Support for state sol of NamedTuple type

[navidcy]

This is an attempt to deal with #212.

[glwagner]

The design requires sol, L, and N to all be NamedTuples with the same names and fields.

The module writers will still provide one function calcN!. However they can use the tuples.

A nice convenience would be that a named tuple value of nothing doesn't error. That will mean that you get correct behavior when typeof(sol.U) isa AbstractArray or when typeof(sol.U) isa Nothing.

Are there any complications with Equation or Problem?

[navidcy]

Equation struct includes dims which would become a tuple of tuples I guess?
But I'm not even sure if Equation.dims is used...

[navidcy]

Also

FourierFlows.jl/src/problem.jl

Line 87 in 20a68da

sol = devzeros(dev, eqn.T, eqn.dims)

should be generalized...

[glwagner]

Also

FourierFlows.jl/src/problem.jl

Line 87 in 20a68da

sol = devzeros(dev, eqn.T, eqn.dims)

should be generalized...

Won't module writers just use

solution = (u=devzeros(...), v=devzeros(...))

why do you need a special utility?

[navidcy]

True.

sol is constructed automatically based on the dimension of L. But I guess I answered myself; if the user prescribed L appropriately the everything should follow through.

[glwagner]

True.

sol is constructed automatically based on the dimension of L. But I guess I answered myself; if the user prescribed L appropriately the everything should follow through.

user: "module writer", I think you mean.

[navidcy]

I'm closing this for now