databroker
databroker copied to clipboard
Add high-level API to get a Serializer
As discussed on the pilot call.
We have gone as a far as concluding that the read and write sides of databroker should be separated (databroker -> for read suitcase/serializer -> for write). We do not want to give people a write handle they are not allowed to us and it is likely that, in the mid-term, the read and write sides may be deeply asymmetric (write to kafka, read from mongo + s3).
We have had vague discussions of adding a db, serializer = get_me_the_things('MAD')
to get them together. This has the advantage that the two sides are generated together, but due to the expected asymmetry in the read/write side this may not actually be a benefit.
On the call @ronpandolfi suggested a databroker.Serializer.named('MAD')
at the top level as a peer to databroker.Broker.named('MAD')
which just has to promise to match on name!