Jan Vaillant

> I'm not very comfortable with adding another way to allow this. D'accord. Fair enough. > Ah I see, you're right. Probably it's not clear enough that the value of...

> Please continue to give helpful feedback, it's very much appreciated! :) All right. Though no guarantees on the "helpfulness". Why not think about the DAG this way: ``` t1...

Great to see the discussion ongoing here. Just some observations while we are porting our model to simlab: - it would indeed be helpful if we could add e.g. a...

> Sorry for the late reply @jvail ! No need to be sorry! Alright - Thank you. I'll try to read the zarr related stuff and try a dev install...

Hi @benbovy , here is a test for some minimal sparse support: https://github.com/jvail/xarray-simlab/commit/107d2e0a2bebff4706043d9708f4fba3e6f2b53d#diff-932f643c875a75e89dbf205d2764a1ac25b260f5c12503543242f833e8359a62 It turns out it is impossible to use sparse with intent='inout' because zarr wont accept it and...

> In the meantime, maybe we could implement some ad-hoc conversion in xarray-simlab, i.e., for sparse arrays having `coords` and `data` saved as two zarr arrays, and then reconstruct the...

Salut @benbovy, if you find time please take a look at this commit/branch: https://github.com/jvail/xarray-simlab/commit/cfa37a69796985beca19ffadcef70d8005ccb1d7 Instead of adding two zarr items - as you suggested - I tried to use the...

> > Unfortunately I could not get it working with the "infamous" mask_and_scale: True setting. Maybe you have an idea? > > Without having looked at your branch yet, my...

> A comment about your implementation: the approach that you use won't scale well if later we want to support other special cases. The [VariableCoder](https://github.com/pydata/xarray/blob/7905c514a12fcbcaaeb634cab94733c7cbdd6ff2/xarray/coding/variables.py#L18) approach used in Xarray would...

> I'm afraid there's currently no workaround other than `mask_and_scale: False`. Apparently setting the zarr fill_value to `None` solves it.