2.  Reverse Mapping

2.1. Customizing Reverse Mapping

OpenJPA includes a reverse mapping tool for generating persistent class definitions, complete with metadata, from an existing database schema. You do not have to use the reverse mapping tool to access an existing schema; you are free to write your classes and mappings yourself, as described in Section 3, “ Meet-in-the-Middle Mapping ”. The reverse mapping tool, however, can give you an excellent starting point from which to grow your persistent classes.

To use the reverse mapping tool, follow the steps below:

  1. Use the schema tool to export your current schema to an XML schema file. You can skip this step and the next step if you want to run the reverse mapping tool directly against the database.

    Example 7.8.  Reflection with the Schema Tool

    java org.apache.openjpa.jdbc.schema.SchemaTool -a reflect -f schema.xml

  2. Examine the generated schema file. JDBC drivers often provide incomplete or faulty metadata, in which case the file will not exactly match the actual schema. Alter the XML file to match the true schema. The XML format for the schema file is described in Section 14, “ XML Schema Format ”.

    After fixing any errors in the schema file, modify the XML to include foreign keys between all relations. The schema tool will have automatically detected existing foreign key constraints; many schemas, however, do not employ database foreign keys for every relation. By manually adding any missing foreign keys, you will give the reverse mapping tool the information it needs to generate the proper relations between the persistent classes it creates.

  3. Run the reverse mapping tool on the finished schema file. If you do not supply the schema file to reverse map, the tool will run directly against the schema in the database. The tool can be run via its Java class, org.apache.openjpa.jdbc.meta.ReverseMappingTool.

    Example 7.9.  Using the Reverse Mapping Tool

    java org.apache.openjpa.jdbc.meta.ReverseMappingTool -pkg com.xyz -d ~/src -cp customizer.properties schema.xml

    In addition to OpenJPA's standard configuration flags, including code formatting options, the reverse mapping tool recognizes the following command line arguments:

    • -schemas/-s <schema and table names>: A comma-separated list of schema and table names to reverse map, if no XML schema file is supplied. Each element of the list must follow the naming conventions for the openjpa.jdbc.Schemas property described in Section 12.1, “ Schemas List ”. In fact, if this flag is omitted, it defaults to the value of the Schemas property. If the Schemas property is not defined, all schemas will be reverse-mapped.

    • -package/-pkg <package name>: The package name of the generated classes. If no package name is given, the generated code will not contain package declarations.

    • -directory/-d <output directory>: All generated code and metadata will be written to the directory at this path. If the path does not match the package of a class, the package structure will be created beneath this directory. Defaults to the current directory.

    • -metadata/-md <class | package | none>: Specify the level the metadata should be generated at. Defaults to generating a single package-level metadata file. Set to none to disable orm.xml generation.

    • -annotations/-ann <true/t | false/f>: Set to true to generate JPA annotations in generated Java classes.

    • -accessType/-access <field | property>: Change access type for generated annotations. Defaults to field access.

    • -useSchemaName/-sn <true/t | false/f>: Set this flag to true to include the schema as well as table name in the name of each generated class. This can be useful when dealing with multiple schemas with same-named tables.

    • -useForeignKeyName/-fkn <true/t | false/f>: Set this flag to true if you would like field names for relations to be based on the database foreign key name. By default, relation field names are derived from the name of the related class.

    • -nullableAsObject/-no <true/t | false/f>: By default, all non-foreign key columns are mapped to primitives. Set this flag to true to generate primitive wrapper fields instead for columns that allow null values.

    • -blobAsObject/-bo <true/t | false/f>: By default, all binary columns are mapped to byte[] fields. Set this flag to true to map them to Object fields instead. Note that when mapped this way, the column is presumed to contain a serialized Java object.

    • -primaryKeyOnJoin/-pkj <true/t | false/f>: The standard reverse mapping tool behavior is to map all tables with primary keys to persistent classes. If your schema has primary keys on many-many join tables as well, set this flag to true to avoid creating classes for those tables.

    • -inverseRelations/-ir <true/t | false/f>: Set to false to prevent the creation of inverse 1-many/1-1 relations for every many-1/1-1 relation detected.

    • -useGenericCollections/-gc <true/t | false/f>: Set to true to use generic collections on OneToMany and ManyToMany relations.

    • -useDatastoreIdentity/-ds <true/t | false/f>: Set to true to use datastore identity for tables that have single numeric primary key columns. The tool typically uses application identity for all generated classes.

    • -useBuiltinIdentityClass/-bic <true/t | false/f>: Set to false to prevent the tool from using built-in application identity classes when possible. This will force the tool to create custom application identity classes even when there is only one primary key column.

    • -innerIdentityClasses/-inn <true/t | false/f>: Set to true to have any generated application identity classes be created as static inner classes within the persistent classes. Defaults to false.

    • -identityClassSuffix/-is <suffix>: Suffix to append to class names to form application identity class names, or for inner identity classes, the inner class name. Defaults to Id.

    • -typeMap/-typ <type mapping>: A string that specifies the default Java classes to generate for each SQL type that is seen in the schema. The format is SQLTYPE1=JavaClass1,SQLTYPE2=JavaClass2 . The SQL type name first looks for a customization based on SQLTYPE(SIZE,PRECISION), then SQLTYPE(SIZE), then SQLTYPE. So if a column whose type name is CHAR is found, it will first look for the CHAR(50,0) type name specification, then it will look for CHAR(50), and finally it will just look for CHAR. For example, to generate a char array for every CHAR column whose size is exactly 50, and to generate a short for every type name of INTEGER, you might specify: CHAR(50)=char[],INTEGER=short. Note that since various databases report different type names differently, one database's type name specification might not work for another database. Enable TRACE level logging on the MetaData channel to track which type names OpenJPA is examining.

