Import the file and use the only function from it to make a correct typing hint from runtime data.

It supports basic types, re.Match, and re.Pattern. For the rest, the behavior is a shot in the dark, depending on the imports in the file.

from typograph import name_type

a_list = [1, 2, None]
print(f'a_list: {name_type(a_list)}')
a_tuple = (1, 2, [3, {1: 1, 2: 'a', 3: [3]}])
print(f'a_tuple: {name_type(a_tuple)}')
a_dict = {1: 1, False: 'a', 3: {1: 1, 2: 'a', 3: [3]}}
print(f'a_dict: {name_type(a_dict)}')

Provide the Python code filenames as its command line arguments to get new files with typing annotations removed. The result promises to be more compatible to Python 3.4.

The new files are of the same names, but stored in a directory specified with -o or --out parameter (“py34” by default) near the original files.

The script mostly works, but not everything is implemented:

• It leaves @ and := operators unchanged, although the former was introduced in Python 3.5 and the latter in 3.8.
• It leaves async statements and await operator unchanged, although they were introduced in Python 3.5.
• It leaves the subclasses of typing classes unresolved, unless they can be substituted by classes from collections.
• It doesn't turn a match statement into a sequence of if, elif, and else, but only removes type annotations in the cases.
• It omits nested formatting options in f-strings.

To see the result, specify -p or --print parameter. Beware, though, that all the output comes with not separator between the content of different files.

python py34.py /usr/lib64/python3.11/tomllib/*.py

It requires QtPy to work.

Until the Python Interface files shipped with PyQt*/PySide* are good enough, the script re-builds them, too.

PyCharm provides awful stubs for Qt for Python (PySide6, PyQt6, PyQt5, PySide2, PyQt4, and PySide). There, the arguments count and the defaults may be wrong, static functions have an inappropriate self argument, and there are no type hints. It looks bizarre, for the correct definitions of the functions are in the docstrings. Finally, the stubs consider Qt Signals to be functions with arbitrary arguments.

The script looks through the stubs located in f'~/.cache/JetBrains/PyCharm*/python_stubs/*/{qtpy.API_NAME}' and in f'~/AppData/Local/JetBrains/PyCharm*/python_stubs/*/{qtpy.API_NAME}'. I don't know the location of the stubs used in macOS.

QtPy has a beautiful way of specifying the version of the Qt bindings in use, while regularizing the differences between the bindings. If you need other bindings than what QtPy defaults to, there are environment variables to alter the behavior.

The script does everything on its own, takes to command line arguments.

python type_qt.py