sympy - numpy conflict?

[Hans]

At first: sympy is a great tool!
I'm a newbie and I tried the following code:

from sympy import *
x = symbols('x')
f = exp(-x)*sin(2*x)
e = diff(f,x)
print e

All worked well!
But adding a numpy import
an error message appeared:

from sympy import *
from numpy import *
x = symbols('x')
f = exp(-x)*sin(2*x)
e = diff(f,x)
print e

Traceback (most recent call last):
File "C:/Python25/discussion.py", line 6, in <module>
f = exp(-x)*sin(2*x)
AttributeError: exp

What happened? How could it be avoided?

Thanks!
Hans

[Ondrej Certik]

Hi Hans!

This is because you are mixing exp from numpy and from sympy. NumPy's
exp doesn't accept SymPy's symbols.

> How could it be avoided?

Use explicit imports, so that you see immediatelly what gets imported.
Or change the sympy and numpy import lines.

Ondrej

[Gael Varoquaux]

On Sun, Jul 20, 2008 at 09:49:30AM -0700, Hans wrote:
> from sympy import *
> from numpy import *
> x = symbols('x')
> f = exp(-x)*sin(2*x)
> e = diff(f,x)
> print e

> Traceback (most recent call last):
> File "C:/Python25/discussion.py", line 6, in <module>
> f = exp(-x)*sin(2*x)
> AttributeError: exp

> What happened?

When importing * from sympy you import "exp", and thus exp is defined as
something that knows about sympy symbols. Whereas when you import * from
numpy you also import "exp", but it is numpy's exp, that is an exp that
knows about arrays. You have overrided sympy's exp with numpy's exp.

> How could it be avoided?

It shouldn't be avoided. The only way to make numpy's exp and sympy's exp
to play well together would introduce cross-dependencies between numpy
and sympy, and we don't want to do that. What you should do is not mix
the namespaces: two differnet functions need to have a different name. In
Python you could do this like this:

from sympy import *
import numpy as np

x = symbols('x')
f = exp(-x)*sin(2*x)
e = diff(f,x)
print e

And remember that numpy function, such as exp, should be called using
"np.exp", rather than "exp".

HTH,

Gaël