What happened:

In a Python environment with xarray 0.18.0 and python-netcdf4 installed, I called xarray.open_dataset("nonexistent"). (The file "nonexistent" does not exist.) xarray threw a ValueError: cannot guess the engine, try passing one explicitly.

What you expected to happen:

I expected a FileNotFoundError error to be thrown, as in xarray 0.17.0.

Minimal Complete Verifiable Example:

python import xarray as xr xr.open_dataset("nonexistent")

Anything else we need to know?:

This is presumably related to Issue #5295, but is not fixed by PR #5296: ValueError is also thrown with the currently latest commit in master (9165c266).

This change in behaviour produced a hard-to-diagnose bug deep in xcube, where we were catching the FileNotFound exception to deal gracefully with a missing file, but the new ValueError was of course not caught -- and the error message did not make the cause obvious. Catching ValueError is a workaround, but not a great solution since it may also be thrown for files which do exist but don't have a recognizable data format. I suspect that other codebases may be similarly affected.

xarray 0.17.0 was capable of throwing a ValueError for a non-existent file, but only in the (rare?) case that neither netCDF4-python nor scipy was installed.

Environment:

What happened:

In an environment with xarray 0.18.0 and dask 2021.05.0 installed, I saved a dataset using to_zarr, opened it again using open_zarr, and called map_blocks on one of its variables. I got the following traceback:

Traceback (most recent call last): File "/home/pont/./dasktest2.py", line 12, in <module> ds2.myvar.map_blocks(lambda block: block) File "/home/pont/loc/envs/xcube-repos/lib/python3.9/site-packages/xarray/core/dataarray.py", line 3770, in map_blocks return map_blocks(func, self, args, kwargs, template) File "/home/pont/loc/envs/xcube-repos/lib/python3.9/site-packages/xarray/core/parallel.py", line 565, in map_blocks data = dask.array.Array( File "/home/pont/loc/envs/xcube-repos/lib/python3.9/site-packages/dask/array/core.py", line 1159, in __new__ if layer.collection_annotations is None: AttributeError: 'dict' object has no attribute 'collection_annotations'

What you expected to happen:

I expected map_blocks to complete successfully.

Minimal Complete Verifiable Example:

```python import xarray as xr import numpy as np

ds1 = xr.Dataset({ "myvar": (("x"), np.zeros(10)), "x": ("x", np.arange(10)), })
ds1.to_zarr("test.zarr", mode="w") ds2 = xr.open_zarr("test.zarr") ds2.myvar.map_blocks(lambda block: block) ```

Anything else we need to know?:

I wasn't sure whether to report this issue with dask or xcube. With dask 2021.04.1 the example runs without error, and it seems that dask PR 7309 introduced the breaking change. But my understanding of xarray's map_blocks implementation isn't sufficient to figure out where exactly the bug lies.

Environment:

What happened:

I tried to use xarray to append a Dataset to a Zarr containing a |S1 (char string) datatype, and received this error:

ValueError: Invalid dtype for data variable: <xarray.DataArray 'x' ()> array(b'', dtype='|S1') dtype must be a subtype of number, datetime, bool, a fixed sized string, a fixed size unicode string or an object

What you expected to happen:

I expected the Dataset to be appended to the Zarr.

Minimal Complete Verifiable Example:

Note: this is not quite "minimal", since it also performs the append using the Zarr library directly and using a |U1 (Unicode) datatype in order to demonstrate that these variations work.

