Database restrictions for EJB query
The EJB query functions must adhere to certain restrictions for databases.
General database restriction
- All of the enterprise beans involved in a given query must map to the same data source. The EJB query does not support cross-data source join operations.
- It is possible that a SQL statement generated by the WAS deployment code generation utility for an ejbSelect EJB query language query returns rows in a result set that consist of null values in all columns.
During run time persistence manager saves the set received as a result from this query. When the application retrieves the primary key of the result bean, persistence manager calls the extractor. The extractor is a method that is an EJB deploy generated class. This method returns a value of 0 for any null column entries. This value is passed back to the EJB container to forward to the application. The EJB container invokes the bean instance with the PK value of 0. This could create a problem, as the end user cannot determine if this bean instance has a null PK or a PK value of 0.
To avoid this, use the IS NOT NULL clause in the finder query to eliminate such null values from the result set.
Specific database restrictions
Different database products place different restrictions on elements that can be included in EJB query statements. Following is a list of those restrictions; check with the database administrator to see if any apply in the environment:
- Certain functions are used in queries that run against DB2 only, because these functions are not supported by other databases. These functions include date and time arithmetic expressions, certain scalar functions including those not listed as portable across vendors, and implied scalar functions when used for mapping certain container managed persistence (CMP) fields. For example, consider mapping an int numeric type to a decimal (5,2) type field. When deployed against a database other than DB2, a finder or select query containing a CMP field with this particular mapping fails, producing a Cannot push down query error message.
- A CMP of type String, when mapped to a character large object (CLOB) in the database, cannot be used in comparison operations because the database does not support CLOB comparisons.
- Databases can impose limits on the length of string values used either as literals or input parameters with comparison operators. These limits can hinder query performance. For example: For DB2 on the z/OS platform, the search "name = ?1" can fail if the value of ?1 at run time is greater than 255 in length.
- Mapping a numeric CMP type to a column containing a dissimilar type can cause unexpected results. Consider the case of mapping the int numeric type to a column of type decimal (5,2). This scenario does not preserve an exact decimal value (for example, the value 12.25) over the course of transfer from the database to the enterprise bean CMP field, and back again to the database. This mapping causes replacement of the initial value with a whole number (in this case, 12). Consequently, to avoid using the CMP field in comparison operations when the CMP field uses a mapping of this nature.
- Some databases do not support a data type that corresponds to the semantics of java.sql.Time. For example: If a CMP field of type java.sql.Time is mapped to an Oracle DATE column, comparisons on time might not produce the expected result because the year-month-day portion of the column value is truncated in the mapping.
- Some databases treat a zero length string value ( '' ) as a null value; this approach can affect the query results. For the sake of portability, avoid the use of zero length string values.
- Some databases perform division between two integer values using integer arithmetic rules, while others use non-integer rules. This discrepancy might not be desirable in environments that use both kinds of databases. For the sake of portability, avoid the division of integer values in an EJB query.
- (iseries) Current releases of UDB DB2 for i5/OS i only support a TIMESTAMP value of the format "yyyy-mm-dd-hh.mm.ss.nnnnnn". This is not compatible with the standard format supported by the java.sql.Timestamp class, which is "yyyy-mm-dd-hh mm.ss.nnnnnn". The TIMESTAMP scalar function should be used to convert a string representation of a java.sql.Timestamp object to a value that can be recognized by DB2 UDB for IBM i.
Related concepts
EJB query language EJB query compatibility issues with SQL Rules for data type manipulation in EJB query
Related tasks
Use EJB query