@emilykl please review

@m-ad Thank you for the PR!

I'm open to something along these lines as a "patch" fix to enable add_vline() and add_hline() to work with datetime axes, although as @nicolaskruchten mentioned, the "correct" fix would be to refactor the code such that we're not trying to do math with data values.

Ultimately, now that shape labels are natively supported in Plotly.js, I think the longer-term fix here is to refactor add_hline() and add_vline() to add annotation_text as a shape label rather than as a separate annotation; that will allow us to skip the step of computing averages of the data values. I mention it here just in case that's a task you'd be interested in -- I've opened issue #5373 to track.

Otherwise, something like what you've proposed here will work as a temporary fix.

There's a few issues with the logic as written; for example, falling back to returning x[0] is incorrect when x[0] and x[1] are not equal. We should simply raise an error in that case because we have no way of calculating an average.

That means cases like the reproducible example in #3065, where the date value is passed as a string, will require a different approach. I'm not sure exactly what the best approach is; trying to parse every value as a datetime seems a bit clunky, but technically speaking it will produce the correct result. Or maybe we just document this limitation and specify that if you are trying to use add_hline() or add_vline() for a datetime axis you must pass the value as a datetime.

@m-ad since we haven't seen any activity on this PR in a while, we may close it in the near future. Let us know if you are still working on it!

@m-ad since we haven't seen any activity on this PR in a while, we may close it in the near future. Let us know if you are still working on it!

@robertclaus I am not currently working on it. This is a is a band-aid-style fix and as @emilykl said, the nicer and more "correct" version would be to implement the changes to add_hline() and add_vline() proposed in #5373. Someone appears to be working on that, which would ideally make this PR unnecessary. That said, if you still like to have my band-aid fix around until #5373 is closed, I'd be happy to finalize this PR.