There are still a handful of things that need (for some value of “need”) doing.
- Implement views for making a nice quick REST API which will map URLS of the
form
/foo/<collection_name>onto DRF views. (This should be done outside of cqrs, in cqrs-renormalize.) - Test automatic derivation of collections.
- Automatically add collections to a backend. (It really grates how we have
this
cqrs.mongo.mongodbhardcoded everywhere. It's fragile, too.) - Implement some form of "tag" functionality, whereby we can have multiple
serializers (and in turn collections) with a tag to discriminate between
them. If no tag is specified, use
None, indicating a default serializer. While it would be possible to change the "which fields do we use?" rules to fall back to a default serializer, I think that is risky behaviour. I suggest completely distinct trees with absolutely nothing shared. It will be easier to reason about that way.