techno weenie - Home

Calendar About Nothing

2008-10-05T23:31:29-05:00

Several months ago, I was listening to one of the Gitsplosion! podcasts. PJ was talking about ant juice, and Chris suggested starting a seinfeld calendar to mark your progress on wearing your Github t-shirts every day for a month. I’d been looking for a cool microapp to build for awhile, and thought I could use this calendar idea. However, I have no idea how to confirm if someone is actually wearing a Github shirt, so I decided to scan public Github feeds.

After I let my microapp sit around unfinished for several months, Kyle whipped up a basic layout, and I launched the Calendar About Nothing.

I also decided to use various frameworks that are unfamiliar to me: Sinatra, Datamapper, and HAML. The code itself isn’t anything special. It’s a big hack that’ll hopefully get refactored as time goes on, and more sources get added. Let me know if you have any ideas of other sources I can check (Gitorious and bitbucket spring to mind). Ideally, you’d be able to take the mythical calendar source API and implement your own private hooks that read private repositories on your own hosted calendars.

Rails Migrations in Git Branches

2008-03-24T23:40:26-05:00

For those of you that don’t ~~stalk~~ follow my other streams:

Migration Buddy

Or, “the inevitable renumbering and conflict-resolution of miggy tardust”

This is a tool to help merge rails branches with conflicting migrations. The basic idea goes like this:

First you migrate down to the migration you were at when you branched.
Renumber any new migrations you created that conflict with any new migrations in the target branch.
Commit the renames.
Merge to the target branch (usually ‘master’).
Migrate back up.

Wow, what a pain, right? (Apologies to Saul Williams, you rock).

I keep threatening, but I may have to replace this blog with a combination twitter / tumblr / github stream.

moving to github

2008-03-04T10:37:48-06:00

I’ve been using git for several months now, waiting for a reason to keep my svn habits around. So far, it’s been pretty smooth sailing. There were some squalls along the way as I unlearned some habits from svn though. I even moved active development of my various open source projects to git. Well, today I’ve moved most of my “active” (and I say that in the loosest of terms) projects to github.

Github is fantastic, but why?. The core repository browser stuff looks and works great, but is nothing revolutionary, to be honest. What’s amazing about Github is how it really brings the social aspect into play. Chris and Tom are showing us all visually how git development is supposed to work. I know I personally had some bing moments once I started pulling in commits from external git repos.

For what it’s worth, Gitorious is another great community of git hackers. They’ve also added merge requests to the mix since the last time I looked at it. The differences with Gitorious are that it’s exclusively geared for open source projects, and that Gitorious itself is open source. That’s wild.

The great thing about git is that it doesn’t matter which you choose. You can easily add a git remote for both, and keep your origin on your own server. I hope we can see some collaboration between the two in the future. I know Github has some neat web hooks powering some backend processing functions. It’d be interesting to be able to initiate fork requests to your git[hub/orious] account by POSTing a public fetch url, or to send a user a pull/merge request by POSTing to a public end point (or emailing a git mailbox).

Naturally, as the author of Warehouse, I have other reasons for wanting to see this happen. I’d like to be able to collaborate with other git hacker networks and keep my data on my own server. The only way this is possible is if there’s a simple API we can all agree on.

At any rate, I firmly believe that fantastic tools like Github and Gitorious will do a lot to drive git adoption. Having clean, visual tools will go a long way in simplifying git for new users.

Update: Oh yea, I have a couple github invites if anyone wants them.

Giles is Proud Not to Understand has_many :through

2008-02-13T03:43:35-06:00

Apparently, the number of exception classes related to has_many :through (HM:T) scared Giles away. I think most of them are there because I put the original ones there. Course, looking at them highlighted does make it seem a bit ridiculous.

HM:T associations are the only associations that actually look at other associations to figure out how to query the database. There aren’t a simple set of conventions to go by that raise simple errors like “couldn’t find the `suspended_users` table”. If you actually look at the associations, they’re not much more than simple ActiveRecordError subclasses with a custom message. It gives you nice, helpful messages like this:


class Post < ActiveRecord::Base
end
class User < ActiveRecord::Base
end
class Topic < ActiveRecord::Base
  has_many :posts
  belongs_to :user
end
class Forum < ActiveRecord::Base
  # commented out bits are things
  # i conveniently 'forgot' in this contrived example.
  #
  # has_many :topics 
  has_many :posts, :through => :topics
  has_many :creators, :through => :topics# , :source => :user
end

Forum.find(:first).posts
# => ActiveRecord::HasManyThroughAssociationNotFoundError: Could not find the association :topics in model Forum

Forum.find(:first).creators
# => ActiveRecord::HasManyThroughSourceAssociationNotFoundError: Could not find the source association(s) :creator or :creators in model Topic.  Try 'has_many :creators, :through => :topics, :source => <name>'.  Is it one of :posts or :user?

