Installation

Note: CGpipe is intended to run on a Unix/Linux workstation, server, or cluster (including macOS).

CGpipe is a Java program that is packaged as either a self-executing fat JAR file, or as an embeddable library. CGpipe is a Java program, but it has been developed to run on *nix-style hosts such as Mac OSX, FreeBSD or Linux. The only pre-requisite is a working installation of Java 1.7 or better. It is untested on Windows.

Step 0: Install Java

Java version 7 or higher can be used. Install Java from your Linux distribution (using apt-get or yum), or see the Oracle site for more details.

Step 1: Download cgpipe

You can download the cgpipe program from the cgpipe releases website. You can save this file anywhere, but it is easiest to use if it is included on your $PATH somewhere, such as in /usr/local/bin or in a personal $HOME/bin directory.

Step 2: Run a test pipeline

Most analysis pipelines you run will be custom written, however you can verify that your cgpipe installation is working with the following script:

[hello.cgp]
#!/usr/bin/env cgpipe

print "Hello from cgpipe!"

This will load cgpipe from your path and execute the above script. Right now the script doesn't do anything other than print a message to the console. If you save this as hello.cgp (and make it executable with chmod +x hello.cgp), you should see the following:

$ ./hello.cgp
Hello from cgpipe!

or…

$ cgpipe hello.cgp
Hello from cgpipe!

We will build from here to demonstate how to make your own pipelines in the next sections.

Step 3: Configure cgpipe for your computing environment

CGpipe can run on a single user workstation, server, or HPC cluster. If you want to run more complex workflows by submitting jobs to a scheduler, it's necessary to configure cgpipe to use your scheduler. In cgpipe, job submission is handled by "job runners".

The currently supported job schedulers are: SBS, SGE, SLURM, or PBS. For more information about available runners, or the possible configuration settings, see "Running jobs".

If no scheduler is configured, jobs will be written as a bash script to stdout.

For information about how to configure cgpipe, see: Configuring cgpipe.