Putting your ducks in a row

For the following examples to work, you will need the follow depedencies:

• Sun’s JNDI File System Service Provider. Follow “Download JNDI 1.2.1 & More” to download File System Service Provider. You will need the fscontext and the providerutil jars both in you classpath. I would like to take this moment and say that Sun needs to get their head out of the mud and join the Maven party, instead of forcing people to navigate their cumbersome site.

• Maven will make you hate less. Manually managing the cadre of jars that Java demands will just chalk you full of spite. While Maven does not alleviate the problem, it does put a purdy bow on it

• A given, you must already have gone through the steps of setting up a JRuby runtime. Steps might be to strong of a word, basically it is unpackage and go.

Configuring Rails to create its own JNDI and Connection Pool

Now it is time to get the show on the road. The following setups will show how to startup a JNDI instance, register a Connection Pool, and then use the same Connection Pool in Rails and in a Spring

Setting up the database.yml

The latest version of activerecord-jdbc already supports using JNDI for connections. By setting additional information in database.yml, jndi_factory_initial and jndi_provider_url, a connection pool can be built and registered into JNDI. The jndi_factory_initial relates to the JNDI property java.naming.factory.initial, which boils down to the JNDI service that is going to be created. The jndi_provider_url relates to the JNDI property java.naming.provider.url, which is the URL to connect to the JNDI service. The standard setting, jndi, is going to be used as the location to store and retrieve the Connection Pool.

… more

I want to expose tons of existing Java code and libs into Rails, but do not want to deal with a segmented world. Everything should be configured and started from one place, this is were a JRuby Service helps. The JRuby runtime will automatically load a class implementing a JRuby service interface (such as org.jruby.runtime.load.BasicLibraryService) when it is packaged in a specifically named jar at a specifically named placed.

Originally I was jumping though hoops trying to call static singletons, but ended up scratching my head how to get access to the Rails runtime. After finding hints from Ola Bini, with some searching and source diving, I found the javadoc for org.jruby.runtime.load.LoadService has the key. (An aside, it would be nice if the JRuby folks published the JavaDocs, even if it changed with every release).

Except from LoadService.java JavaDoc:

How to make a class that can get required by JRuby

First, decide on what name should be used to require the extension. In this purely hypothetical example, this name will be ‘active_record/connection_adapters/jdbc_adapter’. Then create the class name for this require-name, by looking at the guidelines above. Our class should be named active_record.connection_adapters.JdbcAdapterService, and implement one of the library-interfaces. The easiest one is BasicLibraryService, where you define the basicLoad-method, which will get called when your library should be loaded.

The next step is to either put your compiled class on JRuby’s classpath, or package the class/es inside a jar-file. To package into a jar-file, we first create the file, then rename it to jdbc_adapter.jar. Then we put this jar-file in the directory active_record/connection_adapters somewhere in JRuby’s load path.

The short and skinny of this? The package defines the directory and the class name determines the jar name. The package period separators are converted to slashes, so the package com.slackworks would be the directory com/slackworks. For brevity, this example uses the simple package slackworks, for a directory of slackworks. The class name, sans Service and converted from CamelCase to under scores determines the jar name. So the RailsSpringService class translates to rails_spring.jar

… more

We setup each hostname with their own wildcard by default, so that foo.hostname is always an alias (CNAME) for hostname.

sprocket                    IN      A               69.27.234.180
*.sprocket                  IN      CNAME           sprocket

This means that if I want to deploy my new app, The Awesomator, I can use the hostname ‘awesomator.sprocket.slackworks.com’ without having to change anything in DNS.

Then I drop my virtual host config awesomator.conf file, largely stolen from Coda Hale’s config , into Apache’s config directory (/etc/apache2/sites-available/ on Debian, /etc/httpd/conf.d/ on Red Hat). An Apache reload and a few mongrels later and I’m up and running on my new hostname.

… more

After a lot of fruitless Google work (911 is a noisy search term), even checking the unofficial Every Sprint 3g Error Code page, I settled in on trial and error.

It turns out that this is a kind of permissions problem, caused by the “Download Domain” of the new version not matching the old one.

When an application is downloaded the device stores the hostname part of the URL, any updates to the application must come from the same hostname. So if you downloaded it from www.foo.com the first time, you can’t update it from m.foo.com.

