Rail-delegacy: remove the nine grey /editor rail accordions

[WilfordGrimley]

Description

Implements the owner-approved rail-delegacy design (SPEC-rail-delegacy.md + mockup, binding §D token table) for the /editor unified left rail: every one of the nine grey AutofillCollapse sections is removed and its contents folded into designed elements, per the spec's ruled dispositions (§E RD1-RD8).

Disposition -> implementation map

Ruling	Disposition	Implementation
RD1/O1	ONE chip surface - funnel Border/Frame/Treatment chips ARE the implicit-vote surface; separate .achip explicit-vote fieldset (AttributesSection.tsx) scrapped	AttributesSection import/usage removed from DisplayPage.tsx's Rail; funnel chips (FunnelAxisRow/TreatmentChipRow, already implementing suggestedFilterTagNames-driven SUGGESTED state + castImplicitVote on pick) relocated into the shared Filters panel body in SelectVersionResults.tsx
RD2	Sort -> compact Form.Select of the 6 SortByOptions, inline in the SV header row	New .svhead row in SelectVersionResults.tsx (funnel-sort-select), replacing NullableSortByFilter for the rail path
RD3/O4	Jump to Version fully scrapped	Never wired into the new header/panel; GridSelectorFilters' own Jump section stays untouched for the sidebar/modal layout, just never reachable from the rail
RD4/O3	Filters panel tier-conditional: phone = in-rail Collapse; desktop inline rail + tablet drawer = a panel escaping the 380px column/drawer clipping	SelectVersionResults.tsx: phone renders .fpanel.inline in-tree; desktop/tablet renders .fpanel.float + .fscrim via ReactDOM.createPortal(..., document.body) - see FloatFiltersPortalRoot's own comment for why a plain in-tree position:fixed node wasn't enough (LeftRailOffcanvas's position:sticky traps descendant z-index in a local stacking context; caught live via a blocked Playwright click before switching to a portal)
RD5	Print Options + Slot Actions + Report -> one .cstack; Report = small btn-outline-danger expanding to reason chips in place	New ControlStack component in DisplayPage.tsx; ReportCardPanel already implements the expand-in-place behavior verbatim, no changes needed there
RD6/O2	Whole Card-Details metadata block + Download/Favourite -> "More details" disclosure only; rail-head stays lean	RailHeader rewritten: .rhead-row + .detmore toggle + .detbody (CardMetaTable + CardDownloadFavorite)
RD7	Printing id appears ONCE, in D14; RequestedPrintingBadge becomes conditional-on-mismatch only; also required dropping CardMetaTable's own "Canonical Card" row in the rail context (not explicitly listed in the spec's component table but required by the "exactly once" rule - CardMetaTable gained an additive showCanonicalCard prop, default true, unaffected everywhere else)	RequestedPrintingBadge gained additive showOnlyOnMismatch/resolvedPrinting props; CardMetaTable gained additive showCanonicalCard prop
RD8	66px subject-card preview + dashed empty state at the rail top	.subject/.subject.empty in RailHeader, reusing the selected image's own smallThumbnailUrl
item 6	Printing Tags panel hangs off D14, opened on demand	New IdentifyPanel in DisplayPage.tsx, wrapping the existing PrintingTagsBlock (PrintingTagPicker + conditional AttributeVotingPanel) verbatim - no fork
§A drawer flex-height note	.lrail/.lscroll flex trap inside the Offcanvas drawer	Verified NOT applicable to the real app: Bootstrap's stock .offcanvas-body{flex-grow:1;overflow-y:auto} already owns the scroll (no custom .lrail/.lscroll wrapper divs exist here) - confirmed via the new phone bottom-stack-reachability Playwright test, not just asserted

Deviations (documented)

1. Implicit-vote awareness line gated on >=1 active chip, not always-visible as the mockup's decorative demo copy shows - the real awarenessCopy callback names active tags, which has nothing to describe with zero active chips. Existing tested behavior preserved.
2. Active-tag pill row dropped from the SV header (the mockup's own .svhead has no pills row) - chip active state is now shown via each chip's own data-active/data-state attribute inside the Filters panel instead.
3. D21's tier-based axis auto-collapse/auto-open retired - the funnel chips now live inside the same one user-toggled Filters panel as DPI/size/language/tags/NSFW (RD4's unification), so a survivor-count-driven auto-open/collapse would fight the calm, closed-by-default design.
4. DPI/Size/Language/Tags/NSFW fieldsets reuse FilterSettings.tsx unchanged (not re-skinned to the mockup's literal .frange/.ftree/.swtrack token values) - re-styling those shared components would touch the Search Settings modal and GridSelectorModal too, which is out of scope for a rail-only round and would risk drift; flagged as a residual visual-fidelity gap, not silently dropped.
5. CanonicalCardFilter (printings/artists filter) not mounted in the rail's Filters panel - it isn't accounted for anywhere in the spec's §F disposition table (only DPI/Size/Language/Tags/NSFW/sources/attributes are listed as "remaining after Compact the shared filter-panel controls to the approved density #409"), so it's treated as intentionally out of scope for this round, not silently dropped.
6. Subject-preview caption text (mockup's decorative "selected art" label) omitted - no real data backs it in production (the card name is already shown in .idcol).

