134 lines
6.1 KiB
Plaintext
134 lines
6.1 KiB
Plaintext
= 5.3 Migration Guide
|
|
:toc:
|
|
|
|
This guide discusses migration from Hibernate ORM version 5.2 to version 5.3. For migration from
|
|
earlier versions, see any other pertinent migration guides as well.
|
|
|
|
== Background
|
|
|
|
Hibernate 5.3 adds support for the JPA 2.2 specification
|
|
|
|
|
|
== Known changes
|
|
|
|
=== Changes to positional query parameter handling
|
|
|
|
This really breaks down into 2 related changes:
|
|
|
|
* Support for JDBC-style parameter declarations in HQL/JPQL queries has been removed. This feature
|
|
has been deprecated since 4.1 and removing it made implementing the second change, so we decided
|
|
to remove that support. JDBC-style parameter declaration is still supported in native-queries.
|
|
* Since JPA positional parameters really behave more like named parameters (they can be repeated,
|
|
declared in any order, etc.) Hibernate used to treat them as named parameters - it relied on
|
|
Hibernate's JPA wrapper to interpret the JPA setParameter calls and properly handle delegating to
|
|
the named variant. This is actually a regression in 5.2 as it causes
|
|
`javax.persistence.Parameter#getPosition` to report `null`.
|
|
|
|
For JDBC-style parameter declarations in native queries, we have also moved to using one-based
|
|
instead of zero-based parameter binding to be consistent with JPA. That can temporarily be
|
|
reverted by setting the `hibernate.query.sql.jdbc_style_params_base` setting to `true` which
|
|
reverts to expecting zero-based binding.
|
|
|
|
|
|
=== Change in the `@TableGenerator` stored value
|
|
|
|
In order to be compliant with the JPA specification, the sequence value stored by Hibernate 5.3 in the database table used by the `javax.persistence.TableGenerator`
|
|
is the *last* generated value. Previously, Hibernate stored the *next* sequence value.
|
|
|
|
For backward compatibility, a new setting called `hibernate.id.generator.stored_last_used` was introduced, which gives you the opportunity to fall back to the old Hibernate behavior.
|
|
|
|
[NOTE]
|
|
====
|
|
Existing applications migrating to 5.3 and using the `@TableGenerator` have to set the `hibernate.id.generator.stored_last_used` configuration property to `false`.
|
|
====
|
|
|
|
|
|
=== Second-level cache provider SPI changes
|
|
|
|
Hibernate's second-level cache SPI needed to be redesigned to match with expectations between
|
|
Hibernate and the various caching providers. And the SPI did not really clarify the intention.
|
|
These mis-matched expectations certainly had the potential to lead to bugs. Although it was
|
|
originally slated for 6.0 to delay the SPI changes, we decided to back-port the work to
|
|
5.3 to address a number of bugs that could have been avoided with a clarified SPI.
|
|
|
|
The changes also allow the caching providers to perform serious optimizations based on
|
|
the users configuration of domain data caching in Hibernate.
|
|
|
|
Details can be seen on the https://hibernate.atlassian.net/browse/HHH-11356[HHH-11356] Jira issue
|
|
|
|
One potential upgrade concern is any custom `org.hibernate.cache.spi.QueryCacheFactory` implementations.
|
|
`org.hibernate.cache.spi.QueryCacheFactory` was meant as a means to allow service providers the
|
|
ability to define query result caching, generally with more stale-tolerant query result invalidation handling.
|
|
However, the contract also bound it to the old second level cache contracts so they had to be
|
|
updated. Its responsibilities also changed so we also decided to "rename it" to
|
|
`org.hibernate.cache.spi.TimestampsCacheFactory`
|
|
|
|
Another specific change to be aware of is accessing cache entries via `SecondLevelCacheStatistics#getEntries`
|
|
and `NaturalIdCacheStatistics#getEntries`. These methods have been deprecated, however the new
|
|
caching SPI does not really require caching providers to support this. As of 5.3 these methods
|
|
inherently return an empty Map (`Collections#emptyMap`). This has always been something that providers
|
|
did not implement "correctly" anyway
|
|
|
|
|
|
=== Statistics changes
|
|
|
|
The change for HHH-11356 required changes in its consumers. One such consumer is the Hibernate
|
|
Statistics system....
|
|
|
|
|
|
=== Drop hibernate-infinispan module
|
|
|
|
Support for using Infinispan as a Hibernate 2nd-level cache provider has been moved to the Infinispan project so
|
|
the `hibernate-infinispan` module has been dropped.
|
|
|
|
A relocation pom which is pointing to `org.infinispan:infinispan-hibernate-cache` dependency is still generated,
|
|
therefore, avoiding the need of updating any library dependency.
|
|
|
|
[WARN]
|
|
====
|
|
The relocation pom may be dropped in a future release.
|
|
====
|
|
|
|
|
|
=== EnhancementTask changes
|
|
|
|
The API of the `org.hibernate.tool.enhance.EnhancementTask` Ant task was changed, specifically
|
|
the `#addFileset` method was dropped in favor of `#setBase` and `#setDir`
|
|
|
|
See details on the https://hibernate.atlassian.net/browse/HHH-11795[HHH-11795] Jira issue.
|
|
|
|
The main gist is that EnhancementTask was fixed (through a contribution) to actually work with
|
|
`Enhancer` from `BytecodeProvider`. Previously it had not. And part of fixing that required this
|
|
change.
|
|
|
|
|
|
=== Many-to-one association in embeddable collection elements and composite IDs
|
|
|
|
A bug introduced in 4.3 caused many-to-one associations in embeddable collection elements and
|
|
composite IDs to be eagerly fetched, even when explicitly mapped as lazy.
|
|
|
|
This bug does not affect many-to-one associations that are not in a composite ID or embeddable
|
|
collection element.
|
|
|
|
In 5.3.2, this bug was fixed. As a result, such associations will be fetched as specified
|
|
by their mappings.
|
|
|
|
Many-to-one associations mapped by using native HBM xml are lazy by default. In order to keep
|
|
the associations eager in 5.3.2 and later, mappings will need to explicitly specify that
|
|
they are non-lazy.
|
|
|
|
When mapped with annotations, many-to-one associations use `FetchType.EAGER` by default.
|
|
Starting in 5.3.2, if an association is mapped with `FetchType.LAZY`, the assocation will
|
|
be lazily fetched, as expected.
|
|
|
|
See details on the https://hibernate.atlassian.net/browse/HHH-12687[HHH-12687] Jira issue.
|
|
|
|
=== 5.3 -> 6.0 compatibility changes
|
|
|
|
The original driving force behind these series of changes is an effort to be as proactive as possible
|
|
about designing compatibility between 5.3 and 6.0.
|
|
|
|
|
|
==== Type system changes
|
|
|
|
Use of NavigableRole, back-ported from 6.0 rather than plain String |