Tim Hoffmann

Nope, it's a `QuadContourSet`: https://matplotlib.org/api/_as_gen/matplotlib.axes.Axes.contour.html

Not tested, but `set()` claims to accept `lw` https://matplotlib.org/stable/api/contour_api.html#matplotlib.contour.QuadContourSet.set If it does, it'd be reasonable that the constructor does too.

Not sure if that's too much magic, but IMHO tick labels should always "point" to their tick. That would require different alignment depending on the rotation angle. It could be...

This ist mostly "just not implemented". However I except that there can be some corner cases (e.g. the alignment discussion above), where we might want to deviate from simply passing...

While not a replacement, the new rotation modes in #28968 remove the need for explicit alignment control for one of the current most common use cases: fix tick alignment after...

Hi @evelyn-lo thanks for your interest. I'm still unclear whether we should support the alignment. If so, it would have the effect as described [here](https://github.com/matplotlib/matplotlib/issues/13774#issuecomment-479503856), which should explicitly be mentioned...

> The use case that springs to mind for me is the third example here, where we add a bunch of spaces to move the month labels to the right....

If we are touching the existing API, should we reconsider more fundamentally? Currently, `pie()` has two sets of annotations 1. "label", which are strings and by default displayed outside next...

What I've missed in my above thought is that `{}` direct percent-formatting takes case of the factor 100 between relative and percent numbers, but it does not solve the issue...

I've just read your comment on legend. That's actually making it more complicated. ``` plt.pie(x, labels=['a', 'b', 'c']) plt.legend() ``` plots both the outer labels *and* the legend, which is...