Bug in sympy.physics.quantum "qapply"

[asmeurer]

(Note: this was originally reported at #945 )

There seems to be a bug in qapply, both in 0.7.1 and in the current git HEAD.
I was triggering it when trying qapply(vacuum.dual*A) , where vacuum.dual is a custom Bra-class that I'm hacking on, and A is a quantum operator. I am not sure if my Bra-class does indeed conform to what is required for such, but I believe the bug is independent from that. The main requirements are that the two objects are non-commutative things that do not know how to be applied to each other (i.e. lhs._apply_operator(rhs) and vice versa fails). In this case qapply_Mul will continue to try the previous pair in the multiplication via line 166 of qapply.py:

    return qapply_Mul(e._new_rawargs(*(args+[lhs])), **options)*rhs

Now if this was actually the first pair of the multiplication, e._new_rawargs(*(args+[lhs])) will return lhs itself (might this be the bug indeed?). The new call of qapply_Mul will then work on the lhs object, assuming it is Mul, which it isn't. It probably most often goes unnoticed since it has a high chance of bailing out early either due to lhs not having two arguments or one of them being commutative.
The solution should probably be simple, just check for for an empty args at that line.
A testcase will need a lefthandside object that is not commutative and at least two non-commutative arguments. I was unable to find a good example, and my quick try at a simple example Bra-class was not successful.

[oscarbenjamin]

@Costor should this one be reopened?