3D scene embed with PyVista (vtk.js)

[FelipeCybis]

Description

Recently, PyVista introduced a (still in development but not bad) html exporter of its 3D scene. Basically it wraps everything in vtk.js and loads it in an offline viewer.

With jupyter nbconvert it works pretty well, but I did not have the same luck with quarto.

Anyways, here is the code I am using to test this (took out one back quote from the code cells for better rendering here):

---
title: HTML export
execute:
  enabled: true
embed-resources: false
jupyter: python3
---

```{python}
import pyvista
from ipywidgets import widgets
```

```{python}
widgets.IntSlider()
```

```{python}
pyvista.set_plot_theme('dark')
plotter = pyvista.Plotter()


grid = pyvista.ImageData(dimensions=(9, 9, 9))
grid['scalars'] = -grid.x
pl = pyvista.Plotter()
_ = pl.add_volume(grid, opacity='linear', mapper="smart")

pl.show(jupyter_backend="html")
```

## New section

```{python}
print("Hello")
```

The problem here is how the script inside the ipywidgets that is generated by pl.show(jupyter_backend="html") is decoded in quarto, because I can tweak the html after and make it work. For example there is one part of the script that should be ;\"><\/iframe>", but quarto generates ;\">". Then there are also some </script> that apparently should be <\/script>

I am not at all expert in javascript and html so I just back engineered the html outputed by nbconvert to make it work in the html generated by quarto.

What I actually want to know is where the change should be done and what change... Is it a quarto thing when it decodes the ipywidget? is it the string in the ipywidget that should be encoded differently? why nbconvert reads and decodes the string correctly and not quarto? I would love someone help on this :)

Here are all the files I am using
https://github.com/FelipeCybis/trame_quarto_example

Also I am using quarto 1.4.514.

Thanks in advance and thanks for the amazing quarto :)

[mcanouil]

Using latest pre-release version, latest Python and latest modules I have no issues.
Note that it seems, the module requires a running Python server, so you might not be able to publish/share HTML files.

Screen.Recording.2024-01-01.at.14.28.57.mov

[cscheid]

Hm. There might be a funny hack that works. I'll open an issue.

[FelipeCybis]

Great, thanks! Once I have some time I will try to correct this in a fork of pyvista and see if I can propose a PR on their side as well.

[cscheid]

There's not much they'll be able to do. The PR would have to be in ipywidgets, to support base64 encoding of the JSON.

[cscheid]

Tracking here: #8433

[FelipeCybis]

Thanks for that @cscheid !

Just to see if I understood correctly, if you allow me.

The PR you are opening is adressing the same problem that ipywidgets' utils adresses here https://github.com/jupyter-widgets/ipywidgets/blob/b2531796d414b0970f18050d6819d932417b9953/python/ipywidgets/ipywidgets/embed.py#L230-L241, correct?

I tried your PR branch by the way, and works perfectly :)

[cscheid]

Pretty much, yes. If they're escaping their strings explicitly, then I think they should fix the example here to highlight that they expect these strings to be escaped as well: https://ipywidgets.readthedocs.io/en/7.x/embedding.html

But in any case, this fix is coming soon. We're trying to come up with a test that doesn't take ~10MB of space, but it'll be there shortly. Thanks for confirming!