wasmi pauses a run before an instruction the store's fuel can't pay for
and charges nothing for it. A memory.fill or memory.copy is charged for
its whole length, so a pause can leave up to that cost in the store.
Plugin::call replaced the store's fuel with the next slice and compared
the instruction's cost with the rest of the budget alone, so a run whose
budget covered its work could end with Limit("instruction budget").
Each pause now tops the store up to the slice size, keeping what it
held, and ends the run only when the store's fuel and the rest of the
budget together can't pay for the next instruction.
A module's start function runs before vc_run. When it logged or reported progress and then failed, the run's report came back empty. The report now holds what the start function reported.
A plug-in whose manifest gives "api": 2 and lists the new `progress`
permission may import three host functions from module `env`:
vc_progress(done, total), vc_cancelled() -> i32 and
vc_log(level, ptr, len) -> i32. Installing checks every import and
refuses, by name, one the host doesn't offer, one with another type and
one the manifest doesn't allow.
- Progress is clamped and passed on once per tenth of a percent (past
2,000 reports, once per 1,000 changes). A heavy filter's progress
window and the export's progress window show it as a percentage bar,
and their Cancel makes vc_cancelled return 1. A live effect's progress
is not shown.
- A run cancelled before it returns ends as cancelled, and one that
returns after its time budget ran out ends at that budget, so what a
plug-in returns after it sees vc_cancelled changes nothing. This also
applies to a plug-in of API level 1 that finishes after a Cancel or
past its time budget.
- Log text must lie in the module's memory and be UTF-8. Lines go to the
app's log as "<id>: <text>" with hidden characters escaped, at most
1,024 characters a line, 200 lines and 32 KiB a run, and the host reads
at most 1 MiB of log text a run.
- plugin.list and plugin.info give the manifest's `api`; the install
prompt names the new permission in plain words.
Docs: docs/plugins.md, docs/mcp.md, docs/target-app-parity.md.
wasmi compiles each function of a plug-in when it is first called and,
by default, charges the run that calls it 7 instructions for each byte
of the function. The host hands a run its budget in 20 M-instruction
slices and resumes it when a slice runs out. But when a slice runs out
while a function compiles, wasmi returns an error from the call, not an
out-of-fuel pause, and the host reports it as Limit("instruction
budget"). So a plug-in's first run could fail with most of its budget
left, depending on where the slices fell, and the next run, with its
functions compiled, succeeded.
StitchCraft's PES exporter hit it on a design of 98 × 148 mm. In
VectorCraft 0.17.0's host its first export failed with 10 G and with
275 G instructions alike, and its second succeeded, taking under 1 G.
Run on wasmi directly, with the host's slices, the first export failed
in its 36th slice ("required_fuel=174139"), and with compiling free it
finished in 49.
Compiling now costs nothing: wasmi's fuel costs with no cost for
translating or validating, and its default cost for copying. A run pays
for its own instructions only, as many on its first run as on the next.
Compiling every function when the plug-in loads (CompilationMode::Eager)
also fixes it, but loads a 1.4 MB module in 42 ms instead of 10.
The new test fails without the change: a run that takes 1,040
instructions once its functions are compiled failed within them on a
plug-in just loaded.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhvnnnD5fTFupK8WWq35un
A filter plug-in gets 50 M instructions and a minute for each run.
Filters that order a plotter's paths, pack shapes or run a simulation
need minutes, and the app shouldn't freeze while they compute.
A filter's manifest may now ask for "budget": "heavy" (Budget): each run
gets HEAVY_FUEL (1 G instructions) and HEAVY_WALL_TIME_MS (300 s). Only a
filter may ask. plugin.install refuses such a filter without allowHeavy:
true, and plugin.inspect returns the entry a module would have without
installing it. Install Plug-in… and File › Open ask the user and install
on OK; the plug-in store keeps the answer, and folder loads skip heavy
filters with a note, since they ask nobody.
The app runs a heavy filter on a thread of its own (plugin_runs), not in
Background Save's queue, with the status bar and a progress window
(Cancel, Continue in Background) that it shares with Background Export
(export_progress::window). FilterRun splits plugin.run: prepared on the
session, run on any thread, its output handed in, and the next
plugin.run with the same params applies that output instead of running
the plug-in again, so the edit stays one guarded, journaled command. A
result whose document or plug-in changed meanwhile is dropped, and the
status bar says to run it again. A cancelled plug-in run is now
EngineError::Cancelled.
