ActiveLdap, belongs_to, and "invalid filter syntax"

After I started using :adapter => 'net_ldap' in ActiveLdap, my belongs_to associations all broke with when constructing a search filter with Net::LDAP

invalid filter syntax (Net::LDAP::LdapError)

The classes look something like this:

 class Person  'uid', :prefix => 'ou=people', 
       :classes => ['top', 'mozillaAddressBookEntry']
   belongs_to :family, :class => 'Family', 
      :foreign_key => 'dn', :primary_key => 'member'
 class Family  'uid', :prefix => 'ou=families', 
       :classes => ['top','family', 'mozillaAddressBookEntry']
   has_many :people, :class => 'Person', 
      :wrap => 'member', :primary_key => 'dn'

As I found out per this thread, Ruby Net::LDAP apparently can’t handle search filters with “=” or “,” characters, and doesn’t allow escaping with “\3D” or the like.

I switched back to the default Ruby/LDAP adapter, and the belongs_to associations work as expected.

Capifying a Merb application

UPDATE: I was embarrassed that I actually get hits on this post, so I’m adding a stripped down example deploy.rb.

Create a directory named config in the Merb root (not inside dist).

Create your deploy.rb file inside the config directory. Here’s a stripped down version of what I have used:

# ./config/deploy.rb
# Do your merb daemon configuration in ./dist/conf/merb.yml

set :application, "awesome_thingummy"
set :scm_name, "thingummy"
set :repository, "svn://{scm_name}/trunk"
set :checkout, "checkout"

# =============================================================================
# =============================================================================

role :web, ""
role :app, ""

# =============================================================================
# =============================================================================
set :deploy_to, "/var/www/apps/#{application}"
set :user, "deployer"            # defaults to the currently logged in user

# set :scm, :darcs                  # defaults to :subversion
set :svn, "/opt/local/bin/svn"      
set :merb, "/opt/local/bin/merb --merb-root #{deploy_to}/current"

task :spinner do
  run merb

task :restart do
  run "#{merb} -k all"
  run merb

task :update_code do
  run "#{svn} --quiet #{checkout} #{repository} #{release_path}"

task :after_update_code do
  run "ln -s #{shared_path}/log #{release_path}/log"

desc "Deploy the app"
task :deploy do
  transaction do

Use Net::LDAP with ActiveLdap

Prior to version 0.8.2, ActiveLdap required Ruby/LDAP for its underlying connection. This posed an annoyance, if not a problem, as Ruby/LDAP does not exist as a gem. As of 0.8.2, ActiveLdap supports both Ruby/LDAP and the bona fide, gemified Net::LDAP.

The default connection is still Ruby/LDAP, however, and the official docs aren’t abundantly clear on how to make the connections with Net::LDAP. I couldn’t find an answer with Google until after I had puzzled it out myself and was about to write this post. Googling for the answer found for me this post to the ActiveLdap mailing list, which is mostly correct. The value for :adapter should be ‘net_ldap’, not ‘net-ldap’, according to the Constants list at the very bottom of the front page of the Rdocs for ActiveLdap.


 :adapter => 'net_ldap',
 :host => 'localhost',
 :base => 'dc=mycooldomain,dc=com', 
 # doo wa doo shbop

Environment-specific setup in Merb

Merb has production and development environments that act much like their Rails counterparts. While there aren’t separate files for environment-specific configuration, you can get the same effect using a case statement at the bottom of merb_init.rb like so.

This is also a good place to stick database configuration when your dev is different than your prod.

Managing PHP fastcgi with Launchd

Because I’m using PHPLDAPAdmin as a stopgap while I build a custom web directory application, I have gotten soiled a bit by PHP. I’m running everything on a headless Mac mini, and I chose to use Mac OS X as the OS for reasons I won’t go into here.

The front-end web server is nginx, which is proxying to php running as fcgi. Pardon me in advance if/when I get the terminology wrong. I tried using the instructions from the Homo Adminus blog, and while both the Bash script method and the spawn-fcgi method work, I could not figure out how to get launchd to manage them.

I hope that some of you are laughing at me at this point, because the problem was (and may still be) totally centered on my ignorance of the fastCGI usage of the php binary. I was being dense about launchd as well. Launchd doesn’t use spawn-fcgi or a custom shell script as the ProgramArguments. Rather, it completely obviates the need for them. This should have been obvious (a phrase that I promise will be common on this blog).

For the edification of the similarly ignorant, here’s a launchd plist that seems to work.

As I understand it, you set the server IP address and port number with the -b argument to the php binary. Homo Adminus adds the -q option (suppressing HTTP Header output) in his Bash script, so I did as well. To arrange for more than one listening process, you have to set the PHP_FCGI_CHILDREN environment variable.