neworder
neworder is a microsimulation framework inspired by openm++, MODGEN and, to a lesser extent, the python-based LIAM2 tool, and can be thought of as a powerful best-of-both-worlds hybrid of MODGEN and LIAM2. Modellers can define their models in a simple, well-known language, yet benefit from the efficiency of compiled code and parallel execution:
- python module: easy to install and integrate, available on all common platforms
- low barriers to entry: users need only write standard python code, little or no new coding skills required.
- flexibility: models are specified in python code, so can be arbitrarily complex
- data agnosticism: the framework does not impose any constraints on data formats for either sources or outputs.
- reusability: leverage python modules like numpy, pandas and matplotlib.
- reproducibility: built-in, customisable random generator seeding strategies
- speed: the module is predominantly written in optimised C++ and provides fast Monte-Carlo, statistical and data manipulation functions.
- compatibility: operate directly on numpy arrays and pandas DataFrames
- scalability: can be run on a desktop or a HPC cluster, supporting parallel execution using MPI.
System Requirements
neworder requires python 3.6 or above and runs on 64-bit linux, OSX and Windows platforms. In order to take advantage of the parallel execution functionality, the following are also required:
- an MPI implementation, such as mpich, open-mpi or ms-mpi
- the mpi4py package that provides python MPI bindings
but the module will work perfectly well in serial mode without these.
Installation
PyPI
pip install neworder
Conda
conda install -c conda-forge neworder
Docker
The docker image contains all the examples, and should be run interactively. Some of the examples require permission to connect to the host's graphical display.
docker pull virgesmith/neworder
xhost +
docker run -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY -it virgesmith/neworder
NB The above works on ubuntu but may require modification on other OSs.
Then in the container, e.g.
python examples/mortality/model.py
Documentation
To get started first see the detailed documentation here. Then, check out "Hello World" and the other examples.