MultiZarrToZarr: Concatenate along dimensionless variable/coordinate

[mpvginde]

Hi,

I'm trying to combine netCDF-files with 3 dimensions [x, y, leadtime] along a new dimension 'reftime'.
The reftime of each file is included in the netCDF as a dimensionless coordinate as seen here:

Typically I open the multiple files with xarray like this:

def preprocess(x):
    x = x.expand_dims(dim="reftime")
    return x
ds = xr.open_mfdataset(files,preprocess=preprocess)

What would be the best way to use the reftime value of each file as value of the concat_dim?
Do I add the new dimension already when creating single file reference, or do I use some preprocessing or coo_map when using MultiZarrToZar?

Kind regards,
Michiel

[martindurant]

Do I add the new dimension already when creating single file reference, or do I use some preprocessing or coo_map

I would say this is a matter of choice - does the individual dataset gain anything by adding in the extra coordinate information? If no (my suspicion), then using coo_map or a preprocess function in multizarrtozarr is totally fine, probably better the former.

[mpvginde]

Thanks @martindurant, I agree.
Any tips on how you could achieve this?
In other words how can I access the actual value of reftime using the coo_map?
The tutorials show examples for accessing filenames, but it is unclear to me how I can retrieve and use the data of a coordinate of a file as values for the new dimension.

[martindurant]

The mapper variant you want is "data:{var}" (where var is "reftime"), or you can always fall back to the completely general function F(index, fs, var, fn), where you would need to open the zarr dataset in the function and extract the value.

[mpvginde]

Thanks a lot!