Easy way to set the compression level for all dataarrays in a datatree?

[jwbrooks0]

Question Is there an easy/convenient way to set the compression level for all DataArrays in a datatree?
I've been using the solution here when saving xarray.Datasets, but this solution doesn't seem to map conveniently to datatree. Maybe I'm missing something?

Background I often deal with multiple datasets, each on the order of 10 GBs, and this fills up harddrives fast. Typically when I write xarray.Datasets to file, I like to use gzip compression with level around 5, and this tends to reduce the file size by around 50%.

Thanks!

Edit: Or is recommended practice to update the compression info for each DataArray? (The second answer to the above link)

[jhamman]

Or is recommended practice to update the compression info for each DataArray? (The second answer to the above link)

I'll just say that this is the approach I've been taking.

However, I do think there is some room for improvement to the current api. The challenge of course is that the per-variable or per-dataset encoding dictionary needs to map to tree structure. In practice these nested-dict data structures are unruly to work with and I've found myself finding plenty of functionality in the DataArray.encoding approach.

@jwbrooks0 - I'm curious if you have thoughts on a possible api that meets your use case? Do you agree that a nested-dict of encoding parameters is less than ideal from an end-user perspective?

[jwbrooks0]

Particularly with Datasets, I definitely would prefer to just provide a single command/attribute to set the entire thing. I haven't thought through datatrees very carefully yet, but I think the same thing applies. My main goal of compression is to save space on my harddrive and having a single setting seems easier to me.

I also don't really understand why I would ever want to have different compression levels for individual DataArrays in a Dataset or datatree.

For context, I mostly save data as float32/64 and occasionally int64/32.