Add support for modules that are now available in Qt 6.5 #427
Conversation
|
The tests are painful for me, too. Moreover, despite the maintainers claim that they support PyQt5 >= 5.9.0, PySide2 >= 5.12.0, and Qt6 >= 6.2.0, they never actually test the code with the oldest versions. @DaelonSuzuka, would you mind opening a corresponding issue? The PR is not the best place to discuss troubles. |
Why?
Says who? |
|
I've always thought that the PRs are for proposing changes and discussing them. And the Issues are for general discussions and troubleshooting. Forgive me if I'm wrong. I'd prefer not to flood the PR with such useless talk as I'm having now. Although I'm interested in the test as well, I'll follow the discussion of the maintainers and yours silently. |
Unfortunately, I don't have a great answer for you; I at least have generally relied on the CIs to run the full matrix, coupled with a few conda envs with the major Qt/binding versions installed and using the The simplest solution to avoid uninstalling and reinstalling stuff would be just having a directory of venvs each with the different Qt/binding versions you want to test. You would need only two venvs total to respectively test the lower and upper minor version bounds for each binding x Qt major version, since you can switch the binding + Qt major version with with the Beyond that, you could use the https://github.com/spyder-ide/qtpy/blob/master/.github/workflows/test.sh script (probably with a couple tweaks and setting a few env vars) to setup each env and run the tests just as the CIs do. Converting this to use Finally, if the existing CIs aren't adequate to test
I'm not sure what you mean there, sorry. Our CI test matrix includes at least one job each for PyQt5/Qt 5.9, PySide2/Qt 5.12, PySide6/Qt 6.2 and PySide6/Qt 6.2, plus a number of jobs for the upper bound of each (5.15/6.4), and many jobs for the other LTS version/middle bound too (5.12 and 6.3). |
| @@ -8,18 +8,27 @@ | |||
| """Provides QtLocation classes and functions.""" | |||
|
|
|||
| from . import ( | |||
| parse, | |||
There was a problem hiding this comment.
This only works "by accident"/as an internal implementation detail, as parse is in fact imported from packaging, not QtPy; thefore, it should be imported from there instead.
Also, AFAIK our existing tests use the cruder if PYQT_VERSION.startswith("6.5"), but since packaging is required anyway, we may as well use it to be cleaner and more precise, as you do.
There was a problem hiding this comment.
This only works "by accident"/as an internal implementation detail, as parse is in fact imported from packaging, not QtPy; thefore, it should be imported from there instead.
I actually did originally have from packaging import parse but I for some reason I can't remember, I removed it and did this instead.
Also, AFAIK our existing tests use the cruder if PYQT_VERSION.startswith("6.5"), but since packaging is required anyway, we may as well use it to be cleaner and more precise, as you do.
I did see that when I was poking around. I can fix all the startswiths.
There was a problem hiding this comment.
I did see that when I was poking around. I can fix all the
startswiths.
Yeah, thanks—that's probably worth doing in a separate PR from this one, just to be clear.
| if parse(PYQT_VERSION) >= parse('6.5'): | ||
| from PyQt6.QtLocation import * | ||
| else: | ||
| raise QtBindingMissingModuleError(name='QtLocation') |
There was a problem hiding this comment.
If this module was added in Qt 6.5 instead of just in both Python bindings at that same version, then I'm not sure this is the right error, as it is intended for reporting when a particular binding doesn't support a missing module, rather than it not being implemented in Qt itself—not sure how it got overlooked initially, as the current call signature will actually raise an error since it expects binding to be passed. Instead, it would seem to be better to use QtModuleNotInQtVersionError here, with a tweak to either output the full major + minor version automatically (probably best), or pass that version manually.
There was a problem hiding this comment.
If this module was added in Qt 6.5 instead of just in both Python bindings at that same version
This is one of the reasons I need to set up a local environment matrix. I believe that PyQt6 had some of these modules the entire time, and it's only PySide6 that just added them to the distribution, but I won't know for sure until I'm able to step through the versions.
| ) | ||
|
|
||
|
|
||
| @pytest.mark.skipif(not PYSIDE6 or parse(PYSIDE_VERSION) >= parse('6.5'), reason='.') |
There was a problem hiding this comment.
I presume you're going to fill the reason in later?
|
|
||
| @pytest.mark.skipif(not PYSIDE6 or parse(PYSIDE_VERSION) >= parse('6.5'), reason='.') | ||
| def test_qtserialbus_pyside6_below_6_5(): | ||
| with pytest.raises(QtBindingMissingModuleError) as excinfo: |
There was a problem hiding this comment.
| with pytest.raises(QtBindingMissingModuleError) as excinfo: | |
| with pytest.raises(QtBindingMissingModuleError): |
No need for the as if you don't actually use excinfo, no? Or maybe you're planning to?
