Installing Python for BAG

This section describes how to install Python for running BAG.

Installation Requirements

BAG is compatible with Python 3.5+ (Python 2.7+ is theoretically supported but untested), so you will need to have Python 3.5+ installed. For Linux/Unix systems, it is recommended to install a separate Python distribution from the system Python.

BAG requires multiple Python packages, some of which requires compiling C++/C/Fortran extensions. Therefore, it is strongly recommended to download Anaconda Python, which provides a Python distribution with most of the packages preinstalled. Otherwise, please refer to documentation for each required package for how to install/build from source.

Required Packages

In addition to the default packages that come with Anaconda (numpy, scipy, etc.), you’ll need the following additional packages:

  • python-future

    This package provides Python 2/3 compatibility. It is installable from pip:

    > pip install future

    pip also works for pre-downloaded tar file:

    > pip install future-0.16.0.tar.gz

  • subprocess32 (Python 2 only)

    This package is a backport of Python 3.2’s subprocess module to Python 2. It is installable from pip.

  • sqlitedict

    This is a dependency of OpenMDAO. It is installable from pip.

  • OpenMDAO

    This is a flexible optimization framework in Python developed by NASA. It is installable from pip.

  • mpich2 (optional)

    This is the Message Passing Interface (MPI) library. OpenMDAO and Pyoptsparse can optionally use this library for parallel computing. You can install this package with:

    > conda install mpich2

  • mpi4py (optional)

    This is the Python wrapper of mpich2. You can install this package with:

    > conda install mpi4py

  • ipopt (optional)

    Ipopt is a free software package for large-scale nonlinear optimization. This can be used to replace the default optimization solver that comes with scipy. You can install this package with:

    > conda install --channel pkerichang ipopt

  • pyoptsparse (optional)

    pyoptsparse is a python package that contains a collection of optmization solvers, including a Python wrapper around Ipopt. You can install this package with:

    > conda install --channel pkerichang pyoptsparse