问题
I'm using PHP/MySQL...
I'm using interfaces for repositories on my Domain "command" side. This is going well. But I'm stuck on what to do on the "queries" (read) side. Do I make queries each as separate methods grouped in classes by some common similarity I find? Or do I make each query its own class? How should I go about using interfaces to make testing (and easier replacement later)?
Some places I've researched:
- A github php example
- An attempt to find the best decoupled approach to the read-side queries. Examples use C# unfortunately and make it a bit more challenging to grasp.
These solutions are convoluted?
回答1:
We currently use the approach from the second c# link for one of our projects. If you want to use a query facade of some kind, then that's the best one I've come across. What I like most about it is that you create query objects and query handlers, and then a generic query processor does the work of passing the object to the correct handler.
Overall it works well, however it still feels like I'm working harder than I need to. I'm not really sure what this query abstraction is giving me, other than having more to maintain. Testability is often the response to that question, but I for one do not unit test queries. Unit testing in my opinion is best left for code that can potentially break/affect other code. Can a query really break anything that wouldn't be immediately apparent during regular testing? It either works or it doesn't.
If you do have a compelling reason to use a facade, then I recommend grouping the queries that are related into an interface. Such as "IBillingQueries" and "ICustomerQueries".
来源：https://stackoverflow.com/questions/29617433/in-ddd-and-cqrs-for-read-queries-what-is-a-strategy-that-allows-for-interfaces