在toplink中映射nested table
nested table是Oracle所特有的一种实现，作为被Oracle收购了的OR-Mapping工具Toplink，自然也应该对其提供支持。但是在寻找这个的过程中，又一次深深体会到了Toplink文档的稀少（OTN，也只有OTN）以及语焉不详（我就不信有人看着官方文档能调出来的）。为以后起见，这里记录一下吧。
?
假设有以下的nested table:
?
CREATE OR REPLACE TYPE STUDENT_TYPE AS OBJECT( STUDENT_ID NUMBER(4), STUDENT_NAME VARCHAR2(10));CREATE OR REPLACE TYPE STUDENT_LIST IS TABLE OF STUDENT_TYPE;CREATE TABLE SCHOOL(SCHOOL_ID NUMBER(2) primary key, SCHOOL_NAME VARCHAR2(14), STUDENTS STUDENT_LIST) NESTED TABLE STUDENTS store AS STUDENTS_TAB;ALTER TABLE STUDENTS_TAB add CONSTRAINT UK_STUDENTS_TAB UNIQUE(STUDENT_ID);
?
和以下的JO
public class School { private int id; private String name; private List<Student> students; //skip getter, setter}?
public class Student { private int id; private String name; // skip getter,setter }?
那么相应的Descriptor应该这么写
?
?
public ClassDescriptor buildSchoolDescriptor() { ObjectRelationalDescriptor descriptor = new ObjectRelationalDescriptor(); descriptor.setJavaClass(School.class); descriptor.addTableName("SCHOOL"); descriptor.addPrimaryKeyFieldName("SCHOOL.SCHOOL_ID"); DirectToFieldMapping idMapping = new DirectToFieldMapping(); idMapping.setAttributeName("id"); idMapping.setFieldName("SCHOOL_ID"); descriptor.addMapping(idMapping); DirectToFieldMapping nameMapping = new DirectToFieldMapping(); nameMapping.setAttributeName("name"); nameMapping.setFieldName("SCHOOL.SCHOOL_NAME"); descriptor.addMapping(nameMapping); ObjectArrayMapping studentsMapping = new ObjectArrayMapping(); studentsMapping.setReferenceClass(Student.class); studentsMapping.setAttributeName("students"); studentsMapping.setFieldName("STUDENTS"); studentsMapping.setStructureName("STUDENT_LIST"); descriptor.addMapping(studentsMapping); return descriptor; }?
public ClassDescriptor buildStudentDescriptor() { ObjectRelationalDescriptor descriptor = new ObjectRelationalDescriptor(); descriptor.setJavaClass(Student.class); descriptor.descriptorIsAggregate(); descriptor.setStructureName("STUDENT_TYPE"); descriptor.addPrimaryKeyFieldName("id"); descriptor.addFieldOrdering("id"); descriptor.addFieldOrdering("name"); descriptor.addDirectMapping("id", "id"); descriptor.addDirectMapping("name", "name"); return descriptor; }?
相应的测试代码如下
Server server = project.createServerSession(); server.login(); UnitOfWork uow = server.acquireUnitOfWork(); ClientSession cs = server.acquireClientSession(); server.shouldLogMessages(); uow.shouldLogMessages(); uow.setLogLevel(1); School school = new School(); school.setId(1); school=(School)uow.readObject(school); System.out.println(school.getStudents().get(0).getId()); school.getStudents().get(0).setName("NN"); uow.commit();?
?
在这个问题上我花了一整天时间，主要是在官方文档这里完全没有把和核心讲出来。
1，正如论坛jsuther所言，
?
In TopLink if the nested table is of refs, then you use the NestedTableMapping (target class is not aggregate), if it is a nested table or varray of object types, then you use the ObjectArrayMapping (target class is aggregate).
也就是说，所谓的NestedTableMapping，居然只对ref nested table有效，普通的nested table应该用ObjectArrayMapping， 这是何等跳跃的思维啊！
?
2 在Student Descriptor中，一定要有以下两句，不然会报ora-17049 "Inconsistent java and sql object types"
descriptor.addFieldOrdering("id");descriptor.addFieldOrdering("name");?
3.还是在Student Descriptor中，添加DirectMapping时一定要写成以下方式，
descriptor.addDirectMapping("id", "id"); descriptor.addDirectMapping("name", "name");??? 切忌写成以下这样
descriptor.addDirectMapping("id", "STUDENT_ID");?? 这里我没有去看源码，既然addDirectMapping添加的是一个哑元（不要求指定数据库Column name） 又要求指定fieldOrdering，估计在组装对象时是按顺序来的，也就是把STUDENT_TYPE中出现的属性依照addFieldOrdering依次塞给Student上的属性。为了验证这个猜想，把addFieldOrdering的顺序颠倒一下变成
descriptor.addFieldOrdering("name");descriptor.addFieldOrdering("id");?
果然报了java.lang.NumberFormatException: For input string: "NN"，验证了我的猜想