sympy AttributeError: 'Pow' object has no attribute 'sin'

Question

I have read this SO post which says namespace conflict is one reason for this error. I am falling to this error frequently. So, I'd like to learn what exactly is happening here? What is expected by the library?

EDIT: fun = lambda x: 4*x*(np.sin(x**2) - 3)*np.cos(x**2) comes from a test case, so practically I am bound to use it as function 'fun'. Sorry for missing that information. Kindly discuss respecting this constraint.

EDIT2: This is an error reproducing code, not the full script. Task is to calculate differentiation of an input function that can evaluate numpy arrays by using a forward difference approximation with a perturbation ∆=10 −8.

Code:

import sympy
import numpy as np 

# TESTING...
x = sympy.Symbol('x')
fun = lambda x:  4*x*(np.sin(x**2) - 3)*np.cos(x**2)  
print fun
h = 10e-8  #perturbation
print fun(x)
print fun(x+h)
df = (fun(x+h) - fun(x)) / h
print "diff is:", df

Error:

<function <lambda> at 0x000000001068E2E8>
Traceback (most recent call last):

  File "<ipython-input-75-0582d8ebb11b>", line 1, in <module>
    runfile('D:/test_hw3.py', wdir='D:')

  File "D:\anaconda\lib\site-packages\spyder\utils\site\sitecustomize.py", line 866, in runfile
    execfile(filename, namespace)

  File "D:\anaconda\lib\site-packages\spyder\utils\site\sitecustomize.py", line 87, in execfile
    exec(compile(scripttext, filename, 'exec'), glob, loc)

  File "D:/test_hw3.py", line 23, in <module>
    print fun(x)

  File "D:/test_hw3.py", line 20, in <lambda>
    fun = lambda x:  4*x*(np.sin(x**2) - 3)*np.cos(x**2)
AttributeError: 'Pow' object has no attribute 'sin'

kennytm · Accepted Answer · 2017-04-09T14:59:28.600

2

You are supposed to use sympy.sin/cos instead of np.sin/cos. Numpy does not know how to work with sympy expressions.

fun = lambda x: 4 * x * (sympy.sin(x**2) - 3) * sumpy.cos(x**2)

On the other hand, if you must keep np.sin, then you should not send a sympy variable to fun. Instead, df should become a function itself.

fun = lambda x:  4*x*(np.sin(x**2) - 3)*np.cos(x**2)  
h = 1e-8

df = lambda x: (fun(x+h) - fun(x)) / h

# usage:
df( np.array([1.0, 2.0, 3.0, 4.0]) )

BTW, sympy already has a diff function to compute the derivative.

df = sympy.diff(fun(x), x)

edited Apr 09 '17 at 14:59

answered Apr 09 '17 at 14:38

kennytm

510,854
105
1,084
1,005

Can you please review your answer against my edits. Thanks. – devautor Apr 09 '17 at 14:46
@devautor Are you really supposed to use sympy? – kennytm Apr 09 '17 at 14:49
Yes, I am. Do you have alternatives? [Though my requirements won't let me pursue them for this task] – devautor Apr 09 '17 at 14:51
@devautor What is expected from `df`? – kennytm Apr 09 '17 at 14:51
Task is to calculate differentiation of an input function that can evaluate numpy arrays by using a forward difference approximation with a perturbation ∆=10 −8. – devautor Apr 09 '17 at 14:53
@devautor That doesn't sound like sympy needs to be involved – kennytm Apr 09 '17 at 14:54
Then, how should I get an expression/function for differentiation? It should be available to be evaluated on any given 'x'. – devautor Apr 09 '17 at 14:59

score 0 · Answer 2 · answered Apr 09 '17 at 14:39

Function fun uses np.sin and np.cos, which are numpy functions that accept as arguments only data types recognized by numpy. When you do fun(x) you are giving as argument to numpy functions a sympy object, which numpy does not recognize, raising this "cryptic" error message.

Note that when you define fun using the statement lambda x:..., x is treated in the definition as a dummy variable name for a valid data type (even though you have defined x above as a sympy symbol).

sympy AttributeError: 'Pow' object has no attribute 'sin'

2 Answers2

Linked