Re: py-pip dilemma
On 2012/12/15 02:40, Juan Francisco Cantero Hurtado wrote: After of a quick read of the doc and code of pip, I can't find any reference related to the setting of new enviroment variables. Only a few calls for to get and set the variables used by pip. Maybe I've missed something. Actually the problem is not with pip which is correctly getting the paths from distutils.sysconfig for the C compiler, it's numpy which has its own code for the Fortran compiler. We could use something like this... --- numpy/distutils/fcompiler/gnu.py.orig Sat Dec 15 11:10:39 2012 +++ numpy/distutils/fcompiler/gnu.pySat Dec 15 11:10:21 2012 @@ -159,6 +159,8 @@ class GnuFCompiler(FCompiler): '../../../../lib')) if os.path.exists(os.path.join(d2, lib%s.a % self.g2c)): opt.append(d2) + elif sys.platform[:7] == 'openbsd': +opt.append('/usr/local/lib') opt.append(d) return opt A possible fix is to hardcode the paths in __init__.py but I guess isn't a good fix. I don't know. I'm a python newbie :) -- Juan Francisco Cantero Hurtado http://juanfra.info
Re: py-pip dilemma
Daniel Dickman didick...@gmail.com wrote: What ports are you looking at specifically? I'm working to port pandas so we might have some overlap. See here: http://dickman.org/openbsd/patches/pandas/ numpy scipy matplotlib ipython For starters. I know people have been working on ipython but qt4 flavor is holding them back. I use those on daily basis. I use several other modules but these four are the highest priority. Cheers, Predrag The main one I have left to do is matplotlib. If you'd want to test I'm happy send patches over. On Dec 13, 2012, at 10:00 PM, Predrag Punosevac punoseva...@gmail.com wrote: I am soliciting opinions about the safety of py-pip package management system. I am using Python primarily for scientific computing/prototyping. Many of standard scientific python modules in our ports tree are a bit outdated (trying to update some of those ports is on my todo list as I am sure it is on todo list of many people but busy schedule is another story). Inspired by my recent experience with hs-cabal I decided to play with py-pip on one of 5.1 installations which I will update soon. Surprisingly or maybe not I was actually able to update most of my python modules with exception of scipy which fails with linker problem ld: cannot find -lpython2.7 How safe is really using py-pip as a Python package manager? Is it possible to reconfigure py-pip to behave like cabal which installs modules in my home directory rather than in /usr/local/bin? Best, Predrag
Re: py-pip dilemma
On Thu, Dec 13, 2012 at 10:00:28PM -0500, Predrag Punosevac wrote: I am soliciting opinions about the safety of py-pip package management system. I am using Python primarily for scientific computing/prototyping. Many of standard scientific python modules in our ports tree are a bit outdated (trying to update some of those ports is on my todo list as I am sure it is on todo list of many people but busy schedule is another story). Inspired by my recent experience with hs-cabal I decided to play with py-pip on one of 5.1 installations which I will update soon. Surprisingly or maybe not I was actually able to update most of my python modules with exception of scipy which fails with linker problem ld: cannot find -lpython2.7 Export the next variables in your shell before of to run the compilation: LDFLAGS=-L/usr/local/lib CPPFLAGS=-I/usr/local/include But I think is easier just to update the port. How safe is really using py-pip as a Python package manager? Is it possible to reconfigure py-pip to behave like cabal which installs modules in my home directory rather than in /usr/local/bin? Search virtualenv home python. Best, Predrag -- Juan Francisco Cantero Hurtado http://juanfra.info
Re: py-pip dilemma
On 2012/12/14 19:37, Juan Francisco Cantero Hurtado wrote: Export the next variables in your shell before of to run the compilation: LDFLAGS=-L/usr/local/lib CPPFLAGS=-I/usr/local/include But I think is easier just to update the port. That should probably get fixed in the py-pip port..
Re: py-pip dilemma
On Fri, Dec 14, 2012 at 07:07:46PM +, Stuart Henderson wrote: On 2012/12/14 19:37, Juan Francisco Cantero Hurtado wrote: Export the next variables in your shell before of to run the compilation: LDFLAGS=-L/usr/local/lib CPPFLAGS=-I/usr/local/include But I think is easier just to update the port. That should probably get fixed in the py-pip port.. After of a quick read of the doc and code of pip, I can't find any reference related to the setting of new enviroment variables. Only a few calls for to get and set the variables used by pip. Maybe I've missed something. A possible fix is to hardcode the paths in __init__.py but I guess isn't a good fix. I don't know. I'm a python newbie :) -- Juan Francisco Cantero Hurtado http://juanfra.info
py-pip dilemma
I am soliciting opinions about the safety of py-pip package management system. I am using Python primarily for scientific computing/prototyping. Many of standard scientific python modules in our ports tree are a bit outdated (trying to update some of those ports is on my todo list as I am sure it is on todo list of many people but busy schedule is another story). Inspired by my recent experience with hs-cabal I decided to play with py-pip on one of 5.1 installations which I will update soon. Surprisingly or maybe not I was actually able to update most of my python modules with exception of scipy which fails with linker problem ld: cannot find -lpython2.7 How safe is really using py-pip as a Python package manager? Is it possible to reconfigure py-pip to behave like cabal which installs modules in my home directory rather than in /usr/local/bin? Best, Predrag