Filtering and queries

Use source.table with source.where for a table-backed replication. Use source.columns to limit the selected columns when the sink mapping supports that shape.

The table name may identify a connector-supported table-like object such as a view. source.where is appended to the connector’s generated selection, so provide only the predicate rather than the WHERE keyword. source.columns and sink.columns are ordered, comma-separated mappings; keep their counts and types compatible when names or order differ.

mode=complete
source.connect=${SOURCE_CONNECT}
source.table=${SOURCE_TABLE}
source.columns=id,payload
source.where=updated_at > ${SOURCE_CUTOFF}
sink.connect=${SINK_CONNECT}
sink.table=${SINK_TABLE}

For free-form reads, use source.query instead of source.table. Do not combine source.query with the explicit replication.table.{N}.* catalog or with automated incremental watermark tracking, because ReplicaDB cannot infer the required table and merge semantics from an arbitrary query.

mode=complete
source.connect=${SOURCE_CONNECT}
source.query=SELECT customer_id, total FROM reporting_orders
sink.connect=${SINK_CONNECT}
sink.table=${SINK_TABLE}
sink.columns=customer_id,total

A free-form query owns its projection, joins, expressions, and ordering. ReplicaDB streams its result; it does not validate that an expression is portable to the source engine or that the projected values fit the sink types.

Treat filters as source SQL expressions supported by the connector. Use environment substitution for values and validate the resulting query against a read-only account before a production run.