3 ms·
Citing "fast CPU performance" alongside "large community" and "already installed" sounds a bit disingenuous. The version that is fast has a smaller community a
by proveanegative 12y ago
Citing "fast CPU performance" alongside "large community" and "already installed" sounds a bit disingenuous. The version that is fast has a smaller community and is unlikely to be already installed while the one that is already installed and has a large community is not fast.
Like it or not, "Python" means CPython to most people, inducing most library authors, and you run into problems using PyPy if you assume it's "just Python", especially if you have come to rely on Python's easy interfacing with C.
- BuckRogers 12y agoIt wasn't stated alongside. "Most people" aren't ones I want to be around or work with. I come here to associate with those who do their homework. CPython and PyPy are the same Python community as PyPy runs Python code unmodified. PyPy also interfaces just as well with C, inform yourself. Python is a language. Languages don't have speed. Implementations do.
- proveanegative 12y ago>"Most people" aren't ones I want to be around or work with. I come here to associate with those who do their homework. I am not sure what you mean here in the context of what I said. My was simply that most users of Python don't use PyPy, so most library authors don't test against it. >CPython and PyPy are the same Python community as PyPy runs Python code unmodified. Not all of it, as https://bitbucket.org/pypy/compatibility/wiki/Home https://bitbucket.org/pypy/compatibility/wiki/Home will confirm. >PyPy also interfaces just as well with C, inform yourself. Sorry, this is where I did not make my point very clear. I meant to point out that the way in which PyPy interfaces with C is different from the one used by CPython, so libraries that rely on it don't work in PyPy without an extra porting effort.
- BuckRogers 12y agoCTypes and CFFI still work with PyPy. Also, 'not all of it'- well yes, this is a WIP with a ~2 man project/implementation. Not bad. What are you using that's not on the list? The intention isn't so that authors have to test against PyPy.. PyPy's intention is to be as compatible with CPython as possible. Either way, while you attempt to discredit PyPy the fact remains that Python code is and can be very fast for CPU heavy work. There's no way around that other than trying to poke holes.