We don't officially support pandas, so I doubt we are officially going to support polars. However if polars data frames behave properly when using no.asarray then they should work the same as pandas data frames in our methods. Did you test to see if they work?

If you mean df.plot type methods that is totally in pandas court, but if they go that route they should come talk to us.

This is the sort of integration that https://github.com/matplotlib/data-prototype is trying to make easier by pushing the integration point out a level from the core library.

In the current state of Matplotlib, we need things as numpy arrays (and will do that conversion internally via np.asarray as Jody said) because at the bottom we need to actually have all of the array data materialized in our code. If polars supports this it should "just work" and if not, I think that the work needs to happen in polars to support that as it is very very unlikely that we are going to thread a polars dependency through the entire code base.

I am going to close this and track this work in #16402 and #22645 .

Polars works fine so far as I can tell, at least for a simple dataframe. Again, if you want df.plot() that is something to request of polars.

import matplotlib.pyplot as plt
import polars as pl

data = {"a": [1, 2, 3, 4], "b": [3, 4, 5, 6]}
df = pl.DataFrame(data)

plt.plot(df['a'], df['b'])
plt.show()