PySonar2 Code Intelligence for VS Code

PySonar2 Code Intelligence brings PySonar2's whole-project Python type inference and cross-file index into VS Code through a Java language server.

Project links: interactive web demo · VS Code Marketplace · GitHub releases

Install

Install the stable release from the Visual Studio Marketplace, search for PySonar2 Code Intelligence in VS Code, or run:

code --install-extension smallyu.pysonar2-code-intelligence

The extension bundles the PySonar2 server JAR, so no separate server deployment is required. Analysis runs on the machine hosting the VS Code workspace extension. In a Remote SSH workspace, Java and Python must be installed on the remote host.

The extension currently provides:

• go to definition;
• find all references;
• inferred-type and docstring hovers;
• document and workspace symbols;
• conservative semantic diagnostics with cascade suppression; and
• one isolated language-server process per workspace folder.

PySonar2 complements existing Python extensions. It does not provide formatting, debugging, completion, or environment management, so keeping the Microsoft Python extension and Pylance enabled is recommended.

Requirements

• VS Code 1.91 or newer
• Java 11 or newer
• Python 3.10 or newer

By default the extension runs java and python3 from PATH. Use pysonar2.java.command and pysonar2.python.path when those runtimes live elsewhere.

Analysis model

The server indexes the saved files in each workspace folder. After a Python file is saved, it waits briefly for related file events, performs a serialized workspace rebuild, and atomically publishes the new index. Navigation remains available from the previous completed snapshot while indexing.

Unsaved changes are marked as stale in the status bar. Save the document to refresh semantic results. Common virtual-environment, dependency, cache, build, and generated directories are excluded by default. Nested Python package roots and directories containing pyproject.toml, setup.py, or setup.cfg are added to module resolution automatically, including when the VS Code workspace is a larger monorepo.

During a rebuild, the PySonar2 output channel reports the active phase, elapsed time, discovered file count, analysis percentage, and current workspace-relative path. The status bar shows the current analysis percentage; select it to open the detailed output. Progress notifications are throttled so large workspaces remain responsive.

If an unusually large project exhausts the Java heap, first exclude unrelated roots such as tests, scripts, migrations, or sibling repositories. Increase pysonar2.java.maxHeapMb only when the workspace must be analyzed as one unit. Progress lines include current and maximum heap usage to make that decision visible rather than speculative.

For a repository that contains several independent Python surfaces, a focused workspace configuration can avoid indexing code that cannot contribute to application navigation:

{
  "pysonar2.analysis.exclude": ["tests/**", "scripts/**", "alembic/**"],
  "pysonar2.diagnostics.mode": "conservative",
  "pysonar2.diagnostics.maxPerFile": 100,
  "pysonar2.java.maxHeapMb": 2048
}

Exclude only roots whose definitions and references you do not need. The heap override is optional; the optimized snapshot path normally stays well below that limit.

Commands

• PySonar2: Reindex Workspace restarts the per-folder servers and builds fresh indexes.
• PySonar2: Show Output opens status and failure details.

Settings

Changing a PySonar2 setting restarts the language server so the new runtime and file-selection settings take effect consistently.

Run the repository demo

From editors/vscode:

npm install
npm run build
code .

Press F5 and choose Run PySonar2 Extension Demo. The Extension Development Host opens ../../demo_project. Try these flows:

1. Open main.py and hover over DemoApp or report.
2. Use Go to Definition on DemoApp, build_report, or prediction.summary.
3. Use Find All References on Market, Prediction, or weighted_signal.
4. Open the Outline view or run Go to Symbol in Workspace.
5. Save a Python file and watch the PySonar2 status item move from indexing to ready.

You can also open demo.code-workspace to inspect the extension and Python demo side by side.

To verify the packaged Java server over real stdio JSON-RPC without opening VS Code, run:

npm run build
npm run smoke

Build a VSIX

npm install
npm run package

The prepublish step builds the Java project, copies target/pysonar-3.2.0.jar to the extension package as server/pysonar-lsp.jar, bundles the TypeScript client, and produces pysonar2-code-intelligence.vsix. Set PYSONAR_MAVEN_REPO_LOCAL when the build needs to use a non-default Maven dependency cache.

The Marketplace release uses the same VSIX artifact produced by this command. Publishing is handled by the smallyu Marketplace publisher; a local build does not contact or depend on a hosted PySonar2 API.

Current limitations

• Semantic results follow the last saved workspace state; unsaved-buffer overlays are not implemented.
• Rebuilds are whole-workspace rather than dependency-graph incremental.
• Modern Python 3.11-3.14 syntax is indexed, with conservative navigation semantics for exception groups and generic type parameters.
• Type annotations are parsed for navigation but are not yet used as complete inference inputs. Conservative diagnostics avoid treating unresolved annotation and dependency types as definite code errors.
• Web extensions are not supported because the analyzer starts Java and CPython processes and reads a local or remote workspace filesystem.