0.12 Backports #1444
0.12 Backports #1444
Conversation
This apparently will be the default and the option is going away (in favor of opting-in to capture the output). Removing it silences a warning.
|
|
docs/index.rst
Outdated
| @@ -39,14 +39,8 @@ out our :doc:`Getting Started <startingguide>` guide. Development is | |||
| supported by the National Science Foundation through grants AGS-1344155, | |||
| OAC-1740315, AGS-1901712. | |||
|
|
|||
| MetPy follows `semantic versioning <https://semver.org>`_ in its version number. With our | |||
| current 0.x version, that implies that MetPy's APIs (application programming interfaces) are | |||
There was a problem hiding this comment.
Probably want to roll most of this back, with the exception of the updated gallery link.
dopplershift
force-pushed
the
0.12-backports
branch
from
134bb73 to
3e91bba
Compare
tests/test_xarray.py
Outdated
| @@ -114,8 +114,9 @@ def test_convert_units(test_var): | |||
| assert units(test_var.attrs['units']) == units('degC') | |||
There was a problem hiding this comment.
@jthielen Can you sanity check me on this test_convert_units test? I can't see any way that this was correct, at least in the days before we had the integration between xarray and pint working like it does now. It was confusing to see the test suddenly fail, and I want to make sure that I'm right that it should have been failing all along.
There was a problem hiding this comment.
The diff in this test here doesn't seem to be doing anything meaningful, unless something is broken under the hood in assert_almost_equal. Does this diff make a difference between passing and failing?
Since I'm not completely sure what is the direct cause for confusion, here are some notes below:
This particular line
assert units(test_var.attrs['units']) == units('degC')was updated back in #1144 since Pint changed its canonical string for °C from degC to degrees_Celsius, so we can't do string comparison, but comparison of units is still fine.
But, this whole test is no longer valid after Quantity-in-xarray was introduced and after we moved away from in-place unit conversions, hence why this test on master is currently:
def test_convert_units(test_var):
"""Test conversion of units."""
result = test_var.metpy.convert_units('degC')
# Check that units are updated without modifying original
assert result.metpy.units == units.degC
assert test_var.metpy.units == units.kelvin
# Make sure we now get an array back with properly converted values
assert_almost_equal(result[0, 0, 0, 0], 18.44 * units.degC, 2)This test definitely relies on assert_almost_equal automatically converting DataArrays to their corresponding unit arrays, hence why it should have worked fine before this diff, and I don't see what the diff in effect changed.
There was a problem hiding this comment.
Well, it should rely on that. That test did fail on one of the earlier runs. The test passes on 0.12.1. It turns out, though, that the only reason that test passed was because we were swallowing an AttributeError when calling .to() on a DataArray. I think the right way to do this for a bugfix release is to put back the AttributeError clause in check_and_drop_units() since that change wasn't necessary (and I'm not sure where it snuck in).
There was a problem hiding this comment.
Ah, the change to testing crept in on 771cbf0. I'll pop that off and my most recent change to the test.
dopplershift
force-pushed
the
0.12-backports
branch
from
3e91bba to
8f739e8
Compare
This adds a full suite of jobs to handle building docs on Linux, as well as running tests on Linux (using PyPI) and on Windows/Mac using Conda. Instead of relying on what's encoded in setup.py and a bunch of manual additions (Travis) or an environment.yml (AppVeyor), we have requirement and constraint files that get passed/swapped depending on what kind of build we're running. This allows us to pin versions and hopefully let dependabot manage the updates to these so that CI doesn't fail due to the ecosystem updating.
WMO has put older versions like 2014 behind a login screen. Just update the link to the 2018 version which still has the formula.
This one AMS journal DOI resolves to a link that produces 403 errors for the linkchecker. It resolves in the browser, and it's a DOI so should be stable. Just ignore.
Use our new requirement files to determine dependencies. Also remove a bunch of cruft now that we've moved to Ubuntu 18.04 on Travis and also no longer need to worry about building and uploading wheels.
Will update further once we have solidified what we're doing with GitHub vs. Travis.
Thanks for the Windows builds. Also thanks for being one of the bigger pains in my butt--I still can't get rid of the email notifications.
Uses reviewdog (a Go tool) to report flake8 error messages using GitHub API.
Need to see if any of these extras are worth it, but we can leave these linting things as configuration details for the actions. They're also now more accurately reflected in linting_requirements.txt.
