Fix doctests

[keewis]

This is an attempt to fix the examples in our docstrings. With this, the output of:

python -m pytest --doctest-modules xarray --ignore xarray/tests

is

remaining warnings and errors

====================================================================================================== test session starts =======================================================================================================
platform linux -- Python 3.8.5, pytest-6.0.1, py-1.9.0, pluggy-0.13.1
rootdir: ..., configfile: setup.cfg
plugins: cov-2.10.0, env-0.6.2, hypothesis-5.23.11
collected 91 items xarray/conventions.py FF [ 2%]
xarray/backends/api.py F [ 3%]
xarray/coding/cftime_offsets.py . [ 4%]
xarray/coding/cftimeindex.py .... [ 8%]
xarray/coding/strings.py F [ 9%]
xarray/core/accessor_dt.py ... [ 13%]
xarray/core/accessor_str.py . [ 14%]
xarray/core/alignment.py .. [ 16%]
xarray/core/combine.py .. [ 18%]
xarray/core/common.py ....F...... [ 30%]
xarray/core/computation.py ..... [ 36%]
xarray/core/dataarray.py ....................... [ 61%]
xarray/core/dataset.py .....F............. [ 82%]
xarray/core/extensions.py . [ 83%]
xarray/core/groupby.py . [ 84%]
xarray/core/indexing.py FF [ 86%]
xarray/core/merge.py . [ 87%]
xarray/core/nputils.py . [ 89%]
xarray/core/options.py . [ 90%]
xarray/core/parallel.py . [ 91%]
xarray/core/rolling.py .. [ 93%]
xarray/core/rolling_exp.py . [ 94%]
xarray/core/utils.py . [ 95%]
xarray/core/variable.py .. [ 97%]
xarray/plot/utils.py .. [100%]
============================================================================================================ FAILURES ============================================================================================================
___________________________________________________________________________________________ [doctest] xarray.conventions.BoolTypeArray ___________________________________________________________________________________________
048 Decode arrays on the fly from integer to boolean datatype
049 050 This is useful for decoding boolean arrays from integer typed netCDF
051 variables.
052 053 >>> x = np.array([1, 0, 1, 1, 0], dtype="i1")
054 055 >>> x.dtype
Expected:
dtype('>i2')
Got:
dtype('int8')
.../xarray/conventions.py:55: DocTestFailure
_______________________________________________________________________________________ [doctest] xarray.conventions.NativeEndiannessArray _______________________________________________________________________________________
021 022 >>> x = np.arange(5, dtype=">i2")
023 024 >>> x.dtype
025 dtype('>i2')
026 027 >>> NativeEndiannessArray(x).dtype
028 dtype('int16')
029 030 >>> NativeEndiannessArray(x)[:].dtype
UNEXPECTED EXCEPTION: TypeError("unexpected key type: <class 'slice'>")
Traceback (most recent call last):
File "~/.conda/envs/xarray/lib/python3.8/doctest.py", line 1336, in __run
exec(compile(example.source, filename, "single",
File "<doctest xarray.conventions.NativeEndiannessArray[3]>", line 1, in <module>
File ".../xarray/conventions.py", line 44, in __getitem__
return np.asarray(self.array[key], dtype=self.dtype)
File ".../xarray/core/indexing.py", line 1277, in __getitem__
array, key = self._indexing_array_and_key(key)
File ".../xarray/core/indexing.py", line 1269, in _indexing_array_and_key
raise TypeError("unexpected key type: {}".format(type(key)))
TypeError: unexpected key type: <class 'slice'>
.../xarray/conventions.py:30: UnexpectedException
__________________________________________________________________________________________ [doctest] xarray.backends.api.save_mfdataset __________________________________________________________________________________________
1182 compute : bool
1183 If true compute immediately, otherwise return a
1184 ``dask.delayed.Delayed`` object that can be computed later.
1185 1186 Examples
1187 --------
1188 1189 Save a dataset into one netCDF per year of data:
1190 1191 >>> years, datasets = zip(*ds.groupby("time.year"))
UNEXPECTED EXCEPTION: NameError("name 'ds' is not defined")
Traceback (most recent call last):
File "~/.conda/envs/xarray/lib/python3.8/doctest.py", line 1336, in __run
exec(compile(example.source, filename, "single",
File "<doctest xarray.backends.api.save_mfdataset[0]>", line 1, in <module>
NameError: name 'ds' is not defined
