Codebeef

Creating a New Post in Jekyll with Rake

2012-02-11T00:00:00+00:00

Just a quick and dirty rake task that lets you run this:

rake np -- The Post Title

…to set up a new post in Jekyll and open it in Sublime Text 2.

If you prefer another editor, then swap the subl command for the appropriate CLI command for your editor.

require 'rubygems'
require 'optparse'
require 'yaml'

task :np do
  OptionParser.new.parse!
  ARGV.shift
  title = ARGV.join(' ')

  path = "_posts/#{Date.today}-#{title.downcase.gsub(/[^[:alnum:]]+/, '-')}.markdown"
  
  if File.exist?(path)
  	puts "[WARN] File exists - skipping create"
  else
    File.open(path, "w") do |file|
      file.puts YAML.dump({'layout' => 'post', 'published' => false, 'title' => title})
      file.puts "---"
    end
  end
  `subl #{path}`

  exit 1
end

Also available here as a gist.

Industrial Revolutions

2011-08-16T00:00:00+00:00

IE and PDF downloads over HTTPS in Rails

2011-04-22T00:00:00+00:00

So you have a controller that renders PDF’s, and all looks great. You decide to move from HTTP to HTTPS, and everything checks out. You deploy, and suddenly some of your users can’t see the PDF downloads anymore. These users are all running some version of IE - all the way from IE6 up to IE8.

The solution is to set your response headers as follows:

  def show
    inst = Model.find(params[:id])
	
    response.headers['Content-Transfer-Encoding'] = 'binary'
    response.headers['Content-Disposition'] = "attachment;filename=\"#{inst.id}.pdf\""
    response.headers['Content-Description'] = 'File Transfer'
    response.headers['Expires'] = '0'
    response.headers['Pragma'] = 'public'
	
    render # Render your pdf
  end

Diary (2010) by Tim Hetherington

2011-04-21T00:00:00+00:00

Truncating HTML while Preserving Structure

2011-04-15T00:00:00+00:00

Truncating HTML is a bit of a pain - in a recent effory to change the layout of my site, I updated the index page to show the last five posts - truncated, of course.

The problem with doing this is that the most common method of truncating a post is to remove all markup and truncate the remaining string. This would be fine, but for a site with a fair amount of colorized code examples (courtesy of Pigments), it can look a little weird.

A better solution would be to truncate based on the number of characters, yet retain the markup, which is where this cheeky snippet of code comes in.

Requirements

The truncator requires Nokogiri to parse out the HTML string.

Usage

If you’re using Jekyll, add the html_filters.rb file to your _plugins directory - this will give you the helper truncatehtml as a Liquid filter. In your views, you can use this function in the same way as you would use the normal truncate filter:

page.content | truncatehtml: 500

How it works

Given a snippet of HTML, we use Nokogiri to parse it into a tree. This takes care of all of the messiness of dealing with HTML, and gives us back a tree of nodes representing the parsed snippet.

Now that we have the tree of nodes, we can traverse it depth-first. All text nodes are leaf nodes, so when we encounter one, we can count the length of the text.

Once we have all of the text we need, we continue traversing, but instead of counting text lengths, we delete all of the nodes we see after we’ve reached our maximum length.

After that, we have a tree of nodes that represent the truncated tree, and the can output the appropriate HTML.

The Beta - Branch Bacardi

2011-03-30T00:00:00+00:00

Getting Work as a Freelancer

2011-03-21T00:00:00+00:00

I recently quit my day job in order to pursue a career as a freelance Rails developer. I went from having a steady income each month to potentially having no money coming in at all.

This is how I managed to get clients and not die in a ditch during my first few months:

The Big Change

Moving from having a full-time job to a freelancer is hard, but like most difficult tasks, it is ultimately worthwhile.

I decided to take the plunge into freelancing after putting it off for around 12 months - I pretty much committed to it in as short a time period as possible, thereby preempting any possible weaselling out on the part of my reasonable side.

The actual quitting of my job was fairly easy - write a letter, have a meeting then work my month’s notice. Simple.

Then what?

Finding Work

By far the most troubling part of becoming a freelancer (for me, at least) was not knowing where my customers were going to come from. If I can’t see where my next job is, it obviously makes me nervous - how can I pay the bills and mortgage unless I know I have the money coming in?

This is where things got interesting: I put the call out, and almost immediately got too much work to reasonable book in. My previous employer also immediately booked me for a four week contract that would start the day after I finished my notice.

This may not be representative of your own experience, however, it does go to show that work can pretty much come from anywhere. These are a few places that I’ve been able to get jobs so far:

Other Freelancers

Having a network of friends in the same field as yourself (in my case web developemnt) gives you a competitive advantage: If you can’t allocate the time for a project, you can pass it to a friend, creating goodwill on the part of the client you’ve passed on and potentially gaining a reciprocal referral.

Having a network of people - with skills that are complementary to your own - will allow you to provide recommendations if it turns out the client needs additional services like graphic design or copywriting.

It is important to not see other freelancers as ‘the enemy’, and instead to actively seek out new contacts in your field.

Existing Clients

Once you’ve completed a project for a client, and the job went well, you should occasionally give them a nudge to see if they have anything else they want looking at. Discounts for repeat custom can help grease the wheels, and you may end up taking on a retainer if it turns out they do need regular work.

Word of Mouth

Happy clients will recommend you. It is sometimes worth going slightly above and beyond in order to delight your customers, though how far you take this needs to be balanced.

If you do go above and beyond their expectations, this will probably set their baseline expectations higher for future work. Be aware of this if you later need to quote for work with a tight deadline.

Advertising

It seems that the first place people check for Freelancers is often Google, and as it happens, Google AdWords makes it really easy to get a campaign set up to start bringing in interested people.

These types of leads greatly vary in quality, from enquiries along the lines of “Build me a Facebook clone for $500” to the more reasonable “I have a spec, can you help me build it?”.

Keep in mind that getting an enquiry from this avenue may need a little more massaging before it becomes a fully booked-in job, and like all enquiries, they are bordering on meaningless until the first invoice is paid and the job has begun.

Events

If you’re near a major city, there should be a fair few events on that you can get to. In Manchester, there is the local Geekup along with the NWRUG and now Freelance North. Such events are great for getting to know fellow freelancers and developers in a more social setting than IRC.

Twitter

Announcing on Twitter when you go freelance or when you’ve got a bit of slack time coming up is a very good way of making people aware you’re available. Equally, trawling for people looking for freelancers is an excellent way of getting new leads.

When you announce you’re going freelance, try and get other developers (freelance or not) to retweet your announcement. You should notice you get a few new followers - agencies and other freelancers especially.

Conclusion

It’s not hard to find the work once you start making connections - the point is that you need to embrace the social aspect of freelancing to carve out a position. Don’t hang on to work you don’t have time for - give it away and gain karma. Do good work, and don’t forget about your existing contacts.

How I got into iOS Development

2011-01-04T00:00:00+00:00

I recently got an email asking how best to get into iOS development - a request for links, tutorials, and so on. Now, I don’t necessarily know how best to get into developing for iOS, but I know I didn’t have a predefined set of links to run through, or even much reading material outside of some upcoming books, which I didn’t really bother with anyway.

tl; dr:

The method I used to get into iOS dev was to build stuff. Constantly.

More Detail

I’m not sure how much C you know, but when I started in Obj-C, I already had a good grasp of C and C++ and many concepts from these two languages map directly into Obj-C (Obj-C is a superset of C). I’d definitely recommend getting a basic grasp of C before jumping into Obj-C. The only book you need on C is called ”The C Programming Language” by Brian Kernighan and Dennis Ritchie.

There are lots of projects on Github that you can clone and comb through the source of. Try to pick one that has an active and well regarded community - so you don’t begin studying crap unintentionally. As I was building sync stuff into my apps, I first looked through ObjectiveSync and ObjectiveResource (from Ryan Daigle), though dev seems to have slowed somewhat. Facebook’s three20 library is also worth a look, too. If you look on this page, you’ll find lots of links to really good, interesting projects on Github right now.

If you just want to dive right in to iOS development, then the path I took was:

1. Get the single dev plan

If you want the enterprise plan, then do whatever you see fit. It’s a detail that is of little consequence. The enterprise plan will give you more support requests and provisioning for more developers, but if you’re working alone, then the single plan works just fine.

2. Think of a non-trivial app

I chose to build an app for rock climbers - a universal guidebook, and includes things like synchronisation, mapping, image manipulation and location services. It needs to be ambitious enough for you to think you’re not going to be able to do it.

