Picks up the fork's merge of upstream acadrust (geodata decode, DXF
roundtrip newline sanitization, DWG alignment) on top of the current
table/multileader style work.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bump acadrust to pick up the new $CTABLESTYLE / $CMLEADERSTYLE header
variables, then on load seed the ribbon's active table style and the
tab's active multileader style from the document header so the ✓ marks
the style the file actually declares (DXF). Set Current writes the header
field as the round-trip source of truth, and the style-manager staging
snapshot now covers it. sync_ribbon_styles honours the active table style
instead of always falling back to the first one.
DWG still leaves these at "Standard" until acadrust parses them from the
binary header.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- The current-style marker is now a green ✓ in its own fixed-width column
so names stay aligned; rendered as a separate coloured glyph rather than
baked into the label string.
- Resolve the current text / dimension / multiline style from the handle
the DWG header stores into the name the app reads (fix_current_style_names
on load). DXF already provides these as names ($TEXTSTYLE / $DIMSTYLE /
$CMLSTYLE); DWG only stores handles, so the marker previously always fell
back to "Standard". Only overrides when the handle resolves.
Note: table and multileader current styles are not surfaced by the file
reader (no header name/handle exposed), so those markers still default to
the app's active selection until acadrust exposes them.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Every style manager now shows a leading ✓ next to the current style
(current text/dim style, active table/multileader/multiline style), with
the other rows padded so names stay aligned. The marker replaces the
multileader/multiline-only ◀ and is wired for all five managers by
threading the current-style name into each view.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A style manager is now a staged transaction. When it opens, a snapshot of
the style tables, style objects and current-style pointers is taken. Every
New / Copy / Delete / Rename / Set Current / property edit mutates the
document live so the dialog previews the result, but the tab is not marked
dirty, nothing is pushed to undo, and the drawing is not rebuilt.
- Apply commits: one undo entry (restoring the pre-edit state), the tab is
marked dirty, the drawing rebuilds, and the baseline is reset so further
edits stage afresh.
- Closing the window without Apply discards: the opening snapshot is
restored, so none of the changes persist.
Side effects (dirty / undo / bump_geometry) were removed from every style
edit handler and routed exclusively through the commit. Multiline's Apply
is a structural-only commit since its editor has no editable properties.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The list row wrapped a `button` inside the `mouse_area`. The button
captured the press, so the parent `mouse_area` never saw the second
click and `on_double_click` never fired — double-clicking a style name
did nothing.
Drop the inner button: the row is now a plain `mouse_area` carrying both
`on_press` (select) and `on_double_click` (start rename), with selection
shown via the container background. Focus the rename field on start so
the user can type immediately.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The text/dimension/table/multileader/multiline managers were hand-copied,
so their list CRUD kept drifting and breaking: Dimension's New created
nothing (#73), new text styles never reached the ribbon dropdown (#74),
and several were added with a NULL handle (dropped on DWG save, #67).
Collapse the duplicated parts into three shared pieces and leave only the
per-manager property editor bespoke:
- app/style_ops.rs: one StyleKind-dispatched CRUD layer (new / copy /
delete / rename) handling handle allocation, ribbon sync, Table re-key
and name-reference rewrites in a single place.
- ui/style_manager.rs: the window scaffold — uniform toolbar (New / Copy /
Delete on the left, Set Current + Apply on the right), style list and
chrome. Each manager now only builds its editor.
- ui/style_list.rs: the list row — single click selects, double click
starts an inline rename.
New capabilities, now consistent across all five:
- Copy duplicates the selected style.
- Double-click a name to rename it inline.
- Right side is always Set Current + Apply (added Table set-current;
MLine apply is a documented placeholder until its editor is editable).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Large survey drawings (105k entities, multi-million unit extents) could
exhaust all RAM + swap on one core during load and get OOM-killed before
display. Two parser-garbage entities slipped past the corrupt-entity
purge and blew up truck's parameter_division tessellation:
- Near-zero-sweep arcs: the zero-sweep guard only caught sweep < 1e-9,
but a tiny non-zero sweep (e.g. 1.6e-6 rad) still places start/mid/end
within truck's coincidence tolerance and recurses unboundedly. Gate on
arc length (radius x sweep) instead so large-radius small-sweep arcs
(still visible curves) survive while sub-precision arcs drop.
