From: James M. <mu...@pi...> - 2010-07-14 22:51:41
|
Ahh. So, unless you need something that depends on the datetime feature, it is best to avoid Enthought until numpy 2.0. If you need Enthought, you need to build visual yourself to match your version of numpy. As an experiment, I just started going the other way, start with a basic vpython install and then add other packages to it. A download of the most recent scipy 0.7.2 binary works. I didn't try the release candidate of 0.8. A download of the most recent matplotlib (1.0.0) does not, but version 0.99.1 from September does. I think this indicates that the most recent binaries require numpy 1.4.1. I assume Bruce will get around to updating the VPython build to use the most recent numpy at sometime, That is about 95% of what I have used, and it is easy enough that I might be able to require a physics major to do it on their machine. For engineers, it is about 99% of what they would use Matlab for. The best thing about PythonXY is that if you are using more than visual in a course, you can just tell students to get PythonXY, and they have everything you might ever want. With Enthought, you have to make sure you build them a visual that matches the enthought version of numpy. > > Oh ... Enthought consciously chose to stick with 1.4.0 because they wanted to retain the datetime feature. (This was the crux of a long discussion on the numpy list a while back.) They plan to stick with 1.4.0 until 2.0 comes out. |