Волна 2 эпика #674 — у каждого правила один дом, остальные места ссылаются.
DECOR-EDITOR.md ← BACKDROP.md + LIVE-TEXT.md: один документ с нумерованными
разделами (§3 подложка, §5 текст с живыми значениями), на которые теперь
указывают комментарии кода вместо несуществовавших «BACKDROP §2/§3»;
исправлено утверждение, что space-card не рисует декор (он рисует подложку
и картинки декора, но не фигуры, мебель и текст). LIGHT.md ← матрица
настроек света (перевод, тест назван явно: test/devices.test.mjs «issues
84/88»). DEVICE-PRESENTATION.md ← правила «что показывает маркер» из
FILTERING.md (порядок cover → light sources → device role, шторы,
медиаплееры); «в одном pull request» → «в одном коммите». CANVAS.md: §9.5
«Оптимизировать планы» → CONFIG-COMPATIBILITY.md, overlay и планарные грани
Walls → WALL-THICKNESS.md §10–11, таблицы «было/стало» сняты. TESTING-DEMO.md
→ demo/stand/README.md: карта демо-дома и «чего на стенде нет», ручной
чек-лист снят (ручной фазы в процессе нет). ISOMETRIC.md — только текущее;
история Stage 2/4 — docs/adr/570-isometric-stage4-visual-handoff.md.
SUN.md: удалённый контракт фона снят, правило бумаги — в текущем разделе.
UX-MODES.md: декор над заливками, а не «под комнатами»; «hidden isometric»;
follow-up из #3 — все выпущены. Шапки VACUUM, WARM-REMOUNT («Выровнять всё
по сетке» → «Оптимизировать планы»), WALL-THICKNESS, STYLING-HOOKS,
CONFIG-COMPATIBILITY (#33), PDF-EXPORT — без устаревших статусов и планов.
README EN/RU: абзац про пересъёмку скриншотов → CONTRIBUTING.md, RADAR и
PDF-EXPORT в списке документации, RU догнал EN (2.5D, повторное
использование загруженного изображения, STAIRS). Один список канонических
документов подсистем в AGENTS.md и промпте ревьюера (_process.yml).
WALL-THICKNESS.md ссылается на ADR 282.
Правки src/** и validation.py — только пути документов в комментариях.
Issue: #679
User-Visible: no
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
Прямые и винтовые лестницы получили отдельную модель, инструменты редактора, безопасную межэтажную навигацию, вычитание из чистой площади и плоское отображение в 2.5D.
Issue: #663
User-Visible: yes
Two owner reports after v1.57.0, both about coordinates.
=== DEV-B58-01: nothing stops at the old canvas border any more ===
The infinite canvas freed the FRAME and the DRAWING; it did not free the
drag handlers, and both the owner and a user hit that within a day:
"названия комнат и устройства не перетаскиваются дальше старых границ
холста".
Two clamps survived v1.57.0, and the second is the worse one:
* `_pointerMove` (device marker) clamped into `_baseVb()` — the CONTENT
FRAME, with a 0.8 % inset. A marker could never be dragged past the
outline of what was already drawn, so a plan could not be extended by
putting a device where the next room was going to be.
* `_labelMove` (room label) clamped into `_spaceModel().vb` — the
space's STORED `view_box`, which is `[0,0,1,1]` for every plan the
card has ever written. Literally the old square: a room drawn at 2.5
had a name that could not reach its own room.
And one asymmetry: `_decorCommitDraft` and the decor text anchor had no
guard at all, while `_decorMoveUpdate` did — a draft could be born
outside the range the mover then refused to leave.
The rule now is one line: an editor gesture has exactly ONE bound,
`+/-CANVAS_LIMIT`, the same number `validation.py` enforces, and it is a
garbage limit rather than a frame. `clampCanvasR` / `clampCanvasN` in
space-geometry.ts are the only two functions allowed to impose it, and
`_snap()` applies it on the way out, so every gesture that goes through
the snap is bounded by construction.
demo/smoke_drag_bounds.mjs starts from an ORDINARY plan (rooms inside
0..1, so the old clamps really were in the way), drags a marker, a room
name, a decor shape and an opening far past the old square, checks each
arrives, is stored, survives a rebuild and takes the frame with it — and
that a wild drag still parks at exactly 5000 rather than 1e12. Seven of
its eleven facts fail by name on 85263d5.
