Question

I’m using the following code to fetch something from mongo:

class BlockingMongoFetcher
  include MongoConfig

  def initialize
    configure
    @connection = Mongo::Connection.new(@server, @port)
    @collection = init_collection(@connection)
  end

  def fetch(value)
    mongo_cursor = @collection.find({ KEY => value.to_s })

    if mongo_cursor.count == 0
      # do stuff
      return nil
    end

    if mongo_cursor.count > 1
      # do stuff
    end

    mongo_cursor.first
  end
end

init_collection just gets the db and the collection object from the connection.
In the fetch method, I’m using the count method to check, if there are results. Since I got a 0, where there should be 1 item, I added the following code within the gem to the count method of the Cursor class:

if response['n'].to_i == 0
  require "ruby-debug"
  debugger
  puts "stop here"
end

( response = @db.command(command) )

Within the debugger

@db.command(command)['n'].to_i

returns 1. If I call count in the fetch method twice (once without using the output) everything is fine. Am I missing something? Buffer or caching problem? However, this seems not to be deterministic … it only occurs in about 50 % of the runs. Mongodb is 2.0.2 and Ruby 1.9.3p125.

Thanks for your help!

Answer 1

Mhh, unexpected solution:

For inserting test data, I used the following statement within the spec

collection.insert @hash

The insert method has an option :safe, see the API. With the default (false), mongodb stores it async and the code execution continues. This can lead to strange behavior, like the value is not yet in the db, if you query it instantly afterwards. Just use

collection.insert @hash, :safe => true