November 18, 2020

2432 words 12 mins read

mottosso/Qt.py

Minimal Python 2 & 3 shim around all Qt bindings - PySide, PySide2, PyQt4 and PyQt5.


repo name	mottosso/Qt.py
repo link	https://github.com/mottosso/Qt.py
homepage
language	Python
size (curr.)	893 kB
stars (curr.)	657
created	2016-05-27
license	MIT License

Qt.py enables you to write software that runs on any of the 4 supported bindings - PySide2, PyQt5, PySide and PyQt4.

News

Date	Version	Event
Sep 2020	1.3.0	Stability improvements and greater ability for `QtCompat.wrapInstance` to do its job
Jun 2019	1.2.1	Bugfixes and additional members
Jan 2018	1.1.0	Adds new test suite, new members
Mar 2017	1.0.0	Increased safety, backwards incompatible
Sep 2016	0.6.9	Stable release
Sep 2016	0.5.0	Alpha release of `--convert`
Jun 2016	0.2.6	First release of Qt.py

More details.

Project goals

Write once, run in any binding.

Qt.py was born in the film and visual effects industry to address the growing need for software capable of running with more than one flavor of the Qt bindings for Python - PySide, PySide2, PyQt4 and PyQt5.

Goal	Description
Support co-existence	Qt.py should not affect other bindings running in same interpreter session.
Build for one, run with all	Code written with Qt.py should run on any binding.
Explicit is better than implicit	Differences between bindings should be visible to you.

See CONTRIBUTING.md for more details.

Install

Qt.py is a single file and can either be copy/pasted into your project, downloaded as-is, cloned as-is or installed via pip or conda.

# From PyPI
$ pip install Qt.py

# From Anaconda
$ conda config --add channels conda-forge
$ conda install qt.py

Pro tip: Never use the latest commit for production. Instead, use the latest release. That way, when you read bug reports or make one for yourself you will be able to match a version with the problem without which you will not know which fixes apply to you nor would we be able to help you. Installing via pip or conda as above ensures you are provided the latest stable release. Unstable releases are suffixed with a .b, e.g. 1.1.0.b3.
Pro tip: Supports vendoring

Usage

Use Qt.py as you would use PySide2.

import sys
from Qt import QtWidgets

app = QtWidgets.QApplication(sys.argv)
button = QtWidgets.QPushButton("Hello World")
button.show()
app.exec_()

Also see /examples

Documentation

All members of Qt stem directly from those available via PySide2, along with these additional members.

Attribute	Returns	Description
`__version__`	`str`	Version of this project
`__binding__`	`str`	A string reference to binding currently in use
`__qt_version__`	`str`	Reference to version of Qt, such as Qt 5.6.1
`__binding_version__`	`str`	Reference to version of binding, such as PySide 1.2.6

Example

>>> from Qt import __binding__
>>> __binding__
'PyQt5'

Compatibility

Qt.py also provides compatibility wrappers for critical functionality that differs across bindings, these can be found in the added QtCompat submodule.

Attribute	Returns	Description
`loadUi(uifile=str, baseinstance=QWidget)`	`QObject`	Minimal wrapper of PyQt4.loadUi and PySide equivalent
`translate(...)`	`function`	Compatibility wrapper around QCoreApplication.translate
`wrapInstance(addr=long, type=QObject)`	`QObject`	Wrapper around `shiboken2.wrapInstance` and PyQt equivalent
`getCppPointer(object=QObject)`	`long`	Wrapper around `shiboken2.getCppPointer` and PyQt equivalent
`isValid(object=QObject)`	`bool`	Wrapper around `shiboken2.isValid` and PyQt equivalent
`dataChanged(topLeft=QModelIndex, bottomRight=QModelIndex, roles=[])`	`None`	Wrapper around `QtCore.QAbstractItemModel.dataChanged.emit`

Example

>>> from Qt import QtCompat
>>> QtCompat.loadUi

Class specific compatibility objects

Between Qt4 and Qt5 there have been many classes and class members that are obsolete. Under Qt.QtCompat there are many classes with names matching the classes they provide compatibility functions. These will match the PySide2 naming convention.