3. Begin building version 1

This version will be awful, but you’ll learn a lot.

Strip your idea down to a minimum. The first version of Bouldr was read-only, and simply displayed a list of climbs nearby with a detail screen for each entry. It had no sync - it required a connection, no editing and only basic map integration. My code at this point was pretty ropey, but I learned how to put things together, and I could see where I was going wrong when I came back to add additional features.

I didn’t buy any books - the developer resources on Apples site are really good, so long as you actually use them. Additionally, StackOverflow is an amazing source of information - all of the problems you run into will already be covered somewhere on the web, and most likely, there will be a question and answer right there on StackOverflow. I’ve not read any books, so I can’t recommend any.

4. Release

This is important - get your version 1 out. You’ll gather some interest, and have some users you can later get to help you beta test. Note that being able to set up and manage a beta test is not trivial. Once your first version is out, you’ll get feedback.

5. Begin version 2

Don’t leave it at that - dive back into the code and begin adding features that you think are needed. Learn how to use Mixpanel to gather data on what people are actually using. Integrate Hoptoad to get better error reporting. You’ll probably begin to notice how bad the code is in your first version.

This next bit scared me, and took longer than anticipated, but it also taught me a massive amount, and let me improve my codebase:

6. Scrap your first version, and build a new version from scratch

Re-implement all of the functionality, but concentrate on making your code maintainable - don’t add new features. I started building out a test-suite at this point, too.

7. Beta test your next release

Get a beta group together. Gather their UDID’s and send them a beta build. Gather feedback and bug reports, and fix your app up as you need to.

8. Release version 2

It’s important to make sure you’ve covered migrating your data store if you’re using Core Data, and you should test as much as you can to ensure that you’ve not fundamentally broken your app. I’ve submitted a version of the Bouldr app before that contained a crashing bug. Apple won’t necessarily catch problems with your app unless they are obvious. If the upgrade process within your app is complex (migrations, moving data files around, etc…), then it’s vital you test, and test on actual devices.

Keep building out your app and you’ll keep learning. Start new apps, and try implementing individual features as standalone apps to get an understanding of how it works. I built a GPS tracker as a separate app along with a basic line drawing app and image resizing app. These were simple systems I put together just so I could isolate the functionality I was implementing, and build without needing it to fit into any existing design.

Keep building, keep learning, get on StackOverflow and try answering some questions, and you’ll find you pick up the language and SDK with time and effort.

Getting the data with Mixpanel and iOS

2010-12-30T00:00:00+00:00

It’s easy to get used to having detailed statistics when developing web apps, using split tests to determine the better optimisations and finding patterns in usage that lead to your most profitable customers and features. When developing iOS apps, these stats are much harder to collect.

Thankfully, there is a service that helps aleviate this problem: Mixpanel.

Integrating your app with the Mixpanel API is a simple case of grabbing their library, importing the MPLib filder and adding the include to your prefix header:

#import "MixpanelAPI.h"

… initialising the lib in your app delegate:

- (BOOL)application:(UIApplication *)application didFinishLaunchingWithOptions:(NSDictionary *)launchOptions {
  MixpanelAPI* mixpanel = [MixpanelAPI sharedAPIWithToken:@"Your Token"];
	
  [mixpanel registerSuperProperties:[NSDictionary dictionaryWithObjectsAndKeys:
      [[UIDevice currentDevice] systemVersion], @"OS Version",
      [[NSBundle mainBundle] objectForInfoDictionaryKey:@"CFBundleVersion"], @"version", nil
    ]
  ];

  [mixpanel track:@"App Launch"];

  // Override point for customization after app launch    
  [window addSubview:[navigationController view]];
  [window makeKeyAndVisible];

  return YES;
}

… and collecting metrics as you see fit:

MixpanelAPI *mixpanel = [MixpanelAPI sharedAPI];
[mixpanel track:@"Checkin"];

Instead of synchronously sending data back to your Mixpanel, individual events are queued up and sent as a connection is available - good news if your app is going to be used when net connectivity may be poor.

Additionally, you can send properties that allow you to segment your data further. In the above example, I send the OS version and app bundle version with every request. Collecting these additional details will let me determine whether a particular pattern can be attributed to either of these attributes.

An Improved Interface for Migrating to Jekyll

2010-12-27T00:00:00+00:00

I’m a sucker for shiny new services, which means I’ve swapped out blog hosts around half a dozen times. This is the third time I’ve come back to Jekyll, and the importing system hasn’t really changed since I first migrated from self-hosted Wordpress.

My latest migration is from Posterous - a solid blogging system, but lacking in customisation, and with my recent move to self-hosting, I wanted to consolidate my stuff.

There is no easy export option for Posterous - the only method of getting your data out is to use their API.

I set about writing an importer for Posterous, then noticed this ticket asking for a better import CLI.

After a little hacking around, I present: A Blog import command set for Jekyll.

Usage

To import a blog from another service or export file:

CSV

jekyll import csv --file=</path/to/csv.csv>

Typo

jekyll import typo --dbname=<database name> --user=<username> --pass=<password>

TextPattern

jekyll import textpattern --dbname=<database name> --user=<username> --pass=<password> --host=<hostname>

MT

jekyll import mt --dbname=<database name> --user=<username> --pass=<password>

Mephisto

jekyll import mephisto --dbname=<database name> --user=<username> --pass=<password>

Drupal

jekyll import drupal --dbname=<database name> --user=<username> --pass=<password>

Wordpress

jekyll import wordpress --dbname=<database name> --user=<username> --pass=<password>

Wordpress.com

jekyll import wordpressdotcom --file=</path/to/export.xml>

Posterous

jekyll import posterous --user=<email> --pass=<password> --site=<sitename (default is primary site)>

Note that the Posterous importer is required for this to work.

Automated Deploy with Unicorn and Git

2010-12-14T00:00:00+00:00

As part of my recent move to self-hosting my private git repos, I have built a simple Sinatra app that lets me view and browse them.

I’m using Unicorn behind Nginx to serve this little app up, and as I’m self-hosting, I decided to give automatic deployment a try; Capistrano just felt like added overhead in this case.

My ideal deploy path for this app is:

Make changes on master branch
Rebase changes to production branch as required
Push
Post receive hook deploys app, and restarts unicorn

After creating the repo on my server, I added the following script to myapp.git/hooks/post-receive, and ran chmod ug+x post-receive to make the file executable.

APP_ROOT=/path/to/app
UNICORN_PID=$APP_ROOT/tmp/pids/unicorn.pid

# Deploy the updated code
mkdir -p $APP_ROOT

# Only want to export from the production branch
git archive production | tar -x -C $APP_ROOT

if [ -f $UNICORN_PID ]
  then
    kill -USR2 `cat $UNICORN_PID`
  else
    unicorn -c $APP_ROOT/unicorn.rb -D -E production
fi

Note that there are assumptions as to the location of the app and it’s structure. You’ll need a tmp folder in the app root which contains the pid folder. I’ve not included the unicorn config, as it not really related to this post.

Self Hosted Private Git

2010-12-12T00:00:00+00:00

I switched back to Jekyll from Posterous as I prefer having control over my data. Posterous’ auto formatting never really worked out well for me, and I like having the opportunity to customise pages as I see fit rather than having to conform to one particular format.

Anyhow, I’ve ended up with a share on Gandi and decided to take the opportunity to cut my repo hosting bills by self-hosting. At the time of writing, I’ve got a maxed out medium plan, which has become more of a problem as time has gone on - the next larger plan is by request, and will obviously cost more than the $22 I’m paying per month at the moment.

Looking around, the only self-hosting system I could find that had decent documentation was Gitolite.

Gitolite integrates with ssh, meaning there are no daemons running, so there is nothing to monitor or worry about. The installation instructions are thorough and correct, and I was able to get a running installation in under an hour. Additionally, it has permissions that run down to branch granularity, and it can be managed through git. Perfect.

To migrate my repos from Github to Gitolite, I firstly created the repos by updating my gitolite-admin repo (as per the instructions) and pushing those changes. Once I had the new remotes ready to go, I used the following script to automatically migrate:

#! /bin/bash

repos=( repo1 repo2 repo3 )

for repo in ${repos[@]}
do
echo $repo
git clone --bare git@github.com:MattHall/$repo
cd $repo.git
git push --mirror git@your-server.com:$repo.git
cd ..
rm -fr $repo.git
done

‘repos’ contain the names of the repos to be migrated. Note that collaborators, comments and anything outside of git itself will not be migrated.

