Append along an unlimited dimension to an existing netCDF file

[shoyer]

This would be a nice feature to have for some use cases, e.g., for writing simulation time-steps: https://stackoverflow.com/questions/46951981/create-and-write-xarray-dataarray-to-netcdf-in-chunks

It should be relatively straightforward to add, too, building on support for writing files with unlimited dimensions. User facing API would probably be a new keyword argument to to_netcdf(), e.g., extend='time' to indicate the extended dimension.

[Hoeze]

Any updates on this?

[jhamman]

None that I'm aware of. I think this issue is still in the "help wanted" stage.

[mullenkamp]

I would love to have this capability. As @shoyer mentioned, for adding time steps of any sort to existing netcdf files would be really beneficial. The only real alternative is to save a netcdf file for each additional time step...even if there are tons of time steps and each file is a couple hundred KBs (which is my situation with NASA data).

I'll look into it if I get some time...

[thomas-fred]

This would be extremely helpful for our modelling of time varying renewable energy.

[hmaarrfk]

I think I got a basic prototype working.

That said, I think a real challenge lies in supporting the numerous backends and lazy arrays.

For example, I was only able to add data in peculiar fashions using the netcdf4 library which may trigger complex computations many times.

Is this a use case that we must optimize for now?

[hmaarrfk]

Small prototype, but maybe it can help boost the development.

```python import netCDF4 def _expand_variable(nc_variable, data, expanding_dim, nc_shape, added_size): # For time deltas, we must ensure that we use the same encoding as # what was previously stored. # We likely need to do this as well for variables that had custom # econdings too if hasattr(nc_variable, 'calendar'): data.encoding = { 'units': nc_variable.units, 'calendar': nc_variable.calendar, } data_encoded = xr.conventions.encode_cf_variable(data) # , name=name) left_slices = data.dims.index(expanding_dim) right_slices = data.ndim - left_slices - 1 nc_slice = (slice(None),) * left_slices + (slice(nc_shape, nc_shape + added_size),) + (slice(None),) * (right_slices) nc_variable[nc_slice] = data_encoded.data def append_to_netcdf(filename, ds_to_append, unlimited_dims): if isinstance(unlimited_dims, str): unlimited_dims = [unlimited_dims] if len(unlimited_dims) != 1: # TODO: change this so it can support multiple expanding dims raise ValueError( "We only support one unlimited dim for now, " f"got {len(unlimited_dims)}.") unlimited_dims = list(set(unlimited_dims)) expanding_dim = unlimited_dims[0] with netCDF4.Dataset(filename, mode='a') as nc: nc_dims = set(nc.dimensions.keys()) nc_coord = nc[expanding_dim] nc_shape = len(nc_coord) added_size = len(ds_to_append[expanding_dim]) variables, attrs = xr.conventions.encode_dataset_coordinates(ds_to_append) for name, data in variables.items(): if expanding_dim not in data.dims: # Nothing to do, data assumed to the identical continue nc_variable = nc[name] _expand_variable(nc_variable, data, expanding_dim, nc_shape, added_size) from xarray.tests.test_dataset import create_append_test_data from xarray.testing import assert_equal ds, ds_to_append, ds_with_new_var = create_append_test_data() filename = 'test_dataset.nc' ds.to_netcdf(filename, mode='w', unlimited_dims=['time']) append_to_netcdf('test_dataset.nc', ds_to_append, unlimited_dims='time') loaded = xr.load_dataset('test_dataset.nc') assert_equal(xr.concat([ds, ds_to_append], dim="time"), loaded) ```

[espiritocz]

hi - i consider this extremely useful!!!

is your prototype already part of some library (or should we expect it in xr?)

many thanks for the code

[hmaarrfk]

It isn't really part of any library. I don't really have plans of making it into a public library. I think the discussion is really around the xarray API, and what functions to implement at first.

Then somebody can take the code and integrate it into the decided upon API.

[ChrisBarker-NOAA]

Any movement on this? I'd love to have this -- kinda critical for some of my work.

@hmaarrfk seems to have made a start, and it doesn't look too hairy :-)