Installing the newest scipy (0.9) in Ubuntu?

I tried following 开发者_如何学运维the tutorial but after hours of building the ATLAS + LAPACK I got an error in the make install. I tried to download the following 4 libraries and install it still but no dice.

Currently I have installed numpy 1.3 and scipy 0.7.2 from the ubuntu repositories. I need a feature from scipy 0.9 though. Any way (preferably foolproof) I could install it?

ActivePython includes a package manager that allows you to install scipy 0.9 (among numpy, matplotlib, etc.) from PyPM.

pypm install numpy

These packages are built with ATLAS + LAPACK (Linux), veclib (OSX) or Intel MKL / ifortran (Windows).

To install Scipy 0.9, you need to have a newer Numpy installed than 1.3. The oldest Numpy that it will work with is 1.4:

https://github.com/scipy/scipy/blob/maintenance%2F0.9.x/INSTALL.txt

Hopefully upgrading your Numpy install will help!

Otherwise, I second Josh's recommendation for prepackaged distributions. I know several people who use Sage and say it is very easy to get up and running.

Where you able to install the dependencies first?

sudo apt-get install gcc g77 python-dev atlas3-base-dev

And then proceed with the installation of scipy?

I'm a big fan of the Enthought Python Distribution (EPD) to get most of my scientific libraries packaged cleanly in one place:

http://www.enthought.com/products/epd.php

It's free if you're in academia.

There are also other options like Python(x,y) and Sage:

http://code.google.com/p/pythonxy/

http://sagemath.org/

Prepackaged distributions are the closest you'll get to a foolproof way. I have built scipy/numpy from scratch before, but I can't help you without further details.