Imagemagick Error in Textmate

2010-11-26T00:00:00+00:00

If you are getting the "... is not recognized by the 'identify' command" runaround in Textmate, yet your examples pass when run via the command line, then your Textmate shell path might be wrong.

To fix, add the path to ImageMagick (run "which identify" in Terminal) to the beginning of Textmate's Preferences > Shell Variables > PATH entry.

Rerun your examples, and all should be hoopy.

Flight 404 and Kinect

2010-11-22T00:00:00+00:00

Bundler no metadata found problem

2010-09-28T00:00:00+00:00

On installing bundler, and getting into Rails 3.0.0, I began to get the following error when running bundle install:

/Library/Ruby/Site/1.8/rubygems/package/tar_input.rb:113:in `initialize': No metadata found! (Gem::Package::FormatError)

There are sporadic posts about this, but the one that solved the problem for me was the second comment on this issue:

I tracked this down to a corrupt .gem file in rubygems/cache folder. I submitted a patch to the rubygems team to provide better diagnostic output message which identifies the offending file. It still doesn’t address the root cause, but should help people find the problem faster.

At line 111 in /rubygems/package/tar_input.rb, change to:

unless has_meta
  puts "#{io.path} may be corrupt! Delete it and retry the operation"
  raise Gem::Package::FormatError, "No metadata found!"
end

The problem (im my case) was caused by a corrupt .gem file. Adding those lines to tar_input will give you a more meaningful error message, and allow you to delete the offending gem files.

Removing duplicate error messages in Rails

2010-09-04T00:00:00+00:00

If you have a number of validations on a model, then you can quickly end up with something like this appearing in your error blocks:

I'd prefer to see only one error message for each attribute:

This can be achieved by creating an errors_helper.rb file in your apps helpers folder, and pasting the following code (mercilessly modified from this post by Bert over on RailsForum):

module ErrorsHelper

  def error_messages_for(*params)
    options = params.extract_options!.symbolize_keys

    if object = options.delete(:object)
      objects = Array.wrap(object)
    else
      objects = params.collect {|object_name| instance_variable_get("@#{object_name}") }.compact
    end

    count  = objects.inject(0) {|sum, object| sum + object.errors.count }
    unless count.zero?
      html = {}
      [:id, :class].each do |key|
        if options.include?(key)
          value = options[key]
          html[key] = value unless value.blank?
        else
          html[key] = 'errorExplanation'
        end
      end
      options[:object_name] ||= params.first

      I18n.with_options :locale => options[:locale], :scope => [:activerecord, :errors, :template] do |locale|
        header_message = if options.include?(:header_message)
          options[:header_message]
        else
          object_name = options[:object_name].to_s
          object_name = I18n.t(object_name, :default => object_name.gsub('_', ' '), :scope => [:activerecord, :models], :count => 1)
          locale.t :header, :count => count, :model => object_name
        end
        message = options.include?(:message) ? options[:message] : locale.t(:body)
        
        error_messages = objects.sum do |object|
          object.errors.on(:name)
          full_flat_messages(object).map do |msg|
            content_tag(:li, ERB::Util.html_escape(msg))
          end
        end.join.html_safe

        contents = ''
        contents << content_tag(options[:header_tag] || :h2, header_message) unless header_message.blank?
        contents << content_tag(:p, message) unless message.blank?
        contents << content_tag(:ul, error_messages)

        content_tag(:div, contents.html_safe, html)
      end
    else
      ''
    end
  end
  

  ####################
  #
  # added to make the errors display in a single line per field
  #
  ####################
  def full_flat_messages(object)
    message_bucket = {}
    full_messages = []

    object.errors.each do |error|
      if message_bucket[error[0].to_sym].nil?
        message_bucket[error[0].to_sym] = true
        
        full_messages << (error[0]).titleize + ' ' + error[1]
      end
    end
    full_messages
  end

end

Be careful with UIImage imageNamed

2010-09-04T00:00:00+00:00

An iPad app I have been working on recently began being jettisoned by the OS for using too much memory. The annoying this was that there was no sign of leaks, and while running the app through the Allocations Instrument, there was no sign of excessive memory usage.

The app requires the main screen to have a large background image that changed as a UI control is modified by the user.

I checked the usual suspects – leaks, allocations and the stack traces therein. What I found was that there were several large blocks of memory being malloc'ed from within CoreGraphics, and that this memory appeared to have been allocated outside of the scope of the code I’d written:

After a lot of tracing this problem, I found the culprit:

[UIImage imageNamed:@'Someimage']

Calls to this method cause the named image to be loaded into memory and cached. The image will remain in memory for the duration of your apps lifetime, and, if you have enough large images and calls to this function, they will cause the OS to jettison your program.

The Solution

Don’t use [UIImage imageNamed:@‘Someimage’] unless the image is small and you want to use it repeatedly in your app. Instead, the following snippet will allow you to load the image, then release it’s memory once you’re finished:

[UIImage imageWithContentsOfFile:path];

Note that imageNamed is not evil, it just needs some extra thought when using it. If there are enough images of a significant size, then the memory retained by this call may get your program forceably removed by the OS.

WEH

2010-08-19T00:00:00+00:00

This Book is Made (of Rabbits and Lemonade)

2010-08-19T00:00:00+00:00

My favourite track from _why's (poignant) guide to ruby:

Download now or listen on posterous

chapter-2-this-book-is-made.mp3 (4374 KB)

Happy #whyday!

Steph Davies and Friends

2010-08-17T00:00:00+00:00

Lasse Gjertsen

2010-08-12T00:00:00+00:00

Forgot about this. Old(ish), but still awesome.

Maptastic - Formtastic Map Location Selector

2010-07-28T00:00:00+00:00

If you’ve spent time with Rails, you’ll already know that Formtastic is ace, and constructing forms using the SemanticFormBuilder keeps your code neat, and the output sensible and semantic.

More often, I find the need to include a control in my forms that allows the user to find and select a single LatLng, and, up until now, I’ve been adding in hidden fields and creating js that allows me to present a map control.

In an effort to reduce this duplication, I’ve built a Rails plugin for Formtastic that provides a new map_input and multi_input function, which means if you have a lat/lng pair, and you want to let the user fill out these values, all you need do is:

<% semantic_form_for @venue do |f| %>
  <%= f.multi_input :latitude, :longitude, :as => :map %>
<% end %>

Note you’ll also need to include the Google Maps script in your header:

<script type='text/javascript' src='http://maps.google.com/maps/api/js?sensor=true'></script>

Installation

You’ll need Formtastic

Install Maptastic as a plugin:

script/plugin install http://github.com/MattHall/maptastic

Development

This plugin is under development. It’s pretty simple, and patches are very welcome.

The Repo is available on GitHub: http://github.com/MattHall/maptastic

Report bugs here: https://matt.purifyapp.com/projects/maptastic/issues

The Longest Way

2010-07-27T00:00:00+00:00

Beta Testing Features in Rails

2010-07-19T00:00:00+00:00

When building a new feature into a web app, it’s useful to have beta testers use it for a a while before it’s fully launched in order to catch any last minute bugs, to get an idea of whether the new feature feels right and if it actually solves the original problem.

It is also sometimes desirable to have the beta testing process work from within the existing production app, in order for the production data to be used, and to prevent beta users from being isolated from other users. This may not be the case with an application that does not promote interaction between users, and in such cases, it is probably desirable to create a beta deploy of the application.

The following is the method by which I beta test features in my own (Rails) apps – YMMV.

Give certain users beta status

This is fairly simple – either add a boolean beta field to the user model or, if you’re using something to manage roles, add a beta role to the users you’re adding to your beta program. Be sure to protect this attribute:

attr_protected :is_beta

If you’re using roles, then you can segment your beta users, granting access to individual features if necessary.

Hiding actions fron non-beta users

In your application_controller.rb define the following function:

def require_beta
  @current_user.is_beta?
end

In any controllers that have actions that need to be protected, you can then write:

before_filter :require_beta

Hiding UI elements

There will probably be links, forms and UI elements that you’ll need to only display to beta users, so, in your application helper:

def beta(&block)
  yield(block) if current_user.try(:is_beta?)
end

def non_beta(&block)
  yield(block) unless current_user.try(:is_beta?)
end

Then, whenever you have a UI element you only want to appear to beta users, just wrap it in the beta call:

<% beta do %>
  This text will only be shown to logged in beta testers
<% end %>

<% non_beta do %>
  This text will only be shown to normal users and non-logged in users
<% end %>

Flight404

2010-07-16T00:00:00+00:00

If you haven't seen Flight404's stuff yet, you're missing out.