- Desync splines with 100k control points and an invalid knot vector had
no guard at all. Add a Spline arm rejecting the 100k desync signature,
non-finite control points, degree < 1, and bad knot vectors.
Also cap complex-linetype dash expansion (complex_lt::apply_along) at
1e6 pattern repeats, falling back to a solid stroke — a latent OOM of
the same class on map-scale entities carrying a finely-scaled pattern.
Adds regression tests for the arc-length and spline guards.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
PR #80 added the internal SnapType::ObjectPick variant, which left the
exhaustive snap_id match in the settings module non-exhaustive and broke
the build after the merge. Map it to a stable id; it is not a
user-toggleable OSNAP mode, so it never enters SNAP_ORDER or the
persisted set.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
TRIM had no LwPolyline arm in its pick dispatch, so clicking a polyline
segment did nothing (only EXTEND handled polylines). Add trim_lwpolyline:
parameterise the polyline by segment+u, find boundary crossings per
segment, and remove the span containing the click between the adjacent
cuts. A closed polyline needs >=2 cuts and yields the surviving open
arc; an open one yields the surviving piece(s). Fully-surviving segment
bulges are kept; the partial ends become straight. Hover now previews
the trim too.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
POLAR locked the cursor to the nearest angle everywhere, so it could
never point between angles. Snap only when the cursor is within the snap
aperture (in screen pixels) of a polar ray; otherwise leave it free, so
POLAR behaves as if off when pointing away from every angle. The polar
guide line now shows only while actually snapped.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
OTRACK alignment was computed in the wrong plane (it used Z as the
second axis while the model is XY / Z-up), so tracking never aligned
correctly and the aligned point was never fed to a pick.
Rework it: project the cursor onto a ray from a tracking point in the
XY plane — horizontal/vertical by default, or every POLAR increment
when POLAR is on. The aligned point is now used for both the live
preview and the committed click, and a dashed guide shows the ray.
While aligned, a typed distance places the point along the ray from the
tracking point, via the command line and the dynamic-input box (which
shows a single Distance field). The alignment clears with the command.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
OTRACK acquired tracking points by dwell during a command or grip drag,
but only cleared them when OTRACK was toggled off — so the temporary
tracking-point symbols (drawn straight from the point set) lingered in
the view after the command ended.
Drop the tracking points once neither a command nor a grip drag is
active. Acquisition only happens during those, so this covers every
exit path (Esc, cancel, commit) from one place. The tracking vector is
a preview wire already cleared on cancel.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A zero-length dash entry is a dot (e.g. the DOTS pattern), but every
render path treated only positive entries as marks and skipped it, so
dot patterns drew nothing.
- GPU shaders: light exactly the one pixel containing each dot's lattice
centre. The dot grid rotates with the pattern angle while the lit pixel
stays a steady single pixel, so dots no longer flicker at oblique
angles or zoom (an earlier coverage/round test still aliased).
- PDF export (pattern_segments): walk the dash sequence with absolute
phase and emit a short mark for each dot.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three hatch fixes:
- Selected hatches are tinted, but the hatch model cache and GPU upload
were keyed only on geometry_epoch, so Esc cleared the selection set
without repainting — the hatch stayed blue (#71). Key both on
selection_generation too.
- add_hatch stored the boundary entity in local render space, so the
boundary wire re-projected `world_offset` away from the fill on
offset (large-coordinate) drawings. Build it in WCS.
- Clicking a hatch only selected it on its boundary wire, not its fill.
Two causes: visible_hatches_for_click used a narrower block-visibility
test than the render path (fill drawn but not clickable), and
point_in_polygon closed every NaN sub-path — degrading the explicit
2-point edge encoding of a boundary into zero-area loops. Mirror the
render's visibility test and only auto-close sub-paths of >=3 verts.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Removes storm_sewer module and stormsewer crate so this branch is suitable for upstream PR. Plugin host, docs, and CadCommand acquisition hooks remain.
DYN, ORTHO, POLAR, the polar increment, the grid toggle and the
object-snap configuration (OSNAP on/off, active snap modes, OTRACK)
reset to defaults on every launch. Persist them to a per-user
settings.txt, matching the existing recent-files / status-bar stores
(plain key=value, no serialization crate).
