…l size

NEWOBJ aliases to the point-only S-101 VirtualAISAidToNavigation rule, so its
line/area variants errored (drew nothing) and every NEWOBJ point stamped a V-AIS
mark — ignoring the producer's SYMINS instruction (e.g. SYMINS="SY(INFORM01)").
381 of 386 ECDIS-Chart-1 NEWOBJ carry SYMINS (164 TX labels, 91 LS, 80 SY, 27 AC,
17 LC, 2 AP) — that string IS the portrayal.

Resurrect SYMINS02 (was deleted with pkg/s52): parseSYMINS parses the
';'-separated SY/TX/TE/LS/LC/AC/AP ops into SymbolCall/DrawText/StrokeLine/
LinePattern/FillPolygon/PatternFill (reusing formatSubstitute for TE), hooked at
the top of buildFeature so a NEWOBJ-with-SYMINS overrides the V-AIS stream.
NEWOBJ without SYMINS falls back to a dashed magenta new-object boundary;
SWPARE (no SweptArea.lua in the catalogue) falls back to a dashed boundary +
"swept to <DRVAL1>" label.

Fix symbol physical size: DefaultPxPerSymbolUnit divided by 0.35278mm (the 1/72"
point) while the app measures the screen at 0.26458mm (1/96" CSS px), rendering
every symbol ~25% too small. Use 0.26458 so symbols hit their encoded size — the
S-52 size-check SY(CHKSYM01) box now measures ~5mm.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

`make preslib-chart1` (scripts/preslib-chart1.sh + .mjs) renders each S-52 PresLib
"ECDIS Chart 1" panel by reference-plot page number for visual diffing against the
spec (PresLib e4.0.0 Part I §16, doc pages 238-253): extracts the cells, serves a
throwaway server, imports via the normal server-side bake, screenshots each panel
framed to its cell, tears down. Output → testdata/preslib-chart1-out/ (gitignored).

TestS101Audit now also dumps a per-class error breakdown (which classes fail
portrayal and how often) — surfaced the NEWOBJ/SWPARE gaps.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ion scale

Add a "spec" mode (?spec / [spec]): a clean, full-bleed map with every floating
control, the status readout, attribution, load bar and the S-52 scalebar hidden —
for capturing reference-style plots. The scalebar lives in chart-canvas's shadow
root, so it's hidden at the source when an ancestor chart-plotter[-app] has [spec].

Rework the ECDIS-Chart-1 harness to render each page in spec mode at ITS
compilation scale: the viewport is sized to the cell's ground extent / CSCL /
pixel-pitch (1:14 000 harbor pages, 1:60 000 overview), so each panel is captured
full-screen at the scale the legend was drawn for — matching the reference figure.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…hooting out)

The AugmentedRay instruction carries the leg LENGTH's CRS: LocalCRS = display
millimetres (the 25 mm short sector leg), GeographicCRS = a fixed ground distance
in metres (a sectorLineLength or full-VALNMR leg). The parser ignored the CRS and
treated every length as mm, so a GeographicCRS leg of nmi2metres(0.1 NM)=185 m
rendered as 185 mm — ~10× too long, shooting the sector legs off the chart
(visible in the PresLib ECDIS-Chart-1 "Aids and services" panel).

Parse the length CRS into AugmentedGeom.LengthGroundM (metres) vs LengthMM, carry
it onto AugmentedFigure, and in tessellateFigure convert a ground-length leg to
pixels at the tile zoom (like the full-length leg) instead of mm. Sector legs now
render at their correct contained length.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

S-52 TX/TE labels are single-line, but the `text` layer set no text-max-width, so
MapLibre's 10-em default wrapped longer labels (e.g. "Information about chart
display (A,B)") onto a second line — which then collided and was dropped
(text-optional), so the label looked truncated. Set text-max-width: 40 so each
label stays on one line, matching the PresLib reference plots.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…0.6.1.1)

Objects carrying ancillary info (INFORM/NINFOM, or TXTDSC/NTXTDS/PICREP) now get
SY(INFORM01) at their position — a box-on-a-leader "info available" marker. This
is what produces the leader-line callouts on the PresLib "Approved new object
symbols" V-AIS (page 251), whose SYMINS draws only the ring+topmark; the callout
is the §10.6.1.1 indicator, not part of the symbol.

addInformSymbol appends it for any feature with a non-empty info attribute (all
build paths). The bake routes INFORM01 as display priority 8 / category Other
(overriding the host feature's category), so it clears Standard display and only
shows when the mariner enables Other — matching the spec. The Chart-1 harness
enables Other so the spec render shows the callouts.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

… harness

The IHO PresLib reference plots show ALL symbology, so set the harness mariner
explicitly rather than relying on defaults: display category Other on (INFORM01
callouts, "other marks", magnetic variation), data-quality overlay on (the CATZOC
"quality of data" panels), depths in metres (IHO, not NOAA feet), 25 mm short
sector legs + symbolized boundaries (the S-52 defaults the reference uses). Makes
the spec render deterministic and match the reference (e.g. page 243 depths now
read 5.5m/4m and the quality-of-data zones appear).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…nk map)