Last Day Dream

2010-07-15T00:00:00+00:00

Chris Milk

Hacking Through Rails' Multi Parameter Date Handling

2010-07-15T00:00:00+00:00

This code supersedes the code in this post.

Given a date_select, a user might (intentionally or otherwise) neglect to enter a value for one of the selects. Rather than add an error to the instance or default the missing date attributes to the current date, Rails will throw an exception. This is problematic, as it leads to the following sequence of events:

User commits a form with a missing date_select value
Rails throws an exception
User sees a 500

The following code patches Activerecord::Base – adding a validation and messing with #execute_callstack_for_multiparameter_attributes. The end result is that on submitting an invalid date, an error is added to the instance when validations are run:

User commits a form with a missing date_select value
multipart_error_on_attribute gets set to the name of the attribute that contains the invalid date
In the controller, when save is called, the validations fail, and an error is added to the instance
Typically, the user will then see the form rendered and an error message.

This will only work for a narrow set of problems – this code isn’t really fit for stuffing back into Rails, so use with caution.

 

module MultiparameterHack
  def self.included klass
    klass.class_eval do
      @multipart_error_on_attribute = nil

      def validate
        errors.add_to_base "#{@multipart_error_on_attribute.humanize} is invalid" unless @multipart_error_on_attribute.nil?
      end

      undef :execute_callstack_for_multiparameter_attributes
      def execute_callstack_for_multiparameter_attributes(callstack)
        errors = []
        callstack.each do |name, values_with_empty_parameters|
          begin
            klass = (self.class.reflect_on_aggregation(name.to_sym) || column_for_attribute(name)).klass
            # in order to allow a date to be set without a year, we must keep the empty values.
            # Otherwise, we wouldn't be able to distinguish it from a date with an empty day.
            values = values_with_empty_parameters.reject(&:nil?)

            if values.empty?
              send(name + "=", nil)
            else
              value = if Time == klass
                instantiate_time_object(name, values)
              elsif Date == klass
                begin
                  values = values_with_empty_parameters.collect do |v| v.nil? ? 1 : v end
                  Date.new(*values)
                rescue ArgumentError => ex # if Date.new raises an exception on an invalid date
                  instantiate_time_object(name, values).to_date # we instantiate Time object and convert it back to a date thus using Time's logic in handling invalid dates
                end
              else
                klass.new(*values)
              end

              send(name + "=", value)
            end
          rescue => ex
            errors << ActiveRecord::AttributeAssignmentError.new("error on assignment #{values.inspect} to #{name}", ex, name)
            @multipart_error_on_attribute = name
          end
        end
      end

    end
  end
end

ActiveRecord::Base.send(:include, MultiparameterHack)

The Tarantino Mixtape

2010-07-14T00:00:00+00:00

Cheeky video remix by Eclectic Method

MultiParameter Exception Hack

2010-07-14T00:00:00+00:00

When using date_select or datetiime_select form helpers in rails, the valies of the individual time fields are sent separately, and reassembled back into a date within ActiveRecord::Base. Unfortunately, if the user enters an invalid date, then this process fails with a ActiveRecord::MultiparameterAssignmentErrors exception.

The exception occurs before we actually get anywhere near our application code, so it’s by no means a trivial problem, so it’s unfortunately necessary to monkey patch the offending function. The following code represents a substantial hack – there are probably better ways of taking care of this problem, but this is the best I could come up with in the short space of time I had available.

The solution I came to involved reimplementing the execute_callstack_for_multiparameter_attributes function in ActiveRecord::Base, and removing the exception throw. An instance variable (multipart_error_on_attribute) is then used to store the invalid attributes which is checked in a custom validation. If multipart_error_on_attribute contains any number of attribute symbols, then the validation fails, and an error is added to the object.

module MultiparameterHack
  def self.included klass
    klass.class_eval do
      @multipart_error_on_attribute = nil

      def validate
        errors.add_to_base "#{@multipart_error_on_attribute.humanize} is invalid" unless @multipart_error_on_attribute.nil?
      end

      undef :execute_callstack_for_multiparameter_attributes
      def execute_callstack_for_multiparameter_attributes(callstack)
        errors = []
        callstack.each do |name, values_with_empty_parameters|
          begin
            klass = (self.class.reflect_on_aggregation(name.to_sym) || column_for_attribute(name)).klass
            # in order to allow a date to be set without a year, we must keep the empty values.
            # Otherwise, we wouldn't be able to distinguish it from a date with an empty day.
            values = values_with_empty_parameters.reject(&:nil?)

            if values.empty?
              send(name + "=", nil)
            else
              value = if Time == klass
                instantiate_time_object(name, values)
              elsif Date == klass
                begin
                  values = values_with_empty_parameters.collect do |v| v.nil? ? 1 : v end
                  Date.new(*values)
                rescue ArgumentError => ex # if Date.new raises an exception on an invalid date
                  instantiate_time_object(name, values).to_date # we instantiate Time object and convert it back to a date thus using Time's logic in handling invalid dates
                end
              else
                klass.new(*values)
              end

              send(name + "=", value)
            end
          rescue => ex
            errors << ActiveRecord::AttributeAssignmentError.new("error on assignment #{values.inspect} to #{name}", ex, name)
            @multipart_error_on_attribute = name
          end
        end
      end

    end
  end
end

ActiveRecord::Base.send(:include, MultiparameterHack)

One Issue Per Ticket, Please!

2010-07-13T00:00:00+00:00

Bug tracking systems exist to help many people organise their efforts in order to fix a problem or add a feature.

The lifecycle of a bug may go something like this:

Reporter opens a ticket
A developer gets assigned the ticket
The developer completes work on the ticket
The developer marks the ticket as resolved, and assigns it back to the Reporter
The reporter checks the fix, and closes the ticket.

Though there are many versions of this lifecycle, the purpose is to move an issue from new to closed, logging the steps taken on that particular issue.

Problems arise when one ticket contains many issues.

Someone will see that a bug has been worked on or resolved, and on testing the feature will find another semi-related issue that has either snuck in with the work for the original issue, or not been detected up until that point. Spotting this, they will add a comment to the original ticket. This is wrong.

The correct course of action in this situation is to open a new ticket – not to add the problem to the discussion on the original issue. The latter course of action quickly degrades the ticket into an email-like exchange which can cause confusion. If email were a good enough solution, we wouldn’t need bug tracking systems. The only time it is reasonable to add a comment to an existing ticket is when the comment relates only to the original issue.

Showing the Commits You've Yet to Push in Git

2010-07-08T00:00:00+00:00

If you’re using Git, and have set up a local branch that tracks a remote branch, you’ll probably have seen the following message in your git status at some point:

# On branch master 
# Your branch is ahead of 'origin/master' by 2 commits.

This tells us that there are two commits in this branch that don’t exist in the remote branch. To find out which commits these are, we can do the following:

git log origin/master..

Note the double period – it’s not a typo.

This command can be used to compare the branch you’re currently on with any other branch, showing the commits yet to be pushed.

To see commits on the remote branch you have yet to pull, use:

git log origin/master...

I've just asked you to print it for me!

2010-07-08T00:00:00+00:00

Having a common name and an email address comprised of that name means I do occasionally get email that was intended for other people. What I find disturbing is the insistance by some of these people that I am in fact the intended recipient:

From: Anne Hall
To: me
Subject: please print

Hi Mrs. Hall,

I hope you’re having a good summer. I will be retaking my math 2 subject test in november and I have signed up to take the SAT again in november. I’m going to be college tripping in on the east coast in august. I will most likely visit Haverford, Swarthmore, schools around the D.C. area, and other colleges. I haven’t decided completely yet. Right now, everything that we had is still on my list, and I will e-mail Mrs. Mack about the recommendations.

,Chris

From: me
To: Anne Hall
Subject: RE: please print

Hi,

I don’t think this email was intended for me.

Thanks,

Matthew Hall

codebeef.com

From: Anne Hall
To: me
Subject: RE: RE: please print

I’ve just asked you to print it for me!

From: me
To: Anne Hall
Subject: RE: RE: RE: please print

Ok, I’ve printed it. Come round and pick it up.

