Problem

I am working with an arbitrary MxN numpy matrix where each element is a sympy expression, potentially with differing symbols. For the purposes of visualization, let's work with the following matrix test

import sympy as sp import numpy as np a,b,c=sp.symbols('a b c'); test=np.array([[a**2,a+b],[a*c+b,b/c]]); 

When run, test will look like:

In [25]: test Out[25]: array([[a**2, a + b], [a*c + b, b/c]], dtype=object) 

I would like to be able to replace one variable in this array with a number and return a new array that has the same dimensions as test but has replaced the specified variable with it's new value. For example, if I wanted to replace b with 2, the new array should look like:

array([[a**2, a + 2], [a*c + 2, 2/c]], dtype=object) 

Attempt at a Solution

I first tried to use the sympy function subs but I received the following error:

test.subs({b:2}) Traceback (most recent call last): File "<ipython-input-29-a9a04d63af37>", line 1, in <module> test.subs({b:2}) AttributeError: 'numpy.ndarray' object has no attribute 'subs' 

I looked at using lambdify but I believe that it returns a numeric lambda expression which isn't what I want. I need a new symbolic expression, just one that doesn't depend on b anymore. I found some literature in the Wolfram Mathematica documentation under pattern matching that seems to be what I need but I can't figure out how to implement this in Python, or if it's even possible to do so. Any help would be greatly appreciated.

2 Answers

Just use sympy. No need for numpy, at least not for the substitution:

In [117]: import sympy In [118]: a,b,c=sympy.symbols('a b c') In [120]: M=sympy.Matrix([[a**2, a+b],[a*c+b, b/c]]) In [121]: M Out[121]: Matrix([ [ a**2, a + b], [a*c + b, b/c]]) In [123]: M.subs({b:2}) Out[123]: Matrix([ [ a**2, a + 2], [a*c + 2, 2/c]]) 

array is not an expression, therefore it has not the method subs