Settings load on boot and apply to live state; the update wrapper
snapshots them after each message and writes only on change, so a
toggle survives a restart without thrashing the file. LWT is left out
on purpose — it is a drawing header variable, not a UI preference.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
New layers were added to the table with a NULL handle, and the first
patch used the non-advancing `next_handle()` getter, so two new layers
(or a layer and the next allocated object) shared one handle. The DWG
format is handle-based, so the collision dropped the layer on save;
DXF is name-based and survived, which is why a rename appeared to fix
it.
Allocate a unique advancing handle via `allocate_handle()` in LAYER NEW
(panel + command line) and on rename when the source handle is invalid.
Rename now also clones the whole record so colour/linetype/lineweight
are kept, not just the flags.
Adds DWG/DXF round-trip tests covering one and multiple new layers.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
iced has no built-in ellipsis, so long user content (file names, layer
names, paths, property values) wrapped or spilled past fixed-width
panels. Add an `elide` helper (char-budget truncation with a trailing
ellipsis) and apply it to the layer manager, properties panel, document
tabs, recent-files list and the save dialog.
In the layer manager a truncated name reveals its full text in a
cursor-following tooltip on hover, so the column need not be widened.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The current step's prompt was written into the command-line widget
label. Instead, keep it as a normal history line but freeze its
cooldown: the step prompt is pinned so it does not fade while the user
is still on that step. When the step changes the previous line's
cooldown restarts; when the command ends the pin clears and it fades.
`update` is now a thin wrapper that mirrors the active command's prompt
into the command line after each message; the body moved to
`update_inner`.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The save dialog wrote the typed name verbatim, so entering "foo" with
DWG 2018 selected produced an extensionless "foo". The extension was
only added when the format pick-list changed, not on confirm.
Now confirm appends the selected format's extension when the name has
none (empty name falls back to drawing.<ext>); a name that already
carries an extension is left untouched.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Text-input steps (OFFSET distance, POLYGON sides, radii, counts) were
swallowed by the dynamic-input overlay: typed digits landed in the field
buffer and Enter committed a point pick, so the value never reached
on_text_input (issue #61). Now each text step shows one scalar box whose
value routes to on_text_input on commit; non-numeric prompts (names,
keywords) keep going to the command line so mixed text like "Bolt12"
isn't split.
OFFSET reworked: object first, then through-mode by default — the offset
follows the cursor at the perpendicular distance to the object. Typing a
number locks the magnitude (cursor then only picks the side); a value
typed and clicked without Enter is applied before the point pick.
The dynamic-input overlay and the command line now show the active
step's prompt and hold it until the step completes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reports showed structure placement could stall if command-line text entry
didn't register (ribbon-invoked commands, dynamic-input/focus interactions).
Make a canvas click ALWAYS place a structure, using whatever values were typed
so far (defaults for the rest) — so dropping an inlet works regardless of typed
entry. If typing works, enter invert/rim/area/C first, then click to place with
those values. Pipes are pick-first: click START then END structure; the pipe
commits on the second click with the current diameter/n.
Drop the early focus_cmd_input() return in the dispatch arms (the normal
dispatch tail already focuses the command line and records last_cmd).
15 module tests green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The structure/pipe commands took a viewport interaction first and then asked
for typed values. But a point-pick result (NeedPoint) does not re-focus the
command line, so after the click the typed digits went nowhere and Enter hit
on_enter -> Cancel ("Command cancelled.").
Reorder both commands to the OFFSET-style flow: collect typed values first
(command line focused at dispatch and kept focused through text input), then
take the viewport click last and commit from it.
- PlaceStructure: enter invert/rim/area/C, then click the location.
- PlacePipe: enter diameter/n, then click START and END structures; commit
from the second pick using the click points (connectivity stays exact via
structure handles).
- Dispatch arms focus the command line so the first prompt is typeable.
- Reverts the update.rs inject_picked_entity allow-list change (the pt-based
pipe no longer needs the picked entity).
15 module tests updated and green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replaces the .ssn-file workflow with an interactive one driven by the canvas:
- Structures (SS_INLET / SS_JUNCTION / SS_OUTFALL) prompt invert/rim/area/C on
the command line and commit a CIRCLE tagged with STORMSEWER_STRUCT XDATA
[kind, invert, rim, area, C]. Outfalls skip area/C.
