Play 2.0 database model generation for Integration TDD/BDD

2
votes

We're doing our first project using the Play 2.0 framework in Scala. We are wondering if there was one (or more) existing best practices or libraries for generating the test data from the "GIVEN" sections of integration and end-to-end tests?

For example, in the Rails world, one might have a cucumber step GIVEN there are 10 users who like chocolate, which under the hood would use FactoryGirl or a similar library to generate those models in the database; you can then count on the test database being emptied between test runs automatically, and the tests run in a different database instance than your actual application.

In the Scala world using Play 2.0, we are considering using Anorm on MySQL. Should we special case our tests to use an in-memory database (taking care of clearing it), and write code to simply create and persist case classes in our GIVEN blocks?

Along the same lines, will the Play "evolutions" be run automatically to produce the production schema in the testing in-memory database?

Thanks for any help to avoid re-inventing wheels here.

1
scalaplayframeworkplayframework-2.0
@streetpc Yes, adding my own answer below to capture some of what we've learned - ms-tg

1 Answers

2
votes

Since I asked this question, we've settled on the following workflow:

  • Use ScalaQuery 0.10.0-M1 w/ Case Classes for Models
  • Use FakeApplication with an in-memory database for unit and functional tests
  • Use a real, local MySQL database for integration and end-to-end tests
  • As proposed in this question, we are persisting the models for each GIVEN
    • For unit and functional tests, using the H2 in-memory database, each test has a new DB, which is great and fast and isolated from other tests, and no clean-up is necessary
    • For integration and end-to-end tests, using a local MySQL database, the tests share a DB. We are currently using randomized data for each GIVEN, so that tests do not affect each other, and periodically clearing our local databases

This process has worked fairly well for us so far, with a few gotchas that are the subject of other StackOverflow questions (todo: link to those questions here)

In answer to my question above: Yes, the evolutions are run each time when an in-memory database is brought up.