Rails Try with Parameters

2010-07-07T00:00:00+00:00

When dealing with a collection of different objects in Rails, it is often necessary to call a method that may or may not be defined on all of the instances in the collection. For example, formatting the following array of DateTimes will throw an ArgumentError:

[1.day.ago, 1.day.from_now, nil].each {|d| puts d.to_s(:admin)}

The reason being that nil.to_s takes zero arguments.

Trying

Rails provides try, which can be used in association with a list of parameters to attempt to invoke a method on an object:

[1.day.ago, 1.day.from_now, nil].each {|d| puts d.try(:to_s, :admin)}

Splitting an Array in Half in Ruby

2010-07-06T00:00:00+00:00

So I have a collection of objects I need to split into two approximately equal arrays. Luckily, there is the partition method in Ruby:

i = 0
count = User.count
group_one, group_two = User.all.partition { i += 1; i < count / 2 }

The partition method iterates over each object, with the object being pushed to the first array if the block evaluates to true, and vice-versa.

Alternatively, we could use in_groups_of if we’re using Rails:

group_one, group_two = User.all.in_groups_of(User.count / 2)

Variable number of Objects on has_many in Factory_Girl

2010-06-29T00:00:00+00:00

When creating a (factory_girl) factory for an object that has an optional has_many relation, I would like for the factory to create a variable number of objects on the relation. I have the following code, which works, but I suspect there is a better way of doing this:

Factory.define(:product) do |p|
  p.name {Faker::Lorem.words(3)}

  p.images do
    instances = []
    rand(10).times {instances << Factory.build(:image)}
    instances
  end

end

Anyone have any thoughts?

Two Views of my Job

2010-06-25T00:00:00+00:00

Avoiding Short-Circuit Evaluation when Validating in Rails

2010-06-25T00:00:00+00:00

Say you are saving two unrelated models at the same time, and you want to validate them before doing so. The following code is usually the first thought:

if (model1.valid? && model2.valid?)
  # Save and redirect
else
  # Render form again to show errors
end

The problem is that if both models are invalid, then we need to show the error messages for both, however the use of AND will mean that only the first of the two components of the if statement will be executed, leaving model2 unchecked. The user will correct the errors shown in model1 and resubmit, but validation will fail on model2, and the form rendered with error messages a second time.

A better solution is to avoid the short-circuit evaluation:

([model1, model2].map(&:valid?)).all?

This will validate both models, and only return true if they are both valid, meaning that the user will see the errors for both models before resubmitting the corrected form.

Showing the Change History for a File in Git

2010-06-24T00:00:00+00:00

To view a list of changes in a git repo, we can use:

git log

To view the list of changes for a single file:

git log filename

To view the actual patches and commit messages:

git log -p filename

Comparing a DateTime with a Time in Ruby

2010-06-23T00:00:00+00:00

For anyone that comes across this problem:

>> a = Time.now 
=> Wed Jun 23 13:50:35 +0100 2010 
>> b = DateTime.parse(a.to_s) 
=> Wed, 23 Jun 2010 13:50:35 +0100 
>> a 
=> Wed Jun 23 13:50:35 +0100 2010 
>> b 
=> Wed, 23 Jun 2010 13:50:35 +0100 
>> a == b 
=> false 
>> a.to_datetime == b 
=> false

The solution is to use ===

>> a.to_datetime === b 
=> true

From the docs:

Compares dates by Julian Day Number. When comparing two DateTime instances, or a DateTime with a Date, the instances will be regarded as equivalent if they fall on the same date in local time.

Rocks

2010-06-22T00:00:00+00:00

Building a site around rock climbing has meant that my photos folder contains stuff like this.

SSL_requirement and RSpec

2010-02-17T00:00:00+00:00

If you’re using ssl_requirement, and speccing with RSpec, you’ll probably run into a problem whereby the before_filter ensure_proper_protocol redirects during your controller specs. To fix this, you can stub the call to ensure_proper_protocol, and make it return true. Add the following to your spec_helper.rb:

Spec::Runner.configure do |config|
  ...
  config.before(:each, :type => :controller) do
    controller.stub(:ensure_proper_protocol).and_return(:true)
  end
end

Checking for a WiFi Connection in One Line

2009-08-22T00:00:00+00:00

So if you have an app that requires a WiFi network connection in order to reliably synchronize some large-ish data set, you need to notify your user if the connection is not available. Apple provides the Reachability example app that demonstrates how to do this, and you can use the Reachablility class from that app in your own app:

Add the Reachability.h and Reachability.m files to your project.
Include Reachablility.h in the header of the file you’re going to do the check
Add the following line:

[[Reachability reachabilityForLocalWiFi] currentReachabilityStatus] == ReachableViaWiFi

Easy Linking with Polymorphic Associations in Rails

2009-07-24T00:00:00+00:00

Just a quick tip, this one! If you have a model that is polymorphic, say a comment model:

class Comment
  belongs_to :user 
  belongs_to :commentable, :polymorphic => true 
end

and you want to link to the item the comment relates to in the admin system, you can do this:

link_to comment.commentable.title, [:admin, comment.commentable]

Note this assumes all of your commentable models have a title attribute or method, and that you have used normal RESTful routes. The :admin symbol will namespace the link to /admin

Bort updated to Rails 2.2

2008-11-10T00:00:00+00:00

If you don't already know, Bort has now been updated to Rails 2.2. For more information on exactly what has changed, please see Jim's post.

I have also updated my email-as-login fork of bort to include this release.

Please give us your feedback over on our Uservoice page.

A Helper for Testing Access Control with RSpec

2008-10-14T00:00:00+00:00

When testing Controller actions for access restriction, I was building a set of tests for each controller that were in a similar format to this:

describe "access control" do
  it "should prevent access by non-logged-in users"
  it "should prevent access by normal users"
  it "should prevent access by editor users"
  it "should prevent access by admin users"
  it "should allow access by super admin users"
end

Unforunately, this was neither dry, complete, nor very readable, so I have come up with the following helper that allows you to specify the access restrictions for a number of methods: In your spec_helper.rb file, add the following function:

def access_control (code, options={})
  options = {:allow => [], :disallow => []}.merge(options)

  options[:allow].each do |user|
    it "#{code} should allow #{user.to_s}" do
      login_as(user)
      eval code
      response.should_not redirect_to(login_path)
    end
  end

  options[:disallow].each do |user|
    it "#{code} should disallow #{user.to_s}" do
      login_as(user)
      eval code
      response.should redirect_to(login_path)
    end
  end
end

Then in the specs for your controller, call the function as follows:

[ "get :show, :id => 0",
  "get :new",
  "post :create, :setting_value => {:value => 'A new setting', :group => 0}",
  "get :edit",
  "put :update, :id => SettingValue.first.id, :setting_value => {:value => 'changed'}",
  "delete :destroy, :id => SettingValue.first.id"
].each do |action|
  access_control(action, {:allow => [:super_admin], :disallow => [:quentin, :admin]})

This allows you to very quickly build a set of specs that fully describe how access control to your methods should restrict your users:

A Bort Fork - Rails Base Application

2008-09-09T00:00:00+00:00

In order to DRY-up our process of creating new rails projects over at Fudge, we recently created Bort - a base rails application that contains all of the bits that we push together when starting out.

Jim has more about the base system, but there is also a fork that contains all of the bort goodness, with the notable difference that instead of using usernames to authenticate, the users email address is used instead.

Enjoy.

Unterminated String Literal with Asset Packager

2008-05-20T00:00:00+00:00

A recent release of Bouldr threw a bit of a spanner in the works when the excellent Asset Packager began choking when attempting to create our concatenated and compressed javascript file. It turns out that as I am developing in Windows (gah!), I need to run dos2unix in order to ensure that the js files are in the appropriate file format (our server is running CentOS). In order to automate this process, I added the following to our deploy script:

# Need to ensure files are in UNIX format for the asset packager to run correctly 
run "cd #{release_path}/public/javascripts && find . -name '*.js' -exec dos2unix '{}' \\;"

# Run the asset packager to create the production javascript and css files 
run "cd #{release_path} && rake asset:packager:build_all RAILS_ENV=production"

Really Simple History (RSH) Initial URL

2008-04-07T00:00:00+00:00

I recently came across a small problem in RSH whereby on loading a page with a state in the url (e.g. localhost.com/test_page#test) would not trigger the history listener function on it’s initial load. To fix this, I needed to modify my window onload function to include a call to my history listener, as follows:

history_listener: function (newLocation, historyData) { ... },  // Window.onload calls this 

initialize_history: function () {
  dhtmlHistory.initialize();
  dhtmlHistory.addListener(this.history_listener.bindAsEventListener(this));                        

  // Added this line to force a call to my history listener
  this.history_listener(dhtmlHistory.getCurrentLocation()); 
}

This ensures that when the page is freshly loaded, the history listener is called with the contents of our initial location.

