progress with numpy and removal of numpy.py
I have been working to get numfocus benchmarks written for numpy to run on numpypy and pypy https://github.com/numfocus/python-benchmarks The biggest obstacle is actually the need for matplotlib, which is used internally to produce pretty comparisons between different benchmarks. Luckily, Stefan H. Muller did much of what is needed to use matplotlib (non-interactively) with pypy, I have taken over his work and it now lives in two git repos: https://github.com/mattip/numpy on the pypy-hack branch, and https://github.com/mattip/matplotlib master branch. These work, download a nightly pypy, virtualenv, "setup.py install", but there is some missing low-lying-fruit kind of functionality to run the benchmarks, again, in matplotlib. That is the good news, the bad news is that in a perhaps hasty move, I removed lib_pypy/numpy.py since it complicates the numpy install. I fixed tests (one build too late, sorry for the failures), but now people who were used to a nightly pypy being capable of "import numpy" will be disappointed. I did ask on IRC before the change, and got two positive votes, but feel a need for wider consensus. So should we leave the file out, and hopefully encourage people to start trying the pypy-hack branch, or put it back? In an unrelated-but-related issue, should the github repo of numpy be moved to a more pypy-like place (bitbucket.org/pypy)? Matti
On Tue, 2013-10-15 at 07:15 +0300, Matti Picus wrote:
I did ask on IRC before the change, and got two positive votes, but feel a need for wider consensus.
Does that mean that now one only can import NumPyPy as follows import numpypy instead of import numpy ? If yes, what is going to be the way forward? I mean, is the plan such that eventually, you should be able to install PyPy that bundles NumPyPy and then NumPy (when the patches will be accepted upstream, or maybe a PyPy-compatible fork will be released on pypi)? If your github repository works similarly or better than `import numpy` as implemented before, I see no problem with it and will just adapt my scripts; I only feel that this should be better documented, or advertised more prominently. For example, you might consider posting a progress report on the blog which explains the changes and tells how to deal with them... -- Sincerely yours, Yury V. Zaytsev
Hi, On 15/10/13 12:45, matti picus wrote:
and what about cloning the numpy repo into bitbucket/pypy to make it more of a "pypy owned" thing?
I think it's a good idea. I propose the following: 1) we move your repo to bitbucket/pypy/numpypy 2) we package numpypy, so that people can just do "pip install numpypy" 3) once numpypy is installed, we no longer require the ugly "import numpypy"; a simple "import numpy" will just work. 4) for some time at least, we distribute a numpypy.py so that when imported it prints an error message which explain how to get the newer numpypy What do you think?
On 17/10/13 10:59, Antonio Cuni wrote:
Just a note: A repo from github is definately a git repo, and it stays in git format when moved to bitbucket. May not be relevant, I just wanted to mention that. cheers - Chris p.s.: After having worked with both for quite a long time, I now would prefer if pypy and python would use git ;-) -- Christian Tismer :^) <mailto:tismer@stackless.com> Software Consulting : Have a break! Take a ride on Python's Karl-Liebknecht-Str. 121 : *Starship* http://starship.python.net/ 14482 Potsdam : PGP key -> http://pgp.uni-mainz.de phone +49 173 24 18 776 fax +49 (30) 700143-0023 PGP 0x57F3BF04 9064 F4E1 D754 C2FF 1619 305B C09C 5A3B 57F3 BF04 whom do you want to sponsor today? http://www.stackless.com/
participants (6)
-
Antonio Cuni -
Christian Tismer -
matti picus -
Matti Picus
-
Romain Guillebert -
Yury V. Zaytsev