Changelog

armadillo4r 0.9.0

Using Armadillo 15.2.6
as_SpMat() now accepts any sparse matrix class from the Matrix package, including RsparseMatrix and TsparseMatrix subclasses, symmetric (dsCMatrix), and unit-triangular (dtCMatrix) matrices, which are expanded to general form on conversion to SpMat<double>.
as_SpMat() and as_dgCMatrix() now use the Matrix package’s public C API (CHOLMOD bridge) instead of reading S4 slots directly. The bridge is auto-enabled when the Matrix package’s headers are visible at compile time, which happens when the consuming package declares LinkingTo: Matrix (>= 1.6-2) in its own DESCRIPTION. Packages that do not need the sparse-Matrix bridge no longer have to depend on Matrix. The auto-detection can be overridden with -DARMADILLO4R_USE_MATRIX or -DARMADILLO4R_NO_MATRIX.

Adds templates to simplify cube/field data copying between R and Armadillo.
Adds templates to pass lazy Glue<> and Base<> expressions to as_{{ R_TYPE }}().

Using Armadillo 15.2.1 (Medium Roast Deluxe) thanks to @conradsnicta update notification.
Improve complex vectors/matrices copying performance by using memcpy instead of element-wise copying.
Explicit testing with Clang and GCC compiler using C++ 14, 17, 20, and 23 standards (Armadillo 15.2.1 requires at least C++14)