Make altair_viewer and altair_saver work with Altair 5

[mattijn]

As was pointed out in #2686 (comment). Once Altair 5 RC is released altair_viewer and altair_saver won't work anymore as they are using the entrypoints of altair vegalite 4.

Observe current behaviour using altair from GitHub main branch:

import altair as alt
from vega_datasets import data
import altair_viewer

cars = data.cars.url

op_var = alt.param(value=0.1)

chart = alt.Chart(cars).mark_circle(opacity=op_var).encode(
    x='Miles_per_Gallon:Q',
    y='Horsepower:Q',
    color='Origin:N'
).add_params(
    op_var
)

altair_viewer.display(chart)

---------------------------------------------------------------------------
NoMatchingVersions: No matches for version='5.2.0' among ['4.0.2', '4.8.1', '4.17.0'].
Often this can be fixed by updating altair_viewer:
    pip install -U altair_viewer

Originally posted by @mattijn in #2599 (comment)

[joelostblom]

Is there any functionality from altair_viewer or altair_saver that we still would like to be usable from altair 5? For altair_saver, I can't think of anything that isn't already covered by vl_convert. Unless I'm missing something, maybe we should add a note to the altair_saver readme that we encourage users to use vl_convert for altair 5 instead?

For altair_viewer, we don't yet have a way to support the display/show functionality in altair 5 (and some requests to add it in altair-viz/altair_viewer#59). I wonder if we can lift out the browser opening logic from altair_viewer directly into altair and rely on vl_convert to display the chart in a browser tab either online or offline mode. @jonmmease Do you think this would be possible?

[jonmmease]

As I recall, altair_viewer does a bunch of work involving websockets to make repeated chart displays automatically show up in the same browser tab. I'd personally prefer to have each call do chart.show() open a new browser tab. If we use this approach, then I think we can rely on vl-convert to produce the HTML.

I wrote some logic for plotly.py a while back that takes an HTML string and opens it in a new browser tab without writing to a temp file. See https://github.com/plotly/plotly.py/blob/a6662ab48bf0a71f703fac1833a4d7e6c805b805/packages/python/plotly/plotly/io/_base_renderers.py#L655-L709. If this is the behavior we want, then I think it would be pretty straightforward to combine this logic with vl-convert.

[joelostblom]

From my side of things, those solutions sounds great! I think the most important part of this feature is to enable an altair workflow from the terminal; it doesn't have to mimic altair saver exactly. I'm in favor of making the behavior easy to implement/maintain and I don't see a noticeable advantage of reusing the same tab (and if someone really wants to, they could write a chart to file and have that filepath open in a single tab that they refresh). You plotly solution sounds like a good approach to me as well.

[stefaneidelloth]

If you include save/export options in altair, you might want to support *.pdf as well.
#2612
In matplotlib a corresponding command would be chart.savefig(file_path, ...)
https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.savefig.html
In plotly its
https://plotly.github.io/plotly.py-docs/generated/plotly.io.write_image.html

[jonmmease]

Hi @stefaneidelloth, PDF export is supported using chart.save("fig.pdf"). Starting in Altair 5.2.0, this is now based on [vl-convert] rather than altair_saver. vl-convert has no external dependencies on selenium, node.js, or a web browser, so it's a lot less complex to use.

[joelostblom]

I'm closing this since Altair 5.2 brought in PDF export and mentioned above, and since Altair 5.3 the functionality of Altair Viewer is now available in Altair via the vl-convert package instead. See the docs on how to use the browser renderer to display charts via .show(), and the PR #3379 for more details. I believe that means all the functionality from altair_viewer and altair_saver are now completely covered by altair + vl-convert.

Immense thanks to @jonmmease for all your work on vl-convert making it simpler to both save and display altair charts!