so3g 0.1.15

Tools for Simons Observatory work with spt3g_software

Glue functions and new classes for SO work in the spt3g paradigm.

Environment Setup

Before installing SO software, make sure you know what python environment you will be using:

• Are you working on Linux or MacOS?

• Are you using a “python3” executable provided by your OS or one provided by another source (Anaconda, conda-forge / miniforge, homebrew, macports, etc)?

• Are you going to be actively developing so3g or just installing and using it?

After you have determined the answers to these questions, you can set up your working environment.

Using Conda

If you already have a conda installation (a.k.a. conda “base” or “root” environment) that is recent, then you can use that to create an environment. First, verify some info about your installation:

which python3
python3 --version
which conda

Your python version should be at least 3.7.0. Does the location of python3 match the location of the conda command (are they in the same bin directory)? If so, then you are ready. If you do not have conda installed but would like to use it, you might consider installing the “miniforge” root environment (https://github.com/conda-forge/miniforge) which is configured to get packages from the conda-forge channel. Conda-forge provides a large number of software packages built from high quality recipes.

The next step is to create a dedicated conda environment for your SO work. This allows us to create and delete these environments as needed without messing up the root environment:

conda create -n simons # <- Only do this once
conda activate simons

Now install as many dependencies as possible from conda packages. These are listed in a text file in the top of this git repo:

conda install --file conda_deps.txt

Using a Virtualenv

If you are using a python3 provided by your OS or somewhere else, you can work inside a “virtualenv”. This is like a sandbox where you can install packages for working on one project and you can always just wipe the directory and start over if something gets messed up. In general, you should never pip install packages directly to your OS or package manager location (e.g. pip installing as root on linux, or pip installing directly to /usr/local with homebrew). Doing so will create problems for the future on your system. We will create a virtualenv in our home directory:

python3 -m venv ${HOME}/simons # <- Just do this once
source ${HOME}/simons/bin/activate

Now install some basic packages and then all of our requirements:

python3 -m pip install --upgrade pip setuptools wheel
python3 -m pip install -r requirements.txt

Other Python Packages

If you will be using the pointing code in so3g, install pixell and qpoint with pip (regardless of whether you are using a conda env or virtualenv):

pip install pixell
pip install qpoint

Installing Pre-Built Wheels

If you are just using so3g (not modifying or developing it), then you can install the latest release of the package with:

pip install so3g

This command should be used regardless of whether you are working in a conda env or a virtualenv.

Now any time you activate your virtualenv / conda env, you can use so3g.

Building From Source

If you will be developing so3g or want more control over the build, then you should build from source. You will need to install boost, FLAC, a BLAS/LAPACK library, and the spt3g_software package. For this discussion, we will assume that you have git checkouts of spt3g_software and so3g in:

${HOME}/git/spt3g_software
${HOME}/git/so3g

And that you will be installing both spt3g_software and so3g into:

${HOME}/so3g

Adjust the instructions below if your situation is different. We also need to be able to load this install prefix into our shell environment. There are several ways of doing that. The first is to edit your bash shell resource file and add a shell function like this:

load_so3g () {
  # Load the appropriate python environment (edit this!)
  source "${HOME}/simons/bin/activate"

  # Location of our installed spt3g / so3g
  dir="${HOME}/so3g"

  # Prepend our executable search path
  export PATH="${dir}/bin:${PATH}"

  # Get the python major / minor version
  pyver=$(python3 --version 2>&1 | awk '{print $2}' \
    | sed -e "s#\(.*\)\.\(.*\)\..*#\1.\2#")

  # Put our python module into our search path
  export PYTHONPATH="${dir}/lib/python${pyver}/site-packages"

  # Prepend our executable path and shared library search path.
  # These lines are for linux:
  if [ -z ${LD_LIBRARY_PATH} ]; then
    export LD_LIBRARY_PATH="${dir}/lib"
  else
    export LD_LIBRARY_PATH="${dir}/lib:${LD_LIBRARY_PATH}"
  fi
  # These lines are for MacOS:
  #if [ -z DYLD_LIBRARY_PATH ]; then
  #    export DYLD_LIBRARY_PATH="${dir}/lib"
  #else
  #    export DYLD_LIBRARY_PATH="${dir}/lib:${DYLD_LIBRARY_PATH}"
  #fi
}

From a new shell, you can now run “load_so3g” to load your python stack and put your install prefix into your environment. If you use environment modules, see the README and example in the modules/ directory.

Special Note on Conda

If you are building spt3g / so3g from source, it is highly recommended that you install dependencies from OS packages, use the python3 from your OS or package manager, and use a virtualenv as described in the first section. If you use a conda-provided python, then you have two choices:

1. Build boost from source using that python (so that boost-python works correctly).

2. Install the conda package for boost (and other dependencies) and use conda provided compilers to build spt3g / so3g.

Both of these choices are beyond the scope of this README. Below we assume that you are using a virtualenv created with the system (or homebrew / macports) python3.

