Lukas Eder

OK, very curious. This particular query also fails on 2.7.4. The actual query in our integration tests is quite different, which may not have triggered the problem in the past, but now does in 2.7.5. Perhaps this isn't strictly a regression in HSQLDB after all, even if we have a regression in our integration tests.

Regression: General Error in 2.7.5 when using expression in FILTER .. WHERE

Sure, but I wonder what's the point of completely documenting the (potentially future) syntax if it's simply not supported? Or, perhaps, instead of just failing in the parser, you could parse the syntax and throw a "not yet supported" exception of sorts. Some RDBMS do this, e.g. CockroachDB, YugabyteDB, especially when it's about a PostgreSQL syntax that they don't support yet. They usually include a link to an issue in the issue tracker that helps provide feedback and gauge interest. Sure, I'll...

Well, the manual says "HyperSQL does not support deferring constraint enforcement.", but I don't see the point of documenting the syntax and explaining it if it's not supported...

DEFERRABLE constraints don't work

As per my follow up comment, the error is in my thinking. When running the MERGE statement for the first time, there's no match simply because the target table contains no rows.

Egh, stupid 😅 t is empty in the beginning, so there was no match in the inner join

Hmm, I must be misunderstanding something. It seems that most (all?) RDBMS behave this way?