JRuby-Rack is a lightweight adapter for the Java Servlet environment that allows any (Ruby) Rack-based application to run unmodified in a Java Servlet container. JRuby-Rack supports Rails as well as any Rack-compatible Ruby web framework.
For more information on Rack, visit http://rack.github.io/.
JRuby-Rack 1.1.x aims to be compatible with JRuby >= 1.6.4 (we recommend 1.7.x), Generally, any container that supports Java Servlet >= 2.5 (JEE 5) is supported.
The most-common way to use JRuby-Rack with a Java server is to get Warbler. Warbler depends on the latest version of JRuby-Rack and ensures it gets placed in your WAR file when it gets built.
If you're assembling your own WAR using other means, you can install the jruby-rack gem. It provides a method to locate the jar file:
require 'jruby-rack' FileUtils.cp ., '.'
Otherwise you'll need to download the latest jar release, drop it into the
WEB-INF/lib directory and configure the
RackFilter in your application's
web.xml (see following examples).
Alternatively you can use a server built upon JRuby-Rack such as Trinidad with sensible defaults, without the need to configure a deployment descriptor.
Here's sample web.xml configuration for Rails. Note the environment and
min/max runtime parameters. For multi-threaded (a.k.a.
Rails with a single runtime, set min/max both to 1. Otherwise, define the size
of the runtime pool as you wish.
<context-param> <param-name>rails.env</param-name> <param-value>production</param-value> </context-param> <context-param> <param-name>jruby.min.runtimes</param-name> <param-value>1</param-value> </context-param> <context-param> <param-name>jruby.max.runtimes</param-name> <param-value>1</param-value> </context-param> <filter> <filter-name>RackFilter</filter-name> <filter-class>org.jruby.rack.RackFilter</filter-class> <!-- optional filter configuration init-params : --> <init-param> <param-name>resetUnhandledResponse</param-name> <param-value>true</param-value> <!-- true (default), false or buffer --> </init-param> <init-param> <param-name>addsHtmlToPathInfo</param-name> <param-value>true</param-value> <!-- true (default), false --> </init-param> <init-param> <param-name>verifiesHtmlResource</param-name> <param-value>false</param-value> <!-- true, false (default) --> </init-param> </filter> <filter-mapping> <filter-name>RackFilter</filter-name> <url-pattern>/*</url-pattern> </filter-mapping> <listener> <listener-class>org.jruby.rack.rails.RailsServletContextListener</listener-class> </listener>
(Other) Rack Applications
The main difference when using a non-Rails Rack application is that JRuby-Rack
looks for a "rackup" file named config.ru in
WEB-INF/*/config.ru. Here's a sample web.xml configuration :
<filter> <filter-name>RackFilter</filter-name> <filter-class>org.jruby.rack.RackFilter</filter-class> <!-- optional filter configuration init-params (@see above) --> </filter> <filter-mapping> <filter-name>RackFilter</filter-name> <url-pattern>/*</url-pattern> </filter-mapping> <listener> <listener-class>org.jruby.rack.RackServletContextListener</listener-class> </listener>
If you don't have a config.ru or don't want to include it in your web app, you can embed it directly in the web.xml as follows (using Sinatra as an example):
<context-param> <param-name>rackup</param-name> <param-value> require 'rubygems' gem 'sinatra', '~> 1.3' require './lib/app' set :run, false set :environment, :production run Sinatra::Application </param-value> </context-param>
Be sure to escape angle-brackets for XML !
JRuby-Rack's main mode of operation is as a filter. This allows requests for
static content to pass through and be served by the application server.
