Code Monitor: Tells you when you write bad code

codemonitor is a development tool that automatically runs arbitrary commands when files change on disk and displays the output of the commands graphically.

codemonitor can be used for example to automatically run your tests when you save a source file.

I have recorded a demo that you can watch here: http://www.youtube.com/watch?v=oucd1p9cjrI&hd=1

Getting started

codemonitor is written in Haskell and runs on Linux. (The only Linux dependency though is inotify.)

codemonitor is uploaded to Hackage for easy installation. Once you have installed the Haskell Platform ( yum/apt-get install haskell-platform), you can install codemonitor like this:

# You need to have gtk installed first. Install it with you package manager
# before you continue. You might need dev libraries as well to be able to
# install the Haskell bindings.

# The install script depends on programs in this directory, so we must
# export it here. It is a good idea to always put this on your path so that
# cabal always refers to the cabal in ~/.cabal/bin.
export PATH="~/.cabal/bin:$PATH"

# Fore some reason, cabal does not install this dependency, so we have to
# install it manually.
cabal install gtk2hs-buildtools

# Then we can install codemonitor.
cabal install codemonitor

After this, codemonitor should be installed in ~/.cabal/bin/codemonitor.

If you have an old version of cabal, you might need to upgrade it. Before you install codemonitor you need to run this:

# This will install a new cabal in ~/.cabal/bin/, so make sure the path is
# exported when you try to install codemonitor.
cabal install cabal-install

To install a new version of codemonitor, run these commands:

cabal update
cabal install codemonitor

Please contact me if you have trouble installing.

Usage

Specify a configuration

First you need to specify a configuration file that tells codemonitor in which directory to look for file changes and what commands to run. The configuration for codemonitor itself looks like this:

.
tests \.hs$ sh run-tests
lint \.hs$ hlint .

The first line is the path to the directory to watch for file changes.

The rest of the lines are specifications of commands. They have the format "{id} {regexp} {executable} {arguments}".

The id must be unique among the commands.

The regexp is used to only run this command if the file that was changed matches the regexp. (Files ending with .hs in the example.)

Running it

After the application has been installed, you can run it like this, passing the path to the configuration file:

~/.cabal/bin/codemonitor monitor.config

Passing configuration to stdin

You can also pass the configuration directly to stdin. It is useful if you only want a single file to run codemonitor. It can look something like this:

#!/bin/sh
~/.cabal/bin/codemonitor <<EOF
.
tests \.hs$ sh run-tests
lint \.hs$ hlint .
EOF

Known bugs/limitations

Parallel jobs

If a file changes while a job is running, a new job is started immediately. The problem is that the old job is not stopped. So you get two jobs running in parallel. That is very problematic when compiling code for example since the parallel jobs touch the same files.

A workaround for this limitation is to create a shell script that runs the command you want to run, and before it does that, it kills the old version of itself. It can be done with a snippet like this:

pidfile=lastpid.txt
if [ -e $pidfile ]; then
    kill -9 $(cat $pidfile)
    rm $pidfile
fi
echo $$ > $pidfile

... run your command here ...

rm -f $pidfile

Help

Please contact me if you have questions or problems.

rickardlindberg/codemonitor