Are the error messages helpful? I sure think so. But, do we really need all those exception classes? Sure, it’ll keep the error messages DRY.

Uploaded with plasq’s Skitch!

Okay, maybe not. Is it sufficient to have a single ActiveRecordError class (or even a more specific ActiveRecord::AssociationsError)? I could definitely do something like this:


raise ActiveRecord::AssociationsError.new(:has_many, :through_association_not_found, model, reflection)

Whichever way’s the best, I really don’t think it’s good enough validation to miss out on the feature. I can point to a few other more complex areas of the associations code (the crazy eager joins code is why I personally haven’t used :include in any of my apps in over a year).

Timezone awareness in Rails

2008-02-06T12:42:36-06:00

I just committed another piece of the puzzle to Rails’ new time zone support, coming in Rails 2.1. This takes a different approach than my previous attempt at timezone-aware activerecord attributes. The whole idea behind this approach, is that times appear in the local time zone while you work with them, but are persisted to the database in UTC.


>> Time.zone = "Pacific Time (US & Canada)" 
>> @status = Status.find :first
>> @status.created_at
=> Tue, 29 Jan 2008 15:30:09 PST -08:00

If you set the time, it makes any necessary conversions to your current time zone.


>> @status.created_at = Time.utc 2008, 1, 1
>> @status.created_at
=> Mon, 31 Dec 2007 16:00:00 PST -08:00

Multiparameter attributes (say, from a time select box) are changed to the current time zone properly.


>> @status.attributes = {
?>   "created_at(1i)" => "2008", "created_at(2i)" => "1", "created_at(3i)" => "1", 
?>   "created_at(4i)" => "0", "created_at(5i)" => "0", "created_at(6i)" => "0" }
>> @status.created_at
=> Tue, 01 Jan 2008 00:00:00 PST -08:00

Now in our particular app, it’s very important that each user sets up their own timezone. The easiest way, we’ve found, is to take the offset from the browser. This way, the user can move around, hop on a business flight, or play with their system clock, and our app will still function correctly. First, we set a cookie with the timezone offset:


Cookie.set({tzoffset: (new Date()).getTimezoneOffset()});

Then, I wrote a before filter to process this automatically and set the Time.zone value appropriately.


  # The browsers give the # of minutes that a local time needs to add to
  # make it UTC, while TimeZone expects offsets in seconds to add to 
  # a UTC to make it local.
  def browser_timezone
    return nil if cookies[:tzoffset].blank?
    @browser_timezone ||= begin
      min = cookies[:tzoffset].to_i
      TimeZone[-min.minutes]
    end
  end

  def set_timezone
    if logged_in? && browser_timezone \
        && browser_timezone.name != current_user.time_zone
      current_user.update_attribute(:time_zone, browser_timezone.name)
    end
    Time.zone = logged_in? ? current_user.time_zone : browser_timezone
  end

Keep in mind, this does only set the current offset. The offset doesn’t account for things like DST. However, since this will update on each request, it’ll correct itself once the DST status changes.

Huge thanks to Geoff Buesing, the mastermind behind this new TimeWithZone implementation.

Random Rails Tidbits

2008-01-20T23:46:38-06:00

With ActiveSupport::Callbacks being extracted out of ActiveRecord and ActiveSupport::Validations on the way, it looks like Josh and lifo both did a large chunk of the work I was going to do for ActiveModel.

Apparently the Russian Mafia is spreading git propaganda and speeding up the Rails routing implementation. Bravo!

The QuarkRuby folks succeeded in hacking ActiveResource to access Youtube videos. I think there may be an opportunity to tweak ActiveResource under the hood to make these sorts of plugins easier.

Benchmarking ActiveDocument vs ActiveRecord

2008-01-14T10:11:14-06:00

I created a reference rails app with a single controller to benchmark the difference between ActiveRecord and ActiveDocument.

