Is your feature request related to a problem? Please describe.
Testing our images is good! There's a lot of internal code to wrap some matplotlib functionality for image checking.
|
def compare_or_update_plot(f): |
|
"""Build a decorator that wraps a plot test |
|
|
|
The test function must produce a single figure that will be |
|
compared to the expected figure during the test. If the --update |
|
command line option is given, then the newly generated figure |
|
will be save in place of the expected figure |
|
""" |
|
update = config["update"] |
|
baseline = (FIG_RESULT_BASE / f.__name__).with_suffix(".png") |
|
testFile = baseline.with_name("{}-test.png".format(f.__name__)) |
|
# TODO Don't convert to str - needed because MPL for python 3.5 |
|
# doesn't support saving as pathlib.Path instances |
|
baseline = str(baseline) |
|
testFile = str(testFile) |
|
|
|
@wraps(f) |
|
def wrapped(*args, **kwargs): |
|
f(*args, **kwargs) |
|
fig = pyplot.gcf() |
|
if update: |
|
fig.savefig(baseline) |
|
return |
|
fig.savefig(testFile) |
|
res = compare_images(baseline, testFile, tol=config["plot_tolerance"]) |
|
assert res is None, res |
|
remove(testFile) |
|
|
|
# If a test is tagged with this decorator, mark is with the |
|
# plot mark automatically |
|
pytest.mark.plot(wrapped) |
|
|
|
return wrapped |
Describe the solution you'd like
What if we didn't do that? What if we used https://github.com/matplotlib/pytest-mpl instead?
Describe alternatives you've considered
We keep what we have? Doesn't seem super maintainable
Additional context
