Code Like Zell

belongs_to association callbacks fixed in rails 2.3.3

2009-07-20T17:29:00-04:00

Last night I noticed certain callbacks weren’t being called for belongs_to associations in Rails 2.3.2. Here is a simple example to demonstrate:

class Foo < ActiveRecord::Base
  has_many :bars
end  
  
class Bar < ActiveRecord::Base
  belongs_to :foo
end

I overwrote the callbacks in both of these classes to print its name upon being called. Take a look at the output using Rails 2.3.2:

  Bar.create(:foo => Foo.new)

Callback order:

Bar before validation
Bar before validation on create
Bar validate
Bar validate on create
Bar after validation
Bar after validation on create
Foo before save
Foo before create
Foo after create
Foo after save
Bar before save
Bar before create
Bar after create
Bar after save

As you can see, none of the validation callbacks for Foo were called. Lucky for me, Rails 2.3.3 is out today and the problem seems to be fixed. Using the same models:

  Bar.create(:foo => Foo.new)

Callback order:

Bar before validation
Bar before validation on create
Bar validate
Bar validate on create
Bar after validation
Bar after validation on create
Foo before validation
Foo before validation on create
Foo validate
Foo validate on create
Foo after validation
Foo after validation on create
Foo before save
Foo before create
Foo after create
Foo after save
Bar before save
Bar before create
Bar after create
Bar after save

Now they are all called!

Also, it appears that this problem was introduced in Rails 2.3.2. The example above produced the same results with Rails 2.2.2 as it did with 2.3.3.

How To Send Email with Ruby and Gmail

2009-07-08T00:10:23-04:00

I previously posted about sending email from Rails through your Gmail account. This is the first of two posts demonstrating sending and receiving emails from any old Ruby script. For sending emails, I wanted an interface like this:

  gmail = GmailSender.new(:login => 'yourgmail@example.com', 
                          :password => 'yourpass')
                       
  gmail.deliver(:to => 'friend@example.com', 
                :subject => 'Ricky A',
                :body => "ShirtsTasteGood has two rick astley shirts")

This is my implementation: [shrink code]

require 'net/smtp'
require 'rubygems'
require 'tlsmail'
  
Net::SMTP.enable_tls(OpenSSL::SSL::VERIFY_NONE)
  
class GmailSender
  SMTP_ADDR = 'smtp.gmail.com'
  
  def initialize(opts = {})
    unless opts[:login] && opts[:password]
      raise ArgumentError.new("provide login and password") 
    end
    @login = opts[:login]
    @password = opts[:password]
    @domain = @login[/@(.+)$/,1]
  end
  
  def deliver(opts = {})
    tostr = opts[:to].is_a?(Array) ? opts[:to].join(', ') : opts[:to]
    msg = ["FROM: #{@login}",
           "TO: #{tostr}",
           "SUBJECT: #{opts[:subject]}"]
    msgstr = msg.join("\n") << "\n\n#{opts[:body]}"
    smtp_args = [SMTP_ADDR, 25, @domain, @login, @password, :login]
    Net::SMTP.start(*smtp_args) do |smtp|
      status = smtp.send_message msgstr, @login, opts[:to]
      puts status
    end
  end
end

I made this single file into a local gem, so now I’m ready to write command line email hacks to my heart’s content. Now if you are thinking, “why don’t you just use ActionMailer and tlsmail like you did in the Rails example?” Well, this is the best explanation I’ve got. Next up, receiving emails from GMail!

Top search result for "dick gallery"

2009-07-06T00:02:40-04:00

Google webmaster tools shows you the most popular queries that return your pages in the search results. It looks something like this:

I was pretty excited to see we were on the front page for our grape lady t-shirt already! Then I noticed this gem a bit farther down the page, and knew I had to share:

Thank you Google.

Pre-fill forms in development

2009-07-05T00:02:40-04:00

Automated testing is great, but I often want to click through the many pages of my application myself, without slowing down to fill in form data. I use three javascript functions to provide a simple way for me to pre-fill form data. Say I have an order form that takes a first name, last name, and credit card type. Like this:

  <form>
    <input name="order[first_name]" type="text">
    <input name="order[last_name]" type="text">
    <select name="order[card_type]">
      <option value="visa">Visa</option>
      <option value="american_express">American Express</option>
    </select>
  ...
  </form>

I could write a javascript file that I only include in my development environment that would look like this:

  fill('order[first_name]', "Lou");
  fill('order[last_name]', "Zell");
  select('order[card_type]', "American Express");

The javascript functions that enable me to do this:

function find(name){
  n = document.forms.length
  for (var i=0; i < n; i++){
    k = document.forms[i].elements.length
    for(var j=0; j < k; j++){
      if(document.forms[i].elements[j].name == name){
        return(document.forms[i].elements[j])
      }
    }
  }
}
  
function fill(name, content){
  find(name).value = content;
}
  
function select(name, content){
  s = find(name)
  for (var i = 0; i < s.options.length; i++){
    if (s.options[i].innerHTML == content){
      s.selectedIndex = i;
      return;
    }
  }
}

These functions don’t rely on any library, and they also shouldn’t be used in production because they are far from optimized. Yes, I am 100% sure there is a much more efficient way to write them, but they work just fine for what I need them to do.

