Isaac Virshup

Sure. Overall I am much more optimistic about this solution. This is because: * There is buy in from developers of the relevant array libraries * It's less complicated a...

I would be interested in hearing @rgommers take on this, as a maintainer and someone who has been deeply involved in the array-api. ---------- > So if I pull off...

@asmeurer > This is something to bring up on the [array-api repo](https://github.com/data-apis/array-api/issues). :+1:, I will take a little time to formulate this. I think there will also be "things that...

Thanks for the input! I broadly agree with your take @rgommers, with a few comments. I'm going to focus my response around whether scipy.sparse should support the array-api and what...

I've opened https://github.com/scipy/scipy/issues/18915 to discuss updating the sparse array class to adopt the array api as a provider. I am curious what functionality in `sparse` and its submodules could be...

@dorzhey can you reproduce this with the latest scanpy release? I believe it should be fixed there.

In scanpy 1.10, the line referenced is: * https://github.com/scverse/scanpy/blob/214e05bdc54df61c520dc563ab39b7780e6d3358/scanpy/preprocessing/_highly_variable_genes.py#L226 But in 1.9.8 the line is: * https://github.com/scverse/scanpy/blob/6da39f128ecf78cf572f453ee2865d1b901715f3/scanpy/preprocessing/_highly_variable_genes.py#L226 So, to me it seems like the warning would require running an older...

> > should an additional restriction on concatenate_arrays be: (shape % chunks) == 0 for dimensions being concatenated along? > Yes, except the very last one, I suppose. This case...

I've been thinking of something similar. I'm not sure we can get all the flexibility we want out of just two functions, but think a couple classes may do the...

> We can implement these features via a specification, with an aim for multi-language support We definitely want to see multi language support (cc: @manzt) Do you have thoughts on...