Bra-ket products produce 1x1 Qobj

[stared]

As I see bra-ket products, e.g.

c = basis(2, 0).dag() * basis(2, 0)

result in 1x1 Qobjects, not numbers.

(I know why, as it is compatible with matrix multiplication of 1xn times nx1.)

However, it behaves differently

For example, for

basis(2, 0) * c

works fine, but

c * basis(2, 0)

throws an exception.

On paper, once we get a c-number, we shouldn't care if about its order in the multiplication.

Are there any reasons not to automatically convert 1x1 Qobj into a (complex) number?

[jrjohansson]

Thanks for the bug report. Yes, it should be possible to left and right multiply any Qobj with a 1x1 Qobj, so that's a bug.

Regarding returning a C number, it has been discussed in #28. There is no question that psi.dag() * psi is a C number, but in my opinion it should be a C number represented as 1x1 Qobj, not a Python complex. The reason is that arithmetics with Qobj should form an algebra. It can be important that if x and y is a Qobj then we should always have that x_y is a Qobj as well, so that one always can do for example z=x_y and then z.dims or call some Qobj method on z. If z is not a Qobj this breaks. This is my opinion on this matter, but not everyone agreed. See #28.