[BUG]: Plotly doesn't render properly when used in the Colab VS Code extension.

[kevineger]

Description

Hi there 👋, I'm the maintainer of the Colab VS Code extension which we recently launched:

• VS Marketplace
• Open VSX
• https://github.com/googlecolab/colab-vscode

One of our users reported googlecolab/colab-vscode#339. Summarized, today Plotly doesn't render properly in VS Code when using the Colab extension. This requires users to manually configure the default renderer, instead of being able to lean on Plotly's detection.

Steps to reproduce

• Install the Colab VS Code extension
• Create a new Jupyter notebook file
• Click Select Kernel
• Select Colab
• If prompted (first run), sign in to Google
• Author a cell which renders a Plotly plot, e.g.:

  import plotly.express as px
  import pandas as pd
  import plotly.io as pio
  
  data = {'x': [1, 2, 3, 4, 5], 'y': [1, 2, 4, 8, 16], 'text': ['hello', 'world', '!', '!', '!']}
  df = pd.DataFrame(data)
  fig = px.scatter(df, x="x", y="y", text="text", title="Hello World Plotly Express")
  fig.update_traces(textposition='top center')
  fig.show()

• Execute the cell

Expected behaviour: plot is rendered as it does if you set pio.renderers.default = "vscode" before showing the plot.

Actual behaviour: nothing is rendered. You can Copy cell output and see the HTML output that's returned instead. This works in web-based Colab.

Notes

As you'll see in my reply, this seems to be caused by this block in Plotly, where the default_renderer gets set to colab. If I set plotly.io.renderers.default to 'plotly_mimetype' or 'vscode', everything works as intended.

Unfortunately, VS Code doesn't set VSCODE_PID for these remote Colab kernels, so we can't simply re-order the default renderer checks. One possible solution is to extend the Colab check from only looking for import google.colab to also check for COLAB_NOTEBOOK_ID, which I believe (need to double check) is only set for web-based Colab. Feels a bit brittle. Open to suggestions, and happy to send a PR once we reach consensus!