The recommended CockroachDB v22.1 settings for local testing have changed. The recommended
settings are now applied.
The setting `schemachanger.backfiller.buffer_increment` seems to have been removed
(despite still being mentioned in the documentation). We remove it
because it throws an error and this may leave the Cockroach cluster
with some of the recommended settings not applied.
* Switch from the "expression" to "expressionOrPredicate" rule in the HQL grammar where it makes sense as required by some HQL tests
* Cleanup parser rule ordering to allow more keywords in the identifier rule
* Implement literal support for Ordering parser
* Add special AvgFunction as needed by H2, HSQL, DB2, SQL Server and Sybase that casts arguments to double if necessary
* Fix wrong deduplication of order by fragments in case a plural attribute is fetched multiple times
* Implement support for de-referencing any-valued mappings in HQL
* Avoid unnecessary entity subtypes in polymorphic splitted queries if a base type also matches the requested type
* Implement pagination support for polymorphic splitted queries
* Cleanup path part resolving by removing lots of duplicate code
* Aligh HQL parsing expectations to the expected behavior of 5.x
* Add method to `JavaType` that allows determining if a type is can be widened to another which is used for arithmetic type resolving
* Implement validations for fetch owner checking
* Fix issues with the id table creation due to lacking column lengths in the column DDL type
* Fix issues and add some optimizations related to multi-table delete handling
* Add the notion of a special "implicit" alias to avoid generating a unique alias for unaliased or implicit HQL joins
* Properly implement multiple bag fetch validation
* Make sure filter predicates are applied for all plural attribute joins
* Fix some issues with undecidable parameter type inference
* Fix some issues with negated SQM predicates not being converted to the proper SQL AST predicates
* Fix issues with qualifying DML target referencing columns
* Fix `is null` semantics for tuples referring to embeddable types
* Capture necessary details from JdbcValuesMetadata in the cached data to avoid executing a query on cache hit when types should be inferred
* Get rid of special CollectionPropertyNames and writeup a migration guide section for the replacements
* Get rid of unnecessary whitespace and optional keywords in generated SQL
* Handle some type inference related issues with some databases requiring to render casted parameters in some situations
* Ensure SQM model is fully serializable
* Ensure JPA Criteria throws expected exceptions
* Make sure JPA Criteria implementations work properly
* Move jpa.test.callback and jpa.test.criteria packages
* Improve the reuse of SqmPath instances
* Get rid of many raw-types related warnings
* Make Predicate extend Expression and handle SQL rendering/emulation
* Support fetching SqmTuple as array
* Implement treat operator support
Christian Beikov
Karel Maesen
Sanne Grinovero
Andrea Boriero
Steve Ebersole
Yoann Rodière