python: bypass plotnine auto-closing comms

[isabelizimm]

Release Notes

New Features

• N/A

Bug Fixes

• Python - plotnine plots throw: "Error rendering plot to 'Auto' size: RPC timed out..." #6393 Fixes plotnine bug where every plot gives RPC timeout error

QA Notes

pip install plotnine

from plotnine import ggplot, geom_point, aes, stat_smooth, facet_wrap
from plotnine.data import mtcars

(
    ggplot(mtcars, aes("wt", "mpg", color="factor(gear)"))
    + geom_point()
    + stat_smooth(method="lm")
    + facet_wrap("gear")
)

Should create 1 plot with no errors. You can also check that in the logs, the plot comms are being used, rather than display_data

@:plots

[github-actions]

E2E Tests 🚀
This PR will run tests tagged with: @:critical @:plots

^{readme  valid tags}

[isabelizimm]

wip for feedback, but i'll add a test before we merge this

[austin3dickey]

Nice find! This branch seems to fix the weirdness in the linked issue.

[nstrayer]

Works great for me! This feels about as surgical as it could get in this case!

[isabelizimm]

the retina problems were stemming from us overriding plotnine's layout manager! updated so we don't override it

[seeM]

Great catch! Can we use the public method self.figure.get_layout_engine()? Might also need to check that it isn't None. And it might be safer to do something like .startswith('plotnine') in case they move it around

[isabelizimm]

what do we think of always leaving the layout engine alone if it is not None? I understand the sizing won't be exact, but it might be a simple escape hatch to respect user's settings.

[seeM]

Sounds like a better approach! We'd just need to double-check that it doesn't break any existing behavior. Might be covered by e2e tests.

[austin3dickey]

Looks fantastic! what a bug