from Qt import QtCore, QtWidgets, QtCompat
header = QtWidgets.QHeaderView(QtCore.Qt.Horizontal)
QtCompat.QHeaderView.setSectionsMovable(header, False)
movable = QtCompat.QHeaderView.sectionsMovable(header)

This also covers inconsistencies between bindings. For example PyQt4’s QFileDialog matches Qt4’s return value of the selected. While all other bindings return the selected filename and the file filter the user used to select the file. Qt.QtCompat.QFileDialog ensures that getOpenFileName(s) and getSaveFileName always return the tuple.

Environment Variables

These are the publicly facing environment variables that in one way or another affect the way Qt.py is run.

Variable	Type	Description
QT_PREFERRED_BINDING_JSON	str	Override order and content of binding to try. This can apply per Qt.py namespace.
QT_PREFERRED_BINDING	str	Override order and content of binding to try. Used if QT_PREFERRED_BINDING_JSON does not apply.
QT_VERBOSE	bool	Be a little more chatty about what’s going on with Qt.py
QT_SIP_API_HINT	int	Sets the preferred SIP api version that will be attempted to set.

Subset (or “common members”)

Members of Qt.py is a subset of PySide2. Which means for a member to be made accessible via Qt.py, it will need to (1) be accessible via PySide2 and (2) each of the other supported bindings. This excludes large portions of the Qt framework, including the newly added QtQml and QtQuick modules but guarantees that anything you develop with Qt.py will work identically on any binding - PySide, PySide2, PyQt4 and PyQt5. If you need to use such excluded modules with Qt.py, please see QtSiteConfig.py.

We call this subset “common members” and these can be generated by running the build_membership.sh script. The script will output all modules and members of each binding into individual JSON files. These JSON files are then compared and a common_members.json file is generated. The contents of this file is copy-pasted into the _common_members dictionary of Qt.py. Please note that the script will only use the very latest version of our Docker test suite to generate the common members subset, using the most up-to-date set of VFX Platform-stipulated software versions.

:warning: The version of PySide2 used as reference is the one specified on VFX Platform, currently version is 2.0.x. But unfortunately, the version string of PySide2 is not yet properly maintained and the VFX Platform does not specifiy a explicit commit SHA for PySide2. Therefore, it could be difficult to know exactly which PySide2 is running on your system (unless you built it from source). In layman’s terms; as PySide2 is in development and is continuously adding new support for modules, you may see differences between PySide2 built early in the year vs PySide2 built later in the year. The exact commit SHAs of PySide2 used by the Qt.py test suite can be reviewed in DOCKER.md. QtC implemented an alternative way to identify which version of PySide2 you are running. You can read more about that here.

Branch binding-specific code

Some bindings offer features not available in others, you can use __binding__ to capture those.

if "PySide" in __binding__:
  do_pyside_stuff()

Override preferred choice

If your system has multiple choices where one or more is preferred, you can override the preference and order in which they are tried with this environment variable.

$ set QT_PREFERRED_BINDING=PyQt5  # Windows
$ export QT_PREFERRED_BINDING=PyQt5  # Unix/OSX
$ python -c "import Qt;print(Qt.__binding__)"
PyQt5

Constrain available choices and order of discovery by supplying multiple values.

# Try PyQt4 first and then PySide, but nothing else.
$ export QT_PREFERRED_BINDING=PyQt4:PySide

Using the OS path separator (os.pathsep) which is : on Unix systems and ; on Windows.

If you need to control the preferred choice of a specific vendored Qt.py you can use the QT_PREFERRED_BINDING_JSON environment variable instead.

{
    "Qt":["PyQt5"],
    "myproject.vendor.Qt":["PyQt5"],
    "default":["PySide2"]
}

This json data forces any code that uses import Qt or import myproject.vendor.Qt to use PyQt5(from x import Qt etc works too, this is based on __name__ of the Qt.py being imported). Any other imports of a Qt module will use the “default” PySide2 only. If "default" is not provided or a Qt.py being used does not support QT_PREFERRED_BINDING_JSON, QT_PREFERRED_BINDING will be respected.

