v4.0.0

RecursiveArrayTools v4.0.0

Diff since v3.51.0

RecursiveArrayTools v4.0.0

Breaking Changes

• AbstractVectorOfArray{T, N, A} now subtypes AbstractArray{T, N}. Linear indexing A[i] returns the ith element in column-major order (previously returned the ith inner array). Use A.u[i] or A[:, i] for the old behavior.
• length(A) returns prod(size(A)) (total elements), not length(A.u) (number of inner arrays). Use length(A.u) for the old behavior.
• iterate(A) iterates over scalar elements, not inner arrays. Use A.u for column-wise iteration.
• map(f, A) maps over elements, not inner arrays. Use map(f, A.u) for the old behavior.
• Ragged arrays: size(A) reports maximum size; out-of-bounds elements return zero (sparse interpretation). Non-zero-padded ragged arrays are available via using RecursiveArrayToolsRaggedArrays.
• DiffEqArray has new interp and dense fields (new type parameter I).
• Optimized any/all for ArrayPartition moved to RecursiveArrayToolsArrayPartitionAnyAll subpackage to avoid invalidations. Required for GPU arrays.

New Features

• DiffEqArray callable interface: da(t), da(t; idxs=1), da(t, Val{1}) for interpolation.
• Dense plotting: plot(da) with denseplot=true generates smooth interpolated curves.
• Full-featured plot recipe with idxs, phase plots, tspan, plotdensity, plotat.
• RaggedVectorOfArray and RaggedDiffEqArray in RecursiveArrayToolsRaggedArrays sublibrary for non-zero-padded ragged arrays.
• CUDA extension for CuArray(::AbstractVectorOfArray) disambiguation.
• Plotting helpers exported for SciMLBase delegation: interpret_vars, diffeq_to_arrays, solplot_vecs_and_labels.

Merged pull requests:

• BREAKING: Make AbstractVectorOfArray <: AbstractArray (#547) (@ChrisRackauckas-Claude)