Things to be Aware Of

Crashes On Exit

When the Python interpreter leaves a scope (for example when it returns from a function) it will potentially garbage collect all objects local to that scope. The order in which it is done is, in effect, random. Theoretically this can cause problems because it may mean that the C++ destructors of any wrapped Qt instances are called in an order that Qt isn’t expecting and may result in a crash.

However, in practice, this is only likely to be a problem when the application is terminating. For example, it is preferable that any QApplication instance is destroyed only after all widgets are destroyed.

As a way of mitigating this possiblity PyQt5 ensures that the destructors of any module level objects are not invoked when the application terminates. This means that code that follows the pattern below is unlikely to crash on exit:

if __name__ == '__main__':
    app = QApplication(sys.argv)

    w = QWidget()
    w.show()

    app.exec()

Another common pattern (and one that is required when using setuptool entry points) is that the above code in placed in a separate function, typically called main(). This then causes a problem when the function returns as the destructors of the QApplication and QWidget instances may be invoked in the wrong order. To minimise the chances of this happening, the following pattern is recommended:

app = None

def main():
    global app
    app = QApplication(sys.argv)

    w = QWidget()
    w.show()

    app.exec()

if __name__ == '__main__':
    main()

The QWidget destructor may be invoked when main() returns but the module level reference to the QApplication instance will prevent its destructor being invoked at all.

Keyword Arguments

PyQt5 supports the use of keyword arguments for optional arguments. Although the PyQt5 and Qt documentation may indicate that an argument has a particular name, you may find that PyQt5 actually uses a different name. This is because the name of an argument is not part of the Qt API and there is some inconsistency in the way that similar arguments are named. Different versions of Qt may use a different name for an argument which wouldn’t affect the C++ API but would break the Python API.

The docstrings that PyQt5 generates for all classes, functions and methods will contain the correct argument names. In a future version of PyQt5 the documentation will also be guaranteed to contain the correct argument names.

Python Strings, Qt Strings and Unicode

Qt uses the QString class to represent Unicode strings, and the QByteArray to represent byte arrays or strings. In Python v3 the corresponding native object types are str and bytes. In Python v2 the corresponding native object types are unicode and str.

PyQt5 does its best to automatically convert between objects of the various types. Explicit conversions can be easily made where necessary.

In some cases PyQt5 will not perform automatic conversions where it is necessary to distinguish between different overloaded methods.

For Python v3 the following conversions are done by default.

  • If Qt expects a char * (or a const version) then PyQt5 will accept a str that contains only ASCII characters, a bytes, a QByteArray, or a Python object that implements the buffer protocol.
  • If Qt expects a char (or a const version) then PyQt5 will accept the same types as for char * and also require that a single character is provided.
  • If Qt expects a signed char * or an unsigned char * (or a const version) then PyQt5 will accept a bytes.
  • If Qt expects a signed char or an unsigned char (or a const version) then PyQt5 will accept a bytes of length 1.
  • If Qt expects a QString then PyQt5 will accept a str, a bytes that contains only ASCII characters, a QByteArray or None.
  • If Qt expects a QByteArray then PyQt5 will also accept a bytes.
  • If Qt expects a QByteArray then PyQt5 will also accept a str that contains only Latin-1 characters.

For Python v2 the following conversions are done by default.

  • If Qt expects a char *, signed char * or an unsigned char * (or a const version) then PyQt5 will accept a unicode that contains only ASCII characters, a str, a QByteArray, or a Python object that implements the buffer protocol.
  • If Qt expects a char, signed char or an unsigned char (or a const version) then PyQt5 will accept the same types as for char *, signed char * and unsigned char * and also require that a single character is provided.
  • If Qt expects a QString then PyQt5 will accept a unicode, a str that contains only ASCII characters, a QByteArray or None.
  • If Qt expects a QByteArray then PyQt5 will accept a str.
  • If Qt expects a QByteArray then PyQt5 will accept a unicode that contains only Latin-1 characters.

Note that the different behaviour between Python v2 and v3 is due to v3’s reduced support for the buffer protocol.

Historically QString distinguishes between empty strings and null strings. Current versions of Qt treat null strings as empty strings but there may be other C++ code that PyQt5 applications call that maintains the distinction. Consequently PyQt5 will convert None to a null QString. The reverse conversion is not done and both a null and an empty QString will be converted to an empty (i.e. zero length) Python string.

Garbage Collection

C++ does not garbage collect unreferenced class instances, whereas Python does. In the following C++ fragment both colours exist even though the first can no longer be referenced from within the program:

col = new QColor();
col = new QColor();

In the corresponding Python fragment, the first colour is destroyed when the second is assigned to col:

col = QColor()
col = QColor()

In Python, each colour must be assigned to different names. Typically this is done within class definitions, so the code fragment would be something like:

self.col1 = QColor()
self.col2 = QColor()

Sometimes a Qt class instance will maintain a pointer to another instance and will eventually call the destructor of that second instance. The most common example is that a QObject (and any of its sub-classes) keeps pointers to its children and will automatically call their destructors. In these cases, the corresponding Python object will also keep a reference to the corresponding child objects.

So, in the following Python fragment, the first QLabel is not destroyed when the second is assigned to lab because the parent QWidget still has a reference to it:

parent = QWidget()
lab = QLabel("First label", parent)
lab = QLabel("Second label", parent)

Multiple Inheritance

It is not possible to define a new Python class that sub-classes from more than one Qt class. The exception is classes specifically intended to act as mixin classes such as those (like QQmlParserStatus) that implement Qt interfaces.

Access to Protected Member Functions

When an instance of a C++ class is not created from Python it is not possible to access the protected member functions of that instance. Attempts to do so will raise a Python exception. Also, any Python methods corresponding to the instance’s virtual member functions will never be called.

None and NULL

Throughout PyQt5, the None value can be specified wherever NULL is acceptable to the underlying C++ code.

Equally, NULL is converted to None whenever it is returned by the underlying C++ code.

Support for void *

PyQt5 (actually SIP) represents void * values as objects of type sip.voidptr. Such values are often used to pass the addresses of external objects between different Python modules. To make this easier, a Python integer (or anything that Python can convert to an integer) can be used whenever a sip.voidptr is expected.

A sip.voidptr may be converted to a Python integer by using the int() builtin function.

A sip.voidptr may be converted to a Python string by using its asstring() method. The asstring() method takes an optional integer argument which is the length of the data in bytes.

A sip.voidptr may also be given a size (ie. the size of the block of memory that is pointed to) by calling its setsize() method. If it has a size then it is also able to support Python’s buffer protocol and behaves like a Python memoryview object so that the block of memory can be treated as a mutable list of bytes. It also means that the Python struct module can be used to unpack and pack binary data structures in memory, memory mapped files or shared memory.