Dynamic requests only happen for URLs that don't have a corresponding file, much
like many Ruby/Rack applications expect. The (default) filter we recommend
org.jruby.rack.RackFilter, the filter supports the following
- responseNotHandledStatuses which statuses (when a filter chain returns) should be considered that the response has not been handled (default value: "403,404,405") and should be dispatched as a Rack application
- resetUnhandledResponse whether an unhandled response from the filter chain gets reset (accepts values "true", "false" and "buffer" to reset the buffer only), by default "true"
- addsHtmlToPathInfo controls whether the .html suffix is added to the URI when checking if the request is for a static page
- verifiesHtmlResource used with the previous parameter to makee sure the requested static resource exist before adding the .html request URI suffix
The application can also be configured to dispatch through a servlet instead of
a filter, the servlet class name is
Servlet Environment Integration
- servlet context is accessible to any application through the Rack environment
variable java.servlet_context (as well as the
- the (native) servlet request and response objects could be obtained via the java.servlet_request and java.servlet_response keys
- all servlet request attributes are passed through to the Rack environment (and thus might override request headers or Rack environment variables)
- servlet sessions can be used as a (java) session store for Rails, session attributes with String keys (and String, numeric, boolean, or java object values) are automatically copied to the servlet session for you.
Several aspects of Rails are automatically set up for you.
ActionController::Base.relative_url_rootis set for you automatically according to the context root where your webapp is deployed.
Rails.loggeroutput is redirected to the application server log.
- Page caching and asset directories are configured appropriately.
JRuby Runtime Management
JRuby runtime management and pooling is done automatically by the framework. In the case of Rails, runtimes are pooled by default (the default will most likely change with the adoption of Rails 4.0). For other Rack applications a single shared runtime is created and shared for every request by default. As of 1.1.9 if jruby.min.runtimes and jruby.max.runtimes values are specified pooling is supported for plain Rack applications as well.
We do recommend to boot your runtimes up-front to avoid the cost of initializing one while a request kicks in and find the pool empty, this can be easily avoided by setting jruby.min.runtimes equal to jruby.max.runtimes. You might also want to consider tuning the jruby.runtime.acquire.timeout parameter to not wait too long when all (max) runtimes from the pool are busy.
JRuby-Rack can be configured by setting these key value pairs either as context init parameters in web.xml or as VM-wide system properties.
rackup: Rackup script for configuring how the Rack application is mounted. Required for Rack-based applications other than Rails. Can be omitted if a config.ru is included in the application root.
public.root: Relative path to the location of your application's static assets. Defaults to /.
rails.root: Root path to the location of the Rails application files. Defaults to /WEB-INF.
rails.env: Specify the Rails environment to run. Defaults to 'production'.
rails.relative_url_append: Specify a path to be appended to the
ActionController::Base.relative_url_rootafter the context path. Useful for running a rails app from the same war as an existing app, under a sub-path of the main servlet context root.
gem.path: Relative path to the bundled gem repository. Defaults to /WEB-INF/gems.
jruby.compat.version: Set to "1.8" or "1.9" to make JRuby run a specific version of Ruby (same as the --1.8 / --1.9 command line flags).
jruby.min.runtimes: For non-threadsafe Rails applications using a runtime pool, specify an integer minimum number of runtimes to hold in the pool.
jruby.max.runtimes: For non-threadsafe Rails applications, an integer maximum number of runtimes to keep in the pool.
jruby.runtime.init.threads: How many threads to use for initializing application runtimes when pooling is used (default is 4). It does not make sense to set this value higher than
jruby.runtime.init.serial: When using runtime pooling, this flag indicates that the pool should be created serially in the foreground rather than spawning (background) threads, it's by default off (set to false). For environments where creating threads is not permitted.
jruby.runtime.acquire.timeout: The timeout in seconds (default 10) to use when acquiring a runtime from the pool (while a pool maximum is set), an exception will be thrown if a runtime can not be acquired within this time ( accepts decimal values for fine tuning e.g. 1.25).
jruby.runtime.env: Allows to set a custom ENV hash for your Ruby environment and thus insulate the application from the environment it is running. By setting this option to en empty string (or 'false') it acts as if the ENV hash was cleared out (similar to the now deprecated
jruby.runtime.env.rubyopt: This option is used for compatibility with the (deprecated)
jruby.rack.ignore.envoption since it cleared out the ENV after RUBYOPT has been processed, by setting it to true ENV['RUBYOPT'] will be kept.
jruby.rack.logging: Specify the logging device to use. Defaults to
servlet_context. See below.
jruby.rack.request.size.initial.bytes: Initial size for request body memory buffer, see also
jruby.rack.request.size.maximum.bytes: The maximum size for the request in memory buffer, if the body is larger than this it gets spooled to a tempfile.
