tags) maps each tag to its root ancestor as a
RootTag (id, name, color token), keyed by the input tag id. A root maps to
itself; a child is walked to the parentless ancestor via the existing
recursive-CTE findAncestorIds — one CTE per distinct non-root tag (memoized),
plus a single batched findAllById — so a timeline of many letters sharing few
tags costs O(distinct tags) queries, never O(letters). The color is read from
the resolved root's stored token (null when the root has none).
This is the shared enrichment the /zeitstrahl tag chip (#835) and, later, the
Thema buckets (#827) both consume. Unit-tested in TagServiceTest; the
DB-dependent ancestry walk is pinned against real Postgres in
TagServiceIntegrationTest.
Refs #835
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The spine offset (0.5rem phone / 50% desktop) was hard-coded in both
TimelineView's .timeline-axis::before and YearBand's .year-node/.letter-dot,
kept in sync only by a comment. Declare --spine-x once on .timeline-axis
and have the markers consume it by inheritance, so a change to the spine
position moves the markers with it. Add a test that the year-node tracks
the token.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The aria-hidden glyph + sr-only label markup was hand-copied in LetterCard
and YearLetterStrip. Extract a small GlyphLabel component and use it at
both sites so the accessibility idiom has a single owner.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace the flatMap intermediate array plus two filter passes with one
walk over the year bands and the undated bucket. Same counts, no
throwaway allocation.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The "· historisch" register was emitted in all three date branches, with
the dateless branch dropping the leading separator. Render the span pill
or date as a conditional prefix, then a single trailing "· historisch"
span — one render site, consistent separator.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A count of one rendered "1 Briefe" / "1 Ereignisse". Add _singular
companion keys (de/en/es) and select them when the count is exactly one,
following the project's _singular/_plural convention.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A timeline with content in only one category rendered a misleading
"0 Briefe" / "0 Ereignisse" segment. Only push a count segment when its
count is greater than zero; the grouping label and (when present) the
range are unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The provenance token (abgeleitet/kuratiert) was nested inside the
{#if dateLabel} block, so an undated or UNKNOWN-precision event — e.g.
one in the undated bucket — rendered no provenance at all. Compose the
subtitle as an optional "{date} · " prefix in front of the always-present
provenance instead.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The page is already bg-canvas, so the frame's border was the only thing
making it visible; per review it reads cleaner without it. Keep the padded
bg-canvas surface; the timeline sits on the page without a frame line.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The absolutely-positioned spine ::before painted above the in-flow centered
content (density strips, event pills), drawing the line through them. Give
.timeline-axis a stacking context and the spine z-index:-1 so the line is
always background; cards, pills, strips, dots and badges ride on top.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The node marker carried a higher stacking order than the (non-positioned)
badge, so on the centered desktop axis the navy node painted over the white
year digits. Make the badge positioned with the higher z-index; the node now
sits behind the centered pill (which is itself the axis interruption) and
shows only to the badge's left on phone. Guarded by a z-index assertion.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
All sixteen visual-fidelity requirements have a green test (142 timeline +
zeitstrahl + messages tests pass); record the implementation files and the
test that proves each.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Pass the full PageData shape (layout auth fields + timeline) through a
pageData() helper so the route spec is svelte-check-clean; the component
still only reads data.timeline.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The spine now runs mint → navy → slate, matching the spec's life-thread,
using --palette-mint / --palette-navy / --c-tag-slate (no --palette-slate
token exists). Semantic tokens only — no raw hex.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The year badge now centers on the axis at ≥1024px and hugs the left spine
below that (sticky top:4rem preserved), with a navy node marker so it
visibly interrupts the spine. Each letter row gains a connector dot (white
fill, mint ring) on the spine: centered between card and axis on desktop,
on the left spine clear of the indented card on phone. Spine geometry is
commented to track TimelineView's spine so the markers can't silently desync.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The timeline now sits inside a bordered, rounded bg-canvas sheet. Below the
