This page highlights the new features and changes introduced in 3.0 release. It is a high level overview and as such is by no means complete. Consult RELEASE-NOTES.txt file included in each release for the full list of changes, and UPGRADE.txt for the upgrade instructions.
- Distribution Contents Structure
- Object Relational Mapping
- Cayenne Persistence API
- Remote Object Persistence
- Deployment and Runtime
- There are a few changes in the contents of the distribution "lib" folder. For details see this page. Main Cayenne JAR is now called "cayenne-server-x.x.x.jar". We no longer ship the "fat" jar with dependencies. Instead dependencies are included in the separate jars in "lib/third-party" distribution folder.
- CayenneModeler is located in the "bin" folder. Never mind the "lib/cayenne-modeler-x.x.x.jar" - this is just an underlying Modeler library, used in "local JNDI" configuration for instance, not the runnable application.
Stopped supporting DerivedDbEntity - its functionality (aggregate queries) is achieved via EJBQL and other means.
A schema is now available here for the DataMap XML format.
Added a new mapping artifact - Embeddable, that allows to map composite objects to a single table.
Just like with "flattened" relationships, now there is a way to map an ObjAttribute that spans one or more DbRelationships and is mapped to a column of a joined table. An important use of flattened attributes is vertical inheritance mapping, as described below.
Adding flattened attributes to subclasses in a single-table mapping gives us another powerful way to map inheritance - "vertical inheritance mapping", described in more details here.
DataObjects are now allowed to have primitive properties, such as int, boolean, etc. Also Cayenne can now persist the following extra simple property types: java.util.Calendar, java.math.BigInteger, java.io.Serializable, char, java.lang.Character, java.lang.Character, java.lang.Byte.
In addition to mapping a to-many relationship as a List, Cayenne now supports to-many Maps (keyed by target object pk, or an arbitrary unique property), Sets and generic Collections. Note that relationship maps is the only conceptually new type of relationships. Collection and Set are not that different from the earlier List behavior (that was a Collection, and just like Set ensured uniqueness of its objects).
In addition to vanilla Java enums, Cayenne allows enums to provide a custom database value for each enumerated instance by implementing org.apache.cayenne.ExtendedEnumeration. This way a persistent value doesn't have to be tied to the enum Java name or position.
Changed DbAttribute "precision" property to "scale" (as it meant scale all along) and added a separate "precision" attribute to track precision.
Cayenne now requires Java 5 or newer to run. This allowed us to start providing generics-based API, enums and other Java 5 features.
While you can still use DataContext directly, we encourage you to avoid direct dependency on DataContext throughout your code, and instead use org.apache.cayenne.ObjectContext interface. E.g.:
This protects your code from the evolution of the underlying implementation, allows you to customize context behavior by using interface wrappers. There are plans to update all Cayenne API to return ObjectContext everywhere DataContext might be returned.
Never mind the "EJB" in the name. EJBQL queries are a powerful object query mechanism similar to the one supported by the JPA specification. EJBQLQuery is a first-class citizen, just like SelectQuery.
- Significantly simplified the process for fetching DataObjects (vs. DataRows) via a SQLTemplate query with 'setColumnNamesCapitalization' method.
- Added support for fetching scalar and mixed scalar/object results using SQLResult.
- Added support for "path splits". Splits allow to control how the joins are generated for the overlapping path expressions. By default a group of overlapping qualifier paths will produce a single join. To "split" that into separate joins, Cayenne allows to use "aliases" in path expressions, and then bind them to the actual pathes using "SelectQuery.addPathSplitAliases(String path, String... aliases).
Cayenne lifecycle callbacks are similar (but not identical) to callbacks defined in the JPA specification. This is a feature that allows to perform custom operations at certain points in the object lifecycle. For more details see this page.
- Added support for fetch offset (complimentary to the existing "fetch limit").
Implemented DataContext.rollbackChangesLocally(). This was missing in 2.0, limiting the operation of the nested DataContexts.
- pluggable query cache
- a notion of Query cache groups
- automated cache key management for standard queries
- Added missing DB expressions to ExpressionFactory.