Prerequisites on Linux

The easiest approach in this case is to use your OS package manager. For example:

apt install \
libboost-all-dev \
libopenblas-openmp-dev \
libflac-dev

Make sure your python virtualenv is activated. Next, download and install spt3g_software (https://github.com/CMB-S4/spt3g_software). Check the major / minor version of your python (e.g. 3.7, 3.8 or 3.9). We use that information to install spt3g into the correct site-packages directory. Below we assume an install prefix of “${HOME}/so3g” and that we are using python3.9:

cd ${HOME}/git/spt3g_software
mkdir -p build
cd build
cmake \
  -DCMAKE_BUILD_TYPE=Release \
  -DCMAKE_C_COMPILER="gcc" \
  -DCMAKE_CXX_COMPILER="g++" \
  -DCMAKE_C_FLAGS="-O3 -g -fPIC" \
  -DCMAKE_CXX_FLAGS="-O3 -g -fPIC -std=c++11" \
  -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON \
  -DPython_EXECUTABLE:FILEPATH=$(which python3) \
  -DPYTHON_MODULE_DIR="${HOME}/so3g/lib/python3.9/site-packages" \
  -DCMAKE_INSTALL_PREFIX="${HOME}/so3g" \
  ..
make -j 2 install

Prerequisites on MacOS

The so3g / spt3g_software does not seem to run on MacOS when built with the clang++ compiler (unit tests fail with a cereal error). Instead, we will use homebrew to install our dependencies and the latest gcc compiler tools:

brew install \
flac \
bzip2 \
netcdf \
sqlite3 \
boost-python3 \
gcc

Next, download and install spt3g_software. Ensure that your virtualenv is activated. Check the major / minor version of your python (e.g. 3.7, 3.8 or 3.9). We use that information to install spt3g into our virtualenv or conda environment. Below we assume that our environment is in our home directory in a folder called “simons” and that we are using python3.9. We further assume that the homebrew gcc version is called “gcc-11”. Also, this assumes that homebrew is installing things to /usr/local:

cd ${HOME}/git/spt3g_software
mkdir -p build
cd build
cmake \
  -DCMAKE_BUILD_TYPE=Release \
  -DCMAKE_C_COMPILER="gcc-11" \
  -DCMAKE_CXX_COMPILER="g++-11" \
  -DCMAKE_C_FLAGS="-O3 -g -fPIC" \
  -DCMAKE_CXX_FLAGS="-O3 -g -fPIC -std=c++11" \
  -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON \
  -DBOOST_ROOT="/usr/local" \
  -DPython_EXECUTABLE:FILEPATH=$(which python3) \
  -DPYTHON_MODULE_DIR="${HOME}/so3g/lib/python3.9/site-packages" \
  -DCMAKE_INSTALL_PREFIX="${HOME}/so3g" \
  ..
make -j 2 install

Compilation and Installation

To compile and install the so3g package (assuming our same install prefix of $HOME/so3g), we need to point it to the spt3g build directory that we used previously. For example:

cd ${HOME}/git/so3g
mkdir -p build
cd build
cmake \
  -DCMAKE_PREFIX_PATH=${HOME}/git/spt3g_software/build \
  -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON \
  -DPYTHON_INSTALL_DEST="${HOME}/so3g" \
  -DCMAKE_INSTALL_PREFIX="${HOME}/so3g" \
  ..
make -j 2 install

The definition of CMAKE_PREFIX_PATH must point to the build directory for spt3g, because cmake output there will be used to generate best compilation and/or linking instructions for Boost and other dependencies of spt3g/so3g.

Now you can run your “load_so3g” (or similar) command whenever you want to load your python stack and also the so3g install prefix.

Local configuration through local.cmake

Optional, site-specific parameters may be set in the file local.cmake. Lines declaring set(VARIABLE, value) should have the same effect as passing -DVARIABLE=value to the cmake invocation.

To change the destination directory for the installation, add lines like this one:

set(CMAKE_INSTALL_PREFIX $ENV{HOME}/.local/)
set(PYTHON_INSTALL_DEST $ENV{HOME}/.local/lib/python3.7/site-packages/)

To point cmake to the spt3g build directory, add a line like this one:

set(CMAKE_PREFIX_PATH $ENV{HOME}/code/spt3g_software/build)

Testing

The unit tests are not installed with the so3g package, so in order to run them you must have a git checkout of so3g (even if you installed so3g from a pre-built wheel).

After installing the so3g package, you can run the unit tests by passing the path to the test directory to the pytest command:

pytest /path/to/so3g/test

You can run specific tests by calling them directly:

python3 -m unittest /path/to/so3g/test/test_indexed