MethodError: no method matching recursivecopy!(::Float64, ::Float64) about multiscalearrays.jl

[zahachtah]

this seems to fix it:

function recursivecopy!(b::AbstractMultiScaleArray, a::AbstractMultiScaleArray)
    @inbounds for i in eachindex(a.nodes)
        recursivecopy!(b.nodes[i], a.nodes[i])
    end
    @inbounds for i in eachindex(a.values)
        #recursivecopy!(b.values[i], a.values[i])
        copyto!(b.values,a.values)
    end
end

[ChrisRackauckas]

What's the type of values? The better solution is just to add a dispatch to recursivecopy! to that way it does copyto! in this case.

[zahachtah]

At that point its Float64’s

[zahachtah]

If the values is an array of Float64 then this

’’’recursivecopy!(b.values[i], a.values[i])’’’

should dispatch on Float64’s

[ChrisRackauckas]

oh, it should just be recursivecopy!(b.values, a.values).

[zahachtah]

Ah, right.

Also: i tried to use namedTuples instead of an array of float64 but wasn’t sure if 1) it should work or 2) how to change the typeParsmetrization. Any quick ideas? Don’t want to waste your time though

[ChrisRackauckas]

It should work the same way we have tuples working. If not, we'll need an example.

[zahachtah]

this is what I am trying to do

struct scale1{B,V<:NamedTuple{Vararg{<:Array{<:Number}}},P,N<:Tuple{Vararg{<:AbstractMultiScaleArrayLeaf{<:Number}}}} <: AbstractMultiScaleArray{B}
    nodes::N
    values::V
    end_idxs::Vector{Int}
    params::P
end

so I can construct it with

s1A = construct(scale1, (deepcopy(t1), deepcopy(t2)), (a=0.1,b=2.0), [0.1,0.2])

where t1 and t2 are lower level nodes

But this gives type-error

[ChrisRackauckas]

How does it give a type error? Where? I need an example.

[zahachtah]

already when running the above definition of the struct:

TypeError: in Type, in parameter, expected Type, got Type{Vararg{#s37,N} where N where #s37<:(Array{#s36,N} where N where #s36<:Number)}

Stacktrace:
 [1] top-level scope at none:0

[zahachtah]

Closing, don't worry reg the named tuple question, I have to read up on how type parametrizations work. As always, thanks for the great package.

[ChrisRackauckas]

So it had to do with the struct definition?

[zahachtah]

yes. but I also realize I wanted the named tuple as an array for the values which have to be mutable for the ode solver so that doesnt work. I mostly thought it would be neat to be able to write things like

u[:energy] in the differential equations instead of e.g. u[3].

but named tuples are are immutable of course so that doesn't make sense. But I have what I need now I think. Thanks

[ChrisRackauckas]

I'm working towards a solution: https://github.com/JuliaDiffEq/LabelledArrays.jl

[zahachtah]

Oh, that would make it super pretty to write down the equations.

I remember Tom Breloff had this cool macro (I think it was called @with) so that if used you didn't even have to write A.second (he used if for structs) but you could just write second or :second and it would fill out the rest at compile time. I gues sone could do that for

@with a  

that would turn

: second -> a[:second]

making it even prettier. This is awesome. I guess if you have the map between names and indicies the macro could do that right away too....

julia sure is amazing!