=== DEV-B58-02: everything strictly on the grid ===
The owner's suspicion first, answered honestly in docs/CANVAS.md §9.2:
THE GRID STEP DID NOT CHANGE. `_gridPitch = NORM_W / GRID_N = 1000/240`,
both constants, independent of the frame, the view, the zoom, `view_box`
and `cell_cm`; `git log -S` shows neither touched since v1.4.0. So the
move to the infinite canvas did not put any existing element between the
nodes. `gridLevels()` changes what is DRAWN, never what is SNAPPED TO.
What WAS off the grid, and is now fixed:
* auto placements. `defaultPositions`, the `spaceCenter` fallback and
an undragged room label used centroids, which are not nodes for an
odd-sized or polygonal room. This is the likeliest thing the owner
was actually looking at.
* `_decorMoveUpdate` snapped the DELTA, which preserves whatever
off-grid offset a shape already had for ever, one step at a time. It
snaps the resulting anchor now, so one drag is enough.
* `snapToGrid`/`snapR` returned 500.00000000000006 for an exact 500 —
the round trip through a non-dyadic pitch. They are bit-identical on
a node now, so "is this on the grid?" stops answering no.
Openings and split points on a wall are deliberately NOT rounded to a
node — a door on a node but off its diagonal wall is broken geometry.
They are WALL-bound: projected onto the wall, then the offset ALONG it
quantised to the same step (`snapToWall({step,length})`,
`snapPointAlongPoly`). On the axis-aligned, grid-drawn walls the editor
itself makes, the two rules give the same point. The centre magnet is
consulted FIRST, so a wall whose middle is not a node can still hold a
centred window (this is what smoke_opening_measure caught).
Shift now means one thing everywhere: suspend the snap for this gesture.
It keeps its two older meanings (no centre magnet, coarse 15° compass).
=== And why an ACTION rather than a silent migration ===
Old plans may hold coordinates between the nodes. The card does not
round them on update. General settings grow a Grid group with
«Выровнять всё по сетке», which first states how many elements will
move and by how much at most, warns that there is no undo, and only then
writes — one config/set plus the layout updates, in one go.
1. A migration moves the user's data without asking. A house plan is a
drawing; the card has no mandate to redraw it on a version bump.
2. Some elements are off-grid ON PURPOSE — a small decor label nudged
next to an icon, a window on a diagonal wall, a plan traced over a
photo whose scale was never a whole number of cells.
3. A silent migration is unattributable: when a room looks 3 cm wrong
the owner cannot tell whether the card did it or they did.
4. An update that rewrites stored geometry cannot be undone by
downgrading the card. A button can simply not be pressed.
`alignAllToGrid()` (src/align-grid.ts) is pure — it copies its input and
returns the new spaces, the new layout and the report — so the dialog
measures and commits the SAME object and cannot promise one thing and do
another. test/align-grid.test.mjs pins what moves, what does not (a
stray opening with no wall in reach stays put), that a rect's FAR corner
lands on a node too, and idempotency: a second run reports moved 0,
changed false, and deep-equals the first. demo/smoke_grid_snap.mjs does
the same through the DOM plus every by-hand placement.
docs/CANVAS.md §9 carries the whole contract; docs/TESTING.md gains
three manual items. i18n en/ru. The backend is untouched — same
coordinates, same schema.
Audit dev@2c947f4, DEV-2C947-03 (P2). Three rooms in the core plus one dragged
90 canvases out: the frame rejected the stray exactly as §4.1 promises, and
then a perfectly ordinary marker on the main plan came out 90.89x too big and
covered the house. `contentFrame` voted; `iconUnit` did not — it took
`boxOf(every room)`, so the distance to the stray the frame had just thrown
away lived on in the numerator of `iconCqw`.