.../xarray/backends/api.py:1191: UnexpectedException
_______________________________________________________________________________________ [doctest] xarray.coding.strings.StackedBytesArray ________________________________________________________________________________________
199 Wrapper around array-like objects to create a new indexable object where
200 values, when accessed, are automatically stacked along the last dimension.
201 202 >>> StackedBytesArray(np.array(["a", "b", "c"]))[:]
UNEXPECTED EXCEPTION: ValueError("can only use StackedBytesArray if argument has dtype='S1'")
Traceback (most recent call last):
File "~/.conda/envs/xarray/lib/python3.8/doctest.py", line 1336, in __run
exec(compile(example.source, filename, "single",
File "<doctest xarray.coding.strings.StackedBytesArray[0]>", line 1, in <module>
File ".../xarray/coding/strings.py", line 215, in __init__
raise ValueError(
ValueError: can only use StackedBytesArray if argument has dtype='S1'
.../xarray/coding/strings.py:202: UnexpectedException
________________________________________________________________________________________ [doctest] xarray.core.common.DataWithCoords.pipe ________________________________________________________________________________________
578 >>> def adder(data, arg):
579 ... return data + arg
580 ...
581 >>> def div(data, arg):
582 ... return data / arg
583 ...
584 >>> def sub_mult(data, sub_arg, mult_arg):
585 ... return (data * mult_arg) - sub_arg
586 ...
587 >>> x.pipe(adder, 2)
Differences (unified diff with -expected +actual):
@@ -2,6 +2,6 @@
Dimensions: (lat: 2, lon: 2)
Coordinates:
+ * lon (lon) int64 150 160
* lat (lat) int64 10 20
- * lon (lon) int64 150 160
Data variables:
temperature_c (lat, lon) float64 12.98 16.3 14.06 12.9
.../xarray/core/common.py:587: DocTestFailure
_____________________________________________________________________________________ [doctest] xarray.core.dataset.Dataset.filter_by_attrs ______________________________________________________________________________________
5719 ... },
5720 ... coords={
5721 ... "lon": (["x", "y"], lon),
5722 ... "lat": (["x", "y"], lat),
5723 ... "time": pd.date_range("2014-09-06", periods=3),
5724 ... "reference_time": pd.Timestamp("2014-09-05"),
5725 ... },
5726 ... )
5727 >>> # Get variables matching a specific standard_name.
5728 >>> ds.filter_by_attrs(standard_name="convective_precipitation_flux")
Differences (unified diff with -expected +actual):
@@ -2,8 +2,8 @@
Dimensions: (time: 3, x: 2, y: 2)
Coordinates:
+ lon (x, y) float64 -99.83 -99.32 -99.79 -99.23
+ * time (time) datetime64[ns] 2014-09-06 2014-09-07 2014-09-08
+ lat (x, y) float64 42.25 42.21 42.63 42.59
reference_time datetime64[ns] 2014-09-05
- lat (x, y) float64 42.25 42.21 42.63 42.59
- * time (time) datetime64[ns] 2014-09-06 2014-09-07 2014-09-08
- lon (x, y) float64 -99.83 -99.32 -99.79 -99.23
Dimensions without coordinates: x, y
Data variables:
.../xarray/core/dataset.py:5728: DocTestFailure
____________________________________________________________________________________ [doctest] xarray.core.indexing._decompose_outer_indexer _____________________________________________________________________________________
971 -----
972 This function is used to realize the vectorized indexing for the backend
973 arrays that only support basic or outer indexing.
974 975 As an example, let us consider to index a few elements from a backend array
976 with a orthogonal indexer ([0, 3, 1], [2, 3, 2]).
977 Even if the backend array only supports basic indexing, it is more
978 efficient to load a subslice of the array than loading the entire array,
979 980 >>> backend_indexer = BasicIndexer(slice(0, 3), slice(2, 3))
UNEXPECTED EXCEPTION: TypeError('__init__() takes 2 positional arguments but 3 were given')
Traceback (most recent call last):
File "~/.conda/envs/xarray/lib/python3.8/doctest.py", line 1336, in __run
exec(compile(example.source, filename, "single",
File "<doctest xarray.core.indexing._decompose_outer_indexer[0]>", line 1, in <module>
TypeError: __init__() takes 2 positional arguments but 3 were given
.../xarray/core/indexing.py:980: UnexpectedException
__________________________________________________________________________________ [doctest] xarray.core.indexing._decompose_vectorized_indexer __________________________________________________________________________________
893 -----
894 This function is used to realize the vectorized indexing for the backend
895 arrays that only support basic or outer indexing.
896 897 As an example, let us consider to index a few elements from a backend array
898 with a vectorized indexer ([0, 3, 1], [2, 3, 2]).
899 Even if the backend array only supports outer indexing, it is more
900 efficient to load a subslice of the array than loading the entire array,
901 902 >>> backend_indexer = OuterIndexer([0, 1, 3], [2, 3])
UNEXPECTED EXCEPTION: TypeError('__init__() takes 2 positional arguments but 3 were given')
Traceback (most recent call last):
File "~/.conda/envs/xarray/lib/python3.8/doctest.py", line 1336, in __run
exec(compile(example.source, filename, "single",
File "<doctest xarray.core.indexing._decompose_vectorized_indexer[0]>", line 1, in <module>
TypeError: __init__() takes 2 positional arguments but 3 were given
.../xarray/core/indexing.py:902: UnexpectedException
======================================================================================================== warnings summary ========================================================================================================
xarray/core/common.py::xarray.core.common.DataWithCoords.resample
xarray/core/common.py::xarray.core.common.DataWithCoords.resample
xarray/core/common.py::xarray.core.common.DataWithCoords.resample
.../xarray/core/common.py:1230: FutureWarning: 'base' in .resample() and in Grouper() is deprecated.
The new arguments that you should use are 'offset' or 'origin'.
>>> df.resample(freq="3s", base=2)
becomes:
>>> df.resample(freq="3s", offset="2s")
grouper = pd.Grouper(
xarray/core/dataarray.py::xarray.core.dataarray.DataArray.argmax
xarray/core/dataarray.py::xarray.core.dataarray.DataArray.idxmax
.../xarray/core/dataarray.py:4090: DeprecationWarning: Behaviour of argmin/argmax with neither dim nor axis argument will change to return a dict of indices of each dimension. To get a single, flat index, please use np.argmin(da.data) or np.argmax(da.data) instead of da.argmin() or da.argmax().
result = self.variable.argmax(dim, axis, keep_attrs, skipna)
xarray/core/dataarray.py::xarray.core.dataarray.DataArray.argmin
xarray/core/dataarray.py::xarray.core.dataarray.DataArray.idxmin
.../xarray/core/dataarray.py:3987: DeprecationWarning: Behaviour of argmin/argmax with neither dim nor axis argument will change to return a dict of indices of each dimension. To get a single, flat index, please use np.argmin(da.data) or np.argmax(da.data) instead of da.argmin() or da.argmax().
result = self.variable.argmin(dim, axis, keep_attrs, skipna)
xarray/core/dataarray.py::xarray.core.dataarray.DataArray.roll
.../xarray/core/dataarray.py:2974: FutureWarning: roll_coords will be set to False in the future. Explicitly set roll_coords to silence warning.
ds = self._to_temp_dataset().roll(
xarray/core/dataset.py::xarray.core.dataset.Dataset.roll
<doctest xarray.core.dataset.Dataset.roll[1]>:1: FutureWarning: roll_coords will be set to False in the future. Explicitly set roll_coords to silence warning.
-- Docs: https://docs.pytest.org/en/stable/warnings.html
==================================================================================================== short test summary info =====================================================================================================
FAILED xarray/conventions.py::xarray.conventions.BoolTypeArray
FAILED xarray/conventions.py::xarray.conventions.NativeEndiannessArray
FAILED xarray/backends/api.py::xarray.backends.api.save_mfdataset
FAILED xarray/coding/strings.py::xarray.coding.strings.StackedBytesArray
FAILED xarray/core/common.py::xarray.core.common.DataWithCoords.pipe
FAILED xarray/core/dataset.py::xarray.core.dataset.Dataset.filter_by_attrs
FAILED xarray/core/indexing.py::xarray.core.indexing._decompose_outer_indexer
FAILED xarray/core/indexing.py::xarray.core.indexing._decompose_vectorized_indexer
============================================================================================ 8 failed, 83 passed, 9 warnings in 3.73s ============================================================================================

