Reload Your Factories With Spork

Spork is a great tool that can really cut down on the startup time of your tests. This gets more noticable as your code base grows in size. Our current test suite at legitscript takes quite a few seconds to start up, which would be a big impediment to test-driven development without spork.

If spork is used together with factory_girl, \ it helps to load the factories on each test run. If they’re loaded in the pre-fork block, each tweak to a factory would require restarting spork.

The naive approach would be to reload those files just like your models and controllers:

1
2
3
4

Spork.each_run do
  # ...
  Dir["#{Rails.root}/spec/support/*.rb"].each {|f| load f}
end

However, this attempts to redefine existing factories, which will fail with an error:

1
2

Factory already registered: website
  (FactoryGirl::DuplicateDefinitionError)

On top of that, if you have any sequences defined, you would get an error similar to this one:

1
2
3

Exception encountered:
  #<FactoryGirl::DuplicateDefinitionError:
  Sequence already registered: email>

The solution is to explicitly clear out any existing factories and sequences:

1
2
3
4
5
6

Spork.each_run do
  # ...
  FactoryGirl.sequences.clear
  FactoryGirl.factories.clear
  Dir["#{Rails.root}/spec/support/*.rb"].each {|f| load f}
end

Voila! Tweak a factory and keep coding while your tests re-run in the background.

Logging Outbound HTTP Requests

Rails logs a ton of details about incoming HTTP requests in debug mode. But what about outbound requests? Few gems implementing third party APIs bother to log their requests and responses, and hacking their source code is suboptimal. Even for outbound HTTP requests made by your own code, it quickly becomes tedious to sprinkle logging statements all over the place to get a good look at the data.

Metaprogramming to the rescue

Simply override the request method of the Net::HTTP module in ruby to add some logging. Something like this:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

module Net
  class HTTP
    alias_method(:orig_request, :request) unless method_defined?(:orig_request)

    def request(req, body = nil, &amp;block)
      Rails.logger.debug("Request: #{req.method} http://#{@address}:#{@port}#{req.path}")

        if req.method == "POST"
          data = req.body.nil? || req.body.size == 0 ? body : req.body
          Rails.logger.debug("POST Data: #{data}")
        end
      end

      response = orig_request(req, body, &amp;block)
      Rails.logger.debug("Response: #{response.body}")

      response
    end
  end
end

This inserts the details of any outgoing HTTP request and it’s response into the standard rails logger. (Actually, depending on what of the many request methods Net::HTTP offers is used for the request, it may log some stuff twice. See my gem below on how to fix that.)

The somewhat whacky looking way of finding the post data accounts for the different ways in which Net::HTTP::post and Net::HTTP::post_form handle that data internally.

You could take this a bit further. For example, it might be useful to log the TCP connection attempt itself (since if that fails, the request is never sent, and you’re none the wiser) by overriding the (private) connect method.

There’s a gem for that (now)

Since I find myself in this situation regularly, I’ve created a ruby gem, creatively named httplog, that does just that. Here’s some (truncated) sample output from a request made by the simplegeo gem:

1
2
3
4

[httplog] Connecting: api.simplegeo.com
[httplog] Sending: GET http://api.simplegeo.com:80/1.0/context/address.json?address=22201&filter=
[httplog] Status: 200
[httplog] Response: {"query":{"latitude":38.885484,"longitude":-77.099113,"address":"22201"...

Disclaimer: This is my first published gem, your mileage may vary. It’s been working fine for me with ruby 1.9.2 and 1.9.3, but I have not taking extra pains to test it against older versions. If you run into trouble, feel free to open an issue on github.