It does not preserve per-argument type comments about typed-astunparse

[mbdevpl]

Thanks for reporting, will look into that!

[mitar]

(Also see that the docstring does not use """.)

[mbdevpl]

I've verified that the per-argument type comments are indeed not working. I'll fix it.

However, the docstring issue is out of scope for this unparser. There was a similar issue upstream (astunparse): simonpercivall/astunparse#13 however it was closed.

I'm working towards implementing proper docstring unparsing, (as well as arbitrary comment parsing and unparsing) in a different module: https://github.com/mbdevpl/horast however it's not nearly ready yet.

[mbdevpl]

And, there seems to be an issue in typed_ast with parsing per-argument type comments in all cases: python/typed_ast#58 so even if I generate valid code, in some cases I actually cannot verify its validity.

[mitar]

Thanks for looking into this.

[mitar]

(BTW, I used your project here: nvbn/py-backwards#44)

[mbdevpl]

Not minding the issue with typed_ast (hopefully it'll get resolved someday) I added the per-argument type comment unparsing and released https://pypi.python.org/pypi/typed-astunparse/2.0.2

[mitar]

Oho! Thanks!

[mbdevpl]

Oh, and BTW, I just found out that docstrings might be properly stored in AST starting from Python 3.7 - if they will be, then I guess astunparse and typed_astunparse will handle them properly. https://bugs.python.org/issue29463