Tech Notes From The Trenches

How to work with Git submodules and Git for Mac

2012-02-02T18:13:58Z

As much as I love the new Github app (1.1.7) for Mac, it doesn’t work with submodules. This has been confirmed by Github’s tech support.

Hopefully they will graciously solve that problem soon. Until then, here is the workaround.

Whenever you modify anything in a submodule, you will need to commit it and push it:

1
2
3
4
5

$ cd ~/your-project
$ cd path/to/submodule
$ git commit . -m 'I changed X and Y because Z'
$ git push
$ cd ~/your-project

Once you’ve done that, you can sync your main repository with Github for Mac.

That’s it!

For more information on Git, this is a useful page: $ cheat git

Upgrading from Leopard to Lion the painful way

2012-01-31T15:18:23Z

The painful way:

Backup with TimeMachine
Upgrade to Snow Leopard
Upgrade to Lion
Restore essential data from TimeMachine

The faster way:

Backup with TimeMachine
Create a bootable Lion disk
Erase everything
Install over your current disk
Restore from TimeMachine

The painful way takes time.

Backing up 400GB takes time.

Installing Snow Leopard takes time.

Installing Lion takes time.

It’s painful because you should re-install MySQL Server and recover your databases from the TimeMachine backup.

You will need to re-install PostgreSQL. Thankfully it’s just:

brew install postgres
initdb
createuser postgres

It’s painful because you will have to re-install XCode. (or you’ll run into an issue like this trying to install git)

Installing XCode is not just installing it using the App Store, it’s also running Install XCode. Talk about mixed messages.

It’s painful because the version of gcc (LLVM based) won’t be the one you want. Thankfully there is a solution for that too.

Finally you can install RVM and the Ruby versions you want (1.9.2 and 1.9.3 for me)

And you are all set.

Be careful with stub! in RSpec

2012-01-18T14:59:09Z

First of all, you should be careful with any call to stub.

You probably want to use stub just to simulate a behavior, one that you don’t want to test in your current test.

For example: You may want to use stub with a remote service, an API call.

Also, the method is stub not stub! (at least in RSpec 2.6.0)

This will not work but will not fail.

# This won't work
DineroMailIpn::Client.any_instance.stub!(
:consulta_pago).and_return(@response_with_pays)

This is what you are looking for.

# This will work
DineroMailIpn::Client.any_instance.stub(
:consulta_pago).and_return(@response_with_pays)

I don’t know exactly why stub! doesn’t fail, but it wasn’t doing what I was expecting (to actually stub that method and return the mock object I wanted)

The best solution to this warning: Giving a path to render :action is deprecated.

2012-01-05T19:05:29Z

I for one welcome this change and deprecation warning. It means that you should stop using this:

render :action => :index

And, if necessary, you should start using this:

render :index

The code is simpler and prettier. Don’t use render :template, unless you really need to.

If you use vim, you can use this search and replace string:

:%s/render :action => /render /g

Por qué deberías usar `git pull –rebase` en vez de `git pull` solo

2011-12-27T16:13:36Z

Al usar

"git pull --rebase"

Git utiliza git-rebase en vez de git-merge.

git-merge nos genera un commit con todos los cambios entre el branch local y el branch que estamos bajando.

git-rebase va hacia atrás, hasta donde los branches divergieron, hace un reset y empieza a aplicar los cambios progresivamente, hasta llegar al último commit. Esto nos deja un log mucho más fácil de leer que con el commit generado por git-merge.

Así vamos a dejar de ver tantos:

"Merge branch 'master' of github.com:etagwerker/whuffiebank"

Para más información: http://progit.org/book/es/ch3-6.html

Cómo usamos Pivotal Tracker en Ombu Shop

2011-12-19T12:58:18Z

Este es el workflow que usamos en Ombu Shop para seguir las stories con Pivotal Tracker.

Objetivo

Atender todas las tareas pendientes sin que se nos escape nada. En base a la velocidad del equipo, tener una idea de fechas de entrega de cada story.

Prioridades

Se dividen por columnas. Current es más importante que Backlog. Backlog más importante que Icebox. Mientras más arriba en la columna se encuentre una story más prioridad tiene.

1. Toda mejora, bug o tarea debe figurar como una story en Pivotal

Bug: Comportamiento inesperado del sistema. Un error según lo especificado por una story previo.

Requester: La persona que pide la story. El requester es el que finalmente deberá aceptar la story.

Owner: La persona indicada para resolver la story. El requester elige al owner.

2. Antes de empezar, el owner debe estimar la tarea

Para estimar se usan los palitos horizontales a la derecha de la story. Más palitos, más complicado. Los bugs no se estiman.

Ejemplo de una story:

3. Una vez que el owner empieza debe darle Start a la tarea

Esto sirve para calcular la velocidad de nuestro equipo.

4. Una vez que termina debe darle Finish

Esto pondrá a la story lista para darle Deliver.

5. Una vez que se sube a producción debe darle Deliver

Solo se le dá Deliver cuando está en producción.

8. Recién ahí el requester debe darle Accept o Reject

El requester va a ser siempre el responsable de aceptar la story. Si uno le dá Reject hay que ser lo más claro y específico posible para decir por qué no se acepta.

Si la story fue aceptada y estaba relacionada con un cliente, avisarle. Eso queda a cargo del requester

Si la story fue rechazada, el owner va a tener que darle Restart (Ver #3) una vez que vuelva a trabajar sobre la tarea.

Rails 3.1 and PostgreSQL, “Please install the Postgres adapter” hell

2011-12-07T23:07:44Z

I just lost an hour with this error:

1	Please install the postgres adapter: `gem install activerecord-postgres-adapter` (cannot load such file -- active_record/connection_adapters/postgres_adapter) (RuntimeError)

I had to google a lot until I found the answer here, on Matt’s blog: “Please install the postgres adapter”

As usual, it was a stupid mistake in the database.yml:

1
2
3
4
5

development:
adapter: postgres
database: ombu_development
pool: 5
timeout: 5000

The adapter is called “postgresql” not “postgres”

1
2
3
4
5

development:
adapter: postgresql
database: ombu_development
pool: 5
timeout: 5000

I hope this saves you time.

Yet another good reason to keep a gemset per project

2011-12-05T14:22:37Z

I love rvm. It lets you keep your gems organized per project. The combination of .rvmrc and gemset is great for keeping things organized.

Yet another good reason is performance! (From @tenderlove’s blog post)

So, if you still keep all your gems in ‘system’, you should seriously consider switching to rvm gemsets per project.

Pusher: Presence channels are private channels

2011-11-28T13:42:17Z

Channels prefixed with ‘presence-’ don’t need to be also prefixed with ‘private-’

From the Pusher docs:

Presence channels build on the security of Private channels and expose the additional feature of an awareness of who is subscribed to that channel.

So, don’t bother adding the ‘private-’ prefix to a ‘presence-’ prefixed channel because it will be a waste of time.

Having a scope called `parents` breaks acts_as_nested_set

2011-11-27T05:38:50Z

I started to get this weird error with my categories today.

It was sort of ridiculous because I had only changed my scope name from ‘parent’ to ‘parents’!

That was the problem.

acts_as_nested_set, v1.4.4, breaks if you name a scope ‘parents.’

You will get an error like this:

undefined method `const_defined?’ for #<Category:0x112ccfdb8>