Matlab has the issue of not being open source (similarly IDL), but I still see it popping up (though naturally, if I've got any choice it the matter, it's going

Matlab has the issue of not being open source (similarly IDL), but I still see it popping up (though naturally, if I've got any choice it the matter, it's going to be ported to Python). I've also seen new Fortran codebases as well, mainly where the (conceptual, not computational) overhead of using helper libraries isn't worth it. I'd suggest Python (via the various memoryview related PEPs, plus numpy) does provide most of the required information (vs. e.g. Go or Rust or Ruby), so I'm not sure that proving much other than if the value of using a library to handle multidimensional arrays is enough such that the combined value of a more general language plus library beats Matlab or Fortran (likely due to other ecosystem or licensing effects), people will switch.