`iconUnit` now takes `contentFrame(roomItems, { pad: 0 }).core`: the same
main-mass vote, over the same rooms it always used (rooms only is what keeps
the full card and the static card bit-identical), with no padding, because
this is a UNIT and not a viewport. Below MIN_VOTERS nothing is declared an
outlier, so every ordinary plan — and every genuinely wide one, where the
majority veto applies — keeps exactly the unit it had. `defaultPositions`
takes its declump distance from the same call, so the auto-placement spacing
follows without a second rule.
Unit (test/canvas.test.mjs): a far room leaves both the frame and the icon
unit alone, `iconCqw` on the strayed plan equals `iconCqw` on the same plan
without the stray, and a plan that is honestly two canvases wide still scales.
smoke_canvas_frame.mjs measures the rendered badge in px with and without the
far room. Both are red on the parent commit.
Two owner corrections after the infinite canvas.
- --icon-size goes back to being a percentage of the PLAN: a marker
grows and shrinks with the zoom, like everything else drawn on the
plan. The infinite canvas had made it a percentage of the viewport
(fixed pixel size) — the owner looked at it and asked for the
original contract back.
What survives from the canvas work is the NUMERATOR. The old
expression divided by `vb.w`, the stored view_box, which is not a
frame any more; a fixed NORM_W in its place would have shrunk every
marker on a plan drawn past the old square by exactly the factor the
plan is outsized (an invisible dot 50 canvases out). So it is now
`iconCqw() = iconPct * iconUnit(space) * kioskScale / view.w`, one
pure helper both renderers call. `iconUnit` is exactly NORM_W for
any plan that fits the old square — and the editor has never written
anything but `view_box: [0,0,1,1]` — so the rendered size is
bit-identical to the pre-canvas card: measured against the v1.56.0
bundle at a fixed view, both give 3.400 / 3.091 / 6.182 / 12.364 cqw
= 28.52 / 26.11 / 50.22 / 98.44 px. On a plan drawn at 1.5..3.8 the
marker is 26.1 px, the same as on an ordinary plan, instead of the
~11 px a fixed numerator would have given.
The static space-card uses the same helper: it has no zoom, but its
frame is the content now, so a bare iconPct shrank its markers as
the frame tightened. marker.size, the kiosk scales and every
satellite still ride on --dev-size, untouched.
- the icon angle in the device dialog steps by 5 degrees, not 10
(0..355): a marker often has to line up with a wall that is not on a
10-degree grid.
Tests: three unit tests on iconCqw (the legacy expression reproduced
digit for digit, the runaway plan, the no-view fallback); the infinite
canvas smoke's "same pixel size at zoom 1/4/1/3" assert is turned back
into "scales 4x / 1/3 with the zoom" plus a new one that the marker on
the far plan measures the same as on an ordinary one; the angle step
is pinned in smoke_size_angle_parity. docs/CANVAS.md §6 rewritten.
docs/CANVAS.md is the source of truth (owner-approved 2026-08-03): the
normalised square was never a sheet of paper, only a coordinate system,
and users who drew past its edge could not place devices there.
Storage does not change and there is no migration. What changes is what
the renderers DERIVE from it:
- space-geometry.ts gains contentFrame() — one item per drawn object,
a rank-based outlier vote (median centre, 75th-percentile spread,
10x threshold, majority veto) and a fit-everything box beside the
opening view. contentBounds() is now a thin wrapper over it; the old
-25%..125% envelope is gone — it WAS the bug that made a plan drawn
at 1.5..3.0 frame empty canvas.
- spaceFrame()/spaceCenter() make view_box an optional first-frame hint
used only when there is nothing to frame; iconUnit() keeps auto
placement spacing in proportion (NORM_W for anything inside the old
square, so no layout moves); gridLevels() picks a legible grid step.
- validation.py: coordinates ±4 -> ±5000, sizes 0.001..5000, decor
-1..2 -> ±5000, opening length <= 5000. Garbage insurance, not a
frame — a stored 1e100 is still refused.
Units: test/canvas.test.mjs covers the plan past the square, the
outlier (and the three ways NOT to declare one), corruption, empty
space, a lone marker, image plans and the adaptive grid.