If you want to get fancy pants you could create different sets of data to fill a form with, and attach each fill function to a link’s click event. Something like:

  <a id="fill_with_bob">fill with bob's info</a>
  <a id="fill_with_jane">fill with jane's info</a>

  $('#fill_with_bob').click(function(){
    fill('order[first_name]', "Bob");
    fill('order[last_name]', "Jones");
    select('order[card_type]', "American Express");
  });
    
  $('#fill_with_jane').click(function(){
    fill('order[first_name]', "Jane");
    fill('order[last_name]', "Jones");
    select('order[card_type]', "Visa");
  });

I used jquery for that last part. And it’s not tested.

Steer Clear of Paypal

2009-06-23T18:20:40-04:00

If you are considering using Paypal’s Website Payments Pro service for $30 per month, I would think twice. I was warned that their customer service was atrocious, but it’s much worse. If you are currently using them, you already know that every phone number they provide to reach merchant services directly takes you through the same automated system as every other Paypal number. Here are the steps I recommend to get you through to someone that can help, without losing your mind:

1. Take a shot (and one additional shot for each time you are disconnected)
2. Call any of several “direct lines” to merchant services
3. Say “no” when the automated system picks up
4. Say “agent” three times in a row, you will be placed on hold
5. You will get an agent, but they will be in the wrong department. They will inevitably discover that they cannot help you, and will transfer you to the merchant support group (which is the one you wanted in the first place). So save yourself the pain and demand an immediate transfer to merchant support. You will be on hold again for a while.

Good luck!

ShirtsTasteGood.com, YouTube™ Inspired T-Shirts!

2009-06-16T18:20:40-04:00

I am very proud to announce ShirtsTasteGood.com, a collection of t-shirts inspired by our favorite YouTube™ videos. It’s a four man team consisting of Todd Hamilton on artwork and web design, Doug and George Cline on screen printing, and myself on web development. Doug and George ditched their custom screen printing business that they had been running for the last few years, and Todd and I both quit our contracting/freelancing gigs in the city to full time STG. The operation is run out of a barn that Doug and George converted to a print shop near their home in New Jersey. The video that inspired each shirt is embedded on the shirt’s home page, so even if you aren’t interested in buying a t-shirt, you can come hang out and watch some hilarious videos. There is also a suggestions page where we auto-embed any videos that you would like us to turn into shirts! The suggestions page is particularly useful during that 1-3pm slot when you pretend to be working. We hope you enjoy the site, and please feel free to post any suggestions, questions, or comments. Thanks!

The Hudson Raft Project

2009-06-04T18:20:40-04:00

Over the last four years a few friends of mine, led by Dallas Trombley, have attempted to sail from Albany to Manhattan on homemade boats. They now have a site up describing their trials and tribulations. Most of their content is in the form of Wordpress pages rather than posts, so be sure to check out the left nav. It’s a really entertaining read. Good luck this year Dallas, Rob, and Co.

Disable SSL during Rails development

2009-05-26T23:38:00-04:00

Assuming you are using ssl_requirement, to disable SSL in your development environment:

class ApplicationController < ActionController::Base
  include SslRequirement
  
  protected
  def ssl_required?
    RAILS_ENV != 'development' && super
  end
end

Deploying Wordpress with Rails on Passenger

2009-05-20T23:38:00-04:00

I just finished integrating a Wordpress blog into a soon-to-be released Rails app, and I would like to share the steps I took to get it deployed on Phusion Passenger. Getting the permalinks working correctly had me hung up for a bit, so I’ll show you how to get around that.

To get started we can drop the blog right into the public directory of our rails app:

  cd /yourapp/public
  wget http://wordpress.org/latest.tar.gz
  tar xzvf latest.tar.gz
  mv wordpress blog     <--- blog will be at http://yourapp/blog
  cd blog
  mv wp-config-sample.php wp-config.php

Now edit the wp-config.php file (instructions are in the file) and be sure to grant database privileges to the user you specify.