On the v9m (RAZR 2) you can find out what the current “Download Domain” is for a specific application by going to the “Applications” folder under “My Content” and bringing up the “Options” → “Properties” menu for it. The “Download Domain” will be listed at the bottom of the scrollable page.

… more

Tim Kofol took my ideas and ran with them, creating an ingenious way to maintain CouchDB views right in the Ruby code. Witness:

class Post < Blarg::CouchBase 
  couch_accessor :tags

  couch_view :comments_view, %[
    function(doc) {
      if (doc.type  "Comment") {
        map(doc.post_id, doc);
      }
    }
  ]

  couch_view :all_tags_view, %[
    function(doc) {
      if (doc.type  ‘Post’) {
        map(null, doc.tags);
      }
    }
  ]

  def self.all_tags   
    self.all_tags_view.flatten.uniq.sort
  end

  def comments
    comments = self.class.comments_view(:key => self.document_id)
    comments.sort{|a,b| a.created_at <=> b.created_at}
  end
end

With the views relevant to this class defined in the class itself, we need to load them into CouchDB. I added Rake task which finds all views defined in the project and loads them into CouchDB, creating them or replacing older versions.

rake db:views:load

… more

Do you wish you could leverage the power of Selenium? Do you long for it’s “automated web application UI tests”, but are boned by some headless horseman of a server? I have felt your pain. Tons of UI tests, but unable run them on the integration server. Ah, but there is a light at the end of the tunnel (that isn’t convincing the hosting company to attach a monitor)!

Here is the gist of it:

   nohup Xvfb :4 -screen 0 1024x768x24 2>&1 > xvfb.log &
   sleep 3 # give Xvfb a few seconds to sort itself out
   DISPLAY=4 nohup java -jar selenium-server.jar 2>&1 > selenium.log &

Basically, Xvfb creates a virtual X session, without the need for those kludgey monitors and keyboards. The Selenium Remote Control server is started up with the newly created Xvfb display, that is the DISPLAY=4 part. That is that, the Selenium server is now running on the default 4444 port using Xvfb.

… more

When you upload a file to a Rails app, your browser waits while Mongrel buffers the file, then your browser waits some more while Rails processes the fully uploaded file. Normally this “processing” is simply copying the file from where Mongrel put it to where Rails wants it, no big whup. The problem arises when this “processing” is substantially more time consuming, like, say transmitting the file to S3. There’s no guarantee that this will take place in a reasonable amount of time and meanwhile, not only is the user’s browser getting nearer to timing out, but that whole Rails instance is going to block and no one else can use it either. Rails is single threaded, remember?

The solution is to write the file to disk like normal, then spawn a process in the background to take care of uploading it to S3. This is where Bj comes in. Install thusly:

./script/plugin install http://codeforpeople.rubyforge.org/svn/rails/plugins/bj

./script/bj setup

rake db:migrate

Bj is a light weight work queue that uses your app’s database as a store. Requests are put into the bj_job table and run one at a time outside of the mongrel_rails process. Say we have a model UploadFile that uses attachment_fu:

… more

While experimenting with TimeMachine’s unsupported network backup capabilities I rsync’d my host_blah.sparsebundle from one drive to another (without using -E, more on that later). The result was that my .sparsebundle now appeared in the Finder just as a normal folder, instead of with a disk image icon. This also meant that I couldn’t open the image in DiskUtility any more.

The fix is to use SetFile to restore the “Bundle” attribute.

The first time I mounted it afterwards I had to wait for fsck_hfs to complete, since the “com.apple.diskimages.fsck” attribute had been cleared, but otherwise it’s back to it’s old self.

… more

The code for this site is available to the many-eyes beasts of the Interwebs here:

It requires merb and friends, as well as an instance of CouchDB to be running somewhere. There are a few reasons I chose merb over Rails (which is what I generally use professionally) to do this:

• It comes with no pre-packaged ORM. I wanted to use CouchDB, so none of the existing ORMs would really work for me. My previous experiences of trying to rip ActiveRecord out of Rails have been frustrating.
• It has more flexible routing than Rails. I realize there are ways to accomplish the things I’m doing here in Rails, but the merb routes seem clearer to me, less hand-wavy. And perhaps most importantly:
• It’s not Rails. I use Rails all day long, and I wanted to see what else is out there.

… more