|
From: <kh...@ce...> - 2005-11-08 08:12:14
|
On 08.11.2005, at 09:11, Francesc Altet wrote: > Well, I would like to migrate PyTables to scipy.core in the medium > term. > My plan to not bother the users is to start supporting scipy.base by > using the array protocol. That means that, although numarray will > still ... > I really think that the array protocol (that already is supported in > latest Numeric and numarray, bar some small issues) is a Good Thing > (tm) > to consider, specially for the developers. True, but it doesn't solve my problem (I think). Users of my code don't see much of an array interface, it's hidden behind application- oriented abstraction layers. That also means that users don't care much if I use Numeric, numarray, or scipy.core, as long as there are no installation problems. However, my code is necessarily tied to one specific array package because it creates all of its arrays. It also does lots of array operations that go beyond the array protocol. So I have lots of dependencies on the chosen array package in many different modules. My problem is handling the transition period while I test scipy.core and while scipy.core stabilizes. I expect to make releases of my packages during that time, so I need to keep a code base that works with Numeric. Doing development on two code bases in parallel seems like asking for trouble. Konrad. -- ------------------------------------------------------------------------ ------- Konrad Hinsen Laboratoire Leon Brillouin (CEA-CNRS), CEA Saclay, 91191 Gif-sur-Yvette Cedex, France Tel.: +33-1 69 08 79 25 Fax: +33-1 69 08 82 61 E-Mail: kh...@ce... ------------------------------------------------------------------------ ------- |