from __future__ import *

[pyflakes-bot]

Original report by asmeurer (@asmeurer?) on Launchpad:

---

from __future__ import * should be a syntax error.

$pyflakes test.py
test.py:1: 'from __future__ import *' used; unable to detect undefined names
$python test.py
  File "test.py", line 1
    from __future__ import *
SyntaxError: future feature * is not defined

[pyflakes-bot]

Original comment by jwilk (@jwilk?) on Launchpad:

---

It's a regression; pyflakes 0.4 caught this error correctly:

$ pyflakes test.py 
test.py:1: future feature * is not defined
from __future__ import *

In lintian4python I catch this kind of bugs by compile()ing the AST tree.

[pyflakes-bot]

Original comment by florent.x (@florentx?) on Launchpad:

---

This is most probably related to this change:

0.5.0: Convert pyflakes to use newer _ast infrastructure rather than compiler.

We could detect it differently now, of course.

[pyflakes-bot]

Original comment by asmeurer (@asmeurer?) on Launchpad:

---

This is the second (admittedly esoteric) syntax error I've found that Pyflakes didn't catch (the other was https://bugs.launchpad.net/pyflakes/+bug/1293654).

[pyflakes-bot]

Original comment by jayvdb (@jayvdb?) on Launchpad:

---

It is a trivial one liner to detect '*'.
But how do we feel about importing __future__ and ensuring that only valid items are imported? We dont import other packages, even std lib packages, but __future__ is primarily a syntax issue.

The most common mistake I've seen is 'print_statement' being used instead of 'print_function', but other typos/thinkos/etc also happen.
https://bitbucket.org/logilab/astroid/commits/973613a18cccd896c10d6c7081dc818e5581f467

[pyflakes-bot]

Original comment by bitglue (@bitglue?) on Launchpad:

---

I've personally never needed such a feature since any error with __future__ (like syntax errors) is easily caught by just trying to import the module. But __future__ imports can be validated in a way which is fast, and does not have side effects, then I see no harm.

It might be a little tricky to know what things are valid to import, since that changes over time.

I guess we could just have pyflakes look at __future__.all_future_names to get all the valid things for the version of python running pyflakes. There's some opportunity for user mistake there, similar to how people periodically report bugs about pyflakes missing python 3 features when they are running it with python 2. But maybe that won't be a big deal -- I guess the only way to know is to try it and see.

[pyflakes-bot]

Original comment by asmeurer (@asmeurer?) on Launchpad:

---

from __future__ import invalid_name is also raised as a SyntaxError, so I think pyflakes should catch it (IMHO pyflakes should always warn about SyntaxErrors). It's curious that this didn't come up when I did #20. I guess these errors must be in a different file in the CPython source from the one I was looking at.

[pyflakes-bot]

Original comment by jayvdb (@jayvdb?) on Launchpad:

---

Proposed fix: 76dbf3c

(part of #47 , sitting on top of pending PR 43 #43)