Next, you have to edit your apache config file to disable passenger for the blog directory:

<VirtualHost ...>
    ServerName ...
    DocumentRoot ...
    <Location /blog>
       PassengerEnabled off
    </Location>
</VirtualHost>

I found that snippet here.

Restart Apache and point your browser to http://yourapp/blog/index.php to setup the rest of your blog through Wordpress’s web interface. One thing you will probably edit is the default permalink structure (Found at settings > permalink). I prefer to set a custom structure of ‘/%postname%’. Regardless of the structure you use, wordpress will tell you to edit the .htaccess file in your blog directory. For the life of me I could not get .htaccess file to be recognized, so as a workaround I copied the snippet that wordpress specified into my apache config as follows:

<VirtualHost ...>
    ServerName ...
    DocumentRoot ...
    <Location /blog>
       PassengerEnabled off
       <IfModule mod_rewrite.c>
         RewriteEngine On
         RewriteBase /blog/
         RewriteCond %{REQUEST_FILENAME} !-f
         RewriteCond %{REQUEST_FILENAME} !-d
         RewriteRule . /blog/index.php [L]
       </IfModule>
    </Location>
</VirtualHost>

Finally, restart Apache.

A more readable sort by

2009-04-03T18:25:40-04:00

Suppose we want to sort a collection of accounts by their account balance. We could put them in ascending order like this:

  @accounts.sort {|x,y| x.balance <=> y.balance}

Or descending order like this:

  @accounts.sort {|x,y| y.balance <=> x.balance}

To me this isn’t very readable. How about this as an alternative:

  @accounts.sort_by_ascending(:balance)
  # or 
  @accounts.sort_by_descending(:balance)

Much more readable. The implementation is pretty simple, first create the methods:

  module ReadableSortBy
    def sort_by_ascending(method)
      sort {|x, y| x.send(method) <=> y.send(method)}
    end
    def sort_by_descending(method)
      sort {|x, y| y.send(method) <=> x.send(method)}
    end
  end

Then include them in Array:

  class Array
    include ReadableSortBy
  end

You could also make sort_by_ascending! and sort_by_descending! methods that modify the collection itself. Do this by replacing sort with sort! in both cases.

Introducing Nickel

2009-03-29T01:05:40-04:00

A natural language date and time parser, with message extraction

Nickel is a client for Natural Inputs

  sudo gem install lzell-nickel

Then try it out:

  require 'rubygems'
  require 'nickel'
  n = Nickel.query "party this friday at 10pm"
  
  n.message # => party
  n.occurrences.first.start_date # => 20090403
  n.occurrences.first.start_time # => 22:00:00

Many more examples here.

A cheap abstract class in ruby

2009-03-29T01:02:40-04:00

When I create a class that I don’t want to allow anyone to instantiate, I use the following pattern:

  class MakeMeAbstract
    def initialize
      raise "Don't instantiate me!" if abstract_class?
      puts "Instantiated!"
    end
    private
    def abstract_class?
      self.class == MakeMeAbstract
    end
  end

It can’t be instantiated:

  MakeMeAbstract.new rescue puts "I'm abstract"   # => "I'm abstract"

But it can be subclassed, and the subclass can be instantiated:

  class Child < MakeMeAbstract
  end
  Child.new   # => "Instantiated!"

As a full disclaimer, this is a learning experience for me. When subclassing an abstract class I wonder if it would be better to use a mixin instead: I could overwrite mixed in methods just as easily as methods in a parent class. The structure changes from this:

  class Bar
    def status
      print "is full"
    end
  end
  class WetBar < Bar
    def status
      print "is wet, "
      super
    end
  end
  wet_bar = WetBar.new
  wet_bar.status    # => "is wet, is full"

To this:

  module Bar
    def status
      print "is full"
    end
  end
  class WetBar
    include Bar
    def status
      print "is wet, "
      super
    end
  end
  wet_bar = WetBar.new
  wet_bar.status    # => "is wet, is full"

Input is welcome in the comments.

How to backup and restore a mysql database

2009-03-29T00:56:40-04:00

To dump your database:

  mysqldump -u USER -p PASS my_db > my_db.sql

Then to restore:

  mysql -u USER -p PASS my_db < my_db.sql

Or, using gzip for compression, dump with:

  mysqldump -u USER -p PASS my_db | gzip > my_db.sql.gz

And restore with:

  gzip -cd my_db.sql.gz | mysql -u USER -p PASS my_db

Using Gmail with Rails

2009-03-29T00:51:40-04:00

To send email from Rails using your Google Apps account, you will need TLS support:

  sudo gem install tlsmail

Next, edit environment.rb to look like this:

