[AllenDowney]

Link to issue

Closes #3053, which was closed as stale in 2024; it still happens with jupyterlab-plotly 7.1.0.

Description of change

When a kernel updates a displayed figure with update_display_data (for example IPython's DisplayHandle.update), JupyterLab calls renderModel again on the existing renderer, and renderModel returned early because it already had a graph, so the figure never changed. Now it redraws (with Plotly.react, through createGraph) when the figure in the new model differs from the one it drew. It still returns early when the figure is the same, which is what happens when createGraph stores the PNG of the figure in the model.

Demo

The example from #3053, in JupyterLab 4 with plotly 7.1.0:

from IPython.display import display
out = {'application/vnd.plotly.v1+json': {'data': [{'x': ['giraffes', 'orangutans', 'monkeys'], 'y': [20, 14, 23], 'type': 'bar'}], 'layout': {}}}
h = display(out, raw=True, display_id=True)

import time
for y0 in [10, 5, 1]:
    out['application/vnd.plotly.v1+json']['data'][0]['y'][0] = y0
    h.update(out, raw=True)
    time.sleep(0.5)

Before: the first bar stays at 20. After: it goes to 10, 5, then 1.

This also fixes animations in the Octave kernel xeus-octave (its issue #2), which draws figures with this renderer and updates them on each drawnow; checked in JupyterLite with an 8-frame loop: before, the plot stays on the first frame; after, it steps through all eight.

Testing strategy

I didn't find tests for the mime renderer to extend, so I tested by hand: the two examples above, before and after, in JupyterLab 4 and in JupyterLite 0.8, checking the plot's data in the page; and a plot with no updates still renders.

Guidelines

• I have reviewed the pull request guidelines and the Code of Conduct and confirm that this PR follows them.
• I have added an entry to the changelog if needed (not required for documentation PRs).