Kirby's Dreamland

So everybody tells me that Git is cool. I’m used to using Mercurial (which I think is great, by the way) which is also a distributed VCS so I think “how hard can it be?” Silly Kirby, the differences are more than just skin deep!

Git and Hg (Mercurial) have two very different philosophies when it comes to version control and these affect how each of them have implemented core features. Git takes the super-user route and lets you do whatever you want, even if that means you might hurt yourself and hose your repo. While Hg tries to do everything it can to prevent you from modifying your history. This means that a lot of that cool branching stuff that Git does, Hg doesn’t want you to do because it doesn’t want you to make a topic branch that is disposable and deletable since someone may have cloned it from you.

It is worth mentioning that there are a host of optional plugins for Hg that let you do pretty much anything that is possible in Git. But since I’m just learning Git, I’m not going to stick my foot any further down my throat on design philosophies until I’m a little more familiar with it. Which brings me to some things about Hg that i just wish Git had.

The first big differences that I noticed were the lack of auto-complete in Git.  For instance, if you have modified lib/project/widget.rb in your Hg repo, you can go hg ci li<tab> and if the above file was the only one that you’d changed, Hg is smart enough to auto-complete the rest of the filename for you. Awesome!

I haven’t yet found anything comparable in Git although i did come across .git-completion.bash which has amazing support for auto-completing Git commands, options, tags and branches. It also lets you add something like $(__git_ps1 "(%s)") to your PS1 which will show you what your current branch is when you’re in a Git repo. For even more info you can add the following lines to your .bashrc as well:

GIT_PS1_SHOWDIRTYSTATE=1 #... untagged(*) and staged(+) changes
GIT_PS1_SHOWSTASHSTATE=1 #... if something is stashed($)
GIT_PS1_SHOWUNTRACKEDFILES=1 #... untracked files(%)

The next thing that i really, really missed about Hg was the super-short aliases for common commands. Like hg ci for hg commit. Git config to the rescue!

$ git config --global alias.co checkout
$ git config --global alias.br branch
$ git config --global alias.ci commit
$ git config --global alias.st status

That’s it for now. Happy Git-ing

Edit: found another cool one from the great Pro Git Book:

$ git config --global alias.unstage 'reset HEAD --'

Unlike Java, the facilities in Ruby 1.8 (and Rails) for logging exceptions and their associated stack traces are poor.  You might think that something like this (from an ActiveRecord::Base object) would work:

 begin
  # code that raises some exception
rescue Exception => e
  self.logger.info e
end

 begin
   # code that raises some exception
 rescue Exception => e
  self.logger.info e.message
  self.logger.info e.backtrace.join("\n")
 end
 

Thankfully, we live in the age of multi-core processors and excellent tools such as Hydra.  We've got our tests running in about 2 minutes by throwing about 30 cores at them.  This is good.

Another good thing when you've got a lot of tests are factories (check out Machinist 2, it's beautiful).  And factories combined with seed test data (not fixtures!) is simply divine (a great topic for another blog post).

But what happens when you've got a lot of seed data but you don't have a database for each core that's running tests?  Well, the answer to that depends on if you're updating said seed data in your tests.

If you are only reading seed data, here is what the PostgreSQL manual says in section 12.2.1:

Looks like if you have seed data in a DB that's shared by multiple concurrent test runners, any updates to the seed data will cause your test runners to block each other.  Don't change that data!

It's a gem/rails plugin that handles setting belongs_to associations for ActiveRecord models based on a field belonging to the associated model.  I'll do a more thorough write-up of why this plugin is cool and how you go about using it later but i wanted to tell the world about it while i'm still as giddy as a schoolgirl.

Caveat emptor:  All the tests run, but I haven't actually tried using the extracted gem in the project that it was extracted from.  In other words i may have messed up the gem configuration or something.  That will also be verified later.

A big thank you goes out to the rest of the Nulogy guys that agree that contributing to open source is a Good Thing and for the help in writing autocomplete_for.

I was kind of shocked by the gotchas though:

Scared of concurrency?  It's easy! http://www.engineyard.com/blog/2010/concurrency-and-the-aasm-gem

The reason that the new Arel query interface in Rails 3 was written.  Seriously mind-expanding if you follow all the links and references. http://magicscalingsprinkles.wordpress.com/2010/01/28/why-i-wrote-arel/

(Do you see why it's called the Wizard book?)

I just had to give some props to fellow Nulogite, Ian Bailey, for his outstanding suggestion on what I should call this blog.  It's a little ostentatious for me, but he likes "Kirby Superstar" and even made an image to go along with it.

Here are some concrete examples of how 'and' can be more dangerous than '&&' in Ruby: http://www.themomorohoax.com/2008/12/08/and-vs-amperand-in-ruby

I've been going a bit crazy about blocks, Procs and lambdas lately trying to figure out how to jury-rig Machinist (which is an awesome piece of code, by the way) syntax on top of Factory Girl.  Can anyone tell me the difference between lambda's and Procs?  I'll give you a hint: a big part is how returns are handled:

• http://www.robertsosinski.com/2008/12/21/understanding-ruby-blocks-procs-and-lambdas/
• http://moonbase.rydia.net/mental/blog/programming/the-biggest-mistake-everyone-makes-with-closures.html#c1d684a99e55e1e94a68487091111820
• http://www.ruby-forum.com/topic/54096
• http://www.jroller.com/abstractScope/entry/passing_parameters_to_an_instance

Firefox 4 is now in beta and is pretty cool.  They've stolen the tab layout from Chrome and resizeable text areas from Safari, so i don't have to switch browsers.

Nulogy also has a customer in Australia now.  We're going to have to monitor them and tune our pages so they don't suffer from this: http://mike.bailey.net.au/blog/?p=38

I'd also like to recognize fellow Nulogy founder, Jason Yuen, as his wife recently inducted him into the club of Fathers.  Congratulations on successfully reproducing!

That's it for now,