Hello there,

it’s late in Bremen, so I will keep this thing as short as possible: I just wasted the complete day’s work at Tansporter due to a bad preparation… dumb me! (Remember: Benchmark before any optimization!)

I tried to optimize some of our ActiveRecord find statements, installed a whole new column on a table solely for that reason, changed the model to fill this new column right and then changed the actual method I would like to spice up.

It was a 70 line monster of very ugly code. After refactoring it, there were just 40 lines left and it was very cool, clean and beautiful code. The main change: I replaced a find statement which returned a very huuuuge result set, which I had to iterate and do some computing on every element, with 4 very cool queries with just one result, each.

Sounds cool, but unfortunately it isn’t.

After I had a short “wow, that’s great stuff!” experience, I thought it might not be unclever to benchmark the new method and compare it to the old one.
And here comes the trouble: The old method finished in about 0.018 seconds, the new one needed somewhat between 0.54 and 0.4 seconds… What a pitty!

So I switched back to the ugly version, to my rescue the added column could be used elsewhere so my effort was not completely useless.

What I’ve learned: One query with a large result set apparently performs MUCH better than 4 (not absolute trivial) queries with a very small result set, even if you have to iterate through the whole big result set.

Anyway I stumbled upon two cool, small things today that are worth mentioning them:

1st: Get a random “thing” out of the database.

Until today we managed this with a method at the particular model looking somehow like this:

1
2
3
4
5
6
7
8
9
10

  def self.random_good
    good = nil
    while (!good)
      begin
        good = Good.find(rand(Good.maximum(:id)))
        rescue ActiveRecord::RecordNotFound => e
      end
    end
    good
  end

Today I received my copy of The Rails Way and one of the first things I found was the native, MySQL “ORDER BY RAND()” function. Pretty cool. Because database portability doesn’t really matter for us, the same function as above now looks like this:

1
2
3

  def self.random_good
    Good.find(:first, :order => 'RAND()')
  end

2nd: Don’t forget about the power of ActiveRecord’s :include option!

The :include option of a find statement is very powerful! I read about this maybe a year ago, never needed it again and forgot about it. Today I rediscovered the veeeeery nice nesting capability:

1
2

  tickets = Ticket.find(:all, :include => [:user,
      {:operator => {:open_tickets => :user}}])

I can’t really believe it, but after this find statement you can do this:

1

  tickets.first.operator.open_tickets.last.user.nickname

…without ANY extra query! It’s all joined together in the one find query and still runs at a reasonable pace.

Ok, now it’s really time for me to go to bed.

Stay tuned,

Thorben
FEtMab-Team

Bookmarks to:

Hey folks,

just a short update on my Flock usage, I already posted about Flock in late December.

Even after some weeks of browsing Flock seems to be a good companion: Rare crashes on my Mac Book and the memory usage is at least remarkable better than with Firefox 2.x.

I recently gave Firefox 3 Beta 2 a try, too and it was ok. But without any really benchmarking it feels as if Flock is still more reliable and has less memory leaks than Firefox 3, so I decided to switch back to Flock.

This suits my plans very well, as I just gave up my “social network antipathy” and registered a profile on Facebook. Flock is definitely the browser of choice for all this  Flickr, Facebook, delicious stuff. I’m no heavy user of all of that but it’s nice that you can easily share links on delicious, even if you do it only once a day.

So, give Flock a chance, it’s worth it!

Yours,

Thorben
FEtMab-Team

Bookmarks to:

Hi again,

the time has come to “release” our first Rails plugin! *wooooooooah*

It’s mission is really simple: Do things “in” a Rails application periodically, in the background. Ok, sounds like a perfect job for cronjobs? Yes, it is. But let me explain this:

We are just two guys, coding all day long. Server administration is not our favorite sport and we do not have a cool admin guy to keep things going. Due to that, we once had a cronjob running which we didn’t monitor as we should’ve. The end of the story is, that our job was not executed for 2 weeks until we recognized it.

So, cronjobs may be just perfect, but we really don’t want the extra pain of monitoring an extra tool and keep the crontables up to date with every deployment and so on. Our solution:

StupidBackground

Ohhh yes my dear, there are plenty of Ruby and/or Rails tools out there that do that background job thing very well: BackgrounDRb, BackgroundFu and much more cool tools. But while some of these tools just are too much for our needs (BackgrounDRb), the other tools, which are simple enough, do not cover exactly what we wanted to do (BackgroundFu): Periodically call some methods which calculates things and then stores them to the database.

So that’s when StupidBackground comes into play.

