This guide explains how to wrap a Java-based app using the "bash" app
interpreter. Most apps that run existing Java analyses can be ported to run on
the DNAnexus Platform without the need for the Java API bindings. For more
complex apps that need to make API calls in-process with Java, see the full
Java app tutorial.
To get started, follow the Writing DNAnexus Apps in
bash tutorial, which will walk you through
using dx-app-wizard to generate a bash app. The wizard will generate some
boilerplate for your app, including dx download and dx upload commands to
download any file inputs and upload any file outputs that you've declared at
this point.
Then make the following changes to the generated app:
Supply your Java program and invoke it: in the main shell script, where it
says "Fill in your application code here", invoke your program. Here is an
example:
java -jar /opt/myapp/MyApp.jar --input=infile --output=outfile
For this example, you would also copy the .jar file for your app to
resources/opt/myapp/MyApp.jar so that Java can find it.
By default, if you define an input called infile and an output called
outfile, the app wizard adds boilerplate to your app that downloads the input
file to infile in your app's working directory, and expects your app to write
its output to outfile in the same directory (from which it will be uploaded
back to the platform). If your app follows some other convention for its input
or output filenames, then change the filenames in the boilerplate code
(highlighted in bold below).
dx download "$infile" -o infile
# Your Java invocation goes here
# ...
outfile=$(dx upload outfile --brief)
dx-jobutil-add-output outfile "$outfile" --class=file
Install a Java runtime: add the following to your dxapp.json's
execDepends to install OpenJDK's Java 6 runtime before your app begins:
{
...
"runSpec": {
...
"execDepends": [{"name": "openjdk-6-jre-headless"}],
...
},
...
}
If you wish to use Java 7, openjdk-7-jre-headless is also available in the
execution environment.
