I recently found a very neat plugin called Query Reviewer for visualizing slow sql queries in your rails app. When I installed it, I saw this error message

NoMethodError: undefined method `silence‘ for #<Log4r::Logger:0×39593f8>

You will see this error if you are using log4r for logging instead of the default rails logger. This ticket also talks of the same error in a different scenario. To fix the issue, edit the file vendor/plugins/query_reviewer/lib/query_reviewer/mysql_adapter_extensions.rb and replace all instance of @logger.silence with ActiveRecord::Base.silence.

I recently upgraded Muziboo to Rails 2.3 and found out that some of my routes were broken. On digging deeper I found out that I had some routes with two version, one without :format and one with :format going to different controllers/actions. I had the route with :format before the route without :format in the routes.rb file and changing the order fixed the problem. I am not sure why this happened but its been working fine since the change. If you run into routing issues, do try this and let me know if it worked for you.

I am currently working on moving Muziboo’s search from ferret to sphinx. We were using acts_as_ferret (AAF) plugin which provides a very simple rails integration with ferret. AAF binds itself to the active record callbacks and makes index updation painless. However AAF is pretty unstable in production mode and we realized this only lately with increasing traffic. We often had corrupted indexes and page errors as a result of that (since ferret is almost deeply integrated into models, it breaks a lot of functionality apart from search when something goes wrong). Also another problem with ferret is the high memory usage and the fact that with multiple mongrels, you will have to run a drb process to make sure the index is not written by multiple processes and corrupted.

On looking for alternatives, I found out about Thinking Sphinx, a plugin that makes it easy to use sphinx in your rails project. There is some documentation to get you started with TS once you decide to use it and also a peepcode pdf which has some good info. I could get the basic search functionality up and running in no time but had to spend sometime getting some integration going with other plugins. I will detail some of those things in this and next few posts.

Acts as taggable helps you in tagging your objects and we wanted sphinx to search on the tags too. It turns out that its very simple. All you have to say in your define_index block is

1

indexes taggings.tag.name,  :as=>tag_name

Just make sure that you restart sphinx and reindex your data before after adding this in your model(s).

Attachment_fu sanitizes the filenames on uploads to remove any funky character (not 0-9 a-z A-Z, underscore or a period). This is accomplished by the sanitize_filename private method in attachment_fu.rb file

1
2
3
4
5
6
7
8
9
10

def sanitize_filename(filename)
  returning filename.strip do |name|
    # NOTE: File.basename doesn't work right with Windows paths on Unix
    # get only the filename, not the whole path
    name.gsub! /^.*(\\|\/)/, ''
 
    # Finally, replace all non alphanumeric, underscore or periods with underscore
    name.gsub! /[^\w\.\-]/, '_'
  end
end

The shortcut \w is described here as letter or digit; same as [0-9A-Za-z]. However since ruby regex engine has support for unicode, letter means any unicode character. So it will let characters like 爱与希望 remain. This can be a problem if you are passing a filename containing such characters to a flash player. The flash player just won’t play the file!

A quick solution would be to check specifically for 0-9A-Za-z. This can be done by changing the function to

1
2
3
4
5
6
7
8
9
10
11
12

def sanitize_filename(filename)
  returning filename.strip do |name|
   # NOTE: File.basename doesn't work right with Windows paths on Unix
   # get only the filename, not the whole path
   name.gsub! /^.*(\\|\/)/, ''
 
   # Finally, replace all non alphanumeric, underscore or periods with underscore
   #            name.gsub! /[^\w\.\-]/, '_'
   #            Basically strip out the non-ascii alphabets too and replace with x. You don't want all _ :)
    name.gsub!(/[^0-9A-Za-z.\-]/, 'x')
  end
end

Finally this is not a problem if non ascii characters don’t cause any issue in your site.

I was trying to make a payment using PayPal SOAP API today and on a standard rubyworks installation (Apache –> HAProxy –> Mongrels), I was getting a Gateway Timeout Error. On checking the logs, I found out that the payment was in fact going through and rails was generating the thanks page too. My first doubt was that mod_proxy was timing out while waiting for a response. I tried to increase ProxyTimeout to 120 secs but it had not effect. I then tried to send the request directly to port 3001 (haproxy port in rubyworks) and I still saw the timeout happening. On sending the request to a mongrel directly (I tried on port 3002), the payment went smoothly and the thanks page was rendered. I tried to browse around and found this page where they mention that by default, if mongrel does not respond back in 30 secs, haproxy times it out.

The setting can be increased by simply changing the variable srvtimeout to a larger number (say 120000 for 120 secs) in the configuration file /etc/rails/haproxy.conf