1. Find a product-oriented company with a high-performance team building something amazing, or
2. Bootstrap my own tiny consultancy, and / or
3. Build an amazing product.

As luck would have it, I’ve already landed an interview at a company that offers the first option. I’m excited about the interview, but realistic about the possibilities. Because they appear to be a high-performance team, they are guarding the entrance quite closely. My level of intrigue is all that much higher because the interview process is anything but easy and normal. Suffice it to say that you can’t get hired by being a good guy and liberally dropping high-impact buzzwords. You may laugh, but I’ve been hired through such good-guy interviews.

This is a position which features the use of Java as their primary implementation language, so my challenge is to get back into the Java mindset after several months of Ruby development. My friend Mike once supposed that I would never return to Java. Its possible he’s wrong as I could be circling back. Working with Ruby is fun, but my talents and experience position we well for a Java shop and there is more of a job market there too.

But even as I say that, I must admit that looking at Java for the first time in a few months has surprised me. Java hasn’t changed, but it seems that my mind has started to think in Ruby. I am actually alarmed that a language I’ve worked in for years could be subject to this kind of transition. I found myself asking fundamental syntax questions. So I wondered if this had happened to anyone else. Or is this just a by-product of my being 43?

Well, I’m not worried about picking Java up again, but I am picking a problem to code to get my head back into it. I still have a series of Java projects from an interview process I went through in June (yes, I was offered the job), so I’m going to work through one of those. To make it more fun, I may try to integrate it into a JRuby on Rails project and do a local deploy. This will enable me to prepare for the interview and explore the integration possibilities for JRuby and Java.

What happens if I don’t get the job? Well, that’s a real possibility, so I’m planning my bootstrapping activities for my consultancy. Having done recent work for a medium-sized consultancy, I know there is work out there and I know that I can provide high-quality work for a portion of the price that bigger companies are charging. But as every entrepreneur knows, you cannot be successful without customers, so my biggest challenge will be to construct a marketing machine capable of producing a sustainable pipeline of work. For technicians, this is not an intrinsic skill, so I’ve been thinking more about marketing than my technical skills — the latter is the easy part.

In some ways this takes me back a great many years ago when I created a small software company, SportSoft Systems, which created and marketed Fantasy Football software and a few other things. This was back 20 years ago and my product was just about the best out there. Had I only known that the fantasy sports craze was to explode, I would have stuck with it. I did actually make decent money, but the pressures of supporting the product and building in new features every year while being an ethical employee at my primary job caused me to discontinue my efforts.

So for now, I’m circling back — either to the Java world or to a world where I build my own company. Should be fun wherever things lead.

You basically have a method called after_destroy (or whatever), but you don’t readily know what its doing.  A better approach is to write an intention-revealing named method and reference that in the reference to your callback method.  You then can see through the method name what is happening in response to the event.  Further, the method can be used at other times.  This approach restores a bit of maintainability to your code.

I have included an example of this approach below.  You’ll have to visualize the before version — I don’t like posting less maintainable code because sometimes people skim it without reading and take it as good code.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

class Rating < ActiveRecord::Base
  
  # ---- Associations ----
  
  # The rateable that was rated.
  belongs_to :rateable
  
  
  # ---- Callbacks ----
  
  after_destroy :update_ratings
  
  
  # ---- Calculations ----
  
  # Updates the parent's rating count and average rating.
  def update_ratings
    rateable.update_ratings
  end
end

As a simple example, here’s a partial model featuring constants to aid in clear validation specifications:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

class Email < ActiveRecord::Base
  
  MAX_FROM_LENGTH   = 30
  MAX_TO_LENGTH     = 100
  FROM_LENGTH_RANGE = 1..MAX_FROM_LENGTH
  TO_LENGTH_RANGE   = 1..MAX_TO_LENGTH
  TO_FORMAT         = /^[_a-z0-9+.-]+@[_a-z0-9-]+\.[_a-z0-9.-]+$/i
  NOTE_LENGTH_RANGE = 1..DB_TEXT_MAX_LENGTH
  
