issues: 1075765204

This data as json

id node_id number title user state locked assignee milestone comments created_at updated_at closed_at author_association active_lock_reason draft pull_request body reactions performed_via_github_app state_reason repo type
1075765204 I_kwDOAMm_X85AHt_U 6055 Unexpected type conversion in variables with _FillValue 24235303 closed 0     4 2021-12-09T16:26:54Z 2023-09-13T12:40:14Z 2023-09-13T12:40:13Z CONTRIBUTOR      

What happened: When opening a dataset with an int16 variable with the _FillValue attribute, the variable is converted from type int16 to float32. This was originally reported to the TileDB-CF-Py Git repo that contains a TileDB backend for xarray. See TileDB-CF-Py issue #117.

What you expected to happen: I would expect the type to remain the same when applying the _FillValue.

Minimal Complete Verifiable Example:

Original example from TileDB-CF-Py issue #117 using the TileDB backend. ```python import tiledb import xarray as xr import numpy as np

index = tiledb.Dim(name='index', domain=(0, 3)) domain = tiledb.Domain(index) var = tiledb.Attr(name='var', dtype=np.int16) schema = tiledb.ArraySchema(domain=domain, attrs=[var], sparse=False) tiledb.Array.create('dense_array0', schema)

with tiledb.open('dense_array0', 'w') as A: A[:] = np.array([5, 6, 7, 8], dtype=np.int16)

ds = xr.open_dataset('dense_array0', engine='tiledb') ds['var'].dtype ```

NetCDF example with the same behavior: ```python import netCDF4 import xarray as xr import numpy as np

filename = 'temp_file.nc' with netCDF4.Dataset(filename, mode="w") as group: group.createDimension("index", 4) var = group.createVariable("var", np.int16, ("index",), fill_value=-1) var[:] = np.array([5, 6, 7, 8], dtype=np.int16) dataset = xr.open_dataset(filename) dataset["var"].dtype ```

Anything else we need to know?: * I was able to verify the type conversion from int16 to float32 occurs in the conventions.decode_cf_variables call in the open_dataset method of StoreBackendEntrypoint. * I was able to verify the conversion does not happen if mask_and_scale=False. * Note that TileDB is automatically setting a fill value for all dense numerical arrays, and so we are always setting the _FillValue attribute for variables from the TileDB backend.

Environment: I was able to reproduce this with both xarray 0.19.0 and 0.20.1

 
{
    "url": "https://api.github.com/repos/pydata/xarray/issues/6055/reactions",
    "total_count": 1,
    "+1": 0,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 1
}
   completed 13221727 issue

Links from other tables