F-strings!!!

[refi64]

Well, this was a bit more complicated than I thought...

[refi64]

@ddfisher Bu-uuuuump?

[ddfisher]

Sorry for being slow to review this! I was hoping to get to it this week but had some other things take precedence. It's at the top of my list for next week!

[ddfisher]

This PR has made me realize that the typed_ast update process needs to be nailed down a bit better -- I should have something written up about that shortly, but this PR is blocking on that happening, sorry!

[JelleZijlstra]

Any updates here? It would be great to have f-strings supported in mypy before CPython 3.6 final comes out. I can help if necessary.

[mbdevpl]

Python 3.6 is here, and...

Traceback (most recent call last):
  File "/home/user/Projects/python/typed-astunparse/test/test_unparse.py", line 80, in test_files
    tree = typed_ast.ast35.parse(source=original_code, filename=path)
  File "/home/user/.local/lib/python3.6/site-packages/typed_ast/ast35.py", line 46, in parse
    return _ast35._parse(source, filename, mode)
  File "/home/user/.local/lib/python3.6/functools.py", line 286
    args.extend(f"{k}={v!r}" for (k, v) in self.keywords.items())
                           ^
SyntaxError: invalid syntax

What can I do to help?

[yan12125]

@mbdevpl I guess you mix 3.5 and 3.6?

[mbdevpl]

@yan12125 this PR contains changes to ast35 which actually make it a kind of "ast36" (or some intermediate between 3.5 and 3.6), so after merging this PR the error should disappear... As mentioned in #18 , ast35 might eventually be renamed to ast3.

So, am I really mixing 3.5 and 3.6 here?

[yan12125]

functools.py is part of standard Python library and is unrelated to typed_ast. A common reason for syntax error in this file is that you're running the Python 3.5 interpreter and using Python 3.6's library.

How did you run your program?

[mbdevpl]

So, first of all, I compiled Python 3.6.0 from source, and installed using make altinstall to ~/.local - just like I did with Python 3.5.2 some time ago. And then I ran python3.6 setup.py test. With python3.5 my tests pass just like they used to. The test which fails there is about roundtrip parsing and unparsing typed Python ASTs (using typed_ast and typed_astunparse modules).

I think the problem at hand is not that I mix 3.5 and 3.6, but that this PR is still not merged, hence this error. And thus my question, what can I do to help in bringing f-strings to typed_ast?

[gvanrossum]

Patience please. See python/mypy#2622

[mbdevpl]

@gvanrossum Thank you for that reference. Ok, I'll just wait then.

[ddfisher]

Fstrings are now supported. Sorry for the long wait! In the process of reviewing this PR, I realized that porting new Python features onto typed_ast was not the best way to go about things, because pull requests like this are difficult to review, and because typed_ast was turning into a mishmash of Python 3.5 and Python 3.6 syntax (both in parsing and in the produced AST). In lieu of merging this PR, I've instead rebuilt typed_ast on top of the Python 3.6 ast module. I've also formalized the update process for the future, so there shouldn't be such long waits again (see here).

I appreciate your working in putting up this PR, @kirbyfan64 -- sorry it's not getting landed.

[santagada]

no worries, thanks for your work, I can now run mypy again on all my code \o/

[refi64]

I'm just happy that f-strings finally work!! :D