simply-view-image-for-python-debugging

Based on the great work of john-guo, a more general version for the extension.

Features

• Built-in editor for viewing images:

• Support in Jupyter notebooks

• Support view image from expression

  • NOTE when using this feature, the expression is actually executed, so beware of side-effects

• Support plot (matplotlib, plotly) view

  • Currently supports:
    • matplotlib.pyplot.Figure
    • matplotlib.pyplot.Axis
    • plotly Figure (saving backend is needed, see here)

  

• Support Tensor view: pytorch and numpy

  • numpy.ndarray is considered a tensor if it has 4 channels, or 3 channels but it does not pass as an single image. scikit-image is required for this.

• Hover over image variable to see the image shape (sometime it's not visible/easy to find in the general debug hover).

Watch view

Added a watch view, for watching image/plot/tensor variables, while refreshing the image-view at each breakpoint.

• Support custom python expressions (again, beware of side-effects)

Q & A

• Memory blows-up when using the extension.

  It might happen when (very) large, non-numpy array object, is being used. The solution to it is to set the restrictImageTypes setting to true (should be by default).

• Selecting different call-stack frame does not work as expected.

  Use the command: Debug View Python: Update Frame ID.

  Why:

  I couldn't find a way to get the current frame, if it was changed by the user (again and again). So, I've added a command to force VSCode "tell" me the current frame. It's a bit hacky, but it works, I think.