Sympy: Unevaluated dictionary in SymPy

[ritesh99rakesh]

Currently for Binomial distribution under sympy/stats/frv_types.py we have the following code for returning density as dict:

def dict(self):
        n, p, succ, fail = self.n, self.p, self.succ, self.fail
        n = as_int(n)
        return dict((k*succ + (n - k)*fail,
                binomial(n, k) * p**k * (1 - p)**(n - k)) for k in range(0, n + 1))

This forces n to be an Integer. Is there a way to construct a dictionary with symbolic n?

[oscarbenjamin]

If I understand correctly this dict is never exposed directly to users and is used in two different ways:

1. Sometimes it is iterated over for v, p in density.items().
2. Other times it is indexed density[v].

For the second of those it would be possible to make a Python object that supports __getitem__ without being a concrete dict. For example, a density "dict" for the geometric distribution (which has an infinite domain) can be created with:

from sympy import *

class GeometricProbabilityDict:
    def __init__(self, p):
        self.p = p
    def __getitem__(self, n):
        return p*(1-p)**n

p = Symbol('p')
gpd = GeometricProbabilityDict(p)

for i in range(10):
    print(gpd[i])

This wouldn't work if you tried to iterate over it though since that would lead to an infinite loop.

Changes would be needed in FinitePSpace to support something like this.

[czgdp1807]

There is a problem with ArrayComprehension. Dict works fine with list. It can be corrected though. If you are not working on it then I will correct it asap.

[ritesh99rakesh]

Thank you @czgdp1807. Let me give it a try and if I need any help, I will ask you.

[oscarbenjamin]

The issue as described here is misguided. A symbolic Dict will not solve the actual problem since it won't be usable in the ways that FinitePSpace expects.

I suggest to close this issue and open a new one with the title "FinitePSpace doesn't support symbolically sized sample spaces" and the motivating example:

In [21]: from sympy.stats import Binomial, E

In [22]: from sympy import symbols

In [23]: n, p = symbols('n, p')

In [24]: X = Binomial('X', n, p)

In [25]: E(X)
...
ValueError: n is not an integer

[czgdp1807]

New issue has been created. Please close this one.