- SS_PIPE picks a START and END structure (entity-pick), snapping its endpoints
to their centers, prompts diameter/n, and stores connectivity (the two
structure handles) in STORMSEWER_PIPE XDATA [diameter, n, from, to].
- SS_ANALYZE / SS_REPORT / SS_PROFILE rebuild the network from the drawn
entities (data.rs::network_from_entities) and analyze it directly; SS_ANALYZE
overlays flow/HGL labels without redrawing the geometry.
Because the data lives on the entities as XDATA, the network round-trips to
DWG/DXF. One additive line in src/app/update.rs adds SS_PIPE to the
inject_picked_entity allow-list so the command can read the picked structure's
center for snapping.
15 module tests cover the XDATA round-trip, analyze-from-drawing, and the
command step machines.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds a "Storm Sewer" ribbon tab backed by a new in-repo `stormsewer` engine
crate (vendored under crates/). The engine implements standard public-domain
methods — Rational-method peak-flow accumulation, Manning circular-pipe
hydraulics, time-of-concentration / IDF intensity, and an HGL backwater pass
with junction losses — with 24 unit tests and no external dependencies.
Module (src/modules/storm_sewer/):
- Drafting: SS_INLET / SS_JUNCTION / SS_OUTFALL place structure markers and
SS_PIPE chains pipe runs (interactive CadCommands).
- Analysis: SS_ANALYZE / SS_REPORT / SS_PROFILE open a .ssn network file, run
the engine, draw the plan and HGL long-section, and print a report.
- SVG icons; SS_* commands registered for command-line autocomplete.
Integration is additive: build.rs auto-discovers the module and one dispatch
block is added to src/app/commands.rs. 7 module integration tests.
See src/modules/storm_sewer/INTEGRATION.md for the design and the optional
follow-up (drive analysis from canvas geometry via XDATA).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extends the Phase 2.2 per-entity memo to the common modify commands, which
still did a full bump_geometry (whole-model re-tess + block rebuild):
- transform_entities (MOVE / ROTATE / SCALE / MIRROR) marks each
transformed handle dirty, then bump_geometry_no_blocks;
- erase_entities drops the deleted handles from the memo and
bump_geometry_no_blocks;
- copy_entities' new handles are natural memo misses, so just
bump_geometry_no_blocks.
All three are top-level operations (including on inserts) that can't change
a block definition, so keeping the block cache is sound. Each now
re-tessellates only the touched entities and reuses the memoized wires for
the rest, matching the grip / line-commit path.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 2.2. wires_for_block_culled re-tessellated every visible entity on
any geometry bump, so a single-entity edit cost a full model re-tess
(~30ms baseline after the block_epoch fix).
Add a per-entity memo (tess_memo: Handle -> Arc<Vec<WireModel>>) on the
culled Model render path. It's guarded by a hash of the tessellation
parameters (tol / view / anno / world_offset / bg / entered viewport); a
mismatch (zoom, layout switch) clears it. The render path now classifies
the visible set against the memo, materializes hits and tessellates only
the misses in parallel, then stores the misses back.
- bump_geometry clears the memo (structural change — safe default);
- incremental edits drop just the changed handle via mark_entity_dirty
and bump with bump_geometry_no_blocks: line commit (new handle is a
natural miss) and grip commit / Esc-cancel (mark the dragged handle).
The hit-test (view_aabb == None), paper and per-viewport paths bypass the
memo so their differing cull parameters don't thrash it. A single-entity
edit now skips the per-entity geometry math for the entire unchanged set;
the Vec assembly + batched-buffer upload stay O(N).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The block-definition tessellation cache (BlockCache::build — every block
defn) was keyed on geometry_epoch, so every edit rebuilt all block defns.
On a block-heavy drawing that turned the ~30ms visible-wire re-tess into a
~400ms spike at line-commit, grip first-move and grip release.