Installing MySQL Gem - CentOS

2008-01-24T00:00:00+00:00

This one had me for a while: When attempting to install the MySQL gem on my shiny new Slicehost slice, I was getting the folowing error:

Building native extensions.  This could take a while...
ERROR:  Error installing mysql:
ERROR: Failed to build gem native extension.

/usr/local/bin/ruby extconf.rb install mysql -- --with-mysql-include=/usr/include/mysql --with-mysql-lib=/usr/lib/mysql
checking for mysql_query() in -lmysqlclient... no
checking for main() in -lm... yes
checking for mysql_query() in -lmysqlclient... no
checking for main() in -lz... yes
checking for mysql_query() in -lmysqlclient... no
checking for main() in -lsocket... no
checking for mysql_query() in -lmysqlclient... no
checking for main() in -lnsl... yes
checking for mysql_query() in -lmysqlclient... no
*** extconf.rb failed ***
Could not create Makefile due to some reason, probably lack of
necessary libraries and/or headers.  Check the mkmf.log file for more
details.  You may need configuration options.

Provided configuration options:
   --with-opt-dir
   --without-opt-dir
   --with-opt-include
   --without-opt-include=${opt-dir}/include
   --with-opt-lib
   --without-opt-lib=${opt-dir}/lib
   --with-make-prog
   --without-make-prog
   --srcdir=.
   --curdir
   --ruby=/usr/local/bin/ruby
   --with-mysql-config
   --without-mysql-config
   --with-mysql-dir
   --without-mysql-dir
   --with-mysql-include=${mysql-dir}/include
   --with-mysql-lib=${mysql-dir}/lib
   --with-mysqlclientlib
   --without-mysqlclientlib
   --with-mlib
   --without-mlib
   --with-mysqlclientlib
   --without-mysqlclientlib
   --with-zlib
   --without-zlib
   --with-mysqlclientlib
   --without-mysqlclientlib
   --with-socketlib
   --without-socketlib
   --with-mysqlclientlib
   --without-mysqlclientlib
   --with-nsllib
   --without-nsllib
   --with-mysqlclientlib
   --without-mysqlclientlib

Gem files will remain installed in /usr/local/lib/ruby/gems/1.8/gems/mysql-2.7 for inspection.
Results logged to /usr/local/lib/ruby/gems/1.8/gems/mysql-2.7/gem_make.out

It turns out, all I needed was to run the following:

gem install mysql -- --with-mysql-config=/usr/bin/mysql_config

Hope this saves someone the hour or so I spent scratching my head!

Stopping a submit_tag from Sending it's Value

2007-12-12T00:00:00+00:00

OK, so say you have a form as follows:

<% form_tag '/', :method => :get do -%>
<p><%= text_field_tag 'q', params[:q] ||= '' %> <%= submit_tag 'Search' %> 
<% end %>

That form will send the contents of the text box to the page at ‘/’. Unfortunately, it will also send through the value of the submit button, which is ‘Search’ in this case. The url generated will be:

/?q=something&commit=Search

In this case, we would want only the ‘q’ parameter passing through, and in order to do this, all we need do is nullify the name parameter:

<% form_tag '/', :method => :get do -%>
<p><%= text_field_tag 'q', params[:q] ||= '' %> <%= submit_tag 'Search', :name => nil %> 
<% end %>

This will create the following url on submission:

/?q=something

Gem Documentation Server

2007-12-06T00:00:00+00:00

Just a quick tip I recently came across: If you want to see the docs for any of the gems installed on your dev machine, type the following in the command line…

gem_server

…and then browse over to http://localhost:8808/

Rails 2.0 is Almost Here - Learn About It!

2007-12-04T00:00:00+00:00

Railscasts are a Rails developers wet-dream.

Since Ryan created the site, it has been a fount of knowledge for anyone wishing to learn about the latest and greatest tips and tricks. To make things even better, they are free, though PeepCode is also very much worth a look for those of us wanting to glean a little (or a lot) of in-depth knowledge; $9 is a menial sum to pay for what amounts to a full-on lecture!

If you haven't done already, go take a look - there is even a bunch of screencasts dedicated to Rails 2.0!

Browser History in Web Apps

2007-12-03T00:00:00+00:00

Given that the back button has been around for many years, and users have grown accustomed to it's use, it's a shame to see so many web applications neglecting this feature. As an example, I'll pick on Lightbox, a tool that I like, yet one that is flawed in it's disregard for the back button. For example:

You navigate to a page that contains a thumbnailed image
You click the image, and Lightbox shows it centered in your browser, full size
You click back, press the back button on your mouse or whatever

Now, I realise that Lightbox has a close button in the image dialog box, but I'm not looking at the box, I'm looking at the image, and furthermore, my natural path back to the originating page is to press the back button, not to close the Lightbox. I therefore end up at the page I was on before I saw the thumbnail. I've seen this with many users - the back button has been scarred into our mind, and to break it now is frivolous, and will ultimately lead to frustration, and lost users. It's with relief therefore, that I found Really Simple History, a JavaScript library for dealing with this exact problem - to create a JS history system that can run across many browsers, and provide developers with a little hope that our apps will no longer break the most fundamental user assumptions. We tried to get the browser history working in Bouldr, and to an extent, we succeeded. The difference between coding something proprietary, and using an open source solution, though is very great indeed, so without doubt, we'll be refactoring to use this library in the future.

Add your Business to Facebook (for Free)!

2007-11-28T00:00:00+00:00

From a business standpoint, Facebook's killer feature is it's Social Network, which gives even none-techies the ability to share nuggets of information found both on Facebook and the web at large with their friends.

What I didn't realise until recently is that Facebook also allows businesses to set up pages inside the walled garden, that gives FB's users the opportunity to show their support of your business by becoming 'Fans'.

It doesn't cost anything to set up a business page, which comes with a reporting tool that shows you just how many visits / page views / fans you're getting per day, an update tool that allows you to send news items to your fans, and a bunch of advertising tools that you can use to promote your business page.

I set Bouldr up on the 21st November 2007 (check it out!), and already I have around 15 fans, with a slew of visitors coming from Facebook over to Bouldr's main site. If you have a website that could benefit from a little social promotion, I'd suggest taking a look at Facebook's business listing service - it's free, so there is very little risk, really.

Wheel of Life (V16)

2007-11-22T00:00:00+00:00

Just Awesome.

Mike Robertson Sends The Eiffel Tower

2007-11-22T00:00:00+00:00

In protest at Total's involvement in Burma, Mike Robertson solos the Eiffel Tower. (via)

High Fidelity (V13), Caley

2007-11-22T00:00:00+00:00

Andrew Earl sends this Highball problem with perfect style - Amazing to watch!

Helper for Testing Model Validation in Unit Tests

2007-11-21T00:00:00+00:00

Here’s a little snippet that I use to quickly form tests to ensure that my models validation is working correctly from within unit tests:

# In test_helper.rb   
# Give this function an array of invalid values, an array of valid values, the object and   
# attribute name to test, and it will ensure that only valid values are allowed   
def validation_tester ( valid_values, invalid_values, attribute_name, test_object )
  valid_values.each do |val|       
    test_object.[]=(attribute_name, val)
    assert test_object.save, "#{attribute_name} should be valid: #{test_object.[](attribute_name)}"       
    assert test_object.valid?
    assert !test_object.errors.invalid?(attribute_name)
  end

  invalid_values.each do |val|
    test_object.[]=(attribute_name, val) 
    assert !test_object.save, "#{attribute_name} should NOT be valid: #{val}" 
    assert !test_object.valid? 
    assert test_object.errors.invalid?(attribute_name)
  end   
end

Then, use it as follows in your unit tests:

def test_email_validation 
  valid_emails = ['matt@test.com', 'matt.hall@test.com', 'matt@test.ing.com', 'matt.hall@test.ing.com']
  invalid_emails = ['@test.com', 'test.com', 'matt@', 'matt.hall@com', '.matt@test.ing.co.uk', 'matt@@test.com', '', '.@.']
  u = create_user
  validation_tester( valid_emails, invalid_emails, 'email', u)
end

Can anyone suggest any improvements?

