Numpy (http://numeric.scipy.org/) is the new "multidimensional-array object" module for python - it deprecates both Numpy and numarray. Scipy, matplotlib etc., already support numpy and numpy is (hopefully) going to eliminate the confusion over different and sometimes incompatible array modules that exist for Python nowadays. I copied numarray-1.5.0.ebuild to my portage overlay, renamed it to numpy-0.9.2.ebuild and made some changes (basically to make it recognize the installed version of atlas). The ebuild works for me, but I am not sure if it is "correct". In particular, handling of lapack and blas could be improved. I cannot access numpy website right now, so I have kept the LICENSE field as "BSD", but it might be something else.
Created attachment 77364 [details] ebuild for numpy-0.9.2
(In reply to comment #1) > Created an attachment (id=77364) [edit] > ebuild for numpy-0.9.2 Seems to work on amd64 as well, for the few basic tests I ran. TBC nevertheless. Quick question: as scipy is in the sci-libs branch, shouldn't numpy be as well ?
(In reply to comment #2) > Quick question: as scipy is in the sci-libs branch, shouldn't numpy be as well > ? > Well, numarray and Numeric, both of which are python array modules, are in dev-python. Hence I put numpy ebuild in dev-python instead of sci-libs. It's updo the developers to decide where to put the ebuild though---for many of the packages, there is no "definite" single category anyway.
Created attachment 78595 [details] numpy-0.9.4.ebuild Version bump
Hi I've tested this numpy-0.94 ebuild as a replacement for the Numeric package and it works fine. I'm writing a lot of python/Numeric code for my PhD. All the unit test were successfull after upgrading the code to numpy.
added to cc
Adding to portage as dev-python/numpy. thanks for the ebuild. i suppose the depends are the other way around because those ebuilds are actually being changed to use numpy instead.
closing.
