Hi @pijyoi thanks for this PR and tackling this issue!

I'm curious about the changes in all the templates files, for Qt6, we should likely be generating new template files, not modifying the existing ones.

If it would help, I can create new template files and add them to this PR, since they can be a bit annoying to generate.

I'm curious about the changes in all the templates files, for Qt6, we should likely be generating new template files, not modifying the existing ones.

The intention for these fixes is only to get the code "future"-ready. i.e. Minor fixes to convert usage of deprecated and obsolete APIs.

In this case, setWeight() in PySide6 no longer takes an integer, and in any case, the enum value 75 was no longer correct. Since setBold(True) already does the same thing as setWeight(Bold), removing the "weight" property altogether was the simpler solution.

If it would help, I can create new template files and add them to this PR, since they can be a bit annoying to generate.

I did try re-generating a template using pyside2-uic, but this resulted in numerous differences from the existing file, which I thought would clutter the git history.
As the change in the .ui files was minor, I manually edited the auto-generated template files.

In any case, the manual edits are localized to a single commit.

I noticed that examples/VideoTemplate_pyside2.py, compared to the files for the other bindings, imports RawImageGLWidget unconditionally, such that the example will fail to run if pyopengl isn't installed. Any ideas why?

On my local machine, I have manually "generated" all the *_pyside6.py files by making a copy of the corresponding *_pyside2.py files and changing PySide2 to PySide6.

Almost all the examples are able to launch and display something with PySide6

• Remote Plotting example has a "Request timed out" error
  o failure was due to running in a virtualenv (on Windows anyway)
• HDF5 big data example was not tested

MatplotlibWidget.py was not updated.

ScatterPlotItem.py seems to have a PySide2-specific codepath:
https://github.com/pyqtgraph/pyqtgraph/blob/master/pyqtgraph/graphicsItems/ScatterPlotItem.py#L37
I don't know if it should apply to PySide6.

Hi @pijyoi would you like to hop on our slack (so we can communicate outside of issue trackers)? Is interested, feel free to email me, the email is in my github profile.

Hi @j9ac9k , I am good to remain on github for now.

TIL about tools/rebuildUi.py, here I've been regenerating those template files manually, like a chump...

I did try re-generating a template using pyside2-uic, but this resulted in numerous differences from the existing file, which I thought would clutter the git history.
As the change in the .ui files was minor, I manually edited the auto-generated template files.

I've manually made small changes to these template files only to regret it. I've been tempted to migrate the creation of these template files to be part of the CI process ....but for the time being I decided that is more trouble than its worth.

The reason for this change is due to a complete reimplementation of uic. I actually had to regenerate a VideoExample_pyside2 template as part of #1466 (which I know isn't merged yet, but will be shortly).

Should this PR also attempt to get PyQt6 compatibility as well now that its' on PyPi?

Should this PR also attempt to get PyQt6 compatibility as well now that its' on PyPi?

I have tried to make the changes in this PR to be less PySide6 specific.

I attempted to add PyQt6 support but it appears that it is going to require more effort than PySide6. Many enums values are not in their old places.

For example, for the enum TransformationType.TxRotate, in PySide2, PySide6 and PyQt5, it can be accessed as

1. QtGui.QTransform.TxRotate
2. QtGui.QTransform.TransformationType.TxRotate

In PyQt6, it is only accessible from the latter.

https://riverbankcomputing.com/pipermail/pyqt/2020-December/043446.html

Thanks for looking into this @pijyoi

I'll have to figure out how to redo the matrix of test dependencies defined here:

https://github.com/pyqtgraph/pyqtgraph/blob/master/.github/workflows/main.yml#L11-L24

such that PySide6 is added to the test suite. If you have any suggestions, I would say feel free to modify the python-3.9 config such that it uses Qt6.

Also, I'd be curious to get your take on incorporating QtPy (I know they don't currently support Qt6 but they will at some point).

I think PySide6 wheel may have been linked wrongly?

Perhaps? It's easy enough to add apt-get install libopengl-dev as part of the CI process to see what happens. I'll make that change (along w/ the matplotlib export skip) shortly

Sorry for spamming some CI change commits to your PR, feel free to rebase/squash those commits together. Looks like Qt6 effectively needs ubuntu 20.04, github actions was supposed to migrate to that for ubuntu-latest by now, but as they haven't, I've have now specified it. Also, it wasn't libopengl-dev that needed installing, but libopengl0.