With the basemap off, a remote installed cell was just a faint outline box you
had to spot by eye. Add a cell-name label at each coverage box's centroid, capped
at the band's render zoom (like the fill/line) so it vanishes once the chart
itself draws — visible when you're zoomed out and lost, gone when you're there.
Decluttered (drops on overlap). Tap-to-fly-in already worked.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…e + search

Finding an installed cell on a zoomed-out blank map didn't scale: coverage boxes
are per-pack (a single global box over a 7k-cell "import" set) and per-cell would
freeze the map. The scalable answer is search-by-name → fly-to.

Server: a small persistent name→bbox index over the cached cells
(<dataDir>/cells-index.json), backfilled once in the background by reading each
cell's header (baker.ParseCellBytes bounds), refreshed after import. GET
/api/cells now also returns a "bbox" map and accepts ?active=1 — only cells whose
footprint overlaps an ENABLED pack (charts actually on the map), and only those
indexed (an un-indexed cell has no footprint to fly to). No cell→pack table
needed; overlap with enabled-pack bounds defines "active".

Client: the search box's catalog is now the ACTIVE installed cells (name+bbox
from /api/cells?active=1), so typing a cell name finds it and flies to its
footprint — verified flying to AA5C1CDE on a blank map. (Search targets active
charts, not the NOAA discovery catalogue.)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…know a name)

Search-by-name only helps if you know the name. Opening search with nothing
typed now lists the active installed charts — ordered NEAREST to the current view
first (most relevant to where you're looking), capped at 40 with a "N charts —
type to narrow" footer. Triggered on open, so you see your charts immediately and
can click one to fly there, or type to filter. Verified: empty search lists the
14 PresLib cells, nearest-first.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The name→bbox index could drift: build() skipped already-indexed cells (so a
re-imported cell kept stale bounds) and never dropped entries for removed cells.
Now build() reconciles against ENC_ROOT — pruning entries whose cell is no longer
on disk — and an import forget()s the cells it re-caches so the rebuild re-parses
their bounds. (Search results were already remove-safe: serveCells lists live
ENC_ROOT dirs, looking up bbox in the index, so a removed cell never appears even
before the prune.) Add/update/remove all covered; verified by test.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…slib-chart1)

`make s64-pages` (scripts/s64-pages.sh + .mjs) extracts the S-64 ENC TDS, serves
+ imports it, and renders each rendering test in spec mode at the cell's
compilation scale → testdata/s64-pages-out/ (gitignored), for diffing against the
S-64 reference plots. Unlike PresLib Chart 1 (one all-symbology plot), S-64 varies
the mariner per page — §3.1 renders the same area at Base / Standard / Other — and
uses the S-64 setup (safety contour/depth 10 m, symbolized boundaries, metres).
Covers §3.1, 3.2, 3.3, 3.4, 3.6, 3.7/3.7.7, 2.1.1, 5/6/7. §3.9 Polar (beyond
Web-Mercator) and §3.8.5 AML (non-ENC names) are omitted with a note.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Adds a Chart 1 page to the docs that embeds the S-52 PresLib "ECDIS Chart 1"
reference sheet LIVE — one read-only <chart-plotter> widget — and turns it into a
symbol-compliance checker: a list of every reference panel (PresLib §16, pp.
238–253) beside the chart; click one and the widget fitBounds() to that panel
(chrome-aware padding) so you can diff our render against the spec plot. The sheet
is one contiguous synthetic ENC, so it's a single ~1 MB tile bundle and navigation
is just the map camera.

- scripts/fetch-preslib-cells.sh + `make demo-chart1`: fetch the IHO PresLib draft
  (or a local testdata copy), bake the cells to a tile bundle that reuses the demo
  bundle's frontend assets (widget points its catalog= at the Chart 1 manifest).
- CI (docs.yml): cache the source cells + bake the bundle into docs/static/chart1
  alongside the existing Annapolis demo; gitignored, never committed.
- web: expose a public `map` getter on <chart-plotter> so embedders can frame a
  region with MapLibre's own fitBounds (used here, floored at the 1:139 000 SCAMIN
  so the whole-sheet fit never zooms features out of existence).
- README + docs intro: highlight ECDIS Chart 1 rendering as a feature.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…into tests/preslib-chart1

# Conflicts:
#	web/src/plugins/search-box.mjs

…ding

After merging main's physical-scale restage, onReady awaits the catalog load and
then calls addCatalogOverlay — but a setStyle (physical-scale / SCAMIN buckets) can
still be in flight at that point, so addSource threw "Style is not done loading".
The throw happened BEFORE the style.load self-heal listener was registered, so the
overlays never recovered. Bail when !isStyleLoaded() (the style.load handler
re-adds once ready) and guard the initial call so that listener always registers.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>