When database.yml defines more than one top-level key starting with "test",
every standard test task (rails test, rails test:models, rails test:system,
rails test:controllers, …) is automatically rewritten at load time to run
the full suite against each configured database in turn.
Single-database setups are completely unaffected — the wrapping code
activates only when test_configs.size > 1, so existing behaviour is
preserved by default.
Convention: test: is the required primary; test_<name>: adds an adapter:
test_sqlite:
adapter: sqlite3
database: db/fixinme_test.sqlite3
test_pg:
adapter: postgresql
...
Mechanism:
- lib/tasks/test_databases.rake loads after railties/testing.rake (which
defines all test tasks). For each task in the wrapped set it calls
Rake::Task[name].clear_actions and re-enhances with a block that loops
over DB configs, writing a temporary database.yml per database and
running "rails <task_name>" as a subprocess with RAILS_DATABASE_YML set.
- config/application.rb(.dist): reads RAILS_DATABASE_YML and overrides
Rails' database config path before initialisation — no monkey-patching
of the test runner required.
- Adapter gem availability is checked before each run; missing adapters are
skipped with a hint on which bundle group to enable.
- A formatted summary (✓/✗ per database) is printed; exits non-zero on
any failure.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* replace variable names with Hash to avoid invalid Ruby identifiers
* export all values as single-quoted to avoid string interpolation and
treating BigDecimal numbers as Float
* #truncate table instead of #delete_all to avoid foreing_key
constraints errors
Closes#56
