Official documentation describes the following behavior:
https://docs.feldera.com/connectors/sinks/postgresql/
However, the actual behavior is different:
The PostgreSQL column is defined as:
updated_at timestamp DEFAULT now() NOT NULL
The Feldera SQL table does not include a timestamp field in the sink definition.
The sink operation fails with an error indicating that a null value cannot be inserted into the NOT NULL column, even though a default value (now()) is defined.
After debugging, I traced the issue to this code snippet in the sink implementation:
raw_queries.insert = format!(
r#"INSERT INTO "{table}" SELECT * FROM jsonb_populate_recordset(NULL::"{table}", $1::jsonb) ON CONFLICT {on_conflict}"#,
);
Is there a recommended approach to support auto‑generated timestamps (or other default values) without having to explicitly include the column in the Feldera SQL schema or the INSERT statement?
Official documentation describes the following behavior:
https://docs.feldera.com/connectors/sinks/postgresql/
However, the actual behavior is different:
The PostgreSQL column is defined as:
updated_at timestamp DEFAULT now() NOT NULL
The Feldera SQL table does not include a timestamp field in the sink definition.
The sink operation fails with an error indicating that a null value cannot be inserted into the NOT NULL column, even though a default value (now()) is defined.
After debugging, I traced the issue to this code snippet in the sink implementation:
raw_queries.insert = format!(
r#"INSERT INTO "{table}" SELECT * FROM jsonb_populate_recordset(NULL::"{table}", $1::jsonb) ON CONFLICT {on_conflict}"#,
);
Is there a recommended approach to support auto‑generated timestamps (or other default values) without having to explicitly include the column in the Feldera SQL schema or the INSERT statement?