Today, time-series can only be sent to 1 collection; this limits the number of workloads when content routing is required. This should be enhanced so that db and coll can be fields or expressions just like with kafka sinks and normal collections

let e = {
... "$emit": {
... "connectionName": "jsncluster0",
... "db" : "test",
... "coll" : "$coll",
... "timeseries" : {
...  "timeField": "timestamp"
... }
... }
... }


AtlasStreamProcessing> sp.process([s,rr,e])
MongoServerError[CommandFailed]: Failed to start stream processor: Time series $emit to test.$coll failed: The namespace "$coll" is invalid.: generic server error, correlationID = 18074837600d8f1536cc47f6 

Suggested Implementation + Behavior

We will implement this feature incrementally. This ticket will involve being able to specify field path and object expressions for the db and coll fields.

We should gate off specifying expressions for db and coll using a feature flag (enableTSEmitDynamicContentRouting).

In this ticket, we should make a mongocxx.collection instance per input document.

In the followup ticket, we should perform the collection validations that we do today per input document, as well as using a LRU cache to store previously used mongocxx::collection instances per input document