Key the block cache on a separate block_epoch. bump_geometry bumps both
geometry_epoch and block_epoch (safe default — xref, block create,
explode, REFCLOSE, undo, open all keep rebuilding blocks). Operations that
provably can't change a block definition now call bump_geometry_no_blocks,
which bumps only geometry_epoch so the visible wires re-tessellate against
the reused block cache:
- add_entity for a non-Insert/Block entity (line, arc, text, hatch, …) —
scene.add_entity only ever routes to model/paper space, never block
content;
- grip drag hide / commit / Esc-cancel — moving a top-level entity or an
insert never edits a block defn.
Line-commit and grip start/end drop from ~400ms to the ~30ms baseline.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Command-preview / interim / grip-drag wires were appended into the main
wire buffer (all_wires), so every drag frame re-packed and re-uploaded the
entire batched base set — O(N) per move even though only the small overlay
changed.
Split them out: build_primitive keeps `wires` as the cached base only and
passes the overlay as a separate `preview_wires` Arc. The base keeps the
stable tile content id (no per-frame nonce), so its buffer stays resident
and is not re-uploaded during a drag. The overlay uploads to its own small
gpu_preview_wires buffer each frame and draws on top in the wire pass
(same pipeline, same order/depth). The dragged entity still renders
highlighted via its selected-path tessellation.
With the no-re-tess changes, a grip / command drag now costs a single
small overlay upload per move and nothing else.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The grip drag mutates the document live (so grips and the properties panel
track the cursor), which meant Esc left the entity at its dragged position.
Back up the edited entity when the drag starts (the same first-move point
that hides it for the overlay preview) and, on Esc, restore that backup
before un-hiding and re-tessellating — reverting the entity to where it was
when the drag began. A normal release drops the backup and keeps the
dragged geometry.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 2.2 (partial). Dragging an entity's grip called scene.apply_grip on
every mouse move, which bumped geometry and re-tessellated the whole model
per move (plus an O(N) clone of every wire to exclude the edited one from
snapping). On a large drawing that's the 30ms->400ms spike, per frame, for
the duration of the drag.
apply_grip no longer bumps. The drag flow now:
- first move: hide the edited entity from the base tessellation (one
re-tess) and record it in grip_preview_handle;
- each move: apply the grip to the doc (no bump) and show the entity as a
one-entity overlay preview (set_preview_wires) — the base is a cache hit;
- commit on release / Esc: un-hide, drop the preview, one final re-tess.
Because the edited entity is hidden, it's already absent from
hit_test_wires, so snapping runs against the cached set directly — no
per-move clone and no self-snap. Two re-tessellations per drag instead of
one per move.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 2.2 (partial). set_preview_wires / clear_preview_wire /
set_interim_wire called bump_geometry, so every rubber-band frame of a
drawing command invalidated the wire cache and re-tessellated the whole
model — the same 30ms->400ms spike as selection, but while drawing.
Preview / interim wires are an overlay appended to the cached base wire
set in build_primitive; they are not part of the tessellation cache, so
they no longer bump geometry. The base model-tile tessellation stays a
cache hit (static camera, no geometry bump), and the overlay still forces
a GPU wire re-upload through the has_overlay content-id path. Real commits
(commit_entity / transform / copy) bump geometry on their own, so finished
edits and preview-clear still render correctly. Snap/hit-test via
entity_wires_arc also stops re-tessellating per preview frame.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Now that selection/hover are decoupled from tessellation, highlighting the
entity under the cursor only refreshes the GPU xray overlay — no
re-tessellation — so a general rollover is cheap. On an idle ViewportMove
(no active command, no drag) hit-test at the cursor and set_hover_highlight
the result; clear it during commands/drags and on ViewportExit so it
doesn't stick over the ribbon. hit_test_wires is the cached entity-wire
Arc, so the per-move cost is just the existing click_hit projection — the
same the command-hover path already pays.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Picking an entity called bump_geometry, which invalidated the wire cache
and re-tessellated the entire model just to repaint one entity — 30ms to
400ms on a large drawing (visible as a tess-ms spike on the PERF HUD).
Selection highlight is no longer baked into tessellation. wires_for_block_
culled now feeds tessellate_entity an empty selection set, so wires are
always base-coloured and the wire cache is selection-independent. The
highlight is applied in the GPU xray overlay: prepare filters the cached
wires by the live `selected ∪ hover` handle set, recolours them to
WireModel::SELECTED, and batches them — rebuilt only when a new
selection_generation (or the underlying wire content) changes.
