Dot product for QuArray's.

[amitjamadagni]

I was going through the material on Krylov propagator, as it requires the dot of QuArray's I have added this, I am not really sure if we require this , any comments on this would be helpful.

[acroy]

+1 for dot. We should also have this for QuMatrix. Related JuliaLang/LinearAlgebra.jl#206.

[acroy]

The Travis error for 0.4 seems unrelated, but we should have a closer look.

[acroy]

I think it is better to use trace(coeffs(qm1'*qm2)) here. If you want you can also steal trace for QuMatrix from #32. We have to check the issue I linked to if there is a more efficient way to do this dot product.

[amitjamadagni]

@acroy @jrevels I have made an attempt at dot (considering dual vectors as well) and trace, added tests for the same. A review would be helpful.

[acroy]

I would prefer the tests being easier to understand on a first glance, which implies more comments. If one doesn't know about Pauli matrices it is unclear what the tests are doing. I think it is better to say for instance

# Pauli matrices sigmax, sigmay, sigmaz are traceless (see ... )
@assert trace(sigmax) == zero(eltype(sigmax))
@assert trace(sigmaz) == zero(eltype(sigmaz))

and so on.

[acroy]

Apart from the comment on the tests, I think this looks good.

[amitjamadagni]

@acroy edits done. If this is fine, I will squash the commits ?

[acroy]

This line also needs an explanation ...

[amitjamadagni]

@acroy done!

[acroy]

Good. Please squash the commits and then we can merge I would say.

[acroy]

Sorry. I just noticed that you squashed already :-)

[amitjamadagni]

@acroy can we merge this ?

[acroy]

Basically yes. I was just thinking if we should generalize dot to allow for two AbstractQuVectors (and giving an error in the mixed case)?

[amitjamadagni]

@acroy I have tried creating a common dot for AbstractQuVector.

[acroy]

I see, but what I meant was the following

dot{B<:OrthonormalBasis}(vec1::AbstractQuVector{B}, vec2::AbstractQuVector{B}) = ...
dot{B<:OrthonormalBasis}(vec1::DualVector{B}, vec2::DualVector{B}) = ...
dot{B<:OrthonormalBasis}(vec1::DualVector{B}, vec2::AbstractQuVector{B}) = error("Inner product of a dual vector and a vector is not supported.")
dot{B<:OrthonormalBasis}(vec1::AbstractQuVector{B}, vec2::DualVector{B}) = error("Inner product of a dual vector and a vector is not supported.")

and maybe similar for AbstractQuMatrix.

[amitjamadagni]

@acroy the latest commit covers the cases, please do let me know if I am missing something here.

[acroy]

You are checking if the element types of the two vectors are the same, while I simply want to prevent a DualVector and an AbstractQuVector as parameters.

[amitjamadagni]

Sorry my bad I understand what I did here. So what would be the best way out, the initial version or the latest version ?

[acroy]

Either you revert it to your original version and we sort it out later or you just take the code I posted above and replace the ... with the code you provided originally.

[amitjamadagni]

Would be good if we compare the types of vectors i.e., typeof(vec1) == typeof(vec2) which would then generalize to comparing QuArray and CTranspose ?

[acroy]

You don't need to compare the types. dot{B<:OrthonormalBasis}(vec1::AbstractQuVector{B}, vec2::AbstractQuVector{B}) actually applies to all vectors (including dual vectors). Now we have to provide more specific versions to exclude the cases with mixed parameters. All the work will be done by the compiler.

[amitjamadagni]

@acroy a review would be helpful.

[acroy]

@amitjamadagni : Could you please squash and rebase?

[amitjamadagni]

@acroy done !

[acroy]

@amitjamadagni : Thanks!