Series is not simplified to final answer in output in sympy 1.7.1

[Shardul555]

I run the code for this expression in the terminal having installed sympy version 1.7.1 but here the coefficients of y**2 and y are not simplified to final answer. Anyone guide me regarding this issue whether it is done with some purpose or this should be fixed?(Sympy live shell is giving simplified output)

>>> from sympy import *
>>> from sympy.abc import y
>>> var('p_0 p_1 p_2 p_3 b_0 b_1 b_2')
(p_0, p_1, p_2, p_3, b_0, b_1, b_2)
>>> Q = (p_0 + (p_1 + (p_2 + p_3/y)/y)/y)/(1 + ((p_3/(b_0*y) + (b_0*p_2 - b_1*p_3)/b_0**2)/y + (b_0**2*p_1 - b_0*b_1*p_2 - p_3*(b_0*b_2 - b_1**2))/b_0**3)/y)
>>> Q.series(y, n=3)
y*(b_0*p_2/p_3 + b_0*(-p_2/p_3 + b_1/b_0)) + y**2*(b_0*p_1/p_3 + b_0*p_2*(-p_2/p_3 + b_1/b_0)/p_3 + b_0*(-p_1/p_3 + (p_2/p_3 - b_1/b_0)**2 + b_1*p_2/(b_0*p_3) + b_2/b_0 - b_1**2/b_0**2)) + b_0 + O(y**3)
>>> simplify(y*(b_0*p_2/p_3 + b_0*(-p_2/p_3 + b_1/b_0)))
b_1*y
>>> simplify(y**2*(b_0*p_1/p_3 + b_0*p_2*(-p_2/p_3 + b_1/b_0)/p_3 + b_0*(-p_1/p_3 + (p_2/p_3 - b_1/b_0)**2 + b_1*p_2/(b_0*p_3) + b_2/b_0 - b_1**2/b_0**2)))
b_2*y**2

Originally posted by @Shardul555 in #9173 (comment)

[oscarbenjamin]

Can you explain what exactly is the issue?

[Shardul555]

This is the ouput obtained on sympy 1.7.1 version,

>>> Q.series(y, n=3)
y*(b_0*p_2/p_3 + b_0*(-p_2/p_3 + b_1/b_0)) + y**2*(b_0*p_1/p_3 + b_0*p_2*(-p_2/p_3 + b_1/b_0)/p_3 + b_0*(-p_1/p_3 + (p_2/p_3 - b_1/b_0)**2 + b_1*p_2/(b_0*p_3) + b_2/b_0 - b_1**2/b_0**2)) + b_0 + O(y**3)

This is the output obtained on sympy live shell (below),

>>> Q.series(y, n=3)
b_2*y**2 + b_1*y + b_0 + O(y**3)

I am not understanding that when on previous versions(used on sympy live shell) simplified output was obtained then why in the current release, output is not simplified to final expression.

[0sidharth]

git bisect says 5dc58b7 caused this change.

[oscarbenjamin]

CC @sachin-4099

Maybe something like ratsimp could be used with the coefficients of the series.

[sachin-4099]

Maybe something like ratsimp could be used with the coefficients of the series.

Yeah, I guess so. Someone can try and find out the most optimal simplification method which can be used to resolve the issue.

[Shardul555]

Ok so I will try to do it. Thanks for suggestion @oscarbenjamin @0sidharth