Add rudimentary support for python files that use type hints

[DragonMinded]

Pretty much as it says on the tin. I was adding type hints to a small python library that used pscript (https://github.com/DragonMinded/firebeatrtc) and when I went to regenerate the js I ran into lack of support for AnnAssign. This is incredibly basic support, since I don't bother to do anything with AnnAssign nodes that don't actually assign a value (like x: int) aside from raising an exception. I also threw in a hack to allow importing from "typing" as a whitelisted import with similar rationale to the existing whitelisted entries.

Do I have to edit other parsers with similar whitespace or run any tests? I didn't run anything aside from seeing that after this change, pscript successfully generated the js file exactly as I expected it to.

[almarklein]

Thanks for this!

I had to look this up, so adding it for the record: the AnnAssign AST node is an assignment + annotation. E.g.

foo = 3   # results in assign node
foo: int = 3  #  results in AnnAssign node

So this change will work with such code, by simply dropping the annotation.

This looks good to me, but it would be great if a test could be added in this file to make sure that this new feature continues to work. See e.g. this line to run the test only on specific Python versions.

[DragonMinded]

Technically the grammar for AnnAssign allows for annotation without assignment. So, the following:

foo = 3  # results in Assign node
foo: int = 3  # results in AnnAssign node
foo: int  # valid python, results in a type annotation for foo and a blank value in the AnnAssign _ast node

For the third case, I didn't know what to do (should we define the variable, should we propagate type information, should we just ignore the node?) so I throw an exception. Its a much bigger conversation than this simple PR was attempting to address. But yes, your analysis is spot on, it supports these types of assignments by just dropping the annotation. I checked the output before and after adding type hints for the code I cared about and the JS was identical so I was happy.

I can definitely add a test or two.

[DragonMinded]

I changed this to RuntimeError because that's what all other errors in this file were.

[almarklein]

Its a much bigger conversation than this simple PR was attempting to address

Indeed. It would be very interesting to use type information in pscript (see #4), but it would be a huge effort :)

Thanks again for this!