Skip to content
Improve your shell by making it functional through Haskell! (An update to Awkward)
Branch: master
Clone or download
iostreamer-X Merge pull request #10 from BrunoChevalier/master
Allow newer versions of the dependencies too.
Latest commit b6fa9fb May 12, 2017
Type Name Latest commit message Commit time
Failed to load latest commit information.


Improve your shell by making it functional through Haskell! (An update to Awkward)

What is it?

This program is an alternative to 'awk' or at least tries to be, and lets you perform 'awk' like operations through Haskell. So, if you know basic Haskell syntax, you are already set.

Why is it?

As an exercise, I wanted to update Awkward but in Haskell. A few minutes with Awkward and you would notice its drawbacks.

Since Awkward is a shell and that too a minimal at best, a lot of your customized shell's functionality is lost. Hence I decided to implemented FuncShell as an executable program and rely on pipes.


fsh follows the same philosophy as Awkward. Output of most bash commands can be expressed as [[String]] or [String] and one can use standard Haskell functions to modify this data representation.

Additionally, Haskell's pattern matching works really well with this representation. And its lambda functions along with partially applied operator functions ((+3) or (++" append") etc) make the syntax clean.

How does it work?

The command output is relayed to the executable through pipes. That output looks like this:

  PID TTY          TIME CMD
13750 pts/14   00:00:00 bash
25193 pts/14   00:00:03 node
25271 pts/14   00:00:00 sh
25272 pts/14   00:00:00 ps

This is parsed, cleaned and converted to something like this:

[ [ "PID", "TTY", "TIME", "CMD" ],
[ "13750", "pts/14", "00:00:00", "bash" ],
[ "25283", "pts/14", "00:00:00", "sh" ],
[ "25284", "pts/14", "00:00:00", "ps" ] ]

The function you provide, let's say filter (\(pid:_) -> read pid > 9000) is applied over this data structure. The headers are handled appropriately.


wget "" -O fsh && sudo chmod +x fsh && sudo mv fsh /usr/local/bin

You might need to install libgmp-dev for this to work.


After installing, you can use this program through fsh. I aliased it as ']', so my usages look like this.

la |] -p table 'filter (\(_:_:_:_:_:month:_) -> month == "Oct")'

The program expects the input to be received through pipes, always.

Next, it has a few flags. You can use -h to get help, and -p to choose your parser.

fsh supports plugins, which means you can install external modules through cabal(package manager, you can think of it as npm) and get them to work with fsh

And finally, you are supposed to supply a haskell function. The function must take an input of type [[String]] and give an output of the same type.

To sum up, this is how this program is supposed to be used:

bash command |fsh flags 'haskell function'


You can use the -p flag to choose your parser, depending on the kind of output you are expecting from the command. For example, ls gives a simple list, so the default parser is used and hence there is no need to specify anything. Whereas, the ps command outputs a table, hence you must specify the table parser through -p table.

The parsers mentioned above are built-in. You can also download and install external parsers. To do that you'd need cabal. Once installed, cabal update && cabal install fsh-csv will install the csv parser.

To use it, do echo "a,b,c\n1,2,3" |] -p FSH_CSV 'myHaskellFunction'


Making plugins/external parsers is really easy and involves very little fuss. All you need to do is:

  • Make a library project(Using cabal or something else if you want).
  • Build a function named run with type signature run :: String -> String -> IO (). The first argument is the function, and the second is command output
  • Publish to hackage
  • Rejoice!

I use hint to import run from your module and use it as a local function.

Here is the implementation of TableParser to give you an idea. I used this as a base to make the csv parser.

You can’t perform that action at this time.