omnetpy

OMNeT++ meets python

What does this repository provide?

TL;DR: implement OMNeT++ simulations using python programming language:

no need to know C++
all the power of python is suddenly available
faster prototypes
even using the OMNeT++ IDE (if you wish...)

Ok, I'm in... tell me more...:

Leveraging some design decisions made in the original project, it is possible to have OMNeT++ launch a python interpreter, instanciate classes you wrote in python, and use them as regular cSimpleModule subclasses.

The original code for OMNeT++ needs no change. This repository provides python bindings for C++ APIs (to allow the user subclass cSimpleModule from python) and some C++ headers to have OMNeT++ accept this code.

The project is the result of an academic research, see current state and limitations below.

How can I use it?

Launch a contianer:

you@yourmachine$ docker run --rm \
   -ti -e DISPLAY=$DISPLAY \
   -v /tmp/.X11-unix:/tmp/.X11-unix \
   mmodenesi/omnetpy bash

On the container, launch the omnetpp IDE:

userpp@container$ omnetpp

Follow the guide on how to create simulations with python.
Go deeper.

Who could be interested in this?

You want to build simulations using OMNeT++ but your C++ knowledge is weak.
You want to use OMNeT++ for educational purposes, and your students are more proficient in python than C++.
You want to prototype a model faster; you can rewrite the code in C++ later.
You are interested in making changes to the code and being able to use it without any extra compilation steps.

What is the current state of this project?

Excepting dyna, all OMNeT++ demostrative samples were successfuly ported to python (see pysamples), which means it is fairly usable.
The original goal of the project was to enable python implementation of OMNeT++ simluations, which was largely achieved. For the moment, no further improvements are being implemented (automatic bindings generation, QA, docker images for each release of OMNeT++, etc).

What are the limitations?

Dynamic deletion of modules is not supported (the dyna sample could not be ported).
python code debugging is not supported.
Automatic code generation for messages is not supported (yet, being python so flexible, it's not missed, either).
The python bindings were written by hand (an activity that could be automated) and may have subtle bugs.
The python bindings were written as needed in order to port the OMNeT++ samples. Not 100% of the C++ APIs were ported (though it may be easy to include what is missing).
This is an academic project. Many things were written "just to make it work" and may not scale well.
This project was not heavily used.
The optimization of the docker image (size, layering, etc.) was never a top priority.
Launching graphical applications (OMNeT++ IDE, simluations) from docker was only tested on ubuntu.

Licence

As a derivative work of OMNeT++, this code is distributed under Academic Public Licence version 1.1.

mmodenesi / omnetpy