I simplified the example and added xmin and xmax arguments.

Sure thing. Can you elaborate on what you mean by different axes? I can test axhline on different plots, but I am unsure how to utilize it on different axes as it itself only plots on the horizontal. Below is an image of what I have come up with so far. Thank you for your feedback!

@CozyFrog ah I mean having subplots where the X/Y data are different combinations of datetime vs numbers. Right now you have datetimes on the Y axis with numbers on the X axis, and adding the other permutations of that should exercise these codepaths more fully (I think). Your other PR did this well.

@scottshambaugh I don't think that actually makes sense for this method, honestly... the x axis values are in axes coordinates, not data coordinates (well, unless you hand a transform argument), so really the only value handed to axhline that matters for units is the y value.

I'll defer to you on that @ksunden, I'm admittedly not super familiar with how the code paths for the coordinate transforms work work. If it's drawn directly in axes coordinates they I think you're right, my thought was if the xrange was cast to the data coordinates before being drawn.

Thank you for the discussion @scottshambaugh & @ksunden – how should I move forward with this PR? Does this require any modifications?

Yes, the X values are drawn directly in axes coordinates, and in fact the method explicitly checks that those values do not have units and prevents passing an alternate transform even.

Thus I think we should be good to go here, though we do have a 2-reviewer policy for non-doc changes (And I was the first approver, soo).