by tija

GitHub Readme.md

Heroku buildpack: Python, Numpy, and Scipy

This is a custom Heroku buildpack for Python apps that use NumPy and/or SciPy, powered by pip.

Please open a GitHub for any problems encountered or feature requests. If are using this project and found it useful, please let me know! (Preferably by email: brandon.k.liu@gmail.com). I maintain this project on my spare time, and I am much more motivated to work on it if I know there are people who are benefiting.


This buildpack currently supports:


  • 1.6.2
  • 1.7.2
  • 1.8.1
  • 1.8.2
  • 1.9.0


  • 0.13.3 (compiled against NumPy 1.9.0)
  • 0.14.0 (compiled against NumPy 1.8.1)

This package will also install compiled runtime libraries for BLAS, LAPACK, ATLAS, and Fortran, which are needed by NumPy and SciPy at runtime.

Known Issues and Todos

  1. Currently, SciPy will only work with the specific version of NumPy against which it was compiled (e.g., SciPy 0.14.0 with NumPy 1.8.1).
  2. Scikit-learn can be installed, but may not pass all tests. For example, scikit-learn 0.14.1 installed with NumPy 1.9.0 and SciPy 0.13.3 did not pass all tests. You can see the failed tests in this issue.

If any of these issues are immediately impacting you, please open a Github issue so that I know that they are higher priority items.


For a new app:

heroku create --buildpack https://github.com/thenovices/heroku-buildpack-scipy

For an existing app:

heroku config:set BUILDPACK_URL=https://github.com/thenovices/heroku-buildpack-scipy

You must specify your exact desired version in requirements.txt (e.g., numpy==1.9.0). If no version is specified, the latest version available will be used. At this time, this buildpack does not support requirements of the form numpy>=1.8.


$ mkdir testheroku
$ cd testheroku
$ git init
$ heroku create --buildpack https://github.com/thenovices/heroku-buildpack-scipy
$ echo -e "numpy==1.8.1\nscipy==0.14.0" > requirements.txt
$ git add requirements.txt
$ git commit -m 'Added requirements'
$ git push heroku master


This fork is taken from @dbrgn -- I owe thanks to him, @wyn, and others for setting up a lot of the code here.