Backend: the limits, and that a config from any released version
validates untouched.
The ±4 bound from v1.50.2 measured view_box[2:4] and room w/h with the same
ruler as coordinates, so zero and negative sizes still passed the schema —
and viewBox='0 0 0 0' draws nothing on every client, with the static card
computing aspect-ratio: 0 / 0 on top. _EXTENT now requires strictly positive
sizes with a floor of one thousandth of the canvas (1 render unit — far
below any real room, keeps the maths finite); coordinates stay allowed to be
negative, a crop origin legitimately sits past the edge.
Defensive layer for stores that already hold a broken viewport: spaceModels
falls back to the whole canvas — both cards render from that model, so both
get the fallback — and a legacy rectangle with a negative size reads as the
same rectangle drawn from the other corner.
Also: the room settings button is the bottom row of the room card, and the
room name renders in the same spot in view and plan modes (owner's request,
committed earlier on dev).
- HP-1501-01: v1.50.1 bounded layout positions and left room rectangles,
polygon vertices, view_box and opening coordinates on bare _finite — the
same absurd-magnitude failure, one schema over. _GEOM (±4) covers them all
now, opening angles get ±360. And because a store may already hold such a
vertex from before the door existed, contentBounds applies its canvas
envelope to room geometry exactly as it does to device positions: the
point renders where it is, the frame ignores it, a space of nothing but
absurd points falls back to the whole canvas.
- HP-1501-02: a repair matching zero positions answered ok/moved:0 and
replaced the one-deep backup with an empty one — a typo right after
repairing the wrong space destroyed the promised way back. Empty match is
nothing_to_repair now: no write, no revision bump, backup intact.
Old test fixtures carried view_box [0,0,100,100] from the render-unit days;
they now use the normalised box the product actually stores.
- HP-1500-02: the stage budget was the absolute document coordinate, so any
tall dashboard content before the card was billed as header and the stage
collapsed to 0px. Measure our own chrome relative to the card plus a
bounded (<=120px) allowance for what the viewport keeps above us; re-measure
on window resize, remove the listener in disconnectedCallback.
- HP-1500-03, both layers: contentBounds opens a near-zero axis (< ~an icon)
up to a 200-unit floor and ignores extra points outside a canvas envelope
(-25%..125%) for FRAMING purposes only; the server bounds layout coordinates
to +-4 — any finite float used to pass, and one 1e100 hid the plan from
every viewer. A thin real room keeps its tight frame; the gate sensor past
the edge still stretches it.
- HP-1500-01: no automatic double-transform — a correct layout and a stranded
one are indistinguishable, and guessing wrong corrupts good data. Explicit
admin command houseplan/geometry/repair: dry_run previews, the backup rides
the same store write, undo restores, and routine layout writes now preserve
unrelated store keys instead of eating the backup.
Tests: contentBounds guards (unit), layout coordinate bounds + repair
lifecycle (harness), card-below-content smoke. Inventory: 139 / 49 / 42 / 64.
Owner's batch (committed to dev earlier today, released here):
- devices count as content for the default zoom;
- the editor no longer shifts the plan — the stage measures its own top
instead of assuming 118px of header;
- zoom goes out to 0.4x, centred.
From the review:
- HP-1490-01: the square-canvas migration wrote two stores in sequence, and
the first write deleted the aspects the second needed — a crash between
them stranded the layout in the old coordinates with nothing able to
finish it. The intent {space: old aspect} is durable now: saved to the
layout store before anything moves, cleared by the same write that stores
the migrated layout, each half idempotent behind its own trigger. The
update event fires only after both halves are on disk. Proven at the exact
crash boundary by a harness test that fails the layout write once.
- HP-1490-02: check_quota and the file write were two executor jobs with
nothing between them, so N parallel uploads all measured the store before
any of them wrote. One job under a dedicated upload_lock now — narrower
than write_lock on purpose, a directory scan must not stall config saves.
A failed write reserves nothing.
- HP-1490-03: the content frame fed pan, zoom, clamp AND pointer maths, so
the editors were boxed into yesterday's drawing. Edit modes measure from
the full square; mode switches refit rather than carry a view clamped
against the wrong base.
