CPN IDE for VS Code
Run CPN IDE — the editor and simulator for Coloured Petri Nets — inside VS Code, with nothing to install
beforehand. The extension ships the CPN IDE backend and UI, downloads a portable Java 17 runtime on first use, and opens
.cpn models, or text models (.petri.json), in an editor tab.
Install
Install CPN IDE from the VS Code Marketplace (Extensions view, search "CPN IDE"), or download the VSIX for your
platform from the GitHub releases and run
Extensions: Install from VSIX….
Examples
examples/ has four small models that validate cleanly and are simulated in CI:
| Model |
What it shows |
01-simple-flow |
Three UNIT tokens move from Start to Done; the net stops after 3 steps |
02-producer-consumer |
A counter produces items 1..5 into a buffer (guard [n <= 5]) and a consumer takes them out |
03-mutual-exclusion |
Two processes share one lock, so at most one is ever in Critical |
04-hierarchy |
A Worker subpage behind a substitution transition, with a fusion place shared by both pages |
Open the generated .cpn, click the ▶ button in the editor title bar, then validate and step the simulation.
The sources are examples/*.petri.json + *.sml; the .cpn files are generated by node scripts/gen-examples.mjs
(--check fails if they are out of date).
Text models
A model can be written as text: NAME.petri.json describes pages, places, transitions and arcs (see
schema/petri.schema.json), and NAME.sml holds the CPN ML declarations (color sets,
variables, functions). They compile to CPN Tools XML.
- The standard declarations (
UNIT, BOOL, INT, P_HIGH, ...) are added automatically; declare the same name to override one.
(*! Block *) or (*! Outer / Inner *) in the .sml starts a named block of declarations in the CPN Tools tree.- Positions (
pos) are optional; nodes without one are laid out automatically.
- Opening a
.petri.json shows the net in CPN IDE; saving there writes back to the JSON. Convert to Text Model turns an existing .cpn into this format, and Export .cpn goes the other way.
- Compile errors show up as diagnostics while you edit. When the backend is running and no CPN IDE panel is open, saved models are also type-checked by the simulator.
Simulation speed
The bundled UI waits 500 ms between steps of a multi-step run; the backend needs a few ms. Set cpnIde.simulation.stepDelayMs (default 100) and cpnIde.simulation.animate to speed it up, and cpnIde.debug to bring back the UI's verbose console output. The settings apply when an editor is (re)opened. "CPN IDE: Run Simulation (No UI)" runs a model without the UI and prints the final marking to the "CPN IDE" output channel.
| Platform |
Local simulator |
Notes |
| Linux x64 |
yes |
needs 32-bit glibc (libc6:i386 on Debian/Ubuntu, glibc.i686 on Fedora) |
| Windows x64 |
yes |
|
| macOS |
no |
the CPN simulator is a 32-bit binary; point cpnIde.backendUrl at the Docker backend |
Simulator daemon and security
The CPN simulator runs as a separate daemon (cpnmld, fixed port 20001) that executes the SML code of a model. The
extension patches the bundled daemon to listen on 127.0.0.1 only, starts it in a private per-user temp directory
(<globalStorage>/tmp/<pid>), and stops it, with its workers, when the backend stops or exits. Because the port is
fixed, only one daemon can run per machine: a second VS Code window (or another local user) reuses the daemon already
listening on 20001, and the "CPN IDE" output channel warns when that happens. Daemons left by a crashed VS Code are
cleaned up on the next start. Calls that create a net (POST /api/v2/cpn/init) go through a small local proxy that runs
them one at a time, because two at once can hang the upstream backend.
Docker backend
docker run -p 8080:8080 ghcr.io/alefire99/cpn-ide-backend:latest
Contributing
See CONTRIBUTING.md. We use Git Flow: work on feature/* branches, open PRs into develop.
License
AGPL-3.0, same as upstream CPN IDE. See THIRD_PARTY_NOTICES.md.
