print statements get incorrectly mangled by the autocall feature under Python 2.6

[Robert Kern]

For example, from Gaël:

(python 2.6, IPython 0.9.1):

In [1]: print 1, 2, 3
------> print(1, 2, 3)
(1, 2, 3)

The problem arises because Python 2.6 provides a __builtins__.print() function for migration to Python 3.x. However, statements like "print(1,2,3)" still get parsed as a print of a tuple rather than printing threeobjects with separating spaces. A __future__ import can enable the 3.x syntax.

http://docs.python.org/whatsnew/2.6.html#pep-3105-print-as-a-function

A fix is a bit tricky, since autocall should transform such lines if the __future__ import has been made. This might be difficult to check, though.

[Beni Cherniavsky (cben)]

Still happens as of 0.10 (shipped in Ubuntu Karmic).

A side effect of this bug is breaking completion under emacs when autocall is enabled.
(Emacs parses the "-------> print(...)" line as if it was the output.)
Attaching patch to ipython.el that fixes this (and will also work into 3.0 in the future).

[Fernando Perez (fdo.perez)]

Mmh:

- Beni, I've applied your patch to 0.10.1 and trunk-dev, thanks.

- As for the real problem with print, I'm not sure what to do: I've spent some quality time with the Python sources, trying to write some ctypes-based utilities to fish into the compiler state struct (see compiler.c and future.c in the Python tree), but no success yet. It's not the type of code I'm familiar with so it's likely to be slow/hard going, and I don't have the time for more on this now. Ultimately I'm leaning towards suggesting that people simply put in their config file

from __future__ import print_function

and be done with it! That way they start getting used to print as a function everywhere, in preparation for 3.0, they can learn the syntax of the function, and IPython will happily put the parens in there for them.

It seems like a better way to address this than to sink days of work into an obscure tool to spelunk into the compiler internals at runtime to fish a __future__ flag out...

Thoughts?

If people agree on this, closing the bug will just require adding the text above to the docs/faq.

[Fernando Perez (fdo.perez)]

This is really an annoying bug, but I'm a bit lost on how to approach it. For example, we could hardcode a check on 'print' being on the line to avoid triggering the autocall behavior, but that would break legitimate use of autocall for users of print as a function...

Any suggestions welcome.

[Robert Kern (robert-kern)]

In LineInfo.ofind(), when ifun is "print", check if the compiler has recorded the "from __future__ import print_function":

import __future__

def ofind(self, ip):
    ...
    if self.ifunc == 'print':
        if not ip.shell.compile.compiler.flags & __future__.CO_FUTURE_PRINT_FUNCTION:
            return {'found': False, 'obj': None, ...}
    ...

[Fernando Perez (fdo.perez)]

On Thu, Apr 22, 2010 at 11:46 AM, Robert Kern
<email address hidden> wrote:
> In LineInfo.ofind(), when ifun is "print", check if the compiler has
> recorded the "from __future__ import print_function":
>
> import __future__
>
> def ofind(self, ip):
>    ...
>    if self.ifunc == 'print':
>        if not ip.shell.compile.compiler.flags & __future__.CO_FUTURE_PRINT_FUNCTION:
>            return {'found': False, 'obj': None, ...}

Aha thanks! I looked around for this flag by reading the compiler
source code, but I failed to find this.

OK, with this info in hand, we can fix this damned bug quickly, which
is awfully irksome.

Cheers,

f

[Fernando Perez (fdo.perez)]

Thanks to Robert's tip, this is now fixed:

- For 0.10.1: http://bazaar.launchpad.net/~ipython-dev/ipython/0.10.1/revision/1224
- For trunk: http://bazaar.launchpad.net/~ipython-dev/ipython/trunk/revision/1237

[Fernando Perez (fdo.perez)]

Fixed 2.5 and below incompatibitliy in r1225 for 0.10.1 and r1238 for trunk.

[Brian Murray (brian-murray)]

Fernando - So is creating a patch out of revision 1224 the only bit we need to fix this in Ubuntu for Lucid?

[Fernando Perez (fdo.perez)]

Hi Brian,

On Fri, Apr 23, 2010 at 10:19 AM, Brian Murray <email address hidden> wrote:
> Fernando - So is creating a patch out of revision 1224 the only bit we
> need to fix this in Ubuntu for Lucid?

Yes! I hope to release 0.10.1 in the near future officially, so you
can update to that later on.

Cheers,

f

[Launchpad Janitor (janitor)]

This bug was fixed in the package ipython - 0.10-1ubuntu1

---------------
ipython (0.10-1ubuntu1) lucid; urgency=low

  * cherrypicked rev 1224 and 1225 from lp:ipython/0.10.1 to fix (LP: #414967)
    """Fix autocall misfiring on print for python 2.5 and newer."""
 -- Markus Korn <email address hidden> Fri, 23 Apr 2010 20:16:11 +0200