Your Friends are Pokemon

2007-08-28T00:00:00+00:00

I first noticed that I was a slave to Facebook when I was talking to a friend of a friend. We talked casually for a while, in the normal 'getting to know you' way, eventually touching on whether each other has a Facebook account. This seemed perfectly natural, however, on reflection, I could see that I had actually been working towards this question for the entire conversation; I wanted to breech The Facebook Threshold, and add this person as a friend.

This was not the first time I had considered the possibility that I have been trawling my friends friend lists just to increase the friend count on my own profile, rather the first time I have accepted this seemingly innocuous yet worrying prospect.

Speaking to my own friends, I find that this trend continues: many of those with more than fifty friends have actually collected people that they don't consider to be friends with any more - people from their pre-school days that are now 25 years old or more, and that they haven't spoken to or seen in the intervening period.

Maybe it's just our social vanity getting the best of us, but escaping from this arms race feels almost impossible from the inside.

How to Use a Table from Another Database for a Model in Rails

2006-12-15T00:00:00+00:00

The problem is simple; I wanted a method by which a model in my app could use a table from another MySQL Database (on the same server). The solution is equally simple:

  class MyModel &lt; ActiveRecord::Base
    set_table_name "other_db.table_name"
  end

Where other_db is the database name, and table_name is the table you want to use.

Selling our Soul (in the Best Possible Way)

2006-12-13T00:00:00+00:00

Building sites that both support and derive value from a community is one of our great memes at the moment, and one we have fully embraced when building Bouldr. One of the biggest problems, however, is how to break even; servers, bandwidth and development all cost money, and someone has to pay.

Thanks to the legions of advertising networks that have grown out of this rich demand for an easy monetisation method, there are plenty of opportunities to quickly and cheaply monetise a site (so long as you don't end up unintentionally breaking it, that is!).

It is important to note that we have been very concious with Bouldr that it should never contain anything that is of no use or consiquence to our users - banner advertisements make any UI look tarnished, and ad networks can often end up showing adverts that are not relevant to the user.

Bouldr serves climbers from around the world - we are a niche service, and we would like to keep it free for our users. Our niche is so well defined that we decided that a better way to organise the monetisation side of things was not to use a separate advertising publisher service, but to develop our own. In this way, we are able to moderate the advertisers we accept in order to prevent our users seeing irrelevant ads, and we can integrate them into our site in as many varied ways as we like.

Bouldr's Advertising network is not ready just yet, but we are well on the way, and hope to be launching soon.

5 Ways to Turn Buyers Off

2006-12-13T00:00:00+00:00

Once you have set up your business, spent time and money promoting your services and products, you’ll need to think about how to convert your visitors into customers. I am going to look at methods by which you can improve your conversion rates by discussing some of the reasons that people will not make a purchase after visiting your advertisement or site.

1. Getting the wrong people to your site

If you are targeting the wrong market sector with your advertising, then it really doesn’t matter how successful in drawing the crowds it is – they are not going to be interested in your products or services.

Be as specific as possible. If your market is a niche market, then put advertisements in magazines or on Web sites that serve your niche. Decide on your target audience, and bombard them with advertising. You will find that targeted advertising will pay dividends more than generic advertising – if you add your business to our directory, then make sure you add it to relevant categories. Adding your new Website Optimisation service to a ‘Website Hosting’ directory will get you very few interested parties. Add your business to a relevant directory, and you’re target market will find you!

2. Poor content quality

People will not read everything on your Web site or marketing material. Most people skim read when browsing, and as such, need content that is easy to process and digest.

Always ensure that you have spellchecked your copy before posting it to your site; misspelled or grammatically incorrect passages will instantly give your visitors a bad impression of your business, after all, if you can’t be bothered to check your own marketing material for correctness, then your products and services are probably going to suffer from the same lack of professionalism.

3. Not describing products or services fully

Visitors will want to find out more about your products and services before they buy. By not explaining the features or specifications in full, you are leaving questions unanswered in the minds of your customers. If these questions are important enough, there is a good chance that your visitor will look elsewhere for answers, and end up purchasing from your competitors.

Make sure that you cover as much detail as possible when describing your products – ideally, you should aim to be able to explain enough about the product that the potential customer has no need to research further. They should be ready and willing to make a purchase based on the information you have presented.

4. Not providing clear prices

There is no better way to lose the trust of potential customers than to provide obfuscated or inaccessable pricelists. If a customer is looking at your prices, then there is a good chance that they are either ready to make a purchase or are comparing prices. If you can present the price of your product or service on a page of your site in a way that allows a user to determine what it is within six seconds, then you're doing well. Any longer, and you risk overloading the user with options, decisions and ultimately turning them away.

5. Confusing navigation

Navigation is arguably the single most important aspect of any website; if no-one can find what they want, how are they going to buy from you?

Getting navigation right, however, is no simple feat - books have been written on the subject that can cover it to a far deeper level than I can in a short post such as this. Furthermore, there are an endless number of sites dedicated to disseminating the best practices in navigation design that would make far better reading. For example:

Summary

I have discussed some of the reasons as to why people may come to your site and not make a purchase. If you can avoid some of these common mistakes, then you are effectively reducing the number of barriers standing between a browsing customer becoming a paying customer.

One final word of advice: Take a look at your site now – can you see any other potential barriers that may prevent a visitor from finding what they want and spontaneously purchasing it? If so, look at how you can remove these barriers – each time you do, you are effectively opening your doors further, and allowing more people to make purchases.

5 Tips for Good Advertising Copy

2006-12-13T00:00:00+00:00

Advertising copy is the meat of your marketing campaigns – the actual information presented to your customers in order to get them to buy your products rather than a competitor’s.

If you create great advertising copy, your advertising campaigns can only benefit, bringing more custom to your business, however, creating bad copy can waste both time and money – it can even give your company a bad reputation in your target market’s eyes.

What I present here are a few points that I feel are vital to get right in order to produce marketing copy that is effective. These few points can be applied to many areas of marketing, whether it be creating a marketing brochure, a banner advertisement for a website or a point-of-sale display.

1. Get to the point

Most people will not read more than a single sentence or paragraph, so when it comes to writing advertising copy, it pays to get to the point as quickly as possible. Rambling on with details that have already been covered elsewhere in your advertisement will distract your potential customer, and in the time it has taken for them to read ‘Here at Syndicated System Solutions, our priority…’ they will have lost interest, and more than likely, will have moved on to the next item on the page.

Start your advertisement with a benefit of your product or service – something that will catch their eye, for example: ‘Kyotee is a leading UK business directory’. At this point, we have set the precedent for the visitor to read on – Why is Kyotee a leading directory service? What do they offer? What can they do for me?

Now that we have gained the interest of our potential customer, we now must provide the filler – a rich sentence or two describing the unique parts of our products and services that set us apart from the crowd.

2. Offer up

Your product or service is worth talking about, unfortunately, you will normally only have a finite amount of space or time in which to tell your prospective customer about it.

Keep to the point – tell your customer how your product can save time or money, provide them with customers, increase sales margins, decrease costs, make a cup of tea in seconds – give the key features or reasons as to why they should come to you rather than a competitor. This is also a good place to add any current offers, sales or special deals you might have. Give the customer a reason to buy from you.

Always keep your advertising copy focused on the thing you are selling. Do not deviate to tell the customer that you have been in business 15 years if it won’t matter to them, but above all that, make sure you finish your copy with the most important statement – the Call To Action.

3. Tell the Customer What To Do

A Call To Action is a crucial part of most advertising copy; A Call To Action tells the user what to do next. Our advertisement copy ends with such a phrase: “Submit your advertisements today, and gain valuable exposure!” We want the user to sign up for our service so they can begin advertising, however, if we hadn’t included this important sentence, then we risk leaving the user in limbo. We have got them excited about our service, we tell them how we are unique, and then we walk away? No.

In order for an advertisement to be compelling, the user needs be told what they should do next in order to get to the thing that you are selling. Writing a good Call To Action will allow you to get your customers in line to make a purchase, but there is still one last detail to take into consideration: What happens when a customer does what you told them to do?

4. Land it Right

You have told your customer about your new, fantastic service or product, and they click on your web link or pick up the phone and dial the number in your advert. Unfortunately, the phone number you gave was a digit wrong or your Web site is not yet finished. Disaster!

There is a very high probability that your potential customer won’t try to get through to you a different way – they will just hang up and go to the next advert in the list. At best, they will forget about their experience in trying to get hold of you, but at worst, they will associate your business with that bad experience, and avoid you in the future.