  # ---- Attributes ----
  
  # Attributes protected from mass-assignment (as through forms) 
  attr_protected :sent_at
  
  
  # ---- Validations ----
  
  validates_length_of   :from, :within => FROM_LENGTH_RANGE
  validates_length_of   :to,   :within => TO_LENGTH_RANGE
  validates_format_of   :to,   :with   => TO_FORMAT
  validates_length_of   :note, :within => NOTE_LENGTH_RANGE

As you can see the model defines constants to enforce length maximums and in turn the ranges. You could argue that the range constants are less useful since you have to refer back to the constants section to see them, but that’s a matter of taste. I like having them.

With constants like these properly placed in your model, you can take advantage of them in your migration. Here’s a partial example:

1
2
3
4
5
6
7
8
9
10

class CreateEmails < ActiveRecord::Migration
  def self.up
    create_table :emails do |t|
      t.column :emailable_id, :integer, :null => false
      t.column :to,         :string,  :null => false, :limit => Email::MAX_TO_LENGTH
      t.column :from,       :string,  :null => false, :limit => Email::MAX_FROM_LENGTH
      t.column :note,       :text,    :null => false
      t.column :sent_at,    :datetime
    end
  end

The advantage here is that your model and your migration are now in synch for field maximums. If that’s not enough, then consider what you can now do in your test code:

1
2
3

def test_invalid_with_long_to
    assert_value_for_attribute_invalid(Email.new, :to, filled_str("X", Email::MAX_TO_LENGTH + 1))
  end

So now, if you change your mind and decide you need to lengthen your fields, you need only change your model constants. Your migration could be rerun or you could add an additional migration which applies the new field length. And lastly, your test code doesn’t have to change.

I think this is simple, yet very much worth doing.

• [FEATURE] Added timeout support for geocoder web service calls
• [FEATURE] Added proxy support for geocoder web service calls
• [FEATURE] Added support for SQL count queries — they work like the finders
• [BUG FIX] Fixed range finders to support exclusive ranges
• [BUG FIX] Fixed configuration to use class accessors instead of constants (that were being reinitialized)
• [BUG FIX] Eliminated bogus IP address returned from IP geocoder when the request IP is 127.0.0.1
• [BUG FIX] Fixed GeoLoc’s hash method and added an alias to_hash
• [REFACTOR] Require statements to be cleaner all the way around

Thanks to the following who reported bugs that got fixed or who submitted patches that were used in part or in their entirety:

• Norbert Crombach
• Justin French
• Hoan Ton-That
• Jason Seifer

Thanks to everyone for your feedback, patches, and compliments. If I didn’t name you explicitly, it means that your thanks by name will come later.

Since I am in a GeoKit development mode right now, let me or Andre know if there are other features, bug fixes, or patches I should be aware of. I know that we had been planning to modify our configuration approach to be compatible with YM4R, but that’s been delayed indefinitely. They use a YAML file and we use the environment configs — and there really wasn’t a compelling reason to mess with it. Also, no UK geocoding yet — didn’t receive too much demand.

Please review the README for updates and take a look at the new environment config template under the plugin’s assets directory to see how you should config GeoKit. And by all means, let me know if there are issues that I’ve inadvertently created for you.

The next question is: would you consider productizing it to enable others to have online resumes? The answer is: maybe. If I was convinced that I could find enough paying customers, I would. However, I’m not sure I could do that. I will say that it is nice to be able to point recruiters and potential employers to the site for resumes, samples, and testimonials.

Another question is: what’s up with the odd classes in some of the markup? Well, that’s microformats. I am tinkering with hResume, hReview, and hCard. You can download a Firefox extension which alerts you when microformats are present on a page and on my site you can see microformats on about half the pages. For now, I’m probably about the only one who thinks this is cool, but such is the life of an early-adopter.

