4 ms·
CPython is the past only if you conveniently ignore the entire Scientific Python community, which is a huge use case for Python and depends heavily on C-extensi
by mangecoeur 11y ago
CPython is the past only if you conveniently ignore the entire Scientific Python community, which is a huge use case for Python and depends heavily on C-extensions. Frankly saying a transition to PyPy is "done" when NumPy support is not complete is laughable. Huge Python shops (we're talking NASA, CERN, and other institutions on that scale) depend on a Python stack that's heavily reliant on a huge range of things, including the C-API, that aren't even close to mature in PyPy.
- lmm 11y agoNumPy proves that you can do everything that you'd need to do without the CPython C API, no? The fact that other things used in scientific Python only offer a C API interface and not cffi seems like a "legacy"-shaped problem.
- mangecoeur 11y agoOnly if you label basically everything as "legacy", which is really stretching the definition. Just because someone invented a New Thing, doesn't mean all the Other Things are suddenly "legacy" overnight. Fact is, PyPy may be "fast", but in science CPython is in effect much faster because you can use the massive and optimised scientific stack including python, pandas, hdf5, etc etc