CI says no 😆

looks like pg.time() has some sort of name conflict that needs addressing. Would have to look into the other failure a bit more.

thanks for hte PR @fbordignon ... pyqtgraph has had some minor annoyances involving freezing applications such as cx_freeze and pyinstaller. Another one that gets me frequently is the use of template files:

pyqtgraph-pyside2_515-py38 ❯ find pyqtgraph -iname "*template*.py" -type f
pyqtgraph/imageview/ImageViewTemplate_pyqt6.py
pyqtgraph/imageview/ImageViewTemplate_pyside2.py
pyqtgraph/imageview/ImageViewTemplate_pyside6.py
pyqtgraph/imageview/ImageViewTemplate_pyqt5.py
pyqtgraph/flowchart/FlowchartCtrlTemplate_pyside6.py
pyqtgraph/flowchart/FlowchartCtrlTemplate_pyside2.py
pyqtgraph/flowchart/FlowchartCtrlTemplate_pyqt5.py
pyqtgraph/flowchart/FlowchartTemplate_pyqt6.py
pyqtgraph/flowchart/FlowchartTemplate_pyqt5.py
pyqtgraph/flowchart/FlowchartTemplate_pyside6.py
pyqtgraph/flowchart/FlowchartTemplate_pyside2.py
pyqtgraph/flowchart/FlowchartCtrlTemplate_pyqt6.py
pyqtgraph/GraphicsScene/exportDialogTemplate_pyside2.py
pyqtgraph/GraphicsScene/exportDialogTemplate_pyside6.py
pyqtgraph/GraphicsScene/exportDialogTemplate_pyqt5.py
pyqtgraph/GraphicsScene/exportDialogTemplate_pyqt6.py
pyqtgraph/graphicsItems/PlotItem/plotConfigTemplate_pyside2.py
pyqtgraph/graphicsItems/PlotItem/plotConfigTemplate_pyqt6.py
pyqtgraph/graphicsItems/PlotItem/plotConfigTemplate_pyside6.py
pyqtgraph/graphicsItems/PlotItem/plotConfigTemplate_pyqt5.py
pyqtgraph/graphicsItems/ViewBox/axisCtrlTemplate_pyqt5.py
pyqtgraph/graphicsItems/ViewBox/axisCtrlTemplate_pyside6.py
pyqtgraph/graphicsItems/ViewBox/axisCtrlTemplate_pyside2.py
pyqtgraph/graphicsItems/ViewBox/axisCtrlTemplate_pyqt6.py
pyqtgraph/canvas/TransformGuiTemplate_pyside6.py
pyqtgraph/canvas/TransformGuiTemplate_pyside2.py
pyqtgraph/canvas/TransformGuiTemplate_pyqt5.py
pyqtgraph/canvas/CanvasTemplate_pyqt6.py
pyqtgraph/canvas/TransformGuiTemplate_pyqt6.py
pyqtgraph/canvas/CanvasTemplate_pyqt5.py
pyqtgraph/canvas/CanvasTemplate_pyside2.py
pyqtgraph/canvas/CanvasTemplate_pyside6.py
pyqtgraph/console/template_pyqt5.py
pyqtgraph/console/template_pyside2.py
pyqtgraph/console/template_pyqt6.py
pyqtgraph/console/template_pyside6.py

The freezing applications seem to miss these (due to how we handle their imports); so I'm definitely in support of making pyqtgraph a bit more robust in that regard.

Yeah, I've just finished working around the missing template files:

import pyqtgraph
pg_folder = Path(pyqtgraph.__file__).parent
templates = pg_folder.rglob("*pyside2.py")
sources = [str(w) for w in templates]
destinations = ["lib"+str(w).replace(str(pg_folder.parent), '') for w in sources]
for a in zip(sources, destinations):
    include_files.append(a)

...

build_options = {
    "packages": [],
    "includes": [],
    "excludes": [],
    "zip_exclude_packages": "*",
    "zip_include_packages": "",
    "include_files": include_files,
    "include_msvcr": True,
    "build_exe": "../build",
}

It will copy as .py files and not .pyc, but it works for now.

Do you think it is Ok to include this hack into the cx_freeze example or is it too hacky?
Need to modify the pattern to get all the templates, as you did on your 'find'.

Do you think it is Ok to include this hack into the cx_freeze example or is it too hacky?

Sounds like par for the course here, I think that would be a welcome addition!

I'm in the middle of trying to wrap up (and expand) PR 1796 so I'm not sure how much help I'll be with dealing with the CI errors this PR has caused for the coming days/week.

I rebased master onto this branch; and I modified PlotAutoRange.py to use perf_counter (which it should have been doing anyway); and the test suite passes...but I'm not sure we want to use this change. We recently merged #1846 authored by @pijyoi that aimed to reduce the namespace clutter, and this diff, as advertised would sort of undo that change.

That said, freeze applications do need a way of knowing to grab stuff; and we shouldn't be dependent on on people tweaking hooks/config files to make sure they package correctly with greater ease.

I do want to get this working a bit better, and I suspect there is a way we can do this w/o polluting the top level namespace too much.

Actually, I'm curious if #1846 may have fixed this issue in general. @fbordignon could you try and verify it's still a problem based on the 0.12.2 release (or current master branch)?

defining __all__ for a few files will go a long way, here, I bet

I rebased master onto this branch; and I modified PlotAutoRange.py to use perf_counter (which it should have been doing anyway); and the test suite passes...but I'm not sure we want to use this change. We recently merged #1846 authored by @pijyoi that aimed to reduce the namespace clutter, and this diff, as advertised would sort of undo that change.

