How do I install an updated scipy on the dotCloud platform?

Question

Adding "scipy" to my requirements.txt file results in a number of errors. What else do I need to do? I see things like:

Running setup.py egg_info for package scipy
10:19:24 [www]     /home/dotcloud/env/build/scipy/scipy/__init__.py:97: UserWarning: Numpy 1.5.0 or above is recommended for this version of scipy (detected version 1.3.0)
10:19:24 [www]       UserWarning)
10:19:24 [www]     Running from scipy source directory.
10:19:24 [www]     blas_opt_info:
10:19:24 [www]     blas_mkl_info:
10:19:24 [www]       libraries mkl,vml,guide not found in /home/dotcloud/env/lib
10:19:24 [www]       libraries mkl,vml,guide not found in /usr/local/lib
10:19:24 [www]       libraries mkl,vml,guide not found in /usr/lib64
10:19:24 [www]       libraries mkl,vml,guide not found in /usr/lib
10:19:24 [www]       NOT AVAILABLE

etc.

(this question and answer are being migrated from answers.dotcloud.com)

Answer 1

The problem is that the version of numpy installed on the dotCloud Python service is too old, so you need to add the following to your requirements.txt file before scipy to upgrade to a newer version.

numpy==1.6.1

scipy also requires some system packages that aren't installed by default libatlas-base-dev and gfortran in order to build it correctly. You can add these in a systempackages section of your dotcloud.yml file, like this:

systempackages:
    - libatlas-base-dev
    - gfortran

Once you scipy to your requirements.txt and the systempackages to `dotcloud.yml, repush. This should fix your issue, and scipy will be installed.

It is also important to note that it will take a little while to build all of those items so the first push might take a few minutes, so don't be surprised if it is taking longer then usual.

(original answer from Ken Cochrane)