DCProgs in virtual environments

I have had trouble getting the virtual env to correctly find the dynamic library (liblikelihood.so/dylib)

When using virtualenv-wrapper there is a number of postactivate/postdeactivate and so on scripts that we can make set the LD_LIBRARY_PATH and remove it again. (DYLD_LIBRARY_PATH on OSX)

Something like the following should work.

In postactivate to set DYLD_LIBRARY_PATH

export _OLD_DYLD_LIBRARY_PATH="$DYLD_LIBRARY_PATH"
export DYLD_LIBRARY_PATH="$VIRTUAL_ENV/lib:$DYLD_LIBRARY_PATH"

In postdeactivate to remove the setting again

if ! [ -z ${_OLD_DYLD_LIBRARY_PATH+x} ] ; then
    DYLD_LIBRARY_PATH="$_OLD_DYLD_LIBRARY_PATH"
    export DYLD_LIBRARY_PATH
    unset _OLD_DYLD_LIBRARY_PATH
fi

Jupyter

If you have a global IPython4/Jupyter notebook setup you can create a kernel that runs with the global install of the notebook but inside the virtualenv. In the virtualenv you want to do pip install ipykernel

Next you have to define a kernel.json file that tell Jupyter how to run the kernel. This can be done mostly automatically see http://ipython.readthedocs.org/en/latest/install/kernel_install.html From the virtualenv running something like

python -m ipykernel install --user --name dcprogsenv --display-name "DCProgs Env"

should do the trick. Next we probably have to add the DYLD_LIBRARY_PATH to the kernel spec. See below:

The following will tell you where to install the kernels files. They go into a kernels subdirectory of the following:

jupyter --data-dir
/Users/jhn/Library/Jupyter

I.e in my case the kernel is defined in /Users/jhn/Library/Jupyter/kernels/dcprogstest/kernel.json and you can add logo files there too if you wish named logo-32x32.png and logo-64x64.png

The kernel file look like the following:

{
 "language": "python",
 "display_name": "DCProgs Python 3",
 "argv": [
  "/Users/jhn/Envs/dcprogstest/bin/python",
  "-m",
  "ipykernel",
  "-f",
  "{connection_file}"
 ],
  "env": {
      "DYLD_LIBRARY_PATH": "/Users/jhn/Envs/dcprogstest/lib/"
 }
}

Here you have to modify the path to python inside the env and set the DYLD_LIBRARY_PATH same as above