add --QuadratureFunction-- QuadratureRule #176
Conversation
docs/src/solvers/IntegralSolvers.md
Outdated
| @@ -12,10 +12,12 @@ The following algorithms are available: | |||
| - `CubaDivonne`: Divonne from Cuba.jl. Requires `using IntegralsCuba`. Works only for `>1`-dimensional integrations. | |||
| - `CubaCuhre`: Cuhre from Cuba.jl. Requires `using IntegralsCuba`. Works only for `>1`-dimensional integrations. | |||
| - `GaussLegendre`: Uses Gauss-Legendre quadrature with nodes and weights from FastGaussQuadrature.jl. | |||
| - `QuadratureFunction`: Accepts a user-defined function that returns nodes and weights. | |||
There was a problem hiding this comment.
I'm not so sure of the naming on this, and also it sounds more like a function than an algorithm. Maybe GeneralQuadrature?
There was a problem hiding this comment.
Or QuadratureRule, i.e. similar to TrapezoidalRule?
There was a problem hiding this comment.
I think GeneralQuadrature would sound too similar to this work https://github.com/JoshuaTetzner/GeneralizedGaussianQuadrature.jl
There was a problem hiding this comment.
That's a repo with 0 stars, I'm not worried about people knowing that. I'm more worried that people know ODEFunction, SDEFunction, NonlinearFunction, BVPFunction, etc. as a standard in SciML for the SciMLFunction interface https://docs.sciml.ai/SciMLBase/stable/interfaces/SciMLFunctions/, but here it's breaking the rule.
I think QuadratureRule sounds good.
|
This looks great. Only some bikeshedding on the naming and I think this is good to go. |
|
Would it be good to also support vector-valued integrands? Right now you assert |
lxvm
changed the title
lxvm
changed the title
lxvm
changed the title
Thanks, I removed the check. |
|
I renamed to |
| maxiters = nothing) | ||
| prob = build_problem(cache) | ||
| if isinplace(prob) | ||
| error("QuadratureRule does not support inplace integrands.") |
There was a problem hiding this comment.
Any reason it doesn't? That seems like it should be possible. I guess open an issue on this as something to do later?
There was a problem hiding this comment.
We don't know the element type of the output array to allocate. This will be fixed by the IntegralFunction PR
Hi,
This pr adds a
QuadratureFunctionalgorithm that allows the user to pass in a function that returns nodes,x, and weights,w, and computessum(w .* f.(x)). This could serve as a backend for other integration algorithms we would like to support, such as the trapezoidal rule and FastGaussQuadrature.jl rules.