I'm okay with moving forward with using dynamic initializers instead of the attributes. I agree that it's simpler after all.

EDIT: so I took a look at the changes and why the CI fails—it's because, as I thought, simply updating the database initializer to avoid the need for a chain of dependency (by instantiating manually the Connection) is not flexible enough. We use a different Connection in tests, so we can't just instantiate PDOConnection from within DatabaseInitializer. We need to resolve it from the container (by its interface), but then the right DatabaseConfig will not be used in the underlying classes.

That's easily fixed: one initializer can register multiple singletons. It can do whatever it wants with the container. The #[Singleton] attribute is only there for short-hand convenience.

Well, if you manage to fix it using this approach, it's awesome. I'll document the storage functionality after we merge this 👌

Pull Request Test Coverage Report for Build 14724150077

Details

• 23 of 23 (100.0%) changed or added relevant lines in 7 files are covered.
• 23 unchanged lines in 5 files lost coverage.
• Overall coverage decreased (-0.1%) to 80.147%

Totals
Change from base Build 14721086338:	-0.1%
Covered Lines:	11849
Relevant Lines:	14784

💛 - Coveralls