A couple of notes:

test_reload was failing on pyside2/python3.9, whatever the reason it was failing there is likely present on pyside6.

For the image tests, we probably want to add PySide6 to this line here:
https://github.com/pyqtgraph/pyqtgraph/blob/master/pyqtgraph/tests/image_testing.py#L260

Is it possible to run the tests directly from the development directory without installing it?

I can run those under the tests subdir

$ pytest pyqtgraph/tests/test_reload.py

but those under the graphicsItems subdir will complain about "No module named pyqtgraph"

$ pytest pyqtgraph/graphicsItems/tests/test_ROI.py

I think PySide6 wheel may have been linked wrongly?

Perhaps? It's easy enough to add apt-get install libopengl-dev as part of the CI process to see what happens. I'll make that change (along w/ the matplotlib export skip) shortly

https://bugreports.qt.io/plugins/servlet/mobile#issue/PYSIDE-1461

Is it possible to run the tests directly from the development directory without installing it?

I can run those under the tests subdir

$ pytest pyqtgraph/tests/test_reload.py

but those under the graphicsItems subdir will complain about "No module named pyqtgraph"

$ pytest pyqtgraph/graphicsItems/tests/test_ROI.py

In the case of test_reload the pyqtgraph directory is added to the path, so you're able to import the pyqtgraph module by directory name

pgpath = os.path.join(os.path.dirname(pg.__file__), '..')
pgpath_repr = repr(pgpath)

The other directories don't attempt to do the same, and assume the pyqtgraph module is installed.

Is there a reason you're trying to run tests w/o installing the library? When making a bunch of changes, I often install pyqtgraph in editable mode pip install -e . that way differences from branches, take affect right away.

Is there a reason you're trying to run tests w/o installing the library? When making a bunch of changes, I often install pyqtgraph in editable mode pip install -e . that way differences from branches, take affect right away.

Oh, I didn't know about "pip install -e"
Normally, my "install" of pyqtgraph is to just copy the pyqtgraph/pyqtgraph subdir.

Is there a reason you're trying to run tests w/o installing the library? When making a bunch of changes, I often install pyqtgraph in editable mode pip install -e . that way differences from branches, take affect right away.

Oh, I didn't know about "pip install -e"
Normally, my "install" of pyqtgraph is to just copy the pyqtgraph/pyqtgraph subdir.

Highly recommend using editable installs as it best simulates an end-user interaction with the library. There is one difference I've found with editable installs vs. non-editable pip installs; and that is accessing the example app can be different.

@pijyoi I think I know how to fix that git error that comes up on the macOS builds, I'm going to try and get the fix merged shortly... if CI here comes back all green after that, this PR ready to merge?

if CI here comes back all green after that, this PR ready to merge?

Yes, ready for merging.

if CI here comes back all green after that, this PR ready to merge?

Yes, ready for merging.

Sorry, forgot one last requested change, can you update the Qt Bindings test matrix in README.md, to add another row:

| Qt-Bindings    | Python 3.7         | Python 3.8         | Python 3.9         |
| :------------- | :----------------: | :----------------: | :----------------: |
| PySide2-5.12   | :white_check_mark: | :x:                | :x:                |
| PyQt5-5.12     | :white_check_mark: | :x:                | :x:                |
| PySide2-5.15   | :x:                | :white_check_mark: | :white_check_mark: |
| PyQt5-5.15     | :x:                | :white_check_mark: | :x:                |
| PySide6-6.0    | :x:                | :x:                | :white_check_mark: |

As well as the supported Qt binding up higher.

That's the last change I got...

@pijyoi do you have strong opinions if this should be squashed or merged with all the commits?

Well, at the very least, the commit that adds the generated files should be separate.

But otherwise, I have already tried to squash related things together. I try to keep each commit to only perform one type of fix, so that it can be cherry-picked or reverted. i.e I have a mild preference to keep the commits separate.

I have a mild preference to keep the commits separate.

Good enough for me.

Woot