Luke Zappia

Just in case you were wondering I noticed this when trying to find the cause of this issue https://github.com/lazappi/clustree/issues/14 in my clustree package. It's probably not a normal use case...

So admittedly this is mostly a philosophical objection, it's probably unlikely to cause any computational problems. The main practical issue I have is converting between Python and R where this...

> Hey! Just to chime in, I believe plotting functions also expect categoricals and I've had errors from other functions as well about obs columns not being categorical. I think...

> I'm pretty sure I took your position back when this behaviour was added. The closest issue I could find was [theislab/anndata#115](https://github.com/theislab/anndata/issues/115), but I remember having a longer discussion with...

Just noticed this is almost an exact duplicate of #1985 (check the issues but not PRs 🤦🏻 ). Feel free to close this if you want to keep the older...

Made the `.size` change and added a test. Not entirely sure I have done the test correctly so let me know if that needs adjusting.

Sorry, I forgot about this. Not sure why the CI was falling. The only errors I get locally were due to problems with matching images.

It looks like the same `AssertionError: Error: Image files did not match.` error I was getting locally from some of the spatial tests. I haven't touched this so not sure...

See also #1985

@azodichr I made some changes to remove the extra `set.seed()` calls. I think this solves the issue you had but I might have not quite understood. I also added tests...