Support module:entry_point in addition to script.py on command line

[callegar]

Is your feature request related to a problem? Please describe.

I would like to be able to use pyinstaller on a module that has an entry point without the need to create a wrapper script.

Describe the solution you'd like
I would like to be able to say pyinstaller -F mymodule:main -n mytool rather than pyinstaller mytool.py which implies the need to create mytool.py.

Describe alternatives you've considered
N/A

Additional context
N/A

[bwoodsend]

Doesn't mymodule have a __main__.py submodule which runs the entry-point function? If it does, just run PyInstaller on that. If it doesn't then add it so that people who don't have their scripts/bin directory in PATH are able to run the entry-point (via python -m mymodule).

[callegar]

It doesn't, because the package has multiple entry points. Furthermore, I do not like very much the idea of adding __main__.py, because I would not like to touch the package and because the package might need to get a __main__.py in the future, exposing a "default" entry point different from the one I need now.

The fact is that setuptool, the pyproject standardization, poetry, setuptools, etc., all now tend to support the idea of packages with (multiple) entry points declarable via pyproject.toml (e.g., via a [tool.poetry.scripts] section in poetry, or a more standard [project.scripts] entry otherwise). In all cases the syntax is module:callable (or package.module:callable, if the module is inside a package). Would be great to have this syntax supported in pyinstaller too, for homogeneity, and for being able to use pyinstaller on packages one would prefer not to modify.

[rokm]

On one hand, this seems like a non-trivial change to save the user from creating a two-line entry-point script for their frozen application.

On the other hand, having this would allow us to get rid of that horrible traversal of parent directories when entry-point .py file is in a package (has __init__.py next to it), and just raise an error instead.

So while this is not getting an outright rejection from me, it is, as far as I am concerned, quite low on the priority list.