Try OpenEdge Now
skip to main content
ABL Reference
ABL Syntax Reference : CREATE statement

CREATE statement

Creates a record in a table, sets all the fields in the record to their default initial values, and moves a copy of the record to the record buffer.

Data movement


CREATE record [ FOR TENANT tenant-expression ]
[ USING { ROWID ( nrow ) | RECID ( nrec ) } ] [ NO-ERROR ]
The name of the record or record buffer you are creating.
To create a record in a table defined for multiple databases, you might have to qualify the record's table name with the database name. See the Record phrase reference entry for more information.
FOR TENANT tenant-expression
This option is useful only for a multi-tenant database, and primarily one with a connection identity that has super tenant access. If the user has a super-tenant connection identity and you do not specify this option, the record you create is owned by the effective tenant. If you do specify this option, you create a record owned by the regular tenant identified by tenant-expression.
If the user has a regular-tenant connection identity, and you specify this option, tenant-expression must match the tenancy of the connection identity. Otherwise, the statement raises ERROR.
If tenant-expression evaluates to an integer, the value must be a valid tenant ID for a regular tenant or zero (0) for the default tenant. If tenant-expression evaluates to a character string, the value must be a valid tenant name for a regular or "Default" for the default tenant. Otherwise, the statement raises ERROR.
If record belongs to a table that is not multi-tenant enabled, ABL raises a compiler error.
USING { ROWID ( nrow ) | RECID ( nrec ) }
Supported only for backward compatibility.
Suppresses ABL errors or error messages that would otherwise occur and diverts them to the ERROR-STATUS system handle. If an error occurs, the action of the statement is not done and execution continues with the next statement. If the statement fails, any persistent side-effects of the statement are backed out. If the statement includes an expression that contains other executable elements, like methods, the work performed by these elements may or may not be done, depending on the order the AVM resolves the expression elements and the occurrence of the error.
To check for errors after a statement that uses the NO-ERROR option:
*Check the ERROR-STATUS:ERROR attribute to see if the AVM raised the ERROR condition.
*Check if the ERROR-STATUS:NUM-MESSAGES attribute is greater than zero to see if the AVM generated error messages. ABL handle methods used in a block without a CATCH end block treat errors as warnings and do not raise ERROR, do not set the ERROR-STATUS:ERROR attribute, but do add messages to the ERROR-STATUS system handle. Therefore, this test is the better test for code using handle methods without CATCH end blocks. ABL handle methods used in a block with a CATCH end block raise ERROR and add messages to the error object generated by the AVM. In this case, the AVM does not update the ERROR-STATUS system handle.
*Use ERROR-STATUS:GET-MESSAGE( message-num ) to retrieve a particular message, where message-num is 1 for the first message.
If the statement does not include the NO-ERROR option, you can use a CATCH end block to handle errors raised by the statement.
Some other important usage notes on the NO-ERROR option:
*NO-ERROR does not suppress errors that raise the STOP or QUIT condition.
*A CATCH statement, which introduces a CATCH end block, is analogous to a NO-ERROR option in that it also suppresses errors, but it does so for an entire block of code. It is different in that the error messages are contained in a class-based error object (generated by the AVM or explicitly thrown), as opposed to the ERROR-STATUS system handle. Also, if errors raised in the block are not handled by a compatible CATCH block, ON ERROR phrase, or UNDO statement, then the error is not suppressed, but handled with the default error processing for that block type.
*When a statement contains the NO-ERROR option and resides in a block with a CATCH end block, the NO-ERROR option takes precedence over the CATCH block. That is, an error raised on the statement with the NO-ERROR option will not be handled by a compatible CATCH end block. The error is redirected to the ERROR-STATUS system handle as normal.
*If an error object is thrown to a statement that includes the NO-ERROR option, then the information and messages in the error object will be used to set the ERROR-STATUS system handle. This interoperability feature is important for those integrating code that uses the traditional NO-ERROR technique with the newer, structured error handling that features error objects and CATCH end blocks.


The following example creates a record in the order file for each pass through the loop and then updates the record. It also creates an order-line record.
  CREATE Order.
  UPDATE Order.OrderNum Order.CustNum
    VALIDATE(CAN-FIND(Customer OF Order), "Customer does not exist")
    Order.CustNum Order.OrderDate.
    CREATE OrderLine.
    OrderLine.OrderNum = Order.OrderNnum.
    UPDATE OrderLine.LineNum OrderLine.ItemNum
      VALIDATE(CAN-FIND(Item OF OrderLine), "Item does not exist")
      OrderLine.Qty OrderLine.Price.
This procedure adds Orders and OrderLines to the database. Because the user supplies an order number when updating the order record, that order number is assigned (=) to the OrderNum field of the OrderLine record when the OrderLine record is created.


*When you run procedures that create large numbers of records (for example, during initial data loading), the process runs much faster if you use the No Crash Protection (-i) parameter. See OpenEdge Deployment: Startup Command and Parameter Reference for more information on startup parameters. Back up your database before you use this parameter.
*After you create a new record with CREATE, the AVM waits to write the record to the database. The creation often happens after one of the following:
*The AVM is about to release the buffer, which can happen as a result of an explicit RELEASE or VALIDATE statement, the reading of a new record into the buffer, the record going out of scope, or the end of the transaction.
*You set a LOB field.
*You set a field that an index is based on.
*All partition fields must be set by the time a record is created in a partitioned database.
Suppose you have an Order table partitioned on OrderDate, but with a global index on OrderNum. If OrderDate has an initial value of the Unknown value (?), the following code will produce an error:
OrderNum = NEXT-VALUE(Next-Ord-Num).
OrderDate = TODAY.
After the assignment of the OrderNum field, the AVM will try to index the new OrderNum in the OrderNum global index. To do that, it needs the partition of the new Order record. To get the partition, the AVM needs to use the partition column OrderDate, which is still the Unknown value (?). This will produce an error message stating that OrderDate is not in any defined partition. You can fix this by reversing the order of the assignments, grouping the statements into one ASSIGN statement, or setting an initial value for the OrderDate column. If you use one ASSIGN statement, the actual create of the record won't occur until after all the specified field assignments are gathered. Therefore the OrderDate column will already have the correct value by the time the partition for the new record needs to be determined.
Note: Even if all of the partition fields have initial values, an application might be inefficient if the underlying record creation happens before the partition fields are set to something other than the initial values. In this case, the AVM may have to move the record to a new partition once the initial value of the partition field is replaced.
*The CREATE statement causes any related database CREATE triggers to execute. All CREATE triggers execute after the record is actually created. If a CREATE trigger fails (or executes a RETURN statement with the ERROR option), the record creation is undone.
*When specifying the FOR TENANT option, the AVM looks up tenant-expression in the database with a share lock. The AVM waits 60 seconds to get the share lock and raises ERROR if it fails to obtain the share lock in that amount of time. The AVM releases the share lock immediately after successfully fetching the row. This share lock is released even if the statement is called while in the scope of a transaction.

See also

BUFFER-CREATE( ) method, INSERT statement, NEW function (record buffers)