CODEWITHSUNDEEP
pythonsympy
lhk
lhk

Reputation: 30116

sympy: show all symbolic variables that have been substituted in an expression

Is it possible to reveal all symbolic variables that are involved in a sympy expression?

Here is some sample code that could be potentially quite confusing:

from sympy import *
from sympy.stats import Normal, sample, variance

sigma_eps = symbols('sigma_eps')
eps = Normal("eps", 0, sigma_eps)

sigma_any = eps + 1
sigma_1 = (eps+1).subs({sigma_eps:1})
sigma_10 = (eps+1).subs({sigma_eps:10})

If you print sigma_any, sigma_1 or sigma_10 they all look the same. They will all tell you that their value is eps + 1. But actually, the values are completely different distributions.

Is it possible to reveal what random variables have been used and already substituted for an expression?

Upvotes: 0

Views: 114

Answers (1)

smichr
smichr

Reputation: 19093

It is a general rule (not always followed) that the string form of an expression should be copy and pastable to recreate the object. This case is an exception and an issue at https://github.com/sympy/sympy/issues could be opened.

You can tell they are not the same if you inspect the srepr` forms:

>>> srepr(sigma_1)==srepr(sigma_10)
False

Also, free_symbols does not show what you want, but the difference between atoms before and after substituion will reveal what has been changed:

>>> sigma_any.free_symbols
{eps}
>>> sigma_any.atoms()
{0, 1, eps, sigma_eps}
>>> sigma_1.atoms()
{0, 1, eps}

Note that sigma_eps is missing in the last output.

Upvotes: 1

Related Questions