New problem with substitution and paranthesis handling

[ttxtea]

Something has changed in the way substitutions work in sympy version 1.1.1. on python3.4.2

In [3]: sympy.S('R*F').subs({'R':'8314.4*mjoule/(mole*kelvin)','F':'96484.6*(coulomb/mole)'})
Out[3]: 8314.4*mjoule/(mole*kelvin)*96484.6*(coulomb/mole)

In [4]: sympy.S('R/F').subs({'R':'8314.4*mjoule/(mole*kelvin)','F':'96484.6*(coulomb/mole)'})
Out[4]: 8314.4*mjoule/(mole*kelvin)/96484.6*(coulomb/mole)

with sympy 1.0 on python2.7.9 this yielded

In [3]: sympy.S('R*F').subs({'R':'8314.4*mjoule/(mole*kelvin)','F':'96484.6*(coulomb/mole)'})
Out[3]: 802211558.24*coulomb*mjoule/(kelvin*mole**2)

In [4]: sympy.S('R/F').subs({'R':'8314.4*mjoule/(mole*kelvin)','F':'96484.6*(coulomb/mole)'})
Out[4]: 0.0861733375067109*mjoule/(coulomb*kelvin)

The update seems to have wronged something.

[jksuom]

It looks like SymPy is doing exactly what is prescibed:

>>> srepr(S('R*F'))
"Mul(Symbol('F'), Symbol('R'))"
>>> srepr(S('R*F').subs({'R':'8314.4*mjoule/(mole*kelvin)','F':'96484.6*(coulomb/mole)'}))
"Mul(Symbol('8314.4*mjoule/(mole*kelvin)'), Symbol('96484.6*(coulomb/mole)'))"

Before substitution, S('R*F') is a Mul expression with two symbols as arguments. The names of the symbols are Python strings and the substitution will replace them with new (long) strings. This should probably not be regarded as an error.

The values in the dictionary, '8314.4*mjoule/(mole*kelvin)' and '96484.6*(coulomb/mole)', are Python strings. To obtain SymPy objects, the quotes should be omitted.

[ttxtea]

But the division is wrong, isn't it?? It should be

/(coulomb*mole)

not the same as the multiplication. The srepr is fine

sympy.srepr(sympy.S('R/F').subs({'R':'8314.4*mjoule/(mole*kelvin)','F':'96484.6*(coulom
   ...: b/mole)'}))
Out[4]: "Mul(Symbol('8314.4*mjoule/(mole*kelvin)'), Pow(Symbol('96484.6*(coulomb/mole)'), Integer(-1)))"

but souldn't Pow((coulomb/mole),Integer(-1)) should produce mole/coulomb.

[ttxtea]

Sorry for the messy example with the "Nernst equation". Maybe this is clearer: In sympy 1.1.1.

In [9]: sympy.S('1/a').subs({'a':'2*d/c'})
Out[9]: 1/2*d/c

but

In [10]: sympy.srepr(sympy.S('1/a').subs({'a':'2*d/c'}))
Out[10]: "Pow(Symbol('2*d/c'), Integer(-1))"

In sympy 1.0

In [3]: sympy.S('1/(a)').subs({'a':'2*d/c'})
Out[3]: c/(2*d)

and

In [4]: sympy.srepr(sympy.S('1/(a)').subs({'a':'2*d/c'}))
Out[4]: "Mul(Rational(1, 2), Symbol('c'), Pow(Symbol('d'), Integer(-1)))"

[ttxtea]

Note that it is probably the repr() function that causes the problem. In isympy, where the output is based on srepr it is fine. So maybe the bug is somewhere in the __repr__ function?

A workaround is to use additional parenthesis. like

sympy.S('1/a').subs({'a':'(2*d/c)'})

But the backwards compatibility to sympy 1.0 is not given and projects like brian2 have problems with old model definitions.