The xarray global option keep_attrs (introduced in #2482 ) is not respected for unary operators.

MCVE Code Sample

python import xarray as xr x = xr.DataArray([1, 2, 3], attrs={"A": "B"}) with xr.set_options(keep_attrs=True): y = ~x print(x.attrs, y.attrs)

Expected Output

I expect

{'A': 'B'} {'A': 'B'}

Problem Description

I get:

{'A': 'B'} {}

I get the same for the other unary operators +x, -x, and abs(x).

Versions

Tested with latest xarray master (see below for details).

Opening this NetCDF file works fine with the default engine, but fails with AttributeError with the h5netcdf engine:

MCVE Code Sample

Data available from EUMETSAT: https://www.eumetsat.int/website/home/Satellites/FutureSatellites/MeteosatThirdGeneration/MTGData/MTGUserTestData/index.html --> ftp://ftp.eumetsat.int/pub/OPS/out/test-data/Test-data-for-External-Users/MTG_FCI_Test-Data/ --> uncompressed

```python

import xarray f = "/path/to/.../W_XX-EUMETSAT-Darmstadt,IMG+SAT,MTI1+FCI-1C-RRAD-FDHSI-FD--CHK-BODY--L2P-NC4E_C_EUMT_20170410114434_GTT_DEV_20170410113925_20170410113934_N__C_0070_0067.nc" ds = xarray.open_dataset(f, engine="h5netcdf") ```

Expected Output

No output at all.

Problem Description

Results in AttributeError:

Traceback (most recent call last): File "mwe4.py", line 3, in <module> with xarray.open_dataset(f, engine="h5netcdf") as ds: File "/media/nas/x21324/miniconda3/envs/py37e/lib/python3.7/site-packages/xarray/backends/api.py", line 535, in open_dataset ds = maybe_decode_store(store) File "/media/nas/x21324/miniconda3/envs/py37e/lib/python3.7/site-packages/xarray/backends/api.py", line 450, in maybe_decode_store use_cftime=use_cftime, File "/media/nas/x21324/miniconda3/envs/py37e/lib/python3.7/site-packages/xarray/conventions.py", line 570, in decode_cf vars, attrs = obj.load() File "/media/nas/x21324/miniconda3/envs/py37e/lib/python3.7/site-packages/xarray/backends/common.py", line 123, in load (_decode_variable_name(k), v) for k, v in self.get_variables().items() File "/media/nas/x21324/miniconda3/envs/py37e/lib/python3.7/site-packages/xarray/backends/h5netcdf_.py", line 156, in get_variables (k, self.open_store_variable(k, v)) for k, v in self.ds.variables.items() File "/media/nas/x21324/miniconda3/envs/py37e/lib/python3.7/site-packages/xarray/core/utils.py", line 402, in FrozenDict return Frozen(dict(*args, **kwargs)) File "/media/nas/x21324/miniconda3/envs/py37e/lib/python3.7/site-packages/xarray/backends/h5netcdf_.py", line 156, in <genexpr> (k, self.open_store_variable(k, v)) for k, v in self.ds.variables.items() File "/media/nas/x21324/miniconda3/envs/py37e/lib/python3.7/site-packages/xarray/backends/h5netcdf_.py", line 120, in open_store_variable dimensions = var.dimensions File "/media/nas/x21324/miniconda3/envs/py37e/lib/python3.7/site-packages/h5netcdf/core.py", line 114, in dimensions self._dimensions = self._lookup_dimensions() File "/media/nas/x21324/miniconda3/envs/py37e/lib/python3.7/site-packages/h5netcdf/core.py", line 98, in _lookup_dimensions for axis, dim in enumerate(self._h5ds.dims): AttributeError: 'Datatype' object has no attribute 'dims'

Output of xr.show_versions()

The following code should probably fail somewhere else than with an AssertionError triggered by to_netcdf:

``` $ cat mwe.py

!/usr/bin/env python3.6

import numpy import xarray

x = xarray.DataArray( [1, 2, 3], dims=["X"], coords={"X": numpy.zeros(shape=3, dtype="M8[ms]")})

x.to_netcdf("/tmp/test.nc") $ python3.6 mwe.py Traceback (most recent call last): File "mwe.py", line 11, in <module> x.to_netcdf("/tmp/test.nc") File "/dev/shm/gerrit/venv/stable-3.6/lib/python3.6/site-packages/xarray/core/dataarray.py", line 1351, in to_netcdf dataset.to_netcdf(args, *kwargs) File "/dev/shm/gerrit/venv/stable-3.6/lib/python3.6/site-packages/xarray/core/dataset.py", line 977, in to_netcdf unlimited_dims=unlimited_dims) File "/dev/shm/gerrit/venv/stable-3.6/lib/python3.6/site-packages/xarray/backends/api.py", line 573, in to_netcdf unlimited_dims=unlimited_dims) File "/dev/shm/gerrit/venv/stable-3.6/lib/python3.6/site-packages/xarray/core/dataset.py", line 916, in dump_to_store unlimited_dims=unlimited_dims) File "/dev/shm/gerrit/venv/stable-3.6/lib/python3.6/site-packages/xarray/backends/common.py", line 244, in store cf_variables, cf_attrs = cf_encoder(variables, attributes) File "/dev/shm/gerrit/venv/stable-3.6/lib/python3.6/site-packages/xarray/conventions.py", line 1089, in cf_encoder for k, v in iteritems(variables)) File "/dev/shm/gerrit/venv/stable-3.6/lib/python3.6/site-packages/xarray/conventions.py", line 1089, in <genexpr> for k, v in iteritems(variables)) File "/dev/shm/gerrit/venv/stable-3.6/lib/python3.6/site-packages/xarray/conventions.py", line 734, in encode_cf_variable var = maybe_encode_datetime(var) File "/dev/shm/gerrit/venv/stable-3.6/lib/python3.6/site-packages/xarray/conventions.py", line 585, in maybe_encode_datetime data, encoding.pop('units', None), encoding.pop('calendar', None)) File "/dev/shm/gerrit/venv/stable-3.6/lib/python3.6/site-packages/xarray/conventions.py", line 293, in encode_cf_datetime assert dates.dtype == 'datetime64[ns]' AssertionError ```