    • -customizerClass/-cc <class name>: The full class name of a org.apache.openjpa.jdbc.meta.ReverseCustomizer customization plugin. If you do not specify a reverse customizer of your own, the system defaults to a PropertiesReverseCustomizer. This customizer allows you to specify simple customization options in the properties file given with the -customizerProperties flag below. We present the available property keys below.

    • -customizerProperties/-cp <properties file or resource> : The path or resource name of a properties file to pass to the reverse customizer on initialization.

    • -customizer./-c.<property name> <property value> : The given property name will be matched with the corresponding Java bean property in the specified reverse customizer, and set to the given value.

    Running the tool will generate .java files for each generated class (and its application identity class, if applicable), along with JPA annotations (if enabled by setting -annotations true), or an orm.xml file (if not disabled with -metadata none) containing the corresponding persistence metadata.

  4. Examine the generated class, metadata, and mapping information, and modify it as necessary. Remember that the reverse mapping tool only provides a starting point, and you are free to make whatever modifications you like to the code it generates.

    After you are satisfied with the generated classes and their mappings, you should first compile the classes with javac, jikes, or your favorite Java compiler. Make sure the classes are located in the directory corresponding to the -package flag you gave the reverse mapping tool. Next, if you have generated an orm.xml, move that file to a META-INF directory within a directory in your classpath. Finally, enhance the classes if necessary (see Section 2, “ Enhancement ”).

Your persistent classes are now ready to access your existing schema.

2.1.  Customizing Reverse Mapping

The org.apache.openjpa.jdbc.meta.ReverseCustomizer plugin interface allows you to customize the reverse mapping process. See the class Javadoc for details on the hooks that this interface provides. Specify the concrete plugin implementation to use with the -customizerClass/-cc command-line flag, described in the preceding section.

By default, the reverse mapping tool uses a org.apache.openjpa.jdbc.meta.PropertiesReverseCustomizer . This customizer allows you to perform relatively simple customizations through the properties file named with the -customizerProperties tool flag. The customizer recognizes the following properties:

  • <table name>.table-type <type>: Override the default type of the table with name <table name>. Legal values are:

    • base: Primary table for a base class.

    • secondary: Secondary table for a class. The table must have a foreign key joining to a class table.

    • secondary-outer: Outer-joined secondary table for a class. The table must have a foreign key joining to a class table.

    • association: Association table. The table must have two foreign keys to class tables.

    • collection: Collection table. The table must have one foreign key to a class table and one data column.

    • subclass: A joined subclass table. The table must have a foreign key to the superclass' table.

    • none: The table should not be reverse-mapped.

  • <class name>.rename <new class name>: Override the given tool-generated name <class name> with a new value. Use full class names, including package. You are free to rename a class to a new package. Specify a value of none to reject the class and leave the corresponding table unmapped.

  • <table name>.class-name <new class name>: Assign the given fully-qualified class name to the type created from the table with name <table name>. Use a value of none to prevent reverse mapping this table. This property can be used in place of the rename property.

  • <class name>.identity <datastore | builtin | identity class name>: Set this property to datastore to use datastore identity for the class <class name>, builtin to use a built-in identity class, or the desired application identity class name. Give full class names, including package. You are free to change the package of the identity class this way. If the persistent class has been renamed, use the new class name for this property key. Remember that datastore identity requires a table with a single numeric primary key column, and built-in identity requires a single primary key column of any type.

  • <class name>.<field name>.rename <new field name> : Override the tool-generated <field name> in class <class name> with the given name. Use the field owner's full class name in the property key. If the field owner's class was renamed, use the new class name. The property value should be the new field name, without the preceding class name. Use a value of none to reject the generated mapping and remove the field from the class.

  • <table name>.<column name>.field-name <new field name>: Set the generated field name for the <table name> table's <column name> column. If this is a multi-column mapping, any of the columns can be used. Use a value of none to prevent the column and its associated columns from being reverse-mapped.

  • <class name>.<field name>.type <field type> : The type to give the named field. Use full class names. If the field or the field's owner class has been renamed, use the new name.

  • <class name>.<field name>.value: The initial value for the named field. The given string will be placed as-is in the generated Java code, so be sure it is valid Java. If the field or the field's owner class has been renamed, use the new name.

All property keys are optional; if not specified, the customizer keeps the default value generated by the reverse mapping tool.

Example 7.10.  Customizing Reverse Mapping with Properties

java org.apache.openjpa.jdbc.meta.ReverseMappingTool -pkg com.xyz -cp custom.properties schema.xml

Example custom.properties:

com.xyz.TblMagazine.rename:             com.xyz.Magazine
com.xyz.TblArticle.rename:              com.xyz.Article
com.xyz.TblPubCompany.rename:           com.xyz.pub.Company
com.xyz.TblSysInfo.rename:              none

com.xyz.Magazine.allArticles.rename:    articles
com.xyz.Magazine.articles.type:         java.util.Collection
com.xyz.Magazine.articles.value:        new TreeSet()
com.xyz.Magazine.identity:              datastore

com.xyz.pub.Company.identity:           com.xyz.pub.CompanyId