That said, freeze applications do need a way of knowing to grab stuff; and we shouldn't be dependent on on people tweaking hooks/config files to make sure they package correctly with greater ease.

I do want to get this working a bit better, and I suspect there is a way we can do this w/o polluting the top level namespace too much.

Actually, I'm curious if #1846 may have fixed this issue in general. @fbordignon could you try and verify it's still a problem based on the 0.12.2 release (or current master branch)?

Hy @j9ac9k. thanks for the update. The master branch at a472f8c does not fix the issue. The error message when freezing the application still is:

pyqtgraph\functions.py', line 21 in <module>
from . import debug, reload
ImportError: cannot import name 'debug' from partially initialized module 'pyqtgraph' (most likely due to a circular import)

@ksunden, defining __all__ for the affected classes can solve this? The goal is to prevent namespace pollution and allow for from .debug import * for example?

@fbordignon , is it actually necessary to import * from debug in order to make cx_freeze work again? Or is the main thing to ensure that there is an import in __init__.py ?

I think the point of #1846 was to more carefully consider what needs to be part of the main pg.(???) namespace and what can be called as pg.module.(???).

The debug module comes with a huge number of functions/classes, but these would normally be used as e.g. prof = debug.Profiler(), so that import debug (rather than from debug import *) would be sufficient.

But I don't know if that is enough to make cx_freeze happy. :)

@fbordignon Thanks for following up. Can you provide some step-by-step instructions to replicate the issue as part of some minimum example? In case you need a dummy app to try and freeze, may I recommend using one of the examples?

As I said, I am motivated to fix this, but I need to try and find a way of doing so without polluting the namespace. There is likely a solution here that will work for all cases.

A possible alternate solution would be to use a single generated file (with a minor edit) usable for all bindings.
For that, the PyQt6 generated file is the only one that is suitable as it is the only one that uses the long enum spellings.

The minor edit required would be changing

from PyQt6 import QtCore, QtGui, QtWidgets

to: (number of dots varies)

from ...Qt import QtCore, QtGui, QtWidgets

Turns out the original issue with cx_freeze and statements like from . import debug seems to be fixed on the latest version of cx_freeze. The problem was that those statements were not adding the modules to the dependencies and they were not frozen to the app. I did not find the issue related to that in cx_freeze, but it is working now.
Now the frozen app works without even the original fix merged on #1303 I've pushed a commit reverting the issue.

Oops, I was testing the issue and accidentally pushed changes to SignalProxy.py as well. May not change anything but if needed I can revert it.

Hi @fbordignon if I'm reading the PR and your comment right, the only diff you're proposing is removing this bit from __init__.py ?

from .ThreadsafeTimer import *

Yes, this change from .ThreadsafeTimer import * was merged with issue #1303 and it is no longer needed for newer cx_freeze versions. This current issue #1801 is fixed by the cx_freeze update to 6.8.

Implementing this change would break things like...

import pyqtgraph as pg

...

obj = pg.ThreadSafetimer()

@outofculture you were using this very object in ACQ4, I imagine this change would impact you?

@fbordignon does cx_freeze 6.8+ work w/ pyqtgraph as it is on master (or 0.12.3) as is?

@fbordignon does cx_freeze 6.8+ work w/ pyqtgraph as it is on master (or 0.12.3) as is?

Yes it does. The removal I am proposing is only to revert the changes made on issue #1303 and completely optional. If it is to break something I think we can close this issue and do nothing.

@fbordignon does cx_freeze 6.8+ work w/ pyqtgraph as it is on master (or 0.12.3) as is?

Yes it does. The removal I am proposing is only to revert the changes made on issue #1303 and completely optional. If it is to break something I think we can close this issue and do nothing.

Cool, I'm leaning on closing this PR as things are working. Merging the change as-is could have some breakages I think as the previous change added an object to the namespace... but I'll wait to hear from people that actually use the ThreadSafeTimer to chime in 👍🏻

Hi @fbordignon I finally got ahold of some folks that I thought might be impacted by this PR, they gave it a thumbs up. I see we've created some merge conflicts for you. Would you mind if I resolved the conflicts and then merged this?

@j9ac9k I just did, but I believe the changes I've proposed were on master already. Please close this PR without merging since the change that remains does nothing.

I still see ThreadSafeTimer being imported in __init__.py. Regardless, I can close out the PR. Thanks for the follow up here, I'm glad to hear cx_freeze is working w/ pyqtgraph.

Ok, Maybe I messed things on the merge with master. I did not see I was overwriting the from .ThreadSafeTimer import *
Let me give another go

Yeah, the only change is that line now on the diff to master. It does not impact a lot because there is an all defined inside ThreadSafeTimer so I believe it is OK to leave it and do not merge. ThreadSafeTimer.py is fairly small anyways so no harm.

The whole ordeal was me trying to merge imports defined at 6eeb0b5

from .debug import *
from .reload import *

which were needed to freeze an app on a previous version of cx_freeze but were polluting the namespace.
Now it is everything working properly with newer versions of cx_freeze.

Sorry for the very delayed follow up; really appreciate your prompt responses. Please don't hesitate to reach out of pyqtgraph is not meeting some criteria regarding being frozen and such (or any other issue w/ the library) 👍🏻

Nothing to be sorry about, your responses were always on point. It is me that should be thanking you guys for such a great lib!