It’s sooo stupid, that even 5 year olds should immediately know how to use it. But just in case your mind is somehow different to that of a 5 year old, I’ll try to explain it for you anyway:

1. Install it

script/plugin install \ http://svn.fetmab.net/svn/plugins/stupid_background/tags/0.1.1

2. Code your worker

Just create a new class in “lib/workers” which inherits from “StupidBackground::Worker” like this:

1
2

  class FiveYearOldFighter < StupidBackground::Worker
  end

Now just add some methods to this. When these methods are called by StupidBackground they are in a complete Rails environment created by “script/runner“. So feel free to use any models or whatever classes you like to use within your application. Like this:

1
2
3
4
5
6
7
8
9

  class FiveYearOldFighter < StupidBackground::Worker
    def fight(quantity, victory_chant)
      puts victory_chant if FiveYearOld.find(
          :all, :limit => quantity, 
          :order => "badass_rank DESC").all? {|five_year_old|
              five_year_old.fight
          }
    end
  end

Last open question: How can we configure how often we would like to fight 5 year olds? It’s easy:

1
2
3
4
5
6
7
8
9
10
11
12
13
14

  class FiveYearOldFighter < StupidBackground::Worker
 
    #THE NEXT TWO LINES ARE THE MAGIC!
    call(:fight).every(1.hour).with(20, "What a fight!")
    call(:fight).every(10.minutes).with(5, "Eaaasy one!")
 
    def fight(quantity, victory_chant)
      puts victory_chant if FiveYearOld.find(
          :all, :limit => quantity, 
          :order => "badass_rank DESC").all? {|five_year_old|
              five_year_old.fight
          }
    end
  end

To stay trained we now fight five 5 year olds every ten minutes and every hour we’re in the mood to fight 20 of them at the same time! Cool.

3. Start the background magic

Just one last thing to do:

rake background:start

And if you fought enough 5 year olds:

rake background:stop

Oh, and if you’ve forgotten if you’re right now fighting some 5 year olds try this:

rake background:status

That’s it. Pretty simple, but it is enough for our needs. Maybe this plugin is of some use for you.

Behind the scenes

StupidBackground uses the Daemonize library from Travis Whitton, tiny but great! This fact also means, that StupidBackground will not work on Windows machines.

The background behavior is realized via a stupid script/runner call, it’s not high sophisticated but again: it fits our needs.

The execution of the tasks happens in an non sophisticated way, too: StupidBackground just determines how long it is until the next task needs to be run, then sleeps this periods and then runs the job.

So if you have one job which should run in 5 seconds and another job which should run in 6 seconds, the first job runs in 5 seconds and when it’s finished the second job runs. So the intervals which you specify in a worker are not accurate enough for brain surgery… You have to see them as rough intervals which works, you guessed it already: great for us.

Any worker is just instantiated once! So if you would like to share informations between two jobs in the same worker you easily can. But be aware of unattended side effects, when one method saves information in an instance variable and another job reads it, even if it should not! The simplest way to get around this is to outsource any logic from your worker and then write two workers which both use the outsourced code.

If you would like to overwrite the initializer of the worker ensure that your initialize method takes one argument and passes this to the super initializer:

1
2
3
4

  def initialize(method)
    #your stuff
    super(method)
  end

Somehow like this.

Summary

• Periodically, task execution in a Rails application, in the background.
• No gems needed.
• No extra YAML or any other configuration. Anything is configured within your workers.
• One instance per worker, NOT per job.
• No threading if two jobs should run at the same time! Just waiting for one to finish, then starting the other.

Ok, I hope this is of some help for some of you. Feel free to add comments.

Yours,

Thorben
FEtMab-Team

Bookmarks to:

Hey folks,

I’m back from vacations. It was great, even if there was some snow shortage. We nearly all managed to get over the icy, artificial snow… nearly. On our third day’s last run my girlfriend felt unlucky and hurt her knee so bad she might needs an surgery. But she’s alright so this sounds a little bit more scary as it actually is.

Nevertheless our development is gaining some momentum this week as I get more and more back to work. The weeks before Christmas were awful. I missed no chance to get distracted from work and did not really made some progress.

But now things seems to go smoother and it is as the fun would come back to my work. Don’t know why, but it’s definitively a good thing.

I promised you some tech is coming back to this block with the presentation of our first two Rails plugins but this has to wait for some more days, but it’s coming… I promise you.

For now I can tell you something about where we are right now with the game:

I’m really looking forward to finish all of the pathfinding code this week, which means we’re really done with all of the “heavy” code. All what comes up next is not that heavy but not less important interface and minor gameplay work. That said, I hope to start the first private beta testing in February.. wow. It’s really exciting to see the project grow, even more with such a small team as we are.

Ok, before I spent anymore time telling you how cool we are, I should get back to Textmate and get this last pathfinding issues ultimately out of the door.

See you later!

Thorben
FEtMab-Team

Bookmarks to: