A new backend called none. For this backend, no plotting actually happens, but instead, dictionaries with the type of plotting function and arguments used is returned. The equivalent to axes/figures is a list which is extended with the different artists.
There is also an option to raise an error in case artist_kws is present. This is used when testing to ensure we only use properties that are part of the common interface
Tests with hypothesis. These use the "none" backend to check multiple combinations of input arguments and ensure there are no edge cases that don't do what is expected.
These tests are a bit more like unit tests for plot_xyz functions, and should not replace the tests in test_plots.py which are more integration tests going from actual user input to plotting with the chosen backend.
An extra weekly scheduled job to run extended tests with hypothesis. Even with the none backend, the combinations of input arguments are extremely large, so by default not many cases are run, so each hypothesis test takes several seconds at most. This extra weekly job takes >10 seconds per plot_xyz function and opens an issue if there is any test failure.
adds dependabot to keep action versions up to date
fixes a couple issues found in plot_dist and plot_forest when testing with hypothesis
This PR adds several things:
none
. For this backend, no plotting actually happens, but instead, dictionaries with the type of plotting function and arguments used is returned. The equivalent to axes/figures is a list which is extended with the different artists.artist_kws
is present. This is used when testing to ensure we only use properties that are part of the common interfacetest_plots.py
which are more integration tests going from actual user input to plotting with the chosen backend.none
backend, the combinations of input arguments are extremely large, so by default not many cases are run, so each hypothesis test takes several seconds at most. This extra weekly job takes >10 seconds per plot_xyz function and opens an issue if there is any test failure.📚 Documentation preview 📚: https://arviz-plots--41.org.readthedocs.build/en/41/