Deprecate to_html()? Or make it work with Vegafusion?

[joelostblom]

What happened?

I accidentally mixed up to_html() and save and noticed that when using to_html with the vegafusion data_transformers is enabled, I run into this error:

ValueError: When the "vegafusion" data transformer is enabled, the 
to_dict() and to_json() chart methods must be called with format="vega". 
For example: 
    >>> chart.to_dict(format="vega")
    >>> chart.to_json(format="vega")

import altair as alt
import pandas as pd

alt.data_transformers.enable('vegafusion')

source = pd.DataFrame({
    'a': ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I'],
    'b': [28, 55, 43, 91, 81, 53, 19, 87, 52]
})

alt.Chart(source).mark_bar().encode(
    x='a',
    y='b'
).to_html()

What would you like to happen instead?

Instead of hardcoding to_dict() without args and mode='vegalite', we could allow to pass arguments for these parameters in the to_html() method. The mode could just be named mode and the dict option could either just be format or dict_kwds similar to what we have for json_kwds. It's possible to override the mode via embed_options but I think an explicit argument would be clearer.

Another option would be to deprecate it. It has been mention in the past that it is no longer used, but not sure if it was reincarnated after that or just never removed #420 (comment)

Which version of Altair are you using?

main

[jonmmease]

I think we can dynamically choose the mode based on whether VegaFusion is enabled. The easiest thing might be to reimplement to_html, in terms of save, which already does this.

But this does raise the question of why we have this method in the first place!

[mattijn]

With a few more code reincarnations and a bit of generation, we can resolve this related issue #3189.

[mattijn]

In line with what I said in #3189, I would be in favor of keeping this method and make it works with vegafusion.

[joelostblom]

Sounds good to me!

[NickCrews]

I use .to_html() to save multiple charts into a template that holds multiple charts in different tabs, so it would be great for me if we could keep it around!