CODEWITHSUNDEEP
pythonarraysmatlabnumpy
andreasdr
andreasdr

Reputation: 4054

When a string is used as an input into numpy.mat(), certain characters are ignored

Coming to scipy/numpy from MATLAB, I regard the input syntax to numpy.array as overly parenthesis-heavy, e.g.

import numpy as np
import math
np.array([[1, 2], [math.sqrt(3), 4]])

The following MATLAB-like syntax therefore appeals to me:

np.array(np.mat("1, 2; math.sqrt(3), 4"))

However, this particular example apparently results in the array

array([[ 1. ,  2. ],
       [ 0.3,  4. ]])

It seems that numpy.mat() ignores some characters in its input string: in this case "math" and "sqrt". Furthermore, it seems that it is not possible to use variables in the input string. Why is that? And is there a way to use a MATLAB-like, paranthesis-thrifty input method that isn't "broken" in this way?

Upvotes: 0

Views: 203

Answers (1)

HYRY
HYRY

Reputation: 97281

You can use eval() and sys._getframe() to write a function, here is an example:

import numpy as np

def m(s):
    import sys
    frame = sys._getframe(1)
    return np.array([eval(item.strip(), frame.f_globals, frame.f_locals) 
                         for item in s.split(";")])

def f():
    a = 1
    x = 3.14
    print m("1,2,3;np.sin(a),np.sqrt(2),x")

f()

Upvotes: 3

Related Questions