On Jan 31, 2011, at 3:10 AM, elitalon wrote:
Some questions come to mind: do you recommend to make a commit after
applying a migration? Isn’t a migration some kind of commit itself?
While in general putting generated artifacts into a repo isn’t usually a
good idea, I always keep things like the schema in source control, so I
do a commit after each migration is run. Personally I’ve been committing
after creating the migration script and then after running it. It might
make more sense to treat that as an atomic unit of work and make it a
single commit, though.
I also know that some people begin with the view, then develop the model
and in the end implement the controller, so what do you recommend?
Start with the cucumber/capybara test. Then create a route, then the
simplest controller action and view to make the test pass. If there is
any complexity to any of the above, write an rspec test first, then
write the code to make it pass. If the controller has too much
complexity, either get the tests passing and then refactor to the model,
or possibly start with tests for a model object and then call it from
the controller. For example, I needed to geocode addresses the other day
and after writing the cucumber test, route and controller I immediately
added a model object called location to wrap that complexity and rspec
tests (with fakeweb and vcr) to ensure it worked as expected.