James Britt, Logan Barnett, David Koontz
Rawr is a packaging solution for JRuby applications. Rawr comes in two pieces:
rawrcommand that creates a Java
mainfile in your application and a configuration file that will be used by Rawr to build the final jar;
- a Rake task file that you can include into your project's Rakefile to automate the creation of the packaged application.
rawr install rake rawr:jar java -jar package/jar/your_jar_file.jar
- JRuby >= 1.7.0
sudo gem install rawr --source http://gems.neurogami.com
Note: You probably do not want to use
sudo if you are using a Ruby installed using
rvm or something similar.
Configuring Rawr to compile ruby files into class files breaks packaging.
The purpose of Rawr is to take a pre-defined set of files and package them to make a self-contained executable Java application.
Rawr will generate a Java file,
Main.java, that serves as the starting point of a Java-executable
This file in turn loads your Ruby code and executes it via JRuby.
The presumption is that all the files needed for your program are included in the final bundling generated by Rawr.
Your code should not count on loading files or libraries that depend on any gems being available.
If you want to use code provided by a gem you need to unpack the gem files into your project directory and configure
build_configuration.rb to include them.
You might also want to alter
$LOAD_PATH) so that your extracted and bundled gem files are first.
When distrubuting a rawr-packaged program you need to keep in mind that you may not be able to control the end-user environment.
If you are loading gem files you need to consider that an end-user might have a different version of the gem installed.
MUNGING THE LOAD PATH
The Monkeybars library provides a file,
manifest.rb, that serves to prepare the execution environment.
It sets up some load paths based on whether the code is running from a jar or not, and attempts to strike rubygems from the load path to ensure that a program is self-contained.
(Rawr was created out of a need to package Monkeybars programs.)
Current versions of Ruby load
rubygems by default. The rubygems library alters the default
require method such that if you try to load a file and it is not found an attempt is made to locate it among the installed gems. If that suceeds then that gem path gets added to
$: and the file gets loaded.
require method is redefined, but first it is aliased (so that the method itself is still available).
There are two ways you can thwart this
module Kernel alias require gem_original_require end
module Kernel def require path gem_original_require path end end
In either case you will need to have this code in your program before any other code attempts to load a gem file.
Rawr 1.7.0 had some changes to how the application jar is assembled. Many versions ago something changed such that when the application jar-building processing was collecting the files indicated by
build_configuration.rb it was stripping some paths. For example your project folder might have files in
lib/ruby/ but those files would end up in the root of the generated jar. For assorted reasons this did not seem to break anything, likely because these files were still available via
$:. But this was just a coincidence.
Worse, if you had
src/foo.rb your generated jar file would only get one of them. This is bad.
Version 1.7.0 changes how such paths are handled and the behavior should be what it was way back in the early days of
If you find that this new version is breaking your existing applications please report this. It shouldn't (based on testing) but there may be cases where it does it and it would be good to know why. If your program breaks with this version it may be that you need to change how to add pats to
$: or how you reference files when calling
Rawr 1.6.6 has rb source compilation turned off by default but compilation is working again since
Rawr 1.6.5, if you want to turn it on in
Rawr 1.6.0 added support for compiling Mirah source code.
There was already code in place for
duby files, but a) duby morphed in Mirah, and b) the compilation command is somewhat different.
There's a new
build_configuration.rb option to define the root folder for your Mirah files, and the resulting compiled
.class files end up where any
.java files would go.
Rawr 1.4.2 introduced the use of Brian Marick's user-choices library to handle initial configuration properties.
What this means in practice is that there are multiple ways to configure how
rawr handles the
You can use command-line arguments much as before, or use a configuration file (
~/.rawr), or environment variables.
Or all of them; you can mix and match.
For example, if you have a preferred name or location for the main Java class then you might want to stick that in the config
file or some environment variables to avoid having to pass them as command-line arguments on each invocation of
# in ~/.rawr local_jruby_jar: /home/james/JRUBY_JARS/1.7.3/jruby-complete.jar
Please read the docs for
user-choices, but one key thing to know is the precedence for options.
Any option value passed on the command-line overrules any previous value.
Any option defined in an environment variable overrules the value in a config files.
Option values in the config file will be used so long as they are not overridden by the above conditions.
Rawr is released under the Ruby License.
Feed your head.