Reputation: 1123
I'm trying to use sequences to generate incremented IDs for my tables in DB2. It works when I send SQL statements directly to the database, but when using ebean the statement fails. Here's the field in Java:
@Id
@GeneratedValue(strategy = GenerationType.SEQUENCE, generator = "TABLENAME_IDNAME_TRIG")
@SequenceGenerator(name = "TABLENAME_IDNAME_TRIG", sequenceName = "TABLENAME_IDNAME_SEQ")
@Column(name = "IDNAME")
private Long id;
Here's the column in SQL (From TOAD):
Name Data type Not Null Default Generated Bit Data Scope Identity
IDNAME INTEGER Yes No No
And here's the sequence definition in SQL:
CREATE OR REPLACE SEQUENCE SCHEMA.TABLENAME_IDNAME_SEQ
AS INTEGER CACHE 50 ORDER;
And the trigger:
CREATE OR REPLACE TRIGGER SCHEMA.TABLENAME_IDNAME_TRIG
NO CASCADE BEFORE INSERT
ON TABLENAME
REFERENCING
NEW AS OBJ
FOR EACH ROW
BEGIN
SET obj.IDNAME=NEXT VALUE FOR SCHEMA.TABLENAME_IDNAME_SEQ;
END;
What is the issue with my annotations here? As a(n important) side note - when I set GenerationType to AUTO, TABLE, or IDENTITY, it works, even though it shouldn't, because I'm also using this object to represent a parallel oracle table which also uses sequences for ID generation.
Edited to include error message:
javax.persistence.PersistenceException: Error getting sequence nextval
...
Caused by: com.ibm.db2.jcc.am.SqlSyntaxErrorException: DB2 SQL Error: SQLCODE=-348, SQLSTATE=428F9, SQLERRMC=NEXTVAL FOR SCHEMA.TABLENAME_IDNAME_SEQ, DRIVER=4.19.49
EDIT 2: The specific Sql statement that is failing is:
values nextval for QA_CONNECTION_ICONNECTIONI_SEQ union values nextval for QA_CONNECTION_ICONNECTIONI_SEQ union values nextval for QA_CONNECTION_ICONNECTIONI_SEQ
Which is SQL generated by Ebean. This is a smaller version of the real statement, which is repeated 20 times, so I'm guessing something screws up when generating the caching query.
EDIT 3: I believe this might be a bug in Ebean's use of DB2 sequences. This function generates SQl that returns an error for me when used with db2
public DB2SequenceIdGenerator(BackgroundExecutor be, DataSource ds, String seqName, int batchSize) {
super(be, ds, seqName, batchSize);
this.baseSql = "values nextval for " + seqName;
this.unionBaseSql = " union " + baseSql;
}
EDIT 4: Based on this SO link I think it is a bug. Can't insert multiple values into DB2 by using UNION ALL and generate IDs from sequence The correct class probably looks like this? Though I haven't ever tried building the library, so I couldn't test it. Time to learn how to open a defect I guess.
public class DB2SequenceIdGenerator extends SequenceIdGenerator {
private final String baseSql;
private final String unionBaseSql;
private final String startSql;
public DB2SequenceIdGenerator(BackgroundExecutor be, DataSource ds, String seqName, int batchSize) {
super(be, ds, seqName, batchSize);
this.startSql = "values "
this.baseSql = "(nextval for " + seqName);
this.unionBaseSql = ", " + baseSql;
}
public String getSql(int batchSize) {
StringBuilder sb = new StringBuilder();
sb.append(startSql);
sb.append(baseSql);
for (int i = 1; i < batchSize; i++) {
sb.append(unionBaseSql);
}
return sb.toString();
}
}
Upvotes: 1
Views: 769
Reputation: 1123
Temporary workaround for those interested: in ebean.properties, set
ebean.databaseSequenceBatchSize=1
Upvotes: 1