jruby.rack.response.dechunk: Set to false to turn off response dechunking (Rails since 3.1 chunks response on
render stream: true), it's on by default as frameworks such as Rails might use
Rack::Chunked::Bodyas a Rack response body but since most servlet containers perform dechunking automatically things might end double-chunked in such cases.
jruby.rack.handler.env: EXPERIMENTAL Allows to change Rack's behavior on obtaining the Rack environment. The default behavior is that parameter parsing is left to be done by the Rack::Request itself (by consuming the request body in case of a POST), but if the servlet request's input stream has been previously read this leads to a limitation (Rack won't see the POST paras). Thus an alternate pure 'servlet' env "conversion" is provided that maps servlet parameters (and cookies) directly to Rack params, avoiding Rack's input parsing.
jruby.rack.filter.adds.html: deprecated use
addsHtmlToPathInfofilter config init parameter. The default behavior for Rails and many other Ruby applications is to add an .html extension to the resource and attempt to handle it before serving a dynamic request on the original URI. However, this behavior may confuse other servlets in your application that have a wildcard mapping. Defaults to true.
jruby.rack.filter.verify.resource.exists: deprecated use
verifiesHtmlResourcefilter config init parameter. If
jruby.rack.filter.adds.htmlis true, then this setting, when true, adds an additional check using
ServletContext#getResourceto verify that the .html resource exists. Default is false. (Note that apparently some servers may not implement
getResourcein the way that is expected here, so in that case this setting won't matter.)
There are often cases where you need to perform custom initialization of the Ruby environment before booting the application. You can create a file called META-INF/init.rb or WEB-INF/init.rb inside the war file for this purpose. These files, if found, will be evaluated before booting the Rack environment, allowing you to set environment variables, load scripts, etc.
For plain Rack applications, JRuby-Rack also supports a magic comment to solve the "rackup" chicken-egg problem (you need Rack's builder loaded before loading the config.ru, yet you may want to setup the gem version from within the rackup file). As we ship with the Rack gem bundled, otherwise when executing the provided config.ru the bundled (latest) version of Rack will get loaded.
Use rack.version to specify the Rack gem version to be loaded before rackup :
# encoding: UTF-8 # rack.version: ~>1.3.6 (before code is loaded gem '~>1.3.6' will be called)
Or the equivalent of doing
bundle exec rackup ... if you're using Bundler :
# rack.version: bundler (requires 'bundler/setup' before loading the script)
JRuby-Rack sets up a delegate logger for Rails that sends logging output to
javax.servlet.ServletContext#log by default. If you wish to use a different
logging system, configure
jruby.rack.logging as follows:
servlet_context(default): Sends log messages to the servlet context.
stdout: Sends log messages to the standard output stream
slf4j: Sends log messages to SLF4J. SLF4J configuration is left up to you, please refer to http://www.slf4j.org/docs.html .
log4j: Sends log messages to log4J. Again, Log4J configuration is left up to you, consult http://logging.apache.org/log4j/ .
commons_logging: Routes logs to commons-logging. You still need to configure an underlying logging implementation with JCL. We recommend using the logger library wrapper directly if possible, see http://commons.apache.org/logging/ .
jul: Directs log messages via Java's core logging facilities (util.logging).
For those loggers that require a specific named logger, set it with the
jruby.rack.logging.name option, by default "jruby.rack" name will be used.
Checkout the JRuby-Rack code using git :
git clone git://github.com/jruby/jruby-rack.git cd jruby-rack
Ensure you have Maven installed. It is required for downloading jar artifacts that JRuby-Rack depends on.
Build the .jar using Maven :
the generated jar should be located at target/jruby-rack-*.jar
Alternatively use Rake, e.g. to build the gem (skipping specs) :
jruby -S rake clean gem SKIP_SPECS=true
You can not use JRuby-Rack with Bundler directly from the git (or http) URL
gem 'jruby-rack', :github => 'jruby/jruby-rack') since the included .jar file
is compiled and generated on-demand during the build (it would require us to
package and push the .jar every time a commit changes a source file).