In this post, we're going through the installation of IHaskell on GNU/Linux (Ubuntu 14.04 in my case, though it should be pretty similar on other ditributions) step-by-step. We'll see how the use of Stack will simplify the whole process, and how to get all the dependencies right!

Install Stack

From Stack official documentation, for Ubuntu 14.04:

sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 575159689BEFB442
echo 'deb http://download.fpcomplete.com/ubuntu trusty main' \
    |sudo tee /etc/apt/sources.list.d/fpco.list
sudo apt-get update && sudo apt-get install stack -y

You should now be able to run the stack executable in a shell:

stack --help

Being up-to-date is important, so setup your Stack global configuration ~/.stack/global/stack.yaml:

flags: {}
resolver: lts-4.0
packages: []
extra-deps: []

Tell Stack to setup everything (download package list, install latest GHC release, etc.):

stack setup

Note that you can lookup the latest release of stackage at anytime on this page.

Install ZeroMQ latest version

If you try to install IHaskell directly, you'll get an error since Ubuntu 14.04 ships with an older version of ZeroMQ. We need a more recent release. There are several ways to install it:

1. Compile it from source
2. Use Nix

We're going to compile it from source. I'll follow the way of the official README:

# Compiling from source:
git clone git@github.com:zeromq/zeromq4-x.git libzmq
cd libzmq
./autogen.sh && ./configure && make
sudo make install
sudo ldconfig

That will do the trick! If you don't like to install packages in global scope, feel free to install it in a user folder (Add the --prefix=LOCATION directive to ./configure, and change LD_LIBRARY_PATH and PATH in your shell configuration accordingly).

Install Jupyter

EDIT: Following Florian's comment, it appears that IPython is now officialy named Jupyter, so we might as well install it in our virtualenv. The only change in the instructions is to replace pip install ipython by pip install jupyter (Note that IPython is a dependency of Jupyter).

IHaskell requires a recent version of Jupyter, so we need to install it ourselves. There are several options:

1. Use pip and install it globaly (pip install jupyter)
2. Use pip and install it in a virtualenv (This is what we will do here)
3. Use nix (You're on your own)
4. Use conda (conda update jupyter)

If you're on a fresh install: sudo apt-get install python-virtualenv python-dev ncurses-base

With virtualenv:

virtualenv venv-ihaskell
source venv-ihaskell/bin/activate

With virtualenvwrapper:

mkvirtualenv ihaskell
workon ihaskell

Now we install Jupyter:

pip install jupyter

Install IHaskell

Nothing simpler, just use Stack:

stack build ihaskell

You may want to install extra packages to enhance IHaskell capabilities. Here are the ones supported by Stackage:

• ihaskell-aeson
• ihaskell-blaze
• ihaskell-charts
• ihaskell-diagrams
• ihaskell-rlangqq
• ihaskell-magic
• ihaskell-juicypixels
• ihaskell-hatex
• ihaskell-basic

Some others are not in the LTS snapshot of Stackage, but could be useful in the future. I don't know about the current maturity of these packages:

• ihaskell-widgets (conflict with current version of singletons)
• ihaskell-parsec (conflict with current version of aeson)
• ihaskell-plot

Setup IHaskell

If everything went well, you should have Jupyter (latest version) and ihaskell installed properly. The last step is to install the IHaskell Kernel into Jupyter.

stack exec ihaskell -- install

Now run the notebook server and enjoy:

stack exec jupyter -- notebook

Open your browser and go to: http://localhost:8000

What's next

Thanks to Stack, the install process went very smoothely (appart from some dependency we had to install ourselves). Using IPython with a Haskell Kernel I believe writing about Haskell will be a much pleasant experience since it's really easy to export a notebook either in HTML or Markdown.