Mul._eval_subs broken in a number of ways

[Andreas Klöckner]

Hi all,

(Before I start--one question: Google Code bug tracker or Github bug tracker--which is the right one?)

I'm writing because I could use some expert help sorting through my findings in this bug:

https://code.google.com/p/sympy/issues/detail?id=3262

I think I've diagnosed most of the things that *cause* the breakage, but the fix isn't obvious, because the code seems to assume things that then turn out to be untrue. The symptom is this:

>>> x = sp.Symbol("x")

>>> xx = -sp.I*x

>>> print xx._eval_subs(-x, sp.Symbol("y"))

y

The causes (as far as I was able to figure out) are:

- First, I.as_base_exp() returns (-1, 1/2). I doubt this is wise--see next point.

- breakup() within Mul._eval_subs() then adds up exponents of (-1), yielding (-1)**(3/2), which sympy justifiedly rejects as undefined.

- That error really never comes to pass, though, because Mul._eval_subs() throws away the coefficient (and all its powers) if it was able to multiplicatively_extract() them. Unfortunately, it assumes that the entire 'coefficient' is contained in the first arg of the Mul, which is untrue in this case.

Any help would be much appreciated.

Thanks,

Andreas

[Sergiu Ivanov]

On Sun, May 13, 2012 at 6:34 AM, Andreas Klöckner
<li...@informa.tiker.net> wrote:
>
> (Before I start--one question: Google Code bug tracker or Github bug
> tracker--which is the right one?)

SymPy Issues are generally managed on the Google Code bug tracker
(from what I know).

Sergiu

[Andreas Klöckner]

On Saturday, May 12, 2012 11:34:34 PM UTC-4, Andreas Klöckner wrote:

Any help would be much appreciated.

For reference, there seem to be a few more bugs affecting CSE in the presence of complex numbers. I'll post more when I've got them tracked down.

Andreas