Add missing comma and article in error message when attribute values have the wrong type.

I tihnk this change is sufficiently minor that no documentation or whatsnew changes should be necessary.

Code Sample, a copy-pastable example if possible

```python

Your code here

da = xarray.DataArray(arange(5, dtype="i2")) print(da.sum(dtype="i4").dtype) ```

Problem description

The result is int64. This is a problem because I asked for int32.

Expected Output

Expected output int32.

Output of xr.show_versions()

When an xarray installation does not match a released version ,it has a version number like 0.9.1-28-g769f120.

This violates PEP 440, which leads to multiple problems:

• pip install --update will revert xarray back to 0.9.1, because it does not recognise that 0.9.1-28-g769f120 > 0.9.1
• packages with an xarray dependencies will be considered not satisfied. Running a script through setuptools load_entry_point fails with pkg_resources.ContextualVersionConflict: (xarray 0.9.1-28-g769f120 (/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages), Requirement.parse('xarray>=0.8')

Instead, the version number above should be written as 0.9.1+r10345 or so, which would satisfy PEP 440 and not cause problems with pip or setuptools.

Trying to index a Dataset using an index array of dtype uint64 yields a ValueError. int64 works fine. See below:

``` In [13]: import xarray

In [14]: ds = xarray.Dataset({"A": (("x", "y"), arange(5*6).reshape(5,6))})

In [15]: ds[{"x": numpy.array([0], dtype="int64")}] Out[15]: <xarray.Dataset> Dimensions: (x: 1, y: 6) Dimensions without coordinates: x, y Data variables: A (x, y) int64 0 1 2 3 4 5

In [16]: ds[{"x": numpy.array([0], dtype="uint64")}]

ValueError Traceback (most recent call last) <ipython-input-16-4cf23af0967e> in <module>() ----> 1 ds[{"x": numpy.array([0], dtype="uint64")}]

/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages/xarray/core/dataset.py in getitem(self, key) 722 """ 723 if utils.is_dict_like(key): --> 724 return self.isel(**key) 725 726 if hashable(key):

/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages/xarray/core/dataset.py in isel(self, drop, indexers) 1147 for name, var in iteritems(self._variables): 1148 var_indexers = dict((k, v) for k, v in indexers if k in var.dims) -> 1149 new_var = var.isel(var_indexers) 1150 if not (drop and name in var_indexers): 1151 variables[name] = new_var

/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages/xarray/core/variable.py in isel(self, **indexers) 547 if dim in indexers: 548 key[i] = indexers[dim] --> 549 return self[tuple(key)] 550 551 def squeeze(self, dim=None):

/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages/xarray/core/variable.py in getitem(self, key) 377 dims = tuple(dim for k, dim in zip(key, self.dims) 378 if not isinstance(k, integer_types)) --> 379 values = self._indexable_data[key] 380 # orthogonal indexing should ensure the dimensionality is consistent 381 if hasattr(values, 'ndim'):

/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages/xarray/core/indexing.py in getitem(self, key) 467 468 def getitem(self, key): --> 469 key = self._convert_key(key) 470 return self._ensure_ndarray(self.array[key]) 471

/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages/xarray/core/indexing.py in _convert_key(self, key) 454 if any(not isinstance(k, integer_types + (slice,)) for k in key): 455 # key would trigger fancy indexing --> 456 key = orthogonal_indexer(key, self.shape) 457 return key 458

/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages/xarray/core/indexing.py in orthogonal_indexer(key, shape) 78 """ 79 # replace Ellipsis objects with slices ---> 80 key = list(canonicalize_indexer(key, len(shape))) 81 # replace 1d arrays and slices with broadcast compatible arrays 82 # note: we treat integers separately (instead of turning them into 1d

/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages/xarray/core/indexing.py in canonicalize_indexer(key, ndim) 66 return indexer 67 ---> 68 return tuple(canonicalize(k) for k in expanded_indexer(key, ndim)) 69 70

/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages/xarray/core/indexing.py in <genexpr>(.0) 66 return indexer 67 ---> 68 return tuple(canonicalize(k) for k in expanded_indexer(key, ndim)) 69 70

/dev/shm/gerrit/venv/stable-3.5/lib/python3.5/site-packages/xarray/core/indexing.py in canonicalize(indexer) 63 'array indexing; all subkeys must be ' 64 'slices, integers or sequences of ' ---> 65 'integers or Booleans' % indexer) 66 return indexer 67

ValueError: invalid subkey array([0], dtype=uint64) for integer based array indexing; all subkeys must be slices, integers or sequences of integers or Booleans ```