With xlim = ylim = (-0.001, 1.001) and setting aspect=1, I'm getting plots like this:

The curve gets cropped, and not clearly distinguishable from the spines. I think we need at least (-0.01, 1.01), see https://gist.github.com/Charlie-XIAO/ecd5173fe0daa9c49f442153ef19e7d1. Let's wait for maintainers to decide.

Thanks for the PR, could you please document your change for version 1.3 under doc/whats_new/v1.3.rst?

Looking at this PR and #26366, it seems that we need to merge PR together. It will simplify the code: By setting first x- and y-lim then, we can directly set the aspect to 1 and avoid using get_data_ratio.

So let's close one of the 2 PRs and change slightly the scope.

With xlim = ylim = (-0.001, 1.001) and setting aspect=1, I'm getting plots like this:

We should indeed take care of the offset to apply. It might be dynamic in case we want the despine option or not. One thing sure, we want it as little as possible but we don't want to crop the line on the display.

Closing in favor of #26356