As part of my work I needed an implementation of extended EOF analysis and decided to update the PR from arulalant. I implemented the suggestions you made there, namely moving the ExtendedEof classes into their own namespace and removing the inheritance from the Eof class. The only implemented methods are those that I needed myself; I am pretty sure some that would make sense with extended EOF are missing, but those could also be added later.
There's only an interface class for xarray as that is what I am working with.
Still open is a good way to deal with dask arrays, as dask's svd algorithm expects one chunk in the spatial dimensions and rechunking is rather expensive. Maybe svd_compressed is an option here, but I haven't looked into it yet.
Unfortunately I don't have a reference dataset for testing purposes, atleast none that isn't created with this code.
As part of my work I needed an implementation of extended EOF analysis and decided to update the PR from arulalant. I implemented the suggestions you made there, namely moving the ExtendedEof classes into their own namespace and removing the inheritance from the Eof class. The only implemented methods are those that I needed myself; I am pretty sure some that would make sense with extended EOF are missing, but those could also be added later.
There's only an interface class for xarray as that is what I am working with.
Still open is a good way to deal with dask arrays, as dask's svd algorithm expects one chunk in the spatial dimensions and rechunking is rather expensive. Maybe svd_compressed is an option here, but I haven't looked into it yet.
Unfortunately I don't have a reference dataset for testing purposes, atleast none that isn't created with this code.