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
If we move from_netcdf and to_netcdf into InferenceObjects with a fallback that instructs the user to load NCDatasets to use these functions, then we can place the implementations in an extension. There seem to be some trade-offs to this approach, and it's not clear either way whether it should be preferred to our current subpackage approach.
The text was updated successfully, but these errors were encountered:
An even better approach would be to just overload functions in the two files. e.g. for loading an InferenceData from a NetCDF file, we could just define the overload convert_to_inference_data(::NCDataset). The problem is writing to a NetCDF file. The NCDatasets API only provides functionality for iteratively writing data to a NCDataset, not a way to write an object to one.
If #16 is implemented, then a better solution would be to support an NCDataset being a store for an InferenceData.
By comparison, NetCDF.jl has ncwrite and ncread that we could perhaps overload.
If we move
from_netcdf
andto_netcdf
into InferenceObjects with a fallback that instructs the user to load NCDatasets to use these functions, then we can place the implementations in an extension. There seem to be some trade-offs to this approach, and it's not clear either way whether it should be preferred to our current subpackage approach.The text was updated successfully, but these errors were encountered: