-
Notifications
You must be signed in to change notification settings - Fork 416
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Cleanup warnings from tests #3255
Conversation
I like this PR a lot! :) I would say though I was confused by the title of "Silence Warnings", which implies to me a bunch of |
cf52348
to
26ee990
Compare
Ugh, annoying. I used our "new" matplotlib version checking functions to avoid a deprecation warning in declarative, but that's in |
What about putting the version checking of matplotlib in |
@kgoebber My local draft has a commit that looks for a specific attribute on a class in Matplotlib. This avoids the location of that function as well as incurring a hard dependency on |
4101659
to
41627a0
Compare
With this we bump minimum scipy to 1.8.0, which is the first version that has |
3b2cac8
to
a7c0b5b
Compare
These aren't necessarily correct, but fully fixing/eliminating them out of the advection call is difficult.
We want a Quantity instead.
pytest-mpl handles it for figures we compare, but we need to manually close other figures we generate.
Matplotlib 3.8 deprecated the collections attribute on ContourSet.
Avoid passing Quantity instances to matplotlib plots that don't support them in the declarative interface.
Python 3.12 deprecated UTC tz-naive methods utcnow() and utcfromtimestamp(). This adds a filter for one import-time warning from dateutil. Our approach here maintains using tz-naive objects afterwards for user-accessible objects.
This allows the tests to pass properly on a "minimum" config on my mac M2.
These are inevitable when you do mod/remainder on a nan, but we're perfectly happy to get nan back.
These are a natural consequence of masked data/trying to trigger error conditions.
This avoids needing to interpolate an LCL below our bottom pressure.
These aren't planned to be addressed until well after numpy 2.0 is released.
Most of this is fixed by passing filenames, rather than file objects, to pandas/numpy so that they have ownership.
Adds a few missing files that were triggering downloads in tests.
This is to avoid a warning from setuptools_scm.
Moves to 0.21.0, 1.4.0, and 0.17, respectively. These are on the edge of our window but they help avoid deprecation warnings.
b10fe83
to
b3b2c3f
Compare
The fixup in Unidata#3255 inadvertently added unit parsing to all code paths in interpolate_to_slice. This breaks any usages that passed in data with unknown units. Since interpolate_to_slice has no need to do any unit handling, avoid this unless we're actually given a quantity.
The fixup in Unidata#3255 inadvertently added unit parsing to all code paths in interpolate_to_slice. This breaks any usages that passed in data with unknown units. Since interpolate_to_slice has no need to do any unit handling, avoid this unless we're actually given a quantity.
Description Of Changes
Working to get rid of all the warnings that are spewed out by our tests.