Lastly, I’ve had a couple of recruiters ask about the effects found on the site (e.g. samples, about – tag cloud). That’s some Javascript and AJAX just for the heck of it. Its kinda cool to refer a recruiter to my site when they are asking if I’ve ever done any AJAX. So, not hugely useful, but mildly demonstrative of my skills.

I have surmised that my ability to code in Ruby is not of much value to the Java community. There’s just not much crossover between the two. So while I like coding in Ruby and Rails, it would appear that I won’t be paid to do so anytime soon.

The first thing to get straight is that YOU are the product. Prospective employers are your customers. And as a result, you need a marketing plan. Your marketing plan must begin with the usual self-assessments which eventually lead to your main marketing vehicle — the resume. In some ways, you could almost consider your resume as the product.

Next comes the distribution problem. How do you get your resume in front of those who you wish to become your buyers? In today’s world, you get your resume and profile up on all the job boards that are relevant to your profession. However, reentering the same data over and over is such a pain. To ease your pain, use SimplyHired which is a site that will enable you to enter your resume data once and then blast it out to sites of your choosing — a real time saver. Just be sure that you get the information right because after SimplyHired does its thing, you will have to manage individual accounts at all the sites.

As an added touch, I created my own website to promote my efforts. The website serves as a visual reinforcement of my technical skills. Its well-designed, uses semantic standards-based markup, is CSS-based, and uses progressive enhancement in a couple areas to improve the user experience (printing, tag cloud AJAX). Its not an over-the-top tour de force in technology, but is just enough to impress the target audience — prospective employers and recruiters. Because some of you will go there, I readily disclaim that its been soft-launched. Expect to find minor cross-browser issue and some content yet to be provided. You techies are a tougher audience than my target audience!

So when you flip the switch on this machine, you will get assaulted with emails and opportunities. These are your customers banging on the door. To organize yourself, get a Highrise account from 37Signals. Highrise is a mini-CRM application and let’s face it, you will be talking to people, exchanging email with people, and more or less manage the sales process. Highrise is perfect for this and is free for limited use.

Once you actually start interacting with people you’ll have to get pretty savvy at determining who to spend your time with. I have received very generic requests to call recruiters and I don’t get involved further unless I am able to get a compelling story up front. Be prepared, you will also hear from people who will disregard your skills, experience, and relocation preferences entirely. I am deleting most emails on that basis alone.

So far, things are going well. I think that treating this situation like a product campaign has been beneficial. I already have some solid leads and opportunities that I’m following up on. And this is Day 3.

I’m pretty sure that I sound like a Father telling his kid to turn the music down…and so at least I am self-aware.

FEATURE SUMMARY

This plugin provides for easy externalized configurization in a YML file:

• Common and environment-specific configuration contained in a single YML file.
• Capistrano task to preserve configuration file from release to release.

USAGE
Place configuration elements common to all environments in the common area as demonstrated below:

1
2
3

# Configuration for ALL environments.
common: &common
  app_name: your_app_name_here

Place environment-specific elements under appropriate environment as shown below:

1
2
3
4
5

# Configuration for PRODUCTION environment.
production:
  <<: *common
  google_api: some_long_api_key_here

To obtain the values from anywhere in your code, simply access the attributes of the environment like so:

1
2
3

puts AppConfig.app_name
puts AppConfig.google_api

DEPLOYMENT

To preserve the configuration values across releases, call the symlink_app_config task from within the after_update_code task.

1
2
3

task :after_update_code, :roles => :app do
  symlink_app_config
end

To update the app.yml file for the first time or for subsequent times, the task update_app_config can be used as below:

cap update_app_config

This takes the app.yml file from the config directory of the application and uploads it to the release path on the server.

You can get it here:

script/plugin install http://dev.billeisenhauer.com/svn/projects/plugins/appconfig

Hope someone finds this useful.

I decided I would make an attempt to deploy CruiseControl.rb to one of the hosts that I have — and do so with Capistrano. For the most part, it was pretty painless, but I did run into a gotcha or two.

