Skip to main content

Forward-time population genetic simulation in Python

Project description

This is the README for fwdpy11, which is a Python package for forward-time population genetic simulation. It uses fwdpp as its C++ back-end.

Build status





Development branch:


Python code style:


  • Pickle-able population objects
  • Parallel computation via multiprocessing or concurrent.futures.
  • Custom temporal samplers to analyze populations during a simulation may be written in pure Python.
  • Flexible interface for simulating models with multiple populations.


The following must be present on your system:

  • GSL. This is a C library. It is available via conda. fwdpy11 requires version 2.3 or greater.
  • pybind11. This should be installed conda as appropriate for your system, or via your system’s package manager or manually. See note below.
  • cmake. This should be installed by conda or your favorite package manager.

fwdpy11 also uses fwdpp, which is included as a submodule.


The C++ modules are built using cmake, which requires that pybind11’s cmake macros are visible. Installing pybind11 via pip does not install the macros. However, installs using conda, apt-get, or manual installation from source will install both the Python module and the cmake macros.


GPLv3 or later (See COPYING)

Suppored Python version

fwdpy11 is written for Python 3. We will not modify the package to be compatible with Python 2.7.


Building from the git repository:

git submodule init
git submodule update
python build_ext -i
python -m unittest discover tests

Using pip on macOS and Linux (or pip3 as appropriate for your system):

pip install --upgrade fwdpy11

It is possible that the cmake macros to detect the GSL can fail to detect the correct version. Issues like this are a basic weakness of cmake. I’ve seen this in conda environments, where the macro prefers the system version over the newer version in the environment. To “fix” this, give it a hint:

GSL_ROOT_DIR=/path/to/gsl python3 build_ext -i


On Apple’s macOS, we strongly encourage that you use conda with their compiler packages:

conda install clang_osx-64 clangxx_osx-64

Installing these packages will mean that you can get away from the relatively old versions of these compilers that ship with Xcode. However, you do need to add the following flag when building the package:

On macOS versions prior to “Mojave”:

CONDA_BUILD_SYSROOT=/ python3 build_ext -i

For later versions, you may omit the environment variable prefix.

Do the same for a pip install from the source directory.


We have heard positive reports of using fwdpy11 on Windows 10 with the Ubuntu subsystem installed. For such a system, you may use a Linux conda installer and then install fwdpy11 via bioconda.

The developers do not have access to this platform, but we are keen to hear of any issues.


We use the GitHub “release” mechanism to make stable versions available. However, GitHub releases to not include the sub-modules, meaning that the releases themselves cannot be used for installation. (A related irony is that the Zenodo DOI for the releases are somewhat meaningless.)

To install a specific release:

  1. Use pip (see above). This is the recommended approach if you do not use conda.
  2. Install from bioconda. This is the recommended approach.
  3. Clone the repo, checkout the release, and update submodules:
git clone
cd fwdpy11
git submodule init
git submodule update

The latter method is probably the least appealing.

We have a strict policy of putting releases on PyPi and bioconda. If there is a release on PyPi but not on bioconda, then that is because we identified a bug and pushed a new release before the bioconda build happend. It happens. That’s life.

Enabling code profiling

By default, fwdpy11 is compiled with aggressive optimizations to help reduce the library size. One side effect is that it becomes impossible to accurately profile the code. To override these defaults:

python build_ext -i --enable-profiling


The package should not be installed with profiling enabled. This method of building is for developers who need to accurately profile the C++ back-end. Also note that only the main package is affected. Building the unit test modules is not affected.

Enabling debugging symbols in the C++ code

python build_ext -i --debug

Debug mode disables all compiler optimizations, allows C-like assertions, and generated debug symbols.


Never install the package compiled in debug mode! First, things will run much more slowly. Second, triggering an assertion will cause the Python interpreter to crash. These assertions exist as a brute-force method to help developers quickly identify bugs.

Enabling assertions in the C++ code

The fwdpp library code uses C’s assert macros in several places. These are disabled by default. However, it can be useful to enable them when hacking the code. To do so, you must manually set your compiler flags with cmake:


When compiling this way, fwdpy11 makes some extra checks that will throw RuntimeError if they fail. The fwdpp back end also makes extra checks. If those fail, abort will be called, which will crash the Python interpreter. Thus, compiling with this option is a “serious debugging mode only” option.

Enabling aggressive debugging of C++ STL templates using GCC

Use the following flags to enable an “extreme” debugging mode of the C++ standard template library:


Static analysis using clang-tidy

It is sometimes useful to go through the code and to a “static” analysis to look for problems. The clang-tidy tool is especially useful. For example:

find fwdpy11/src/ -name '*.cc' | xargs -I {} clang-tidy -checks='performance-*' {}  -- -I/usr/include/python3.7m -I./fwdpy11/headers/fwdpp -I./fwdpy11/headers


fwdpy11 is available through bioconda for Linux and for macOS:

conda install -c bioconda fwdpy11

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for fwdpy11, version 0.8.0
Filename, size File type Python version Upload date Hashes
Filename, size fwdpy11-0.8.0.tar.gz (1.9 MB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page