I've found out the hard way that boxing/unboxing of PDF primitives to Python is _really_ expensive, so that my workflow has been counter-intuitively quite a lot

I've found out the hard way that boxing/unboxing of PDF primitives to Python is _really_ expensive, so that my workflow has been counter-intuitively quite a lot slower than with PyPDF2.