search

dflook / python-minifier

Transform Python source code into its most compact representation
MIT License
558 stars 41 forks source link

`--remove-literal-statements` doesn't work #24

Closed phorward closed 3 years ago

phorward commented 3 years ago

Hi!

First of all, thank you very much for writing this software. It is very useful to decrease the size of Python's standard library when exposing it for web-apps running with Pyodide, a full CPython 3.8 compiled to web-assembly and running in the browser.

Anyway, it seems I found a bug, because the --remove-literal-statements does not seem to have any effect on the provided sources. These two calls return exactly the same result. Docstrings are everywhere.

$ pyminify  0.py >a.txt 
$ pyminify --remove-literal-statements 0.py  >b.txt

Attached are the results I got. I'm on pyminify 2.4.1 as installed from PyPI, the used Python version for execution is 3.9.1, but I also tested it with 3.8.2 with the same result.

0.py => this is the __init.py__ of Python 3.8.2's collections standard library package a.txt b.txt

phorward commented 3 years ago

Supplement: After a few further tests, I found out that the docstring from other modules seems to be removed properly - except in the case with the 0.txt above. (__init__.py from cpython/Lib/collections)

dflook commented 3 years ago

Hi @phorward, thanks for creating an issue.

It looks like this is because the problem modules use the global __doc__ name or .__doc__ attributes, which currently disable all literal removal. This doesn't always work correctly, but cautiously leaves docstrings in if unsure.

I do plan to make this more flexible, see #11.

phorward commented 3 years ago

@dflook Thanks for the quick answer! Yes, it seems that this module uses __doc__, but with a different context, so it shouldn't break. But it's OK for me, and I'm looking forward so that #11 is getting fixed.