There was a problem hiding this comment.
I was considering inspecting the actual error in the test. Do you have an opinion on whether that's worth doing?
There was a problem hiding this comment.
Perhaps, yeah, at least the custom attributes of the error object (as opposed to the message, which is considered a user-facing implementation detail subject to change and a test of which would be too fragile).
| import pytest | ||
| from qtpy import ( |
There was a problem hiding this comment.
Split third party imports from local imports (same with the others)
| PYQT_VERSION, | ||
| PYSIDE_VERSION, |
There was a problem hiding this comment.
Instead of importing both of these separately, and particularly if this is related to the Qt version rather than the binding version, you could just import QT_VERSION and check that instead.
There was a problem hiding this comment.
I actually tried that first, and my editor insisted that QT_VERSION didn't exist.
This is also one of the modules that I need the environment matrix for. I think that PyQt6 had this module earlier than PySide6 did.
| # ----------------------------------------------------------------------------- | ||
| # Copyright © 2009- The Spyder Development Team | ||
| # | ||
| # Licensed under the terms of the MIT License | ||
| # (see LICENSE.txt for details) | ||
| # ----------------------------------------------------------------------------- |
There was a problem hiding this comment.
We should at least be consistent about whether we're adding this header to the tests—either to all newly added ones (ideal), or else none.
There was a problem hiding this comment.
Is this the correct copyright header? Various files have slightly different headers and some of them also list specific individuals, so I've never been sure what header I should be adding.
There was a problem hiding this comment.
I was going to comment on that, but I didn't want to be too nitpicky, heh.
The up to date version of the canonical header (for new files, at least) is in the License Guidelines on the our governance repo, i.e.:
| # ----------------------------------------------------------------------------- | |
| # Copyright © 2009- The Spyder Development Team | |
| # | |
| # Licensed under the terms of the MIT License | |
| # (see LICENSE.txt for details) | |
| # ----------------------------------------------------------------------------- | |
| # ----------------------------------------------------------------------------- | |
| # Copyright (c) 2023- QtPy contributors | |
| # | |
| # Released under the terms of the MIT License | |
| # (see LICENSE.txt in the project root directory for details) | |
| # ----------------------------------------------------------------------------- |
The reason a number of files list different individuals is because a lot of the code was combined from multiple predecessor projects (see the Readme section that touches on a bit of this history), plus various other historical contributions. New files should just use the above, but it is important to retain the existing copyright notices in files that have them for both legal and ethical purposes.
Co-authored-by: C.A.M. Gerlach <[email protected]>
Actually, your comment helps a lot. Thanks for the input(as always!).
I've tried to use
At least as an experiment, I'm considering building a big 'ol matrix with all the major versions of each binding, and then run an audit of the entire history of Qt. I think I will start with
I definitely wouldn't advocate rebuilding the existing CI system, and I don't believe the audit tool needs to be a part of the standard CI workflow. |
I'm sorry, I haven't noticed that PyQt5 5.9 is tested only on Windows. I've looked through Ubuntu tests and assumed that the versions used are identical for all the platforms. I must have had a bad day yesterday, and I apologize for the noise I made here. |
I actually used it successfully just earlier today; you can check if it is by running the Just to note, the binding selection logic is currently kinda jankily structured and could really use a refactoring, which is something I'm planning to do if and when the next major version rolls around (since it would result in some non-trivial behavior changes in certain edge cases).
That would be super cool, at least IMO! Would love to see that. At least, you could run the supported versions from 5.9 - 5.15 and 6.2-6.5.
Yeah, the matrix is pretty darn complex, even if it is mostly single-sourced now in one declarative config file.
No worries on my side and hope you're having a better day today; I did get a little worried it might put either you or @DaelonSuzuka off from contributing as you've both been very helpful in recent QtPy releases but hopefully that has smoothed out. Cheers! |
PySide6 v6.5.0 released last week, allegedly adding the
QtTextToSpeech,QtSerialBus, andQtLocationmodules.This PR adds the completely missing
QtSerialBus, and adds version-dependent imports toQtTextToSpeechandQtLocation.This is currently a draft because I haven't gotten all the tests written yet, and I wanted to ask if there's any advice for running the full matrix of tests locally. It's way too much work to shuffle different Qt bindings and versions in and out of my local venv. I've used tox in another project, and even though it worked, it wasn't especially enjoyable.
Edit: Additionally, if there's a reasonable way to run this matrix, then I might be motivated enough to revisit my previous automatic-qt-library-auditor from PR #344, and set that up as a runnable tool. This would be useful in keeping up with further changes in the Qt ecosystem.