heading a sub-line composes the year range, the letter and event counts
(from timelineMeta), and the static "Gruppierung: Datum" — joined by " · "
so the range drops out when there are no year bands and the whole line is
absent for an empty timeline. Semantic tokens only; AA-legible text-xs.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A pure timelineMeta() returns the year range (first/last band, null when
there are no bands) and the letter/event totals across all year bands plus
the undated bucket — the single place these counts are computed.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The "Ohne Datum" section now renders inside a dashed-bordered surface box
whose heading reads "Ohne Datum · {count}", matching the spec's .undated
treatment. The kind/type dispatch (events as pills/bands, letters as cards)
is unchanged; the section stays absent when there are no undated entries.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The dense-year strip count now carries the ✉ glyph (aria-hidden + sr-only
"Brief"), and beneath the sparkline a "Monats-Dichte" caption sits between
the two endpoint month labels (Jan/Dez {year}) at the ≥10px micro-axis
floor, localized via the shared month formatter. The ≥44px keyboard-
focusable "Briefe anzeigen" expand toggle is preserved unchanged.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A HISTORICAL band's subtitle now carries the visible "historisch" register
inline as plain text: "{date} · historisch", or — for a RANGE — after the
existing 1914–1918 span pill (whose Zeitraum aria-label is unchanged). The
descriptor is a text node, never a second pill. Every WorldBand is
historical, so the suffix also trails an undated band on its own.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A present LetterCard title now reads "✉ {title}" with an aria-hidden glyph
and an sr-only "Brief" label rendered as sibling nodes — never interpolated
into the escaped user title, which keeps its own pre-line span for
multi-line OCR text. No title → no glyph, no label (the row still shows
sender → receiver and the date). An XSS regression pins the no-{@html}
contract: an HTML-bearing title renders verbatim as text.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The derived/curated pill subtitle now reads "{date} · abgeleitet" or
"{date} · kuratiert", keyed off entry.derived so a reader sees both the
date and whether the event was derived from Person data or curated. Only
the single provenance token ships; the spec sheet's "· persönlich" /
"· SEASON" annotations stay out (already covered by the ★ glyph + mint
border and not production UI).
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Seven new timeline_* keys feeding the upcoming chrome: the header meta
line (grouping label + events count), the event-pill provenance token,
the ✉ sr-only label, the world-band "historisch" suffix, and the strip
density caption — present in de/en/es with matching key sets, pinned by a
new parity test.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Trace the visual-fidelity requirements end to end before implementation,
matching the timeline-feature convention of landing RTM rows in the first
commit of the branch. Status: Planned; flipped to Done as each task lands.
Refs #833
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The submitted type comes from EventTypeSelect's own hidden input; EventForm's
`type` $state was only read to seed `value=` and the onchange reassignment
had no downstream reader. Inline the seed expression and pass markDirty
directly, removing the second source of truth.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A RANGE event with a blank end date passed validateEventForm and reached
the backend, which 400s with a generic INVALID_DATE_RANGE mapped to "end
must not be before start" — wrong for a missing end date, and shown only as
a top-of-form alert. Validate it before the API call and surface a dedicated
event_editor_end_date_required message on the end-date field via a new
DatePrecisionField endDateError prop (defaults '', so the document form is
unchanged).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
preservedFormFields echoed back only title/type/description/pickers, and
EventForm seeded dateIso/precision/endDateIso solely from `event` (undefined
on /new). So a no-JS validation-error reload silently dropped the entire
When-section the curator had entered, while every other field survived.
Echo eventDate/precision/eventDateEnd in the fail payload and seed the date
controls from `form` ahead of `event`.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
titleError/dateError short-circuited on the server fail payload
(`form?.titleError ?? …`, `form?.dateError ?? ''`), so after a fail(400)
the red border and message stuck until the next submit even once the user
typed a valid value. Derive both from the current field value instead: the
server error still seeds the message, but a non-empty title/date clears it
immediately.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The delete <form> combined onsubmit={confirmDelete} with use:enhance.