  Rails::Initializer.run do |config|
    config.gem "tlsmail"
    ...
  end
  Net::SMTP.enable_tls(OpenSSL::SSL::VERIFY_NONE)

Then add this to environments/development.rb:

  config.action_mailer.delivery_method = :smtp
  config.action_mailer.smtp_settings = {
    :address => "smtp.gmail.com",
    :port => "25",
    :domain => "your_domain.com",
    :user_name => "your_login@your_domain.com",
    :password => "your_password",
    :authentication => :login
  }

Test it out by sending an email to yourself, create the model mailer.rb:

  class Mailer < ActionMailer::Base
    def test
      from "your name <your_login@your_domain.com>"
      recipients ["your_login@your_domain.com"]
      subject "working?"
      body "sending from rails"
    end
  end

Finally, from script/console:

  Mailer.deliver_test

Use inject for nicer output in IRB

2009-03-29T00:45:40-04:00

Sometimes you want to inspect an array in irb by iterating through it and printing out some key values, for example:

  events.each do |event|
    puts event.title
  end

But if the array is large, you will first have to scroll up through the returned array to get to your desired printout. We can prevent that by using inject instead:

  events.inject do |x, event|
    puts event.title
  end

Now nil will be returned instead of the events array, and your desired output is in view.

Preventing default event behavior with Javascript

2009-03-29T00:34:40-04:00

with examples using jQuery and Prototype

Let’s say we want to disable all the links on a page, using jQuery we just have to return false from the click handler:

  jQuery(function(){
    jQuery('body a').click(function(){
      return false;
    });
  });

Using Prototype, we pass the event instance to Event.stop():

  document.observe("dom:loaded", function(){
    $$('body a').each(function(a, num){
      $(a).observe('click', function(e){
        Event.stop(e);
      });
    });
  });

With everyday Javascript, we again return false from the click handler:

  window.onload = function(){
    elements = document.getElementsByTagName('a');
    for (var i = 0; i < elements.length; i++){
      elements[i].onclick = function(){
        return false;
      }
    }
  }

Ruby Bitwise Notes

2009-03-29T00:18:40-04:00

Start with an integer:

  x = 1

Shift by two bit positions:

  x <<= 2

OR with 1:

  x |= 1

AND with 14 (hex E):

  x &= 0xE
  # is the same as:
  x &= 14

Get binary representation:

  x.to_s(2)

Get hex representation:

  x.to_s(16)

XOR (exclusive OR) with 1

  x ^= 1

Vim single line move when :set wrap is on

2009-03-29T00:02:40-04:00

If you are editing a long paragraph with word wrap on, you may find that ‘j’ and ‘k’ jump farther than you would expect. This is because ‘j’ is moving to the next real line instead of the next wrapped line. To make vim behave more intuitively (at least for me, and probably you since you are here), put this in your vim config file (~/.vimrc in my case):

  function! ScreenMovement(movement)
     if &wrap
        return "g" . a:movement
     else
        return a:movement
     endif
  endfunction
  onoremap   j ScreenMovement("j")
  onoremap   k ScreenMovement("k")
  onoremap   0 ScreenMovement("0")
  onoremap   ^ ScreenMovement("^")
  onoremap   $ ScreenMovement("$")
  nnoremap   j ScreenMovement("j")
  nnoremap   k ScreenMovement("k")
  nnoremap   0 ScreenMovement("0")
  nnoremap   ^ ScreenMovement("^")
  nnoremap   $ ScreenMovement("$")

Thanks to rick_h in #vim

Rails 2.2 migrations without the timestamp

2009-03-28T23:56:40-04:00

For Rails 1 style migrations, set this in environment.rb:

  config.active_record.timestamped_migrations = false

Note this will not work before Rails 2.2. Thanks to mikeg1a in #rubyonrails.

Edit through FTP with TextMate and Cyberduck

2009-03-28T23:51:40-04:00

I do a bit of freelance php work where I need to edit files through FTP. Textmate does not support this very well out of the box, and at first it seemed like I would have to open and edit each file in its own window. But by combining this tip by subtlegradient with this tip by ciaranwal.sh we can use tabbed editing, project drawer, and cmd+t to switch files. The steps:

1. In Cyberduck’s preferences make sure external editor is set to TextMate. Also, it helps if “Double click opens file in external editor” is set to true.

2. Open up a file from your FTP project that is close to the top level directory of your project, you’ll see why in a minute.

3. In the TextMate window that popped up, hit ctrl+shift+o (Bundles > Shell Script > Open Terminal). The terminal will open at the temporary location where the file you are editing is.

4. cd .. back to the top level directory of your project and enter: mate .

Now any subsequent files you open using Cyberduck will be added as tabs in the current TextMate project.