What a filter keeps with the document (#3720) goes with the run: it is
given _context.documentData when the run is prepared, and the
documentData of its output is applied with its objects, in the same undo
step, when the run is handed in (FilterRun::finish).
Tests: plugins/tests/abi.rs (100 M loop steps out of a light budget,
within a heavy one; folders skip heavy filters), manifest.rs, engine
tests_plugins.rs (consent, inspect, the store, hand-in, a handed-in
run's document data, cancel) and
ui-egui tests_plugin_runs.rs (installing asks, no Preview, a background
run as one undo step, a stale result kept out, the progress window).
CRAFT_UI_HEAVY_FILTER_WASM=<folder> writes a test filter to try in the
app. Its strings and messages have rows in each complete catalog
(COMPLETE_MESSAGES).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhvnnnD5fTFupK8WWq35un
A plug-in could filter objects or reshape them as a live effect, but
nothing it made could leave the app: pen plotters, cutting machines and
laser cutters read formats VectorCraft doesn't write. An exporter is a
third kind of plug-in ("kind": "exporter", with the file "extensions"
it writes), in the same sandbox and with the same ABI v1 exports.
- It adds a write-only format, plugin.<id>. document.formats lists it
with its plug-in id, and document.export writes it by that id, or by
a path ending in one of its extensions that no built-in format has;
"params" carries its parameters, checked against its manifest as a
filter's are.
- File › Export › Export As lists it after the built-in formats, with
Use Artboards greyed out, since the plug-in writes the whole
document. One with parameters asks for them in a dialog generated
from its schema (pluginExport) once the file is picked.
- The plug-in is given the native file's JSON (format::json_part:
images of 8 KiB or more without their data) and _context {mode:
"export", extension, name}. It returns {"dataBase64", "warnings"?}
or {"error"}. The host writes the file, and nothing when the plug-in
declines or returns bad data.
- An export gets 10 G instructions plus 4,000 per byte and 300 s.
Background Export's Cancel stops it between two slices of that
budget (Error::Cancelled), and a cancelled live-effect run is not
cached as a failure.
- plugin.run and Object › Plug-ins say where an exporter is instead of
calling it a live effect.
engine/tests/plugin_export.rs reads back what an exporter is handed
with wat::echo_exporter, an exporter whose file is its input or its
parameters (its base64 encoder is written in WAT). The document it is
given is document.save's JSON, without the large image's data, and its
parameters arrive defaulted, clamped and with the file's name and
extension. It also covers listing and removal, built-in extensions
winning, selectedOnly, declines, bad base64, missing data, warnings,
and Cancel stopping a plug-in that never returns. It is a test binary
of its own: installed plug-ins are process-wide, and the format-list
unit tests check exact lists. dialogs/tests_plugin_export.rs drives
Export As headlessly; plugins/tests/abi.rs covers the budget, the
parameters and the echo exporter; format/tests/native_v4.rs covers
json_part.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhvnnnD5fTFupK8WWq35un
VectorCraft had an Additional Plug-ins Folder preference but nothing loaded from it. Port
PhotoCraft's wasmi plug-in host (same sandbox, manifest and parameter schema) as the new
vectorcraft-plugins crate (L2), extended for vector documents:
- Object filters read the selected paths and compound paths as JSON (native PathData and
Paint forms: geometry, fills, strokes, gradients, opacity) and return the objects to keep,
change, add or delete; the output is validated in full and applied as one undo step
(plugin.run). A failing plug-in leaves the document untouched.
- Live-effect plug-ins are geometry effects in the appearance stack (effect id
plugin.<id>), re-evaluated (and cached) when the object or parameters change, editable,
exported and baked by Expand Appearance; a missing plug-in leaves the geometry as is.
- plugin.list / info / run / install / remove / reload; the preference folder installs
every *.wasm at startup and when it changes; File > Open of a .wasm installs it (web).
- Object > Plug-ins and Effect > Plug-ins menus with dialogs generated from the parameter
schema (live preview), the Appearance panel's fx menu, docs/plugins.md, an original
example plug-in (Desaturate, a Rust crate outside the workspace) and fuzzing of the
plug-in boundary (engine/tests/plugin_fuzz.rs).
Tool plug-ins and paint-changing live effects are left for later (docs/plugins.md).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
dotCooCoo