First, the host I deployed to uses the Apache / Lighttpd (fastcgi) style of deployment. Unfortunately, Lighttpd just would never start up for me! After way too long, I finally figured out that I had DOS line breaks in my file that I could not see. So I zapped them with dos2unix and all was well. Unfortunately, this ate up a lot of my time as I’m not very skilled at debugging fastcgi issues.

CruiseControl.rb is a Rails app, but not your typical Rails app. It uses no Rails app, but rather stores your project information in a projects directory underneath the app root. Further, its more than just a Rails app, it utilizes additional processes as project builders.

To deploy with Capistrano, you have to preserve the projects directory so that it will persist across your releases. That’s pretty easy to do with symlinks, so that’s a piece of cake. You can use the after_update_code callback to do this.

There are two other use cases that I could see that I would need. Adding a project and starting a builder for that project. To enable those, I added Capistrano tasks for each. Both tasks rely on environment variables and both tasks could probably benefit from a bit more error-checking, but they are good enough for me.

So to start a project, I use:

NAME=GeoKit URL=svn://rubyforge.org/var/svn/geokit cap add_project

to start its builder, I use:

NAME=GeoKit cap start_builder

I added my custom tasks under config/recipes/cruisecontrol.rb and required this in deploy.rb. The custom tasks code is below:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

Capistrano.configuration(:must_exist).load do

  desc "Update symlink for projects directory."
  task :after_update_code do
    run <<-CMD
      rm -rf #{release_path}/projects &&
      ln -nfs #{shared_path}/projects #{release_path}/projects
    CMD
  end

  desc "Add a project to cruise control."
  task :add_project do
    unless ENV['NAME'] && ENV['URL'] 
      raise ArgumentError, "***** You must specify the NAME and URL parameters to add a project. *****" 
    end   
    run "#{current_release}/cruise add #{ENV['NAME']} --url #{ENV['URL']}"
  end
  
  desc "Start a builder for a project."
  task :start_builder do
    raise ArgumentError, "***** You must specify the NAME of the project to build. *****" unless ENV['NAME'] 
    run "#{current_release}/cruise build #{ENV['NAME']} &"   
  end
end

And with that, any of my projects can now be monitored with CruiseControl.rb.

A few caveats.

• I had to physically edit the cruise_config.rb file (on the server) for the project to include my email address and polling interval. No big deal, its a one-time thing.
• If your code is proprietary, you may want to protect the site with
  basic authentication to keep unwanted eyes from seeing your project assets.
• Since your project data is on your server’s file system, you may want to consider a backup strategy if you value your build history.

Suggestions for ThoughtWorks:

• Consider including Capistrano tasks to make it easier — mine are admittedly rough, I’m sure.
• Consider adding some element of authorization by project. For example, I have public plugins, but I also have proprietary code that I do not wish to share. Obviously, its easy enough to deploy a second site, but that’s mildly inconvenient.

I hope this is helpful to continuous integration fans out there trying to get set up!

James wrote that it would be nice to be able to reset the default_units value to enable localization.  Well, actually you can.  :default_units is modifiable through a class accessor.  Also, you can pass an optional parameter in your finders called :units which enables you to override whatever default value you’ve set.

While we’re talking about defaults, I applied a patch submitted by Hoan Ton-That which enables you to specify your preferences for units and for the formula from within the environment files.  So these preferences would appear right along side your API keys for GeoKit.  Its a small feature, but you may appreciate it.  Thanks Hoan!

Otherwise, James mentioned that :include support is lacking with GeoKit.  That’s not entirely true.  I added a passing test case which successfully pulls columns from a belongs_to relationship.  I suspect that the has_many relationships are the problem.  But if your case is covered by the belongs_to relationship, then you are covered.

There is a less-used method called distance_sql which enables you to just get the distance calculation in SQL form.  You would use this if you ever found that GeoKit could not support your exact query requirements.  Its less elegant for sure, but saves you the trouble of having to code the distance calculation.

UK geocoding is still in the roadmap, but both Andre and I have a few other projects and side endeavors going on.