Pure select / deselect / hover paths now call bump_selection instead of
bump_geometry, so a pick refreshes just the overlay — no re-tessellation,
no main wire re-upload (geometry_epoch and wire_content_id are unchanged).
Geometry edits, undo and file-open still bump_geometry and rebuild
everything. Grips/properties refresh through their existing app paths.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Follow-up to 3.3. The selection xray pass still built one buffer + one
draw call per selected wire, so select-all over a large drawing issued a
draw per entity. That pass applies neither scissor nor mesh-edge skip, so
all selected wires merge into one order-preserving run via from_run.
Removes the now-dead WireGpu::new / build per-wire constructors.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 3.3. upload_wires built one GPU instance buffer + one draw call per
WireModel — tens of thousands on a large drawing, a major per-frame CPU /
driver cost. Merge maximal runs of consecutive wires that share scissor +
mesh-edge state into one concatenated instance buffer each (WireGpu::
from_run); the existing draw loop then issues a single draw per run. A 2D
model (uniform None scissor, no mesh edges) collapses to one buffer + one
draw.
Runs are consecutive, never globally regrouped, so the sorted wire order
is preserved exactly — depth-biased overlap and alpha blending both
resolve in submission order, so the image is unchanged. from_run
guarantees instance order (index-ordered collect + sequential flatten).
The WireInstance layout, shader, scissor projection and draw loop are
untouched; only buffer packing changed. Selection overlay stays per-wire.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
build_primitive ran split_face3d_wires every frame — for each wire a
string-parse + document handle lookup + clone, to separate Face3D wires
from the rest — even when a pan had just reused the tessellation. The
split is content-only, so memoize it by the Model-tile wire content id:
a pan-reuse (same id) reuses the split, and the common no-overlay frame
returns the cached `other` Arc with zero clone. Preview/interim frames
clone once to append. Non-tile paths split inline as before.
Completes the cheap-pan path: a pure pan now reuses the CPU
tessellation, the Face3D split, and the GPU wire upload — only the
view_proj uniform and per-frame scissors recompute.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 3.2 (wire pan path). Wire vertex buffers are world-space, so a
camera move changes only the view_proj uniform (uploaded every frame),
never the wire data — yet upload_wires re-ran on every camera_generation
tick, re-packing vertices and re-writing the GPU buffer on each pan.
Stamp every Model-tile re-tessellation with a monotonic content id
(WIRE_CONTENT_GEN), reused when a pan reuses the tessellation. The
pipeline records the resident buffer's id and skips upload_wires when the
incoming id matches. The gate is pulled out of the (epoch, camera_gen)
block so a preview/interim wire change still uploads even with a static
camera; non-tile paths and overlay frames force a fresh id, preserving
current behaviour. Monotonic id sidesteps the ABA hazard a freed/realloc'd
Arc pointer would carry.
Pairs with the pan tessellation reuse: a pan now skips both the CPU
re-tessellation and the GPU wire upload.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 3.1 (partial — pan). The Model-tile wire cache keyed on the exact
camera hash, so every pan re-tessellated the whole tile even though pan
changes neither the curve tol nor the culled set (beyond a shift). Re-key
on (geometry_epoch, pan_invariant_hash, tessellated_region):
- pan_invariant_hash covers rotation + wpp (tol) but excludes the pan
target, so a pure pan keeps it stable;
- the wires were culled to a 1.25x-margin region, recorded with the entry;
- a hit now also requires the tile's current visible rect to still fit
inside that region, guaranteeing every on-screen wire is present.
Result: panning within the existing margin reuses the tessellation (tess
ms -> ~0 on the PERF HUD) with no missing geometry. Zoom, orbit and edits
change the signature/epoch and rebuild as before; the cull margin is
unchanged so miss cost — and zoom behaviour — is identical. GPU still
re-uploads per camera_generation (diff upload is Phase 3.2).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The purge scan tests every entity with is_entity_corrupt, whose
per-vertex finite/extent checks on large polylines dominate. Detection
is pure and read-only, so gather entity refs in one pass and test across
cores with rayon; removal stays serial (remove_entity needs &mut doc).
Cuts the "purge" span in the open-time breakdown on polyline-heavy files.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