It is imperative that the contact details you give in your advertisements are correct. Ideally, if you are advertising a particular service or product, then your web link should go directly to a page focused on it rather than your company homepage or contact form. If you give a phone number, then the person answering the phone should at least be the correct person, and even better, be able to help the customer to make a purchase or take the enquiry further.

5. Be Professional

Spelling, punctuation and grammar are all very important when it comes to writing copy of any type for your business – be it an email, brochure, advertisement or receipt. Lacking in any of these three areas gives a very poor impression of the level of professionalism employed within your business.

If you are not confident in any of these three key areas, then it is better to have your copy proofread and run through a spell-checker than to leave it to chance. Remember that your customers probably do not know you, so all they a judging you on are the words you have chosen to promote yourself with. If these words don’t make sense, then your customer will disappear.

Simple CSS Star Rating Bar

2006-12-12T00:00:00+00:00

Just a quick bit of CSS that we came up with to display a percentage star rating that I'd like to share. This method requires two images - one with all of the stars highlighted, and one with them all greyed out:

Note that these images are both 100 pixels wide (for simplicity) .

CSS

.starCont {
width: 100px;
height: 16px;
display: -moz-inline-box;
display: inline-block;
background-image:url(/images/zero-star.gif);
background-repeat: no-repeat;
background-position: bottom left;
vertical-align:text-bottom;
}
.stars {
display: -moz-inline-box;
display: inline-block;
height: 16px;
background-image:url(/images/five-star.gif);
background-repeat: no-repeat;
background-position: bottom left;
vertical-align:text-bottom;
}

.stars span {
visibility: hidden;
font-size: 1px;
}

HTML

<span class="starCont"><span class="stars" style=""><span>%</span></span></span>

Serving Suggestion

Use as part of a voting system whereby the percentage values are injected into the page via the value of *percentage*. If your star images are larger than 100px in width, then you will need to adjust the value that is calculated by an amount proportianal to the actual width.

How to Execute a Rails Controller Action via. Cron

2006-12-06T00:00:00+00:00

I recently had to create a cron job that controlled the delivery of newsletters to a set of subscribers. Unfortunately, the code required some refactoring, but up until it could be modified, the task still needed firing at specific intervals. I had a little trouble finding much on this subject, so here is the code I eventually used:

./script/runner -e production "app = ActionController::Integration::Session.new; app.get 'account/send_newsletters'"

With any luck, this will help another person save a couple of minutes!

Update

Since Rails 1.2, this method of executing rails controller actions has been throwing an “uninitialized constant ActionController::Integration” exception. Adding the following line to your environments.rb file fixes this problem:

require ‘action_controller/integration’

Thanks, Dimitry Hristov!

SSH Access to StrongSpace from Textdrive

2006-12-05T00:00:00+00:00

I came across a few pitfalls in my attempt to get my Textdrive account talking to my StrongSpace account over SSH without needing to enter my StrongSpace password at the prompt. I am setting out all of the steps necessary to get this working here so that others can avoid these same problems.

Prerequisites

You should be familiar with gaining access to and using BASH on your Textdrive account before you go any further - if you don't then I suggest you go here for more information. I accept no responsibility for any damage or problems you may cause when using this tutorial. Make sure you take adequate backups and check any scripts before you run them.

Step 1: Creating the keys

To get started, we need to create a set of keys:

Log into your Textdrive shell, and navigate to /users/home/USERNAME/.ssh/
Run the following command ssh-keygen -b 1024 -f ss -P '' -t dsa
This will create two files - ss and ss.pub

Step 2: Uploading the public key (ss.pub) to your StrongSpace account

Download ss.pub to your local machine
SFTP to your StrongSpace account
Create a new directory called .ssh under your account
chmod this directory to 700
Upload ss.pub to this new folder
rename it to authorized_keys
chmod authorized_keys to 600

Step 3: Test!

OK, you now have your private key on your Textdrive account, and the corresponding public key on your StrongSpace account. We are good to test.

Go back to your Textdrive shell, and type:

ssh -i ss SS_USERNAME@SS_USERNAME.strongspace.com

You should now log directly into your account without needing to enter a password.

Notes:

If you already have an keys on separate lines.

Script to Automate Wordpress Update

2006-08-01T00:00:00+00:00

If you are a Textdrive user, and you have one or more WordPress installations under your control, you might find the following BASH script to be of use. The script is fairly straightforward in that it manages the backup of your existing WordPress installation and database to StrongSpace, then it automatically downloads and extracts the latest version of WordPress.

Notes:

If you do not have a StrongSpace account, then you will need to modify the script to change the safe storage of backups.
To take advantage of StrongSpace, you need to get the relevant SSH keys in place on both servers in order to be able to log into your StrongSpace account over SSH without a password.
Ensure that you have created both the updates and backups directories on your Textdrive account
WordPress has an upgrade script that sometimes needs to be run after the system software has been updated. This upgrade script may contain more than one step, and require confirmations or input from the administrator. For these reasons, I have not included that step in the automation script. Make sure you check to verify that the upgrade script is run if necessary.

The Script:

#!/usr/local/bin/bash  

# Configuration
HOME=/users/home/username
TEMP_UPDATE_DIR=$HOME/updates
TEMP_BACKUP_DIR=$HOME/backups  

LOCAL_WORDPRESS=$HOME/public_html/blog
WORDPRESS_URL=http://foo.bar.net  

MYSQL_DB=wordpress_database
MYSQL_USER=username
MYSQL_PASS=password  

STRONGSPACE_USER=<a href="mailto:foo@foobar.strongspace.com">foo@foobar.strongspace.com</a>

# No need to change this setting!
LATEST_WORDPRESS=http://wordpress.org/latest.tar.gz

# Prepare the directories
rm -rf ${TEMP_UPDATE_DIR}/*
rm -rf ${TEMP_BACKUP_DIR}/*  

# Backup the current database and WP installation
mysqldump --opt --lock-tables=false --skip-add-locks --user=${MYSQL_USER} --password=${MYSQL_PASS} ${MYSQL_DB} |gzip - &gt;${TEMP_BACKUP_DIR}/${MYSQL_DB}.sql.gz
tar zcf ${TEMP_BACKUP_DIR}/wordpress.tar.gz ${LOCAL_WORDPRESS} &gt;/dev/null  

# Send the backups to StrongSpace
scp -i ${HOME}/.ssh/ss -r ${TEMP_BACKUP_DIR}/* ${STRONGSPACE_USER}:Backups/Wordpress/  

# Get the latest Wordpress /usr/local/bin/curl -s ${LATEST_WORDPRESS} &gt; ${TEMP_UPDATE_DIR}/latest.tar.gz  

# Unzip the wordpress file to a temp location, and update the main wordpress install
cd ${TEMP_UPDATE_DIR}
tar -xzvf latest.tar.gz
cp -R -f ${TEMP_UPDATE_DIR}/wordpress/* ${LOCAL_WORDPRESS}

Save all of that into /users/home/username/scripts, and replace all of the configuration options with your own parameters. You should probably be comfortable using the CLI before you start tinkering with this sort of thing - I'm not responsible for any fuckups on your part. Once you are happy with the script, all you need do is run it, and if all goes well, your installation will be backed up and updated all at once!

Check to make sure that you don't need to run the upgrade script (run it if you do), and then you should be finished!

I'm going to post more information relating to getting your Textdrive account working with your StrongSpace account soon - particularly with regards to automating backups.

Google Analytics

2005-11-15T00:00:00+00:00

Well, I first came across Google's free web analytics program the other day, and so far my impression is mixed. The idea of having a site analysis system provided by the top-notch design and development teams at Google is fantastic, however, it would appear that they have underestimated the popularity of the service somewhat.

I added the tracking code to both Kyotee and Bouldr's holding page yesterday. Google's system recommended that I wait 12 hours for my data to show up (although in the help section, this period was given as 6 hours) - after which, my stats would be updated on an hourly basis. Ace.

Unfortunately, it was not to be - after 15 hours, still nothing. Also, the help system seems to be a little out of sorts - it keeps displaying 404's when I click through links to get to different articles, then when I retry a second later, the link takes me to the correct document.

It's like being given a wonderful present, then realising that it's broken, so you won't really be able to use it for a few days.

Coincidentally, I feel a headache coming on.

The Obligatory First Post

2005-11-10T00:00:00+00:00

Feeling good to finally be jumping on the biggest web Bandwaggon since BB's - though to be frank, I have nothing wothwhile to say just yet...

I have a couple of projects 'in the pipeline' at the mo, so maybe I could use this to make announcements or discuss their features and so on.

We'll see, I guess.