VSCode Array Inspector

A VSCode extension that shows detailed array information (shape, dtype, device) in a sidebar panel during Python debugging sessions.

Features

• Automatic Array Detection: Debug sidebar panel fills up with all JAX, NumPy, and PyTorch arrays/tensors among locals and globals
• Pin Arrays: Pin important arrays to keep them visible across stack frames
• Real-time Updates: Array information updates automatically as you step through code

Usage

1. Start a Python debugging session
2. The Array Inspector panel will appear in the activity bar
3. When you enter a scope, the panel updates with all arrays and tensors in the frame
4. See detailed information (shape, dtype, device) appear in the panel
5. Click the pin icon to keep an array visible even when out of scope

Pinned Arrays

Pinned arrays stay at the top of the Array Inspector and show:

• Current values when the variable is in scope
• "N/A" when the variable is not available in the current frame

Supported Types

By default, the extension supports:

• JAX arrays (objects of type jax.Array, jaxlib.xla_extension.ArrayImpl)
• NumPy arrays (numpy.ndarray)
• PyTorch tensors (torch.Tensor)

Configuration

Configure which types and attributes to display:

{
  "arrayInspector.supportedTypes": [
    "jax.Array",
    "jaxlib.xla_extension.ArrayImpl",
    "numpy.ndarray"
  ],
  "arrayInspector.attributes": [
    "shape",
    "dtype",
    "device"
  ]
}

Development

# Install dependencies
npm install

# Compile TypeScript
npm run compile

# Watch for changes
npm run watch

Testing

1. Open this folder in VSCode
2. Press F5 to launch the Extension Development Host
3. Open a Python file with JAX/NumPy arrays
4. Set a breakpoint and start debugging
5. Check out the debugging sidebar to see the Array Inspector panel

License

MIT