Tripped doc8.
It's not even part of the doc dependencies any more.
This way we can at least tell what the failing line was.
This should be the final part of our new CI setup, handling updates to our pinned dependencies.
Bumps [pyflakes](https://github.com/PyCQA/pyflakes) from 2.1.1 to 2.2.0. - [Release notes](https://github.com/PyCQA/pyflakes/releases) - [Changelog](https://github.com/PyCQA/pyflakes/blob/master/NEWS.rst) - [Commits](PyCQA/pyflakes@2.1.1...2.2.0) Signed-off-by: dependabot[bot] <support@github.com>
Bumps [pytest](https://github.com/pytest-dev/pytest) from 5.4.2 to 5.4.3. - [Release notes](https://github.com/pytest-dev/pytest/releases) - [Changelog](https://github.com/pytest-dev/pytest/blob/master/CHANGELOG.rst) - [Commits](pytest-dev/pytest@5.4.2...5.4.3) Signed-off-by: dependabot[bot] <support@github.com>
Was failing to build CartoPy without NumPy installed first.
We renamed some dependency files, so we need to make sure the hashes for the caches reflect this.
We removed all linting from the test install, so need to remove pytest-flake8. Also remove the totally useless and unused [dev] install target.
Use 1.5.3 since that's what's installable on conda on Windows right now.
Bumps [xarray](https://github.com/pydata/xarray) from 0.15.1 to 0.16.0. - [Release notes](https://github.com/pydata/xarray/releases) - [Changelog](https://github.com/pydata/xarray/blob/master/HOW_TO_RELEASE.md) - [Commits](pydata/xarray@v0.15.1...v0.16.0) Signed-off-by: dependabot[bot] <support@github.com>
-Removing Inoperable 'Say Thanks' Link
Bumps [matplotlib](https://github.com/matplotlib/matplotlib) from 3.2.1 to 3.3.0. - [Release notes](https://github.com/matplotlib/matplotlib/releases) - [Commits](matplotlib/matplotlib@v3.2.1...v3.3.0) Signed-off-by: dependabot[bot] <support@github.com>
This is no longer needed since we're on newer matplotlib versions. Also, now this fixture actually *causes* problems because matplotlib is relying on builtin round returning an int (when not asking for digits), which numpy's does not do.
A few items got deprecated/removed and need updating.
Trying to eliminate false alarm with small decreases in coverage caused by deleting lines.
Since there's no interdependency between the jobs, just make them separate workflows. This helps with the fact that you can't restart individual job instances and can only restart a workflow.
This may or may not be faster than downloading as needed from Natural Earth. It should at least be more reliable.
This allows it to fallback and re-use as much as possible.
Bumps [pytest](https://github.com/pytest-dev/pytest) from 5.4.3 to 6.0.0. - [Release notes](https://github.com/pytest-dev/pytest/releases) - [Changelog](https://github.com/pytest-dev/pytest/blob/master/CHANGELOG.rst) - [Commits](pytest-dev/pytest@5.4.3...6.0.0) Signed-off-by: dependabot[bot] <support@github.com>
Bumps [pytest](https://github.com/pytest-dev/pytest) from 6.0.0 to 6.0.1. - [Release notes](https://github.com/pytest-dev/pytest/releases) - [Changelog](https://github.com/pytest-dev/pytest/blob/master/CHANGELOG.rst) - [Commits](pytest-dev/pytest@6.0.0...6.0.1) Signed-off-by: dependabot[bot] <support@github.com>
RST formatting was not correct.
Need to be able to run on 0.12.x branch
Slight change to rotation calculation results in subtle pixel-level changes, so update the tolerances--the changes aren't big enough to warrant new images.
dopplershift
force-pushed
the
0.12-backports
branch
from
8f739e8 to
c813544
Compare
|
Hey, how long does it take for a release to become available on conda-forge? |
Not long now, conda-forge/metpy-feedstock#41 |
|
Usually a matter of hours. Just getting the package recipe updated (we waited for their update bot to do make the PR to bump it) and the packages to build and show up on the Anaconda CDN. Currently the package is up on anaconda.org so it's a matter of waiting for the CDN to get them (usually about 20 minutes). |
|
Got it, thank you! |
Description Of Changes
Merging in all the PRs that we want on 0.12.2. This is a lot bigger than normal because of the massive CI overhaul.