First, let’s create some posts to populate the database.


# ActiveRecord
> script/performance/benchmarker 100 'Importer.run'
            user     system      total        real
#1      2.080000   1.060000   3.140000 (  4.655729)

# ActiveDocument
> script/performance/benchmarker 100 'Importer.run'
            user     system      total        real
#1      2.060000   0.150000   2.210000 ( 28.470869)

# ActiveDocument (json branch)
> script/performance/benchmarker 100 'Importer.run'
            user     system      total        real
#1      2.020000   0.140000   2.160000 ( 27.215934)

Now, let’s retrieve some posts:


# ActiveRecord
> script/performance/benchmarker 1000 'Post.find(55)'
            user     system      total        real
#1      0.640000   0.050000   0.690000 (  0.685781)

> script/performance/benchmarker 100 'Post.find(:all)'
            user     system      total        real
#1     16.840000   0.080000  16.920000 ( 17.107640)

# ActiveDocument
> script/performance/benchmarker 1000 'Post.find("76d77998-7721-48c6-9ffd-8925efeb8b7e")'
            user     system      total        real
#1      0.490000   0.030000   0.520000 (  0.891832)

> script/performance/benchmarker 100 'Post.find(:all)'
            user     system      total        real
#1      4.860000   0.070000   4.930000 (  8.574478)

# ActiveDocument (json branch)
> script/performance/benchmarker 1000 'Post.find("1b3033bb-7d04-44ff-a1d3-1727d06c603e")'
            user     system      total        real
#1      0.400000   0.030000   0.430000 (  0.820451)

> script/performance/benchmarker 100 'Post.find(:all)'
            user     system      total        real
#1      3.110000   0.060000   3.170000 (  7.097080)

It’s still a bit messy since ActiveDocument is still incomplete. But, you can grab the Battle Royale app I used to test it at:


git://activereload.net/battleroyale.git

Oh, and it turns out that I was not the first or second, but fourth or fifth to start this project. Sebastian started a Google group for all of us.

ActiveDocument

2008-01-14T02:12:20-06:00

This is just a quick post, but I’ve started playing around with thrudb, and naturally, writing a ruby library for it. It’s still incredibly rough. The indexing code is busted since I started testing with a mock manager class instead of the actual thing, so about all it does is saves and fetches records at the moment.

You can get the code from:


git://git.caboo.se/activedocument.git.

Note: I have an experimental json branch and a sample app to pit the 2 data stores against each other.

More on Controller Specs, or Fun with Ruby2Ruby

2008-01-10T13:04:15-06:00

One of the problems with the rspec_on_rails_on_crack (git://activereload.net/rspec_on_rails_on_crack.git) was that the redirection examples had horrible descriptions.

describe FooController, "GET /index" do
  it_redirects_to { foo_path }
end

The route goes in a ruby proc because routes aren’t available from the context of the example group. They’re only available from inside a running example. But this is what the description looks like when running rake spec:doc:

FooController GET #index
  - redirects to #<Proc:0x0224e7d8@./spec/controllers/foo_controller_spec.rb:68>

Solution? Install the ParseTree and ruby2ruby gems and add this to #it_redirects_to:

route.to_ruby.gsub(/(^proc \{)|(\}$)/, '').strip

FooController GET #edit
- redirects to "foo_path(1)"

Controller specs are a drag

2007-12-26T11:30:30-06:00

Ever since Jamis taught us about fat models and skinny controllers, the controller’s importance has dipped a bit (for the best). Controllers should only be testing the basic things, such as the fact that it’s using the params correctly to retrieve and create records, and returning the right response. Any more details about the model are usually exercised in model specs. This can make the examples incredibly repetitive:


it "should set @records when accessing GET /index" do
  get :index
  assigns[:records].should_not be_nil
end

For those not using rspec, you can pretty much replace ‘spec’ with ‘test case’ and ‘example’ with ‘test’. It’s all the same, really.

Rspec tends to encourage that the controller examples be broken up into finer contexts, such as by action. (You can do this with test/unit of course, but it’s not nearly as common). You’re also encouraged to keep to a “single assertion per example” discipline. This way if one assertion fails, the other assertions are still run in other examples. Good practice to follow, but it results in a lot of extra time spent writing more spec examples. Here’s what I came up with: (old and busted)

While soliciting for comments, Yehuda Katz scoffed at all the repetition. Not that I blamed him, it seemed rather silly to write 560 lines of spec code for a 90 LOC controller. We swapped ideas back and forth and wondered if there was some way to define meta spec examples to ease the controller speccing process. Here’s the basic outline we came up with:


it.assigns :forum, :flash => {:notice => :not_nil}
it.redirects_to { forum_path(@forum) }
it.renders :template, :index

As you can see, I managed to cut the controller spec in half, with this new hotness. There are a few minor things there to contribute to back to rails and rspec. Yehuda mentioned he was going go implement the same API for the merb testing harness. Any thoughts on this approach?

Update: fixed a textile error around the “90 LOC…” link. Also, moved this to its own git repository: git://activereload.net/rspec_on_rails_on_crack.git. I also made a quick custom_scaffold generator in git://activereload.net/custom_scaffold.git.

Moving to git

2007-12-26T11:24:52-06:00

Ever since starting my latest client project a few months ago, I’ve started using git. I had the fortune to be trained by a couple git badasses, so it was a pretty smooth transition. At some point, I figured it was time to start porting my own open source projects to git, while keeping the svn repositories current. Here’s a rakefile I worked up to move my own plugins in one command:

# imports all dirs in given SVN dir as a separate git repo in the current directory
# pass SERVER= to specify a remote git server.  it'll add it as a remote for origin and push master:refs/heads/master
task :fetch_plugins do
  raise "need svn repo in SVN env var" if ENV['SVN'].nil?
  ENV['GIT'] ||= 'git'
  repos = %x{svn ls #{ENV['SVN']}}.split
  repos.each do |repo|
    next unless repo =~ /\/$/ && !File.exist?(repo)
    repo.gsub! /\/$/, ''
    FileUtils.mkdir_p repo
    error = false
    Dir.chdir repo do
      begin
        puts "initializing #{repo}..."
        %x{git svn init #{ENV['SVN']}/#{repo}}
        puts %x{git svn fetch}
        if ENV['SERVER']
          puts %x{git remote add origin #{ENV['GIT']}@#{ENV['SERVER']}:#{repo}.git}
          puts %x{git push origin master:refs/heads/master}
          puts %x{git config branch.master.remote 'origin'}
          puts %x{git config branch.master.merge 'refs/heads/master'}
        end
      rescue
        error = $!
      end
    end
    if error
      FileUtils.rm_rf repo
      puts "removing #{repo}: #{error.inspect}"
    end
  end
end

If you read through, it does these basic commands:


mkdir pluginname
cd pluginname
git svn init http://pluginurl
git svn fetch
# get some coffee
git remote add origin git@server:pluginname.git
git config branch.master.remote 'origin'
git config branch.master.merge 'refs/heads/master'
git push origin master:refs/heads/master

The last command pushes the local master branch to a master branch on the remote git server. Once that’s done, you should be able to git push/pull without issues. However, there’s something missing from the .git/config file, and in my git newbieness, I have no idea how else to fix: Update – I know how to fix this now. See the git config commands in the rake task. No more need to modify .git/config manually.

[branch "master"]
        remote = origin
        merge = refs/heads/master

You can setup gitosis on the server to manage the repos. It’s a brilliant piece of software that uses a single ‘git’ user account and public keys to manage access to various repos. The only issue I had with the config file though, was that it required a [repo] entry for every plugin to enable git daemon access. Here’s what I mean:

[group activereload]
writable = gitosis-admin lighthouse warehouse darkroom
members = caged rick

[group mephisto]
writable = mephisto
members = imajes psq svenfuchs caged rick

# enable gitweb 
[repo mephisto]
daemon = yes

You now clone mephisto and the rest of my open source projects with a command like “git clone git://activereload.net/mephisto.git”.

Oh, and merry christmas and all that. I didn’t really intend for this to be “one of those” posts. I’m just having some problems going back to sleep after playing santa (putting out presents, eating the cookies, etc).

Adventures in Rails Debugging

2007-12-18T19:45:08-06:00

After upgrading my app to Rails 2.0, I started getting an error, but only when running a rake task (such as rake db:migrate).


$ rake db:migrate --trace
(in /Users/rick/p/lighthouse/trunk)
** Invoke db:migrate (first_time)
** Invoke environment (first_time)
** Execute environment
rake aborted!
uninitialized constant Technoweenie::ActiveRecordContext
/Users/rick/p/lighthouse/trunk/vendor/rails/activerecord/lib/../../activesupport/lib/active_support/dependencies.rb:266:in `load_missing_constant'
/Users/rick/p/lighthouse/trunk/vendor/rails/activerecord/lib/../../activesupport/lib/active_support/dependencies.rb:453:in `const_missing'
/Users/rick/p/lighthouse/trunk/vendor/plugins/active_record_context/init.rb:1:in `evaluate_init_rb'

So, it’s failing on the first line of my active_record_context’s init.rb file:

ActiveRecord::Base.send :include, Technoweenie::ActiveRecordContext

Where’s the require? I leave it out, relying on ActiveSupport::Dependencies to auto-load the module for me. But, it seems to be failing me in my rake tasks. The fact that it’s working when I run script/server or script/console makes me think it’s some wacky LOAD_PATH issue. So, I try this in the init.rb file:

puts $LOAD_PATH.inspect
puts Dependencies.load_paths.inspect
ActiveRecord::Base.send :include, Technoweenie::ActiveRecordContext

Running both my rake task and script/console don’t show any differences, so that’s a dead end. Next, I turn my attention to Dependencies. The first stack trace showed it failing on line 266 in Dependencies.load_missing_constant. Above the large if statement seems like a good spot for some debugging.

$ rdebug `which rake` db:migrate -n -- --trace

Here are some interesting variables:

qualified_name – “Technoweenie::ActiveRecordContext”
path_suffix – “technoweenie::active_record_context”

That doesn’t look like a normal path, String#underscore should be converting :: to /. Next, I run this in script/server (script/console has no switch to enable the debugger).

$ script/server -u

qualified_name – “Technoweenie::ActiveRecordContext”
path_suffix – “technoweenie/active_record_context”

So, something is messing with String#underscore, but only when rake runs. Interesting… I then do a grep of my vendor directory, looking for ‘def underscore’ and find 3 locations: ActiveSupport, bj, and AWS::S3. bj managed to copy ActiveSupport’s version of the method exactly, but S3 left out the part that converts :: to /:


  # ActiveSupport adds an underscore method to String so let's just use that one if
  # we find that the method is already defined
  def underscore
    gsub(/([A-Z]+)([A-Z][a-z])/,'\1_\2').
    gsub(/([a-z\d])([A-Z])/,'\1_\2').
    downcase
  end unless public_method_defined? :underscore

So, AWS::S3 only adds that method if ActiveSupport isn’t available. Looking at ‘String.included_modules’ while debugging does list ‘ActiveSupport::CoreExtensions::String::Inflections’, so it’s looking like AWS::S3 is loading before ActiveSupport somehow.

Then it hits me, I’m loading it in a rake task! Doh…

Go Daddy Assumes 850,000 RegisterFly Domains

2007-05-29T14:19:36-05:00

Yay.

High profile Beast usage, and Joyent Slingshot ports

2007-05-04T19:27:59-05:00

Joyent made their public release of Slingshot recently, along with a Slingshot Port contest to go with it. Two of the listed entries include Inferno, a Mephisto port, and Brute, a Beast port. Good luck to both entrants!

Today, Matt at 37signals posted a Design Decisions article about the move from punBB to Beast. I love seeing Beast in action.

Where did rick go?

2007-05-01T12:22:11-05:00

I’m finding it difficult to come up with a use for this blog. Most of my work is spent on client work or Lighthouse, so therefore my rails based blogging is at the Active Reload blog. Mephisto has its own blog of course, but a fair amount of you still can’t access it (Registerfly has my domains locked and untransferrable, I may have to bite the bullet and buy a new name.).

I’m also having a lot of fun with my tumblr. Even though I can technically do everything there in Mephisto, Tumblr provides a nice, focused interface for tumbling. I’ve pretty much ceased my twitter activity lately, I can’t imagine it was anything exciting to read. I’d like to turn this blog into more of a dashboard with various feeds for my OSS projects and blogs. It’d give me a chance to beef up that aspect of Mephisto as well…

As for offline activities, I’m teaching my son to be a Guitar Hero and getting him prepped for kindergarten graduation (don’t ask).