Multiline statements not upgraded for f-strings (Py 3.6+) #178

Activity

1. asottile commented on Jul 5, 2019

   

   asottile

   on Jul 5, 2019

   Owner

   More actions

   Please search the issue tracker next time! #55 #49
2. asottile

   closed this as completedon Jul 5, 2019
3. taldcroft commented on Jul 5, 2019

   

   taldcroft

   on Jul 5, 2019

   Author

   More actions

   Sorry I assumed this would be an open issue and didn't see it. Given that this is the third repeat of the issue maybe some less timid option might be handy? It's easy enough for users to fix line length post facto with automated tools.
4. asottile commented on Jul 5, 2019

   

   asottile

   on Jul 5, 2019

   Owner

   More actions

   PRs welcome :)
5. taldcroft commented on Jul 5, 2019

   

   taldcroft

   on Jul 5, 2019

   Last edited by asottile

   Author

   More actions

   @asottile - this package is generally way beyond my level, but I had a look anyway and did spot the nicely labeled timid restrictions. A couple of them look relevant, in particular the bit about not making an f-string longer. But no joy, commenting out the break did not help, and the print statements show for my example code that these clauses are never being reached:

   diff --git a/pyupgrade.py b/pyupgrade.py
   index dd5f23c..5120bd6 100644
   --- a/pyupgrade.py
   +++ b/pyupgrade.py
   @@ -1636,7 +1636,8 @@ def _fix_py3_plus(contents_text):  # type: (str) -> str
                while tokens[j].src != node.attr:
                    # timid: if we see a parenthesis here, skip it
                    if tokens[j].src == ')':
   -                    return
   +                    print('Not timid!!')
   +                    # return
                    j += 1
                tokens[i:j + 1] = [new_token]
    
   @@ -1812,7 +1813,8 @@ class FindSimpleFormats(ast.NodeVisitor):
                        candidate, _, _ = name.partition('.')
                        # timid: could make the f-string longer
                        if candidate and candidate in seen:
   -                        break
   +                        print('Not timid!')
   +                        # break
                        # timid: bracketed
                        elif '[' in name:
                            break

   Any hints? pyupgrade is an awesome package, so it would be nice if it could go the full distance for the hundreds of files that need fixing in astropy.
6. asottile commented on Jul 5, 2019

   

   asottile

   on Jul 5, 2019

   Owner

   More actions

   there's a few things that aren't yet handled:

   • implicit string joins
   • parenthesized strings that are then .format(...)ed
   • repeated names in format strings

   The first two need some new code written for them (similar to the stuff that's happening for the py27 format strings, basically a state machine to collect all the separate STRING tokens that are implicit-joined). I ~eventually want something like this in tokenize-rt (since there's currently code for this both here and in future-fstrings that does the same parsing).

   The second should (?) have been covered by the second part of the patch