Using "from __future__ import division" in my program, but it isn't loaded with my program

Question

I wrote the following program in Python 2 to do Newton's method computations for my math problem set, and while it works perfectly, for reasons unbeknownst to me, when I initially load it in ipython with %run -i NewtonsMethodMultivariate.py, the Python 3 division is not imported. I know this because after I load my Python program, entering x**(3/4) gives "1". After manually importing the new division, then x**(3/4) remains x**(3/4), as expected. Why is this?

# coding: utf-8
from __future__ import division
from sympy import symbols, Matrix, zeros

x, y = symbols('x y')
X = Matrix([[x],[y]])
tol = 1e-3

def roots(h,a):
  def F(s):
    return h.subs({x: s[0,0], y: s[1,0]})
  def D(s):
    return h.jacobian(X).subs({x: s[0,0], y: s[1,0]})
  if F(a) == zeros((2,1)):
    return a
  else:
    while (F(a)).norm() > tol:
      a = a - ((D(a))**(-1))*F(a)
      print a.evalf(10)

I would use Python 3 to avoid this issue, but my Linux distribution only ships SymPy for Python 2. Thanks to the help anyone can provide.

Also, in case anyone was wondering, I haven't yet generalized this script for nxn Jacobians, and only had to deal with 2x2 in my problem set. Additionally, I'm slicing the 2x2 zero matrix instead of using the command zeros(2,1) because SymPy 0.7.1, installed on my machine, complains that "zeros() takes exactly one argument", though the wiki suggests otherwise. Maybe this command is only for the git version. (Thanks eryksun for correcting my notation, which fixed the issue with the zeros function.)

Solution

Both ipython -i command and run -i in ipython interpreter ignore from __future__ import division in print05.py script.

$ cat print05.py 
from __future__ import division
print(1/2)

In ipython console:

In [1]: print 1/2
0
In [2]: run -i print05.py
0.5
In [3]: division
Out[3]: _Feature((2, 2, 0, 'alpha', 2), (3, 0, 0, 'alpha', 0), 8192)
In [4]: print 1/2
0
In [5]: from __future__ import division
In [6]: print 1/2
0.5

execfile and import produce the same result:

>>> print 1/2
0
>>> execfile('print05.py')
0.5
>>> print 1/2
0
>>> from __future__ import division
>>> print 1/2
0.5

from __future__ import division should not have effect on the source code from different modules, otherwise it would break code in other modules that don't expect its presence.

Here, from __future__ import division has effect:

$ python -i print05.py
0.5
>>> print 1/2
0.5
>>> division
_Feature((2, 2, 0, 'alpha', 2), (3, 0, 0, 'alpha', 0), 8192)

The module name in this case is __main__ both inside print05.py and in the prompt.

Here, the first print 1/2 executes in print05 module, the second one in __main__ module so it also works as expected:

$ python -im print05
0.5
>>> print 1/2
0

And here's something wrong:

$ ipython -i print05.py
0.5
In [1]: division
Out[1]: _Feature((2, 2, 0, 'alpha', 2), (3, 0, 0, 'alpha', 0), 8192)
In [2]: print 1/2
0

The docs for __future__ say:

If an interpreter is started with the -i option, is passed a script name to execute, and the script includes a future statement, it will be in effect in the interactive session started after the script is executed.

So It might be a bug in ipython if its -i option tries to emulate the same python option.

Comments

[Sara Fauzia]

Thanks a lot for the explanation. It's unfortunate, because I inputed my function "h" on the ipython commandline, and it has terms with fractional exponents, but I guess I have to hardcode "h" into the program if I don't want to import future division manually.

[jfs]

@Sara Fauzia: if you defined h() function before ..import division then it will use integer division. You could use edit you_module.py to add functions from ipython console.

[Sara Fauzia]

@J.F. Sebastian So using python interactively, as you pointed out, indeeds does work properly, as I just tested. Is this a bug, what happens with ipython?

[Eryk Sun]

@Sara Fauzia: There's a sympy profile that imports future division.

[Sara Fauzia]

@eryksun But the default profile imports everything, and I thought this is eschewed from in an individual program. (Correct me if I'm mistaken, as I am a newbie.)

[jfs]

@Sara Fauzia: I've updated the answer with a quote from __future__'s documentation.

[Sara Fauzia]

@J.F.Sebastian From iPython's 0.11 documentation: "IPython is meant to work as a drop-in replacement for the standard interactive interpreter. As such, any code which is valid python should execute normally under IPython (cases where this is not true should be reported as bugs)." So, unless I've misunderstood, I should file a bug report.

[jfs]

@Sara Fauzia: Even if the behavior is expected for ipython then at least ipython documentation for -i option should mention that it doesn't confirm to python -i expectations, so it is worth to file a bug report. btw, jython (Python implemented in Java) works as expected and it print 0.5. But ipython is not along pypy (Python implemented in RPython) also prints 0.