You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It would be good to fix this problem eventually, either by finding the bug (potentially in h5py) or by finding a way around it. We could potentially just up cast the array from (N,) to (N,1) save and then squeeze the array after loading. That is a little hacky and I would like to avoid it.
Describe the context
You can save this type of array with the .zspy format so in most cases I would just recommend using that.
The text was updated successfully, but these errors were encountered:
This is the issue tracker for the incompatibility found in #193 with saving 1d dask arrays of objects using the h5py package.
The problem appears to be a mis-match related to the size of the h5py.Dataset and the dask array. This is most likely from some casting done by h5py as it doesn't appear in the zarr library. For example https://github.com/h5py/h5py/blob/4c01efa9714db40ffe27a322c4f1ba4635816e44/h5py/_hl/selections.py#L236-L249 casts to make sure that the Dataset and the array are the same shape.
Describe the functionality you would like to see.
It would be good to fix this problem eventually, either by finding the bug (potentially in h5py) or by finding a way around it. We could potentially just up cast the array from (N,) to (N,1) save and then squeeze the array after loading. That is a little hacky and I would like to avoid it.
Describe the context
You can save this type of array with the .zspy format so in most cases I would just recommend using that.
The text was updated successfully, but these errors were encountered: