issues

Hi there, we just wanted to upgrade xarray to version 0.21.0 and our ASV workflow on the CI started failing.

python File "/home/runner/work/pyglotaran/pyglotaran/benchmark/.asv/env/76391772e92136ec87b9940d70226329/lib/python3.8/site-packages/xarray/core/dask_array_compat.py", line 4, in <module> from packaging.version import Version ModuleNotFoundError: No module named 'packaging' The reason is that packaging isn't a direct dependency of xarray.

```console $ johnnydep xarray 2022-01-29 08:28:38 [info ] init johnnydist [johnnydep.lib] dist=xarray parent=None 2022-01-29 08:28:40 [info ] init johnnydist [johnnydep.lib] dist=numpy>=1.18 parent=xarray 2022-01-29 08:28:44 [info ] init johnnydist [johnnydep.lib] dist=pandas>=1.1 parent=xarray 2022-01-29 08:28:47 [info ] init johnnydist [johnnydep.lib] dist=numpy>=1.18.5 parent=pandas>=1.1 2022-01-29 08:28:51 [info ] init johnnydist [johnnydep.lib] dist=python-dateutil>=2.8.1 parent=pandas>=1.1 2022-01-29 08:28:53 [info ] init johnnydist [johnnydep.lib] dist=pytz>=2020.1 parent=pandas>=1.1 2022-01-29 08:28:56 [info ] init johnnydist [johnnydep.lib] dist=six>=1.5 parent=python-dateutil>=2.8.1 name summary

xarray N-D labeled arrays and datasets in Python ├── numpy>=1.18 NumPy is the fundamental package for array computing with Python. └── pandas>=1.1 Powerful data structures for data analysis, time series, and statistics ├── numpy>=1.18.5 NumPy is the fundamental package for array computing with Python. ├── python-dateutil>=2.8.1 Extensions to the standard Python datetime module │ └── six>=1.5 Python 2 and 3 compatibility utilities └── pytz>=2020.1 World timezone definitions, modern and historical ```

The problem why your tests didn't catch this is that packaging is a dependency of pytest itself. ```console $ johnnydep pytest 2022-01-29 08:25:03 [info ] init johnnydist [johnnydep.lib] dist=pytest parent=None 2022-01-29 08:25:07 [info ] init johnnydist [johnnydep.lib] dist=atomicwrites>=1.0 parent=pytest 2022-01-29 08:25:09 [info ] init johnnydist [johnnydep.lib] dist=attrs>=19.2.0 parent=pytest 2022-01-29 08:25:11 [info ] init johnnydist [johnnydep.lib] dist=colorama parent=pytest 2022-01-29 08:25:14 [info ] init johnnydist [johnnydep.lib] dist=iniconfig parent=pytest 2022-01-29 08:25:16 [info ] init johnnydist [johnnydep.lib] dist=packaging parent=pytest 2022-01-29 08:25:18 [info ] init johnnydist [johnnydep.lib] dist=pluggy<2.0,>=0.12 parent=pytest 2022-01-29 08:25:20 [info ] init johnnydist [johnnydep.lib] dist=py>=1.8.2 parent=pytest 2022-01-29 08:25:23 [info ] init johnnydist [johnnydep.lib] dist=toml parent=pytest 2022-01-29 08:25:25 [info ] init johnnydist [johnnydep.lib] dist=pyparsing!=3.0.5,>=2.0.2 parent=packaging name summary

pytest pytest: simple powerful testing with Python ├── atomicwrites>=1.0 Atomic file writes. ├── attrs>=19.2.0 Classes Without Boilerplate ├── colorama Cross-platform colored terminal text. ├── iniconfig iniconfig: brain-dead simple config-ini parsing ├── packaging Core utilities for Python packages │ └── pyparsing!=3.0.5,>=2.0.2 Python parsing module ├── pluggy<2.0,>=0.12 plugin and hook calling mechanisms for python ├── py>=1.8.2 library with cross-python path, ini-parsing, io, code, log facilities └── toml Python Library for Tom's Obvious, Minimal Language ```

We also only saw our ASV workflow fail since ASV only installs the package and all direct runtime dependencies in the venv it runs the benchmark with.

I prepared a dummy PR on my fork to demonstrate that this fixes the issue.

Since the release is only 10h old maybe just do a quick post-release so most users won't even notice.

• [ ] User visible changes (including notable bug fixes) are documented in whats-new.rst

Affected xarray versions 0.20.0+

When dependabot recently made a PR to update xarray to 0.20.0 our integration test CI failed with .... File "/opt/hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/xarray/plot/plot.py", line 447, in line ax.set_title(darray._title_for_slice()) File "/opt/hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/xarray/core/dataarray.py", line 3140, in _title_for_slice v=format_item(coord.values), File "/opt/hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/xarray/core/formatting.py", line 146, in format_item return f"{x:.4}" TypeError: unsupported format string passed to numpy.ndarray.__format__

We use xarray.Dataarray.plot.line(x="<dim_name>") to plot data of shape (N, M) where N and M are >=1.

The crash is caused because the following line

https://github.com/pydata/xarray/blob/402a9b398868bd9bd385210138d6f8ea610c8c40/xarray/core/dataarray.py#L3136

allows arrays of shape () and (1,) to be passed into format_item, where both pass this check

https://github.com/pydata/xarray/blob/402a9b398868bd9bd385210138d6f8ea610c8c40/xarray/core/formatting.py#L145-L146

But f"{x:.4}" only works for arrays of shape () and crashes on arrays of shape (1,)

Before the check was https://github.com/pydata/xarray/blob/a78c1e0115d38cb4461fd1aba93334d440cff49c/xarray/core/formatting.py#L145-L146

Which didn't allow any arrays and caused the float formatting issue.

I hope TestPlot1D was the correct place to put the test since the plot itself is 1D but the data which are plotted have a second dimension with size 1 (naming thing and where to put things... 😅).

• [x] Tests added
• [x] Passes pre-commit run --all-files
• [x] User visible changes (including notable bug fixes) are documented in whats-new.rst

PS: Thanks for the awesome package ❤️