# Try PyQt5 first and then PyQt4 for the Qt module name space.
$ export QT_PREFERRED_BINDING_JSON="{"Qt":["PyQt5","PyQt4"]}"
# Use PyQt4 for any other Qt module name spaces.
$ export QT_PREFERRED_BINDING=PySide2

QtSiteConfig.py

Add or remove members from Qt.py at run-time.

Examples

If you need to expose a module that isn’t included in Qt.py by default or wish to remove something from being exposed in Qt.py you can do so by creating a QtSiteConfig.py module and making it available to Python.

Create a new file QtSiteConfig.py
Implement update_members
Expose to Python

# QtSiteConfig.py
def update_members(members):
    """Called by Qt.py at run-time to modify the modules it makes available.

    Arguments:
        members (dict): The members considered by Qt.py
    """
    members.pop("QtCore")

Finally, expose the module to Python.

$ set PYTHONPATH=/path/to
$ python -c "import Qt.QtCore"
ImportError: No module named Qt.QtCore

Linux and MacOS users, replace set with export

Compile Qt Designer files

WARNING - ALPHA FUNCTIONALITY See #132 for details.

.ui files compiled via pyside2-uic inherently contain traces of PySide2 - e.g. the line from PySide2 import QtGui.

In order to use these with Qt.py, or any other binding, one must first erase such traces and replace them with cross-compatible code.

$ pyside2-uic my_ui.ui -o my_ui.py
$ python -m Qt --convert my_ui.py
# Creating "my_ui_backup.py"..
# Successfully converted "my_ui.py"

Now you may use the file as you normally would, with Qt.py

Loading Qt Designer files

The uic.loadUi function of PyQt4 and PyQt5 as well as the QtUiTools.QUiLoader().load function of PySide/PySide2 are mapped to a convenience function loadUi.

import sys
from Qt import QtCompat

app = QtWidgets.QApplication(sys.argv)
ui = QtCompat.loadUi(uifile="my.ui")
ui.show()
app.exec_()

For PyQt bindings it uses their native implementation, whereas for PySide bindings it uses our custom implementation borrowed from the qtpy project.

loadUi has two arguments as opposed to the multiple that PyQt ships with. See here for details - in a nutshell, those arguments differ between PyQt and PySide in incompatible ways. The second argument is baseinstance which allows a ui to be dynamically loaded onto an existing QWidget instance.

QtCompat.loadUi(uifile="my.ui", baseinstance=QtWidgets.QWidget)

uifile is the string path to the ui file to load.

If baseinstance is None, the a new instance of the top-level widget will be created. Otherwise, the user interface is created within the given baseinstance. In this case baseinstance must be an instance of the top-level widget class in the UI file to load, or a subclass thereof. In other words, if you’ve created a QMainWindow interface in the designer, baseinstance must be a QMainWindow or a subclass thereof, too. You cannot load a QMainWindow UI file with a plain QWidget as baseinstance.

loadUi returns baseinstance, if baseinstance is provided. Otherwise it will return the newly created instance of the user interface.

sip API v2

If you’re using PyQt4, sip attempts to set its API to version 2 for the following:

QString
QVariant
QDate
QDateTime
QTextStream
QTime
QUrl

Rules

The PyQt and PySide bindings are similar, but not identical. Where there is ambiguity, there must to be a clear direction on which path to take.

Governing API

The official Qt 5 documentation is always right. Where the documentation lacks answers, PySide2 is right.

For example.

# PyQt5 adheres to PySide2 signals and slots
PyQt5.Signal = PyQt5.pyqtSignal
PyQt5.Slot = PyQt5.pyqtSlot

# PySide2 adheres to the official documentation
PySide2.QtCore.QStringListModel = PySide2.QtGui.QStringListModel

Caveats

There are cases where Qt.py is not handling incompatibility issues. Please see CAVEATS.md for more information.

