1328837476-hug commented on issue #10443: URL: https://github.com/apache/seatunnel/issues/10443#issuecomment-5603830552
Thanks for the clarification. I reviewed the current `dev` JDBC path at commit `75fd4ed4b2e63579a57b461285997951490c4fe7`. | Area | Current generic JDBC behavior | Ingres compatibility / deviation | |---|---|---| | Driver | SeaTunnel loads the configured driver class through the context classloader, so the driver can remain user-provided. | The current Actian artifact is `com.ingres.jdbc:iijdbc:12.1-4.6.5`, driver class `com.ingres.jdbc.IngresDriver`. Its Maven metadata declares a commercial license with redistribution restrictions. I will not add or bundle this dependency. | | JDBC URL detection | `GenericDialectFactory` accepts unmatched JDBC URLs. The loader removes Generic only when a specific factory matches. | `jdbc:ingres://...` currently falls back to Generic. PostgreSQL accepts only `jdbc:postgresql:`, so Ingres does not enter the PostgreSQL path. | | Source | The generic path loads schema through JDBC `ResultSetMetaData`, using `GenericTypeMapper`, when no catalog factory exists. Query-based reads therefore have a viable code path with a user-installed driver. | Live execution is still unverified. Owner/schema-qualified table paths and identifier case need focused verification. | | Sink | Plain append INSERT uses generic generated SQL and the generic row converter. | Plain append has a viable code path. Native upsert is not currently usable because `GenericDialect.getUpsertStatement()` throws `UnsupportedOperationException`. | | Identifier quoting | Generic returns identifiers unchanged. | Ingres delimited identifiers use double quotes, embedded quotes must be doubled, and delimited identifier case is significant. This is a concrete dialect deviation. | | Table paths | Generic parsing does not model a schema/owner-qualified path. | Ingres treats schema names as object owners and JDBC `Connection.getSchema()` returns `null`; `owner.table` handling needs an Ingres-specific parse/table-identifier override. | | Type mapping | Generic maps the standard JDBC numeric, character, binary, date, time, and timestamp type codes. | Actian documents JDBC 4.2 conversions for those standard types and does not support ARRAY, DATALINK, DISTINCT, JAVA_OBJECT, REF, REF_CURSOR, ROWID, SQLXML, or STRUCT. The metadata codes and behavior of Ingres-specific types such as `INGRESDATE`, `MONEY`, and intervals remain a live-test boundary. | | Upsert | With primary keys and the default `enable_upsert=true`, Generic reaches the unsupported native-upsert method. With a dialect returning `Optional.empty()`, SeaTunnel can use its existing row-existence plus INSERT/UPDATE fallback. | Ingres supports `MERGE`, but I propose not adding speculative native MERGE SQL in the first slice. The initial dialect can explicitly use the existing fallback until real JDBC behavior is verified. | Proposed first implementation, after agreement: 1. Add only an Ingres JDBC dialect/factory inside `connector-jdbc`, recognizing `jdbc:ingres:`. 2. Reuse the generic row converter and standard type mapper. 3. Override only confirmed differences: double-quoted identifier escaping, owner-qualified table paths, and an explicit non-native-upsert fallback. 4. Add focused factory, quoting, path, SQL-generation, and standard type-mapping tests. 5. Add documentation stating that users must supply the commercially licensed driver themselves. Validation boundary: Actian publishes an official `actian/ingres` image, but standalone use requires a valid `license.xml` and explicit license acceptance. I do not currently have a licensed Ingres test instance, so I cannot claim a live JDBC integration result before the PR. I will keep all Ingres-specific types and native MERGE behavior out of scope unless a licensed test environment becomes available. References: - [Actian driver coordinates and license](https://central.sonatype.com/artifact/com.ingres.jdbc/iijdbc) - [Official JDBC URL and driver class](https://docs.actian.com/actianingres/12.0/Connectivity/JDBC_Driver_and_Data_Source_Classes.htm) - [Ingres identifier rules](https://docs.actian.com/actianx/12.0/OpenSQLRef/Restrictions_on_Identifiers.htm) - [Official Docker image and license requirements](https://hub.docker.com/r/actian/ingres) If this evidence and validation boundary are acceptable, I will implement only the narrow dialect/factory slice above—no CDC, standalone connector, catalog redesign, bundled driver, or speculative DDL. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
