How to change the meta dict keys in the result file?

[Yifang-Yuan]

Sorry, it might be a stupid question cz I'm really not familiar with the data structure.
If I want to change the session name or animal name after processing the data, how can I change and save it with the original format?

I can assign new session name by minian_dt["session"]="foo", but when I saves it by minian_dt.to_zarr, the folder structure seems different from what I get from save_minian in the pipeline, and I cannot open it by open_minian.

I also tried save_minian(minian_dt, dpath), but not working, maybe I was doing it not properly, like this, I got the following error:

minian_ds=save_minian(minian_ds.rename("ds"), dpath, meta_dict=dict(session="s9170_210612_A", animal="Data"), overwrite=True)

ValueError: the first argument to .rename must be a dictionary

[phildong]

Hey sorry for the belated reply!

1. there's no built-in function to change the metadata of existing data on disk, since it's a bit to tricky to generalize all the re-writing cases and we'd like to avoid any data loss. The built-in way to correct for this is actually to correct it on-the-fly everytime during load time -- through the post_process argument to open_minian. So you would basically do something like this:

def correct_meta(ds, dpath):
    ds['session'] = 'foo'
    return ds

open_minian(..., post_process=correct_meta)

2. if you do want to correct the data on-disk so you don't have to worry about fixing it everytime, you can easily do so with to_zarr as you noted. Minian dataset is nothing else than a folder containing a bunch of zarr array, and once you put your zarr arrays in a folder you should be able to use open_minian on that folder. btw, you are also not limited to zarr storage backend once you're done with minian, check this out.

3. I'm not sure exactly what happened for your save_minian call, but usually that error happens when the argument (minian_ds in your case) is a xr.Dataset, however save_minian expect a xr.DataArray