- HP-1490-04: Save could outrun the proportions read and ship the previous
file's ratio. Picking a plan clears it immediately; Save awaits the
bounded read and stores 'unknown' over a lie.
- §5: package-lock version synced, duplicated comment removed.
New: smoke_audit_1490.mjs, migration crash-recovery pure + harness tests,
parallel-quota harness test. Inventory: 138 unit / 49 pure / 40 harness / 64
smokes.
Three owner reports:
- The content frame behind the default zoom only looked at rooms, and devices
are allowed to stand outside every one of them — a gate sensor by the fence
was left outside the opening view. contentBounds() takes the marker positions
now, and they count as content even on a space with no rooms at all.
- Entering an editor 'strangely shifted' the plan. The stage height was
100dvh minus a hard-coded 118px of header, and the editor header is ~90px
taller than that: the whole scene slid down by the difference and its bottom
went below the fold. The card now measures where the stage actually starts
(HA toolbar, margins and our header included) and gives it the rest of the
viewport; the measurement is deferred a frame because setting state straight
from a ResizeObserver callback trips the 'undelivered notifications' error,
which smoke_dialog_zombie rightly counts as a page error.
- Zoom stopped at the base fit. The floor is 0.4× now, and zoomed out the
clamp centres the content instead of pinning it to the top-left corner —
with the view larger than the plan there is nothing to clamp against.
New smoke: smoke_zoom_out.mjs (editor keeps the stage inside the viewport,
0.4 floor, centring, the device-stretched frame); a unit test for the extra
points of contentBounds. Not released — the next release goes out after the
v1.49.0 audit.
Owner's batch:
- zoom now opens on what is DRAWN (rooms + 5% margin) for spaces with no
background image; with one the image is the plan and still fits whole. A small
plan on the square canvas no longer opens as a speck.
- swiping between spaces, and the kiosk carousel, slide sideways; honours
prefers-reduced-motion.
- the room settings button reads 'Room settings' and lightens on hover.
- 'curation' is filtering everywhere: UI strings, docs, code.
Checked the yard while I was there: its drawing sits off-centre because it was
drawn that way — before the migration x spanned 0.12..0.54 with 0.12 and 0.46 of
margin. The migration added 0.1465 on each side, symmetrically. Content-fit zoom
makes it moot anyway.
From the v1.47.0 review:
- HP-1470-02: the picker let you delete the plan you had just selected — it is
not in the stored config yet, so the server rightly called it free, and the
save then stored a url with no file. The button is disabled, and since two
clients can do this in either order, config/set now verifies every internal
plan url against the disk under the write lock and answers .
External and legacy urls are not ours to police.
- HP-1470-01: growth is bounded at the door rather than by deleting old files —
that mistake cost real plans twice. check_quota refuses an upload that would
push the store past 256 MB / 200 plans (1 GB / 1000 attachments) or leave less
than 512 MB free. The plan list is capped at 60 newest with a total, and
thumbnails load lazily.
- HP-1470-03: picking a saved plan waited for nothing and stored a fallback
ratio when the signature had not arrived — a square plan came out stretched.
It waits for the signature, binds the result to the dialog that asked, and the
dialog preview is signed too.
- report §5: the last lifecycle comments still described age-based collection.
Not released yet — the owner asked for a release once the batch is done.
A space carried an aspect ratio, and coordinates were normalised against it: x
by the width, y by the height. Every geometric question therefore depended on a
per-space number, and picking a canvas orientation was a decision the user had
no reason to make. The render space is now NORM_W x NORM_W and a plan image is
fitted into it by its OWN ratio, centred — wide plans get margins above and
below, tall ones at the sides.
Migration (geometry_migration.py, pure and unit-tested) runs once at setup under
the write lock. Nothing about a drawing changes: the old box is padded out to a
square and every coordinate re-expressed against it — rooms as rects and
polygons, openings and their lengths, decor, view_box, and the marker positions
in the separate layout store. In render units it is a uniform scale plus an
offset, so angles and proportions are exact. cell_cm is scaled for tall plans,
because the grid pitch is a fraction of the width: without it a wall would
measure less than it does.
is now dropped by the schema rather than accepted — a stale tab sending
it would be sending coordinates from the old normalisation too, and honouring
the field would not make them right.
