Separate repositories for commands and queries

[subhashb]

Repositories are primarily useful on the write-side of an application. They are meant to handle complex relationships within Aggregates, enclose changes in transactions, and represent domain concepts close to the database layer.

The read-side, in contrast, is simpler, closer to database concepts, and custom-fit for the consumption side of the application (like GET APIs or user interfaces). These queries often change with interface requirements. Mechanisms like Graph APIs and tools like PyPika make it easy to reduce the burden on the app itself and allow consumers to specify what they want to fetch. Maintaining queries on the app itself can be a waste of time with rapid-changing requirements.

Should we distinguish between command repositories and query repositories?

This would allow the read side to be completely distinct from the write side and potentially work with a different set of data structures. If we use the CQRS pattern and split the command and query sides, the read side will work with query models anyway.

Follow-up questions:

• What would be the best way to separate the repositories for read and write?
• Should repositories be allowed to query databases?