SvelteKit's enhance ignores event.defaultPrevented, so the DELETE fired on
the bare click — before the dialog was answered — and the post-DELETE
redirect ran regardless of the user's choice. Reading getConfirmService()
lazily inside the handler also threw (Svelte context is init-only), so the
dialog never appeared even with <ConfirmDialog> mounted.
Capture confirm at init and run the gate inside the enhance submit phase,
calling cancel() on "no". Clear dirty in the result callback so the
beforeNavigate guard no longer prompts "unsaved changes" on the post-delete
redirect.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The EventForm onsubmit guard called e.preventDefault() on a blank title,
but use:enhance ignores defaultPrevented (forms.js only bails on cancel()),
so a blank-title Save still fired a network POST. In a component unit test
the resulting update() -> applyAction() dereferenced an undefined root
($set on undefined), surfacing as an unhandled rejection.
Move the guard into the enhance submit phase and call cancel() so the POST
is actually stopped; the server still owns the authoritative fail(400).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The round-2 dirty-guard used an $effect that both read and wrote its `dirtyArmed`
$state, so the self-write re-triggered the effect and `dirty` flipped true on
mount — the beforeNavigate confirm then fired on every navigation away from an
untouched form (caught by the round-3 clean-agent review + the Svelte autofixer,
which flags assigning state inside $effect).
Replace it with the component's existing idiomatic pattern: DatePrecisionField,
PersonMultiSelect, and DocumentMultiSelect each gain an optional `onchange`
callback fired on a real user edit, and EventForm passes `markDirty` to all
three. Now date/precision/end-date and picker add/remove mark the form dirty
exactly like title/type/description already did — no effect, no mount-timing
trap. The new props are optional, so the other consumers (WhoWhenSection, the
document forms) are unaffected. Svelte autofixer: clean.
Addresses PR #832 review (round-3 clean-agent concern).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The beforeNavigate unsaved-changes guard only fired for title/type/description
(their oninput/onchange hooks set `dirty`). Editing only the date, precision,
end-date, or the linked persons/documents left `dirty` false, so a curator could
navigate away and silently lose those edits — defeating the guard for the senior
author audience. Add an $effect that watches those values and marks dirty on any
change after the initial prop snapshot (first run only arms the watcher).
Addresses PR #832 review (round-2 clean-agent concern).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
DatePrecisionField derives the precision select's id from dateInputName, so the
document form's id moved from #metaDatePrecision to #documentDatePrecision (the
name attribute is unchanged). This maintained E2E still queried the old id and
would fail when run. Not CI-gated, but a real silent breakage.
Addresses PR #832 review (round-2 clean-agent out-of-diff finding).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The DatePrecisionField extraction derives element ids from dateInputName, so the
document form's precision/end-date ids changed (metaDatePrecision →
documentDatePrecision, metaDateEnd → documentDateEnd, date-error →
documentDate-error, end-date-error → documentDate-end-error). The name attributes
are unchanged, so form submission is unaffected — but the pre-existing
WhoWhenSection.svelte.test.ts (a separate file from the .spec.ts) still queried
the old ids and was failing 5 assertions in CI's client project. It wasn't in
the PR diff, so the multi-persona review missed it. Re-point the selectors.
Addresses PR #832 review (round-1 clean-agent blocker).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Architect/Developer review suggestion: flag that other WRITE_ALL-gated author
loads (e.g. documents/[id]/edit) still inline the throw-403 guard and can adopt
requireWriteAll so it doesn't diverge. Comment-only.
Addresses PR #832 review (Architect suggestion).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Tighten the review-flagged test gaps (no production change):
- submitting state: the old test only asserted the button was initially
enabled (a tautology). Now stub a never-resolving fetch, click Speichern, and
assert the button gains `disabled` — exercising the double-submit guard
(Decision 8).