Known Problems

Send us a pull-request with known problems here!

Maya and incompatible bindings on PYTHONPATH

Who’s using Qt.py?

Send us a pull-request with your studio here.

Presented at Siggraph 2016, BOF!

Projects using Qt.py

Send us a pull-request with your project here.

Projects similar to Qt.py

Comparison matrix.

Project	Audience	Reference binding	License	PEP8	Standalone	PyPI	Co-existence
Qt.py	Film	PySide2	MIT	X	X	X	X
jupyter	Scientific	N/A	N/A	X
QtPy	Scientific	N/A	MIT		X	X
pyqode.qt	Scientific	PyQt5	MIT	X		X
QtExt	Film	N/A	N/A		X
python_qt_binding	Robotics	N/A	BSD	X	X	X	X

Also worth mentioning, pyqt4topyqt5; a good starting point for transitioning to Qt.py.

Send us a pull-request with your project here.

Developer Guide

Chat with us

Tests are performed on each aspect of the shim.

Each of these are run under..

Python 2.7
Python 3.4
Python 3.5
Python 3.6

..once for each binding or under a specific binding only.

Each test is run within it’s own isolated process, so as to allow an import to occur independently from other tests. Process isolation is handled via nosepipe.

Tests that are written at module level are run four times - once per binding - whereas tests written under a specific if-statement are run only for this particular binding.

if binding("PyQt4"):
    def test_something_related_to_pyqt4():
        pass

Code convention

Below are some of the conventions that used throughout the Qt.py module and tests.

Etiquette: PEP8
- All code is written in PEP8. It is recommended you use a linter as you work, flake8 and pylinter are both good options. Anaconda if using Sublime is another good option.
Etiquette: Double quotes
- " = yes, ' = no.
Etiquette: Napoleon docstrings
- Any docstrings are made in Google Napoleon format. See Napoleon for details.
Etiquette: Semantic Versioning
- This project follows semantic versioning.
Etiquette: Underscore means private
- Anything prefixed with an underscore means that it is internal to Qt.py and not for public consumption.

Running tests

Due to the nature of multiple bindings and multiple interpreter support, setting up a development environment in which to properly test your contraptions can be challenging. So here is a guide for how to do just that using Docker.

With Docker setup, here’s what you do. Please note this will pull down a ~1 GB image.

cd Qt.py

# Run nosetests (Linux/OSX)
docker run --rm -v $(pwd):/Qt.py -e PYTHON=2.7 fredrikaverpil/qt.py:2018
docker run --rm -v $(pwd):/Qt.py -e PYTHON=3.4 fredrikaverpil/qt.py:2018
docker run --rm -v $(pwd):/Qt.py -e PYTHON=3.5 fredrikaverpil/qt.py:2018
docker run --rm -v $(pwd):/Qt.py -e PYTHON=3.6 fredrikaverpil/qt.py:2018

# Run nosetests (Windows)
docker run --rm -v %CD%:/Qt.py -e PYTHON=2.7 fredrikaverpil/qt.py:2018
docker run --rm -v %CD%:/Qt.py -e PYTHON=3.4 fredrikaverpil/qt.py:2018
docker run --rm -v %CD%:/Qt.py -e PYTHON=3.5 fredrikaverpil/qt.py:2018
docker run --rm -v %CD%:/Qt.py -e PYTHON=3.6 fredrikaverpil/qt.py:2018

# Doctest: test_caveats.test_1_qtgui_qabstractitemmodel_createindex ... ok
# Doctest: test_caveats.test_2_qtgui_qabstractitemmodel_createindex ... ok
# Doctest: test_caveats.test_3_qtcore_qitemselection ... ok
# ...
#
# ----------------------------------------------------------------------
# Ran 21 tests in 7.799s
#
# OK

Now both you and Travis are operating on the same assumptions which means that when the tests pass on your machine, they pass on Travis. And everybody wins!

For details on the Docker image for testing, see DOCKER.md.

See CONTRIBUTING.md for more of the good stuff.