Added Normalize to QuArray.

[amitjamadagni]

This is my first attempt at contributing. Though not that significant, please do comment. Thanks.

[acroy]

Thanks for your contribution! It's a good start. Accidentally I thought that adding norms would be a natural next step ...

Regarding your code I have some comments (mostly because we don't have an explicit guide how such functions for QuArrays should work, but I have many implicit ideas ;-)

• Firstly, norms are not so easy. You used Base.norm on the container, which computes the "p-norm". However, for (Hilbert space) operators it is more common to use the Frobenius norm, which you would get via Base.vecnorm. For vectors this doesn't matter, but for N>1 we should probably use the latter. Note that we need a different norm (i.e. trace) for density matrices though.
• I would prefer to have a separate function norm(qa::QuArray) (and probably also for CTranspose).
• For scaling we should provide a mutating function by extending Base.scale! and then implement multiplication and division in terms of that. You will need a copy function first (@jrevels Could you add copy for QuArray and CTranspose?).
• Note that your current normalize! function actually makes a copy of the coefficient array.
• I think coeffs was replaced by rawcoeffs in one of the latest commit?
• Now that we have Travis in place, which runs tests for each PR, we should really try to add tests for each function/operation we add.

Sorry for the long list of comments. Maybe the more general points should go into some "guideline issue".

[amitjamadagni]

@acroy I have worked on the commit. Here are the details :

• Attempt at implementing the norm function which takes QuArray as a parameter.
• Attempt at implementing copy for QuArray and CTranspose.
• normalize, normalize! implemented.

WIP :

• Scaling
• Tests

Doubts :

1. Requirement of rawcoeffs when .coeffs is doing the same job ??

[acroy]

You can always use vecnorm. For N=1 this is the same as norm (sum of absolute values squared of all elements and square root of that).

[amitjamadagni]

My bad !!! Sorry for this. Done editing.

[acroy]

Nice! The point of using rawcoeffs is that you can use it for QuArrays and CTranspose, i.e., norm could take qa::Union(QuArray,CTranspose) and you can simply call rawcoeffs(qa).

[acroy]

You have to use function Base.norm or import Base: norm otherwise norm will only be visible within the module.

[acroy]

If you use rawcoeffs and bases instead of accessing the fields (qa.coeffs) you can replace the arguments qa::QuArray by qa::Union(QuArray,CTranspose) (I think). Don't forget to add norm to the import at the top of arraymath.jl. For the test you can basically use the same approach as Jarrett.

[amitjamadagni]

@acroy is this good to go ?? Any comments would be helpful.

[acroy]

Those imports you can just add to the line above.

[acroy]

Looks good now. Thanks! Just one minor point: could you squash your commits into a single one? It's not really necessary but makes the history more readable.

[amitjamadagni]

@acroy done. Thanks for all the help.