```python import numpy as np import xarray as xr import zarr

def test_append(data_type, zarr_path): print(f"Creating {data_type} Zarr...") ds = xr.Dataset({"x": np.array("", dtype=data_type)}) ds.to_zarr(zarr_path, mode="w")

print(f"Appending to {data_type} Zarr with Zarr library...")
zarr_to_append = zarr.open(zarr_path, mode="a")
zarr_to_append.x.append(np.array("", dtype=data_type))

print(f"Appending to {data_type} Zarr with xarray...")
ds_to_append = xr.Dataset({"x": np.array("", dtype=data_type)})
ds_to_append.to_zarr(zarr_path, mode="a")

test_append("|U1", "test-u.zarr") test_append("|S1", "test-s.zarr")

```

Anything else we need to know?:

I came across this problem when converting some NetCDFs from this dataset to a Zarr, appending them along the time axis. The latest data format vesion (1.4) includes a dimensionless variable crs with type char, which xarray reads as an |S1, causing the error described above when I attempt to append. Replacing crs with a |U1-typed variable works around the problem, but is undesirable since we need to reproduce the NetCDFs as closely as possible. The example above shows that the Zarr format and library themselves don't seem to have a problem with appending byte string variables.

The obvious fix would be to loosen the type check in xarray.backends.api._validate_datatypes_for_zarr_append:

python if ( not np.issubdtype(var.dtype, np.number) and not np.issubdtype(var.dtype, np.datetime64) and not np.issubdtype(var.dtype, np.bool_) and not coding.strings.is_unicode_dtype(var.dtype) and not coding.strings.is_bytes_dtype(var.dtype) # <- this line added to avoid "Invalid dtype" error and not var.dtype == object ):

This change makes the example above work, but I don't know if it would result in any unintended side-effects.

Environment:

What happened:

I used xarray.open_dataset to open a NetCDF file whose time coordinate had the calendar attribute set to Gregorian. All dates were within the Timestamp-valid range.

The resulting dataset represented the time co-ordinate as a cftime._cftime.DatetimeGregorian.

What you expected to happen:

I expected the dataset to represent the time co-ordinate as a datetime64[ns], as documented here and here.

Minimal Complete Verifiable Example:

```python import xarray as xr import numpy as np import pandas as pd

def print_time_type(dataset): print(dataset.time.dtype, type(dataset.time[0].item()))

da = xr.DataArray( data=[32, 16, 8], dims=["time"], coords=dict( time=pd.date_range("2014-09-06", periods=3), reference_time=pd.Timestamp("2014-09-05"), ), )

Create dataset and confirm type of time

ds1 = xr.Dataset({"myvar": da}) print_time_type(ds1) # prints "datetime64[ns]" <class 'int'>

Manually set time attributes to "Gregorian" rather

than default "proleptic_gregorian".

ds1.time.encoding["calendar"] = "Gregorian" ds1.reference_time.encoding["calendar"] = "Gregorian" ds1.to_netcdf("test-capitalized.nc")

ds2 = xr.open_dataset("test-capitalized.nc") print_time_type(ds2)

prints "object <class 'cftime._cftime.DatetimeGregorian'>"

Workaround: add "Gregorian" to list of standard calendars.

xr.coding.times._STANDARD_CALENDARS.add("Gregorian") ds3 = xr.open_dataset("test-capitalized.nc") print_time_type(ds3) # prints "datetime64[ns]" <class 'int'> ```

Anything else we need to know?:

The documentation for the use_cftime parameter of open_dataset says:

If None (default), attempt to decode times to np.datetime64[ns] objects; if this is not possible, decode times to cftime.datetime objects.

In practice, we are getting some cftime.datetimes even for times which are interpretable and representable as np.datetime64[ns]s. In particular, we have some NetCDF files in which the time variable has a calendar attribute with a value of Gregorian (with a capital ‘G’). CF conventions allow this:

When this standard defines string attributes that may take various prescribed values, the possible values are generally given in lower case. However, applications programs should not be sensitive to case in these attributes.

However, xarray regards Gregorian as a non-standard calendar and falls back to cftime.datetime. If (as in the example) Gregorian is added to xr.coding.times._STANDARD_CALENDARS, the times are read as np.datetime64[ns]s.

Suggested fix: in xarray.coding.times._decode_datetime_with_pandas, change ‘if calendar not in _STANDARD_CALENDARS:’ to ‘if calendar.lower() not in _STANDARD_CALENDARS:’.

Environment: