jump to navigation

ActionMailer Timeouts February 22, 2009

Posted by John Dewey in Code, Monkey Patch, Rails.
add a comment

The following gist contains a monkeypatch allowing configurable timeouts on ActionMailer.  Since ActionMailer is blocking, it is nice to set a reasonable timeout, rather than wrapping each of your ActionMailer calls with a Timeout (alarm).  I actually monkeypatch all of ‘net/smtp’, so just rescue ‘Timeout::Error’ where needed.

Subselector, Moneypenny November 1, 2008

Posted by reidmix in ActiveRecord, Code, Database, Example, Monkey Patch, Plugins & Gems, Rails.
Tags: , , , , ,

Building on Josh and Damon‘s idea of Hacking a Subselect in ActiveRecord, I wondered if you could bake this kind of functionality into ActiveRecord.  So Doug and I went digging into the rails code, and came up with a plugin that adds subselects to ActiveRecord which we call Subselector.

So far, it only works on the Hash version of conditions.

On a column you wish to perform a subselect, pass a hash with :in, :not_in, :equals, or :not_equals as the only key.  The value is any of the options you normally would pass to ActiveRecord find.  Notice that we make sure to select a single column with the :select option:

Critic.find(:all, :conditions => { :id => {:in => {:select => :id, :conditions => {:active => true} } } })

Although the example may be contrived, here, we are looking for a critics that are in a set of active critics. The SQL:

select * from critics where id in (select id from critics where active = false)

You can see by default it runs the subselect on the table of outer select.  It gets more interesting you want to run a query on another ActiveRecord model:

Critic.find(:all, :conditions => { :id => {:in => {:model => :rankings, :select => :critic_id, :conditions =>
  {:week => 39} } } })

Here we set :model to :rankings.  Rankings is the ActiveRecord model to perform the find, notice we select the :critic_id column, the SQL is:

select * from critics where id in (select critic_id from rankings where week = 39)

And of course you can always just pass a string as a value to the subselect:

Critic.find(:all, :conditions => { :id => {:not_in => 'select id from critics where active = true' } })

Here’s how subselector can be used with the original example:

Post.find(:all, :conditions => :id => {:in => { :select => :post_id, :conditions => {:blog_id => self.id}, :order => "published_at DESC", :limit => options[:limit] || 10, :offset => options[:offset])} }, :order => "published_at DESC")

UPDATE: Subselector now likes Condition Arrays and Named Bind Variables.

Just pass the hash as a bind variable and specify the type (in/equals) of subselect in the string, make sure to enclose your ‘?’ inside parentheses:

Critic.find(:all, :conditions => ['id in (?)', {:select => :id, :conditions => 'active = true' }])
Critic.find(:all, :conditions => ['id not in (?)', {:select => :id, :conditions => {:active => false} }])
Critic.find(:all, :conditions => ['id in (?)', {:model => :rankings, :select => :critic_id, :conditions => {:week => 39} }])

As you can see, you can format the subselect hash just as above and can specify another model to run the subselect on. If you prefer to use named bind variable hashes, they still work (yay) as you would expect. And you can assign the subselect using them:

Critic.find(:all, :conditions => ['id in (:subselect)', {:subselect => {:select => :id, :conditions => {:active => false} } }])

UPDATE 2: Now with no ActiveRecord breakage

We’ve run the rails ActiveRecord tests without any problems. Let me know if you find any problems.

Railsolver – now with wildcard host resolution September 12, 2008

Posted by John Dewey in Monkey Patch, Plugins & Gems, Rails, Ruby.
add a comment

I was watching the Subdomains Railscast. Ryan listed a few ways to point subdomains at your development system, and I thought Railsolver should support wildcards.

Updates are on github, along with an updated README and bugs :). I also added some Railsolver props on the Railscast page.

This plugin will “hijack” Ruby host(s) resolver aka(resolv-replace.rb), allowing programmatic host file resolution (including wildcard host resolution).™

IE6 Accept Header is Faulty and Makes format.any Suck May 14, 2008

Posted by reidmix in AbstractRequest, ActionController, acts_as_authenticated, Changeset, Code, Example, MimeResponds, Monkey Patch, Plugins & Gems, Rails.
Tags: , , , , , , , , , ,

I’ve been using the awesome acts_as_authenticated plugin as the basis for user login and authorization. Its access_denied method takes advantage of the ActionController::MimeResponds.any method which will be invoked on a login_required before filter:

    def access_denied
      respond_to do |format|
        format.html do
          redirect_to new_session_path
        format.any do
          request_http_basic_authentication 'Web Password'

This code essentially says, if it’s an html request, store the original request and redirect to the login page, any other requests get the 401 Unauthorized header which the user agent can then use to initiate an HTTP Auth. This bit of code only works correctly with Edge Rail’s 8987 Changeset that allows any to act as a catch-all for any request mime-types not already specified.

Except when Internet Explorer 6 comes into the picture and happily sends this
strange and incomplete ‘Accept:’ header:

image/gif, image/x-xbitmap, image/jpeg, image/pjpeg, application/x-shockwave-flash, application/vnd.ms-excel, application/vnd.ms-powerpoint, application/msword, */*

As you can see, it doesn’t specify application/xhtml+xml or even text/html and what’s worse is the subsequent requests to the same page sends the catch-all */*.

What does this all mean?
When IE6 navigates to a page that requires login, instead of redirecting the user to the login page, the user is presented with an HTTP Auth Dialog. What’s worse is that in my application, cookies and other application state for an web-based end-user is set outside of the login_from_basic_auth method — you know — in my login action.

HTTP Auth Prompt by IE6

My guess is when the format.any fix goes out in the next rails release, IE6 users on acts_as_authenticated sites are going to be sad-faced. I’m not entirely sure the correct way to fix this problem but this is how I solved it in my application:

class ActionController::AbstractRequest
  def accepts_with_faulty_header
    @env['HTTP_ACCEPT']='*/*' if @env['HTTP_USER_AGENT'] =~ /msie/i and @env['HTTP_USER_AGENT'] !~ /opera|webtv/i
  alias_method_chain :accepts, :faulty_header

Originally, I thought to patch the Mime::Type.parse function, but decided that I wanted to make sure to look at the user agent for IE6. As you can see, I switch the HTTP_ACCEPT header to the catch-all */* string if the original HTTP_ACCEPT header is a match to the exact (bizarre) IE6 string and the HTTP_USER_AGENT is IE (matches MSIE but not Opera or WebTV). Then I call the original ActionController::AbstractRequest.accepts method.

I’d love to hear your ideas and suggestions — and if others have run into this problem as I didn’t find many users have this exact problem and how they may have solved it.

Update: Based on the conversation in the comments, I’ve re-written the patch to always change the HTTP_ACCEPT header to the catch-all */* if it is an IE Browser.

Run a Capistrano task on a single server in a given role February 7, 2008

Posted by John Dewey in Capistrano, Code, Command Line, Deployment, Monkey Patch.

There are times I want to run a Capistrano task on a single server, rather than every system in the role. Adding this monkey patch to deploy.rb adds this functionality. To utilize this feature append ‘server=IP_ADDRESS’ to the Cap command. Tasks are defined to run on particular role(s), this feature will target the server inside a particular tasks role. It will not execute a task on an arbitrary server.  CJ Kihlbom filled me in and stated this can be accomplished with `cap deploy HOSTS=IP_ADDRESS`.  See his comments attached to this post.

Patching Capistrano to use CVS over SSH September 11, 2007

Posted by reidmix in Capistrano, Code, Command Line, CVS, Monkey Patch, Ruby, SSH.

At work we still use CVS as our versioning repository and most of our systems rely on CVS to build and deploy into production. Currently there is no effort to transition to SVN, so we need to develop a patch to Capistrano to allow our projects to deploy to our development lab. We needed to extend Capistrano handle SSH logins to CVS and handle CVS repository directories.

Overall, both changes needed to be applied to the checkout method in Capistrano’s Cvs class.

We use SSH as the underlying communication mechanism for CVS but there is a difference in the password prompt: SSH’s is capitalized. We just need to make the prompt check case insensitive with the i switch:

if out =~ %r{password:}i

When setting up Capistrano (v1.4) for use in our development lab, the :application configuration is not sufficient to identify our rails project the CVS tree. We are not cool enough to setup or use modules to map into our repository directories, so our projects are usually in some form of webdev/project/appname. To handle this, i create an additional configuration parameter called :project in the deploy.rb:

set :project, "cvs/dir/to/myproj"

And we no longer use the application name for the CVS path, we use the project variable:

project = configuration[:project] || actor.application
command = <<-CMD
  cd #{configuration.releases_path};
  CVS_RSH="#{cvs_rsh}" #{cvs} -d #{configuration.repository} -Q #{op} -D "#{configuration.revision}" #{branch_option} -d #{File.basename(actor.release_path)} #{project};

The hardest part is patching Capistrano. Capistrano’s code gets loaded well after any library or environment gets loaded. The trick is to use the method_added hook to wait for the checkout method to get loaded and then alias the method to the patched one, we need to keep a variable around to check if we’ve patched the method so we don’t end up in an infinite loop:

@@checkout_patched = false
def Cvs.method_added(id)
  if id.id2name == "checkout"
    unless @@checkout_patched
      @@checkout_patched = true
      alias_method :checkout, :checkout_patch

I’ve not checked to see how these changes work or differ in Capistrano 2. But with all these elements together using Capistrano v1.4, I can drop this file in my lib directory or create a plugin with it. Together, here is the full patch which I have in a file called capistrano_cvs_ext.rb:

module Capistrano
  module SCM
    class Cvs < Base
      @@checkout_patched = false
      def Cvs.method_added(id)
        if id.id2name == "checkout"
          unless @@checkout_patched
            @@checkout_patched = true
            alias_method :checkout, :checkout_patch

      def checkout_patch(actor)
        cvs = configuration[:cvs] || "cvs"
        cvs_rsh = configuration[:cvs_rsh] || ENV['CVS_RSH'] || "ssh"

        if "HEAD" == configuration.branch then
            branch_option = ""
            branch_option = "-r #{configuration.branch}"

        # cvs has a root and repository, repository is not the same as application name as it can be a path
        op = configuration[:checkout] || "co"
        project = configuration[:project] || actor.application

        command = <<-CMD
          cd #{configuration.releases_path};
          CVS_RSH="#{cvs_rsh}" #{cvs} -d #{configuration.repository} -Q #{op} -D "#{configuration.revision}" #{branch_option} -d #{File.basename(actor.release_path)} #{project};

        run_checkout(actor, command) do |ch, stream, out|
          prefix = "#{stream} :: #{ch[:host]}"
          actor.logger.info out, prefix
          if out =~ %r{password:}i  # SSH asks for "Password" with a capital P
            actor.logger.info "CVS is asking for a password", prefix
            ch.send_data "#{actor.password}n"
          elsif out =~ %r{^Enter passphrase}
            message = "CVS needs your key's passphrase and cannot proceed"
            actor.logger.info message, prefix
            raise message