Hi all,

I start to play with xarray and zarr and came across something curious : I create a zarr folder and a zarr variable in uint32. When i load this dataset with xarray, it loads in float64. I don't know if it is something expected ?

```python fichier1 = "/tmp/test.zarr"

zh = zarr.open(fichier1, "w")

example = np.zeros(10, dtype=np.uint32) myvar = zh.create_dataset("myvar", shape=example.shape, dtype=example.dtype )

myvar.attrs["_ARRAY_DIMENSIONS"] = ["obs"] # <- without this, the zarr dataset will not be readable by xarray myvar[:] = example

dtype is uint32

zh.myvar.dtype python

dtype('uint32') ```

when reloading with zarr :

```python

dtype is stil uint32

zh = zarr.open(fichier1, 'r') zh.myvar.dtype ```

```python

dtype('uint32') ```

But when loading with xarray :

```python

dtype is float64

ds = xr.open_zarr(fichier1) ds.myvar.dtype ```

```python

dtype('float64') ```

Is it something expected ? Am I missing something ?

link to the notebook created : bad_dtype_zarr_xarray

Environment:

Output of <tt>xr.show_versions()</tt>

INSTALLED VERSIONS ------------------ commit: None python: 3.7.6 (default, Jan 8 2020, 19:59:22) [GCC 7.3.0] python-bits: 64 OS: Linux OS-release: 4.15.0-106-generic machine: x86_64 processor: x86_64 byteorder: little LC_ALL: None LANG: fr_FR.UTF-8 LOCALE: fr_FR.UTF-8 libhdf5: 1.10.4 libnetcdf: 4.6.1 xarray: 0.15.1 pandas: 1.0.3 numpy: 1.18.1 scipy: 1.4.1 netCDF4: 1.4.2 pydap: None h5netcdf: None h5py: 2.10.0 Nio: None zarr: 2.3.2 cftime: 1.1.1.1 nc_time_axis: None PseudoNetCDF: None rasterio: None cfgrib: None iris: None bottleneck: None dask: 2.13.0 distributed: 2.13.0 matplotlib: 3.1.3 cartopy: None seaborn: None numbagg: None setuptools: 46.1.1.post20200323 pip: 20.0.2 conda: None pytest: 5.4.1 IPython: 7.13.0 sphinx: None