- the two redirect-throwing save tests now use `await expect(...).rejects` so a
future missing redirect fails loudly instead of being swallowed by try/catch.
- the YEAR end-date-hide assertion uses getByLabelText('Enddatum') not-present,
symmetric with the RANGE reveal, instead of a raw #eventDateEnd querySelector.
- PersonMultiSelect + DocumentMultiSelect: assert the chip remove button carries
the min-h-[44px]/min-w-[44px] target the rtm cites for REQ-017.
Addresses PR #832 review (Tester + Requirements Engineer test-coverage concerns).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
REQ-011 asks for both required-field errors via per-field aria-invalid, but the
blank-date error was rendered as a standalone <p> after the date card while the
date input's aria-invalid only reflected the client-side malformed-date cue.
DatePrecisionField gains a `dateError` prop: a server error now marks the field
aria-invalid and renders inline under the input (sharing the same error id), and
EventForm drops its detached <p>. While here, migrate the field's two error
texts from hard-coded text-red-600 to the semantic `text-danger` token so they
keep ≥4.5:1 contrast in dark mode (the token remaps; #dc2626 was borderline) —
this also fixes the contrast for WhoWhenSection, the other consumer.
Component test asserts the date input gains aria-invalid on a server date error.
Addresses PR #832 review (Requirements Engineer REQ-011; UI/UX dark-mode contrast).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Decision 6 / REQ-010 promised the pickers survive a fail(400) "including
pre-selected persons/documents", but EventForm only seeded them from
event/initialPersons — never from the fail payload — and the payload carried
only bare ids, which can't rebuild a chip (chips need displayName/title). On
the use:enhance path the in-memory selection survived; on a no-JS full reload
the chips were silently dropped.
Now the save action re-fetches the selected persons/documents by id
(lookupSelections, non-ok swallowed like the prefill path) and returns full
chip data; EventForm seeds the pickers from form.persons/documents ahead of
the seeded event. Extract preservedFormFields() to DRY the four fail payloads;
validateEventForm now returns the error pair and the route owns the fail().
Component test pins the rehydration; the server spec now asserts the fail
payload carries labelled chips, not just ids.
Addresses PR #832 review (Developer + Requirements Engineer concern, REQ-010).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Validate the submitted precision against the DatePrecision allow-list in
parseEventForm (falls back to DAY) so an untrusted token can't flow into
the request body — symmetric with the existing `type` narrowing.
- Parameterize the precision input name via DatePrecisionField's new
precisionInputName prop; the timeline form now submits `precision` instead
of the misleading document-domain `metaDatePrecision`. Document form keeps
the default, so its behaviour is unchanged.
- Capture EventTypeSelect's onchange into EventForm's `type` state so it no
longer goes stale (the submitted value was already correct via the hidden
input; this keeps the local state in sync).
Addresses PR #832 review (#781).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Both curator-event loaders repeated the same null-user + hasWriteAll block.
hasWriteAll already returns false for an anonymous user, so a single
requireWriteAll(locals) helper covers both REQ-002 (null user → 403) and
REQ-003 (no WRITE_ALL → 403) without the redundant pre-check.
Addresses PR #832 review (#781).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A TimelineEvent's DocumentRef carries documentDate but no precision, so
formatDocumentOption hit formatDocumentDate's undefined-precision path and
surfaced the UNKNOWN label instead of the date. Default a missing precision
to DAY so the chip shows the full date; add formatDocumentOption unit specs.
Addresses PR #832 review (#781).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
DocumentOption's metaDatePrecision/metaDateEnd are now optional so a
TimelineEvent DocumentRef (id/title/documentDate only) maps cleanly into a
picker chip — formatDocumentOption already degrades gracefully when precision is
absent. The server specs read fail()'s union data via a small failData() cast
that TS cannot narrow. svelte-check shows zero new errors in the #781 files.
Refs #781
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