where DataWithCoords.pipe and Dataset.filter_by_attrs fail mostly because the order of the coords is not deterministic. Should we sort before formatting the coords?

Does anyone know what ds in save_mfdataset should look like? Also, how do we fix the construction of OuterIndexer and VectorizedIndexer?

• ClosesShould we run tests on docstrings? #3837
• Tests added
• Passes isort . && black . && mypy . && flake8
• User visible changes (including notable bug fixes) are documented in whats-new.rst

Edit: if we merge #4409 before this, we can make sure all doctests pass.

[max-sixty]

This is awesome, thanks a lot @keewis !

Is there a tool to do this or you had to copy and paste all the results?

[keewis]

Is there a tool to do this or you had to copy and paste all the results?

that was a entirely manual process. However, it should be possible to write a script using doctest.DoctestParser, doctest.DoctestFinder and doctest.DoctestRunner (and I would be surprised if there's nothing like that already).

[max-sixty]

that was a entirely manual process. However, it should be possible to write a script using doctest.DoctestParser, doctest.DoctestFinder and doctest.DoctestRunner (and I would be surprised if there's nothing like that already).

Well, thanks for doing all that!

On the tool, I agree that would be great. Here's something I did in a day that aimed to do it for any value rather than doctests. Would be easier and less complicated for doctests: https://github.com/max-sixty/pytest-accept

[dcherian]

shall we merge and deal with the effects of #4409 later?

[keewis]

I tried to fix the remaining doctests that were unrelated to #4409, so those need a review first (especially the StackedBytesArray and indexing doctests).

[keewis]

wow. The RTD check fails due to:

<frozen importlib._bootstrap>:219: RuntimeWarning: numpy.ufunc size changed, may indicate binary incompatibility. Expected 192 from C header, got 216 from PyObject

[max-sixty]

Looks great! I raised an eyebrow about the os.remove but it's limited to a finite set of paths.

Re the warning, we could add an ignore in the doctest prelude?

[keewis]

I raised an eyebrow about the os.remove but it's limited to a finite set of paths.

yeah, well, I didn't want to leave files lying around. Not sure if there's a better way to do that. Maybe if there's a way to have pytest create a temporary directory we could use that to construct the paths?

we could add an ignore in the doctest prelude

I doubt that would fix it (maybe I'm missing something), these are triggered by ipython directives (combining.rst, groupby.rst). I'll try rerunning the RTD check.

[keewis]

the main build is also failing so I guess this is unrelated. There is a numpy release PR in the feedstock, maybe that will fix it?

[keewis]

@max-sixty, I removed the call to os.remove, instead every doctest gets its own temporary directory and by changing to that directory first, any file will be written into it. This seems a bit wasteful because not every doctest needs a temporary directory (actually, the vast majority doesn't), but I couldn't find an option to enable that only for certain tests. The doctests still complete within a few seconds, though.

[max-sixty]

Perfect, nice idea

[max-sixty]

Great! To the extent the RTD build failure is not from this branch, shall we merge this now?

[max-sixty]

Great! To the extent the RTD build failure is not from this branch, shall we merge this now?

(i.e. feel free to merge!)

[keewis]

the failing hypothesis test looks unrelated, so I'm merging.

Edit: it seems RTD was fixed

[max-sixty]

Thanks a lot @keewis ! This is a big step forward!

[keewis]

should we add a CI that checks that we don't regress?

[dcherian]

Sounds good to me. Thanks for working on this.

[keewis]

okay, then let's wait for #4409 which should fix the remaining three doctest failures.

[max-sixty]

For sure re CI!