Dropped tests (with justification)

• DisplayPage.spec.ts: "the Attributes section casts a real tag vote when a chip is tapped" dropped outright (not replaced) - AttributesSection.tsx's useTagVoting/submitTagVote explicit-vote fieldset is scrapped per RD1/O1, not relocated. The funnel chips' own implicit-vote path already has dedicated coverage (SelectVersionSection.spec.ts, SelectVersionResults.test.tsx); explicit attribute voting now lives only in AttributeVotingPanel (inside the identify panel), which keeps its own pre-existing, unmoved coverage.

Two further tests were adapted, not dropped (kept in the report for completeness since their shape changed materially): the two requested-printing-badge tests now assert the new mismatch-only contract (one asserts absence for a matched printing, the other asserts the single .mismatch style for a genuine mismatch) instead of the old plain/degraded two-look contract; the "clicking a collapsed section expands it" / "selecting a different slot resets the accordion" pair now exercise "More details" + the identify panel instead of the retired Attributes accordion.

Measured delta (css-diff harness)

Built a computed-style extraction harness (extract-mockup.js + a throwaway Playwright spec + diff.js, pattern from the prior corrected-bundle round's own harness) comparing the built /editor rail against rail-delegacy-mockup.html's own computed styles, at both 1400px (desktop, float Filters panel) and 390px (phone, inline Filters panel), across every introduced (§D.2) element type plus a sample of inherited (§D.1) ones: rail-head, subject preview, slot/name text, "More details" toggle/body, D14, identify panel band, .svhead, Sort select, Filters toggle, the Filters panel container, the filter fieldset legend, and the control stack + its legend.

Result: 42/42 rows matched, 0 mismatches, 0 skipped (1400px: 26/26; 390px: 16/16).

Verification

• tsc --noEmit: clean (one pre-existing, unrelated marked module error in docsSite.ts, confirmed present before this change).
• npx prettier@2.7.1 --check on every changed file: clean.
• npx jest: 568/568 passing (64 suites) - includes a jest.setup.ts fix (global.matchMedia polyfill gained addEventListener/removeEventListener alongside the deprecated addListener/removeListener pair it already had, since useViewportTier.ts's modern API call was previously an unhandled crash under Jest the moment any component using it mounted - a real, previously-latent test-infra gap this round exposed).
• Playwright, --project=chromium --workers=4, isolated port (never 3000; playwright.config.ts reverted to its original :3000/npm run dev config before this commit - diff is clean): DisplayPage.spec.ts + DisplayLeftRailFidelity.spec.ts + SelectVersionSection.spec.ts = 44/44 passing. Also spot-checked GridSelectorModal.spec.ts + CardDetailedViewModal.visual.spec.ts (shared-component blast-radius check for CardMetaTable's new prop and the sidebar/modal SelectVersionResults layout) - 24/24 passing.
• Screenshots (built app, from an untracked throwaway spec run against this branch, not committed): /home/ubuntu/.claude/jobs/e893dbef/tmp/rail-build/v-1400-rail-allopen-top.png, v-1400-rail-allopen-bottom.png, v-1400-more-details.png, v-1400-filters-float.png, v-390-top.png, v-390-bottom-stack.png.
• Delta reports: /home/ubuntu/.claude/jobs/e893dbef/tmp/rail-build/delta-1400.md, delta-390.md.

New test coverage added

Subject preview + dashed empty state; "More details" contents (and that the printing id is NOT duplicated there); single printing-id-occurrence dedup (both the "no art" and "has art" cases); mismatch badge conditional (present-only-on-mismatch, absent-when-matched); phone in-rail vs desktop/tablet float (float node provably absent at 390px, not just hidden); bottom-control-stack reachability at 390px (the §A phone-reachability hard requirement).

Checklist

• I have installed pre-commit and installed the hooks with pre-commit install before creating any commits.
• I have updated any related tests for code I modified or added new tests where appropriate.
• I have manually tested my changes as follows:
  • Ran the full display-page Playwright suite (44 tests) plus the css-diff harness screenshots/extraction above; did not additionally drive a live browser session by hand beyond what those specs + screenshots cover.
• I have updated any relevant documentation or created new documentation where appropriate.

Open items

1. Docs not updated this round - docs/features/printing-tags.md/grid-selector.md's "Frontend consumer (funnel round)" sections still describe the funnel chips in their pre-rail-delegacy arrangement (always-visible above the grid, not inside a Filters panel); should be updated to reflect RD1/RD4 before/at merge. No prior SPEC-display-left-rail.md-tracking doc file exists in docs/ to supersede (the spec lives only in the job tmp dir passed to this task).
2. Wiki checklist (per CLAUDE.md's "Wiki maintenance" convention): this changes what a USER sees on /editor (rail layout) - add "wiki: /editor left-rail page needs updating for the rail-delegacy layout (accordions removed, More details/identify panel/Filters panel/control stack)" to the merge-time checklist.
3. DPI/Size/Language/Tags/NSFW pixel-token fidelity (deviation 4 above) is a known residual gap versus the mockup's own literal .frange/.ftree/.swtrack values - flagged for a follow-up pass if the owner wants those shared components re-skinned.

Co-Authored-By: Claude Fable 5 noreply@anthropic.com