The demo fixture was migrated with the same transform, so the smokes exercise
the new geometry rather than a square-native fake; six of them needed their
render-space helpers updated and one its click coordinates.
Not released — dev only, per the owner's instruction.
HP-1454-01 (high, release blocker): an uploaded SVG plan opened directly is a
top-level document of Home Assistant's own origin, so a <script> inside it
reaches the session's localStorage and API. Uploading needs write access, which
by default every authenticated user has. SVG responses now carry a sandbox CSP;
only SVG, because a CSP on a PDF can break the browser's viewer and a raster
image has nothing to disable. Verified in Chromium both ways: the script runs
without the header and does not with it.
HP-1454-02: attachment uploads wrote straight to <marker>/<filename>, outside
the config transaction — a cancelled dialog or a rejected save left the stored
url serving new bytes, and every new icon shared one 'new' folder, so two of
them attaching manual.pdf pointed at one file. Uploads take a free name, a new
icon gets a per-dialog staging folder promoted on an accepted save, and
config/set collects superseded and aged-orphan attachments like it does plans.
HP-1454-03: the debounce spaced out the starts of a write, not the writes. A
save slower than 500 ms let the next edit go out with the same expected_rev;
the server accepted the first, rejected the second, and the conflict handler
reloaded over the local copy. Writes are chained now — one in flight, each with
the revision the previous returned.
HP-1454-04: _openPairsCache keyed on room ids and links only, so an aspect
change or a dragged vertex left open boundaries and their glow cuts at old
coordinates. It keys on the rendered model object now — the same invalidation
the model cache already has, not a second strategy. The fingerprint also gained
an O(1) geometry roll-up per room.
HP-1454-05: outer collections were capped, inner ones were not. Limits for
poly points, open_to, controls, pdfs, text and url lengths, plus a total
serialized size cap; legacy is dropped server-side.
HP-1454-06: upload streams to a temp file and downloads use FileResponse, so a
50 MB manual no longer costs ~100 MB of RSS per transfer.
HP-1454-07: spaceModels() dropped room.settings, so the static card ignored the
per-room fill override. HP-1454-08: layout had no revision on point-wise writes
and no event, leaving static cards stale forever; it now keeps a revision,
returns it and fires houseplan_layout_updated. HP-1454-09: repair cleanup only
walked existing spaces, so a deleted space kept its warning. HP-1454-10:
serialize-javascript pinned past two advisories.
Tests: smoke_svg_sandbox (proves both directions), smoke_config_writer and
smoke_render_parity (both verified failing against a v1.45.4 build), six pure
tests for attachment collection and inner limits, four HA-harness tests for the
CSP, non-overwriting uploads, the size cap and layout revisions.
Docs: CHANGELOG.md + CHANGELOG.ru.md + ARCHITECTURE.md + TESTING.md + STATUS.md.
L2 (silent data loss): debounce gains flush()/pending(); _reloadConfigOnly
flushes a pending write and defers while one is in flight; conflict path
forces; failed reload now toasts instead of an empty catch; teardown flushes.
G1 (split corruption): same-edge cuts carve the niche properly instead of
walking the outline twice; partition invariant (parts sum to the original)
rejects anything else; +1 unit test covering 5 niche shapes and both legacy
cut shapes.
B1 (unauthenticated content): plans and marker files move to
HouseplanContentView (/api/houseplan/content/..., requires_auth); only the
card bundle stays static; contentUrl() rewrites legacy URLs on read (no
storage migration); repairs.py accepts both prefixes; +1 unit test.
L3 (dialog zombies): all four save catch-blocks guard against a closed
dialog; the card no longer blanks when a save fails after Esc.
smokes: smoke_save_race, smoke_dialog_zombie; docs (TESTING/CHANGELOG/
ARCHITECTURE incl. the optimistic-UI note) same-commit
