Michael Waskom

Now what to do about it? On the one hand, I'm inclined to say that this is an example of seaborn behaving as expected and the weird behavior is upstream....

> You might wanna write on top of that page in big red letters "DISCOURAGED!". Well, the second paragraph does say > The figure-level functions are built on top of...

If I had to guess, I'd think pandas thinks about the `pd.NA` objects as being downstream from the "nullable dtypes", and so the proper order of operations would be to...

seaborn is not doing the wrong thing from a plotting perspective here, it is just treating the vector as categorical because a) it does not have a numeric dtype and...

> Expected behavior: Setting clip_on=True should result in neither bars or error bars being clipped. It's reasonable to have this expectation, but it's not correct. The kwargs (as the docs...

Perhaps you could quote the relevant part of the docs? Otherwise it’s impossible to fix…

Yes, it is true that other kwargs are passed through to `ax.bar`. If the bar itself extended past the axis limits, it would be visible. But the error bars are...

This has definitely come up before, I think I have articulated some thoughts. Basically, making this *possible* is straightforward; your approach is the right one. The blocker has always been...

> if the user is specifying the figure, they are responsible for setting it up in a "clean" way, and if they don't then results may not be expected Easy...

To be clear — I am in favor of adding this functionality (and it already exists in the new objects interface). I just expect it to take some careful thought...