- Added support for OUTER JOINS
- Added support for split expressions (see "ExpressionFactory.matchAllExp(String path, Object... values)" and "ExpressionFactory.matchAllExp(String path, Collection values)"). This is also related to the qualifier splits for SelectQuery described above.
- Added support for matching against TRUE and FALSE.
- Added support for "long" primary keys (vs. "int" PK used previously)
- Added adapter for SQLite database.
- Added adapter for H2 database.
- Added stored procedure support on MySQL.
- Changed default table type on MySQL to InnoDB, added MySQL FK constraints support in the Modeler.
- Turned on update query batching on SQL Server and Derby.
- Generating classes for DataMaps that contain type-safe methods for all mapped SelectQueries.
- Joint prefetches are switched to OUTER joins, removing the major limitation to their use.
- Disjoint to-many prefetches no longer have a limitation preventing their use if a query qualifier spanned prefetched relationships.
Now object validation works the same way on the ROP client as it did on the server.
Nested contexts can now be created for ROP contexts. Creation method is added to ObjectContext interface.
Saving projects with CayenneModeler v.3.0 will render them uneditable with earlier versions of the Modeler
The Modeler underwent significant changes with the goal to improve usability and look-and-feel.
- Added copy/paste support
- Reworked ObjRelationship mapping dialog
- Added combo auto-completion
- Added a usable and good looking welcome screen
- Added contextual menus
- Added support for flattened attributes
- Added copy/paste support
- Added detection of project file changes on disk
- Added multi-item selection
- Added ability to specify password encryption mechanism
- Added ability to drag and drop cayenne.xml from the filesystem to the Modeler window to open the project.
- Added support for to-many relationship collections mapping.
- Added "search" functionality
- Added lifecycle callback and listener mapping functionality.
- Added support for merging DataMap changes back to the database.
- Removed support for DerivedDbEntity
- Added confirmation dialog for delete actions
- Reopen Last Project on Startup
- Added progress/log view
- Added SQL syntax highlighting.
- Added EJBQL Query mapping capabilities, with syntax highlighting.
- Added support for quoting of SQL Identifiers
- The Modeler can be started from Maven
- Added "Infer Relationships" operation to automatically create DbRelationships based on PK/FK naming conventions.
- Added support for mapping Embeddable objects.
- Added undo/redo support
- Paginated query algorithm is changed, resulting in very significant performance improvements on large data sets (this also improves ROP paginated query performance). This includes faster query times and much smaller memory footprint of an unresolved list.
- Reduced the scope of lock on a shared cache during commit, that should result in improved concurrency.
- Queries using disjoint to-many and joint to-one prefetching now work significantly faster.
- ROP new context creation speed is improved, as EntityResolver is cached on the client after first retrieval.
- ROP uses a "diff compression" algorithm on commit, reducing the amount of data passed around.
- Leak free ObjectContext
- Query lists are cached in an LRU map, so there's no indefinite memory growth when using query caching.
Cayenne no longer has a Log4J dependency. Instead it uses Jakarta commons-logging, as described here. Note that 2.0 Log4J configurations should continue to work, although now it is a user responsibility to bootstrap Log4J - Cayenne will no longer do it for you. E.g. you can do this on startup:
Cayenne now supports optional quoting of SQL identifiers in the generated SQL. This has been a very popular request in the past. Identifier quoting can be turned on per DataMap in the Modeler (look for "Quote SQL Identifiers" checkbox). This will allow to map column names that contain spaces, non-ascii characters, reserved words, etc., i.e. the schemas that could not be mapped with Cayenne before.
DataNode has a new attribute in the Modeler - "Schema Update Strategy". This allows to specify runtime behavior in regards to schema consistency checking. The default behavior is "do nothing" (org.apache.cayenne.access.dbsync.SkipSchemaUpdateStrategy). Alternatively Cayenne can create the schema if none is found or throw an exception if schema inconsistencies are detected.
- Switched "cgen" to "version 1.2" of the template as a default version
- Support for classpath in "cdbgen".
- The Modeler can be started from Maven, that can potentially help in aligning the Modeler version with the project runtime Cayenne version.