I am working on a project that revolves around multiprecision "complex" numbers, specifically it's a Mandelbrot Set-based app, but with a twist that requires decent correspondence between the output of a (fast) C++ py extension module (boost, cython, or other...) and the pure python modules that might want to use it.
Right now, I'm using boost::multiprecision to wrap the MPFR raw type, and yeah if I just wanted to pass an mpfr_t to python that'd be one thing. However, for this app I need to store the C++ module's result as a string which will be interpreted later by a Python module, and needs to give the same number.
BigFloat is supposed to be an MPFR python wrapper, but it doesn't interpret a string literal of an X-precision float exactly the same way as boost::multiprecision's data() method.
Does anybody know of a combination of libraries plus an approach that does result in an exact correspondence between C++ and Python string literals of arbitrary-precision floating-point numbers?
I can provide code excerpts if needed to illustrate the problem, but figured it was arcane enough that if anybody had the answer, they'd know exactly what I mean and how to fix it.
gmpy2did the trick. My C solution is still slightly faster, so I ended up offering the user the option to choose which MP solution they wanted to use. – Antoninaantonino