The STOP-AFTER phrase specifies a time-out value for a DO, FOR, or REPEAT block. This is the syntax:
Syntax
STOP-AFTER expression
The integer expression specifies the number of seconds each iteration of a block has until a time-out occurs. If a time-out occurs, the AVM raises the STOP condition and default STOP condition handling occurs. Use an ON STOP phrase on the block (or an enclosing block) to alter the default STOP condition handling.
If the block iteration completes before the specified time expires, the timer resets to expression for the next iteration. In other words, the timer is limited to the scope of a single block iteration. If a block with a STOP-AFTER phrase encloses another block or calls another block, the timer continues while the inner blocks execute.
If a block with a STOP-AFTER phrase contains a nested block with a STOP-AFTER phrase, then each has a timer in effect. If the outer block timer expires while the inner block is executing, the STOP condition is raised even if the timer for the inner block has not expired.
If the STOP condition is handled and execution resumes within the scope of a block with a STOP-AFTER phrase, no timer is in effect until the next iteration of a block with a STOP-AFTER phrase. In other words, all old timers are dismissed but new timers can now be established.
When the timer expires, the STOP condition is raised on the current statement.
Two important use cases for the STOP-AFTER phrase are to time-limit dynamic queries and to time-limit a procedure call. The following example time-limits a procedure call using a RUN statement:
Use this technique to also make timed calls to class methods and user-defined functions.
The following example is simplified code that lets you try different STOP-AFTER cases.
DEFINE VARIABLE EndlessCount AS INTEGER INITIAL 0.
DO STOP-AFTER 5 ON STOP UNDO, LEAVE:
FOR EACH Customer STOP-AFTER 1:
ASSIGN EndlessCount = EndlessCount + 1.
/* Try a complex operation on a Customer record to use up the timer
in a single iteration and raise the STOP condition in the inner
block */
END.
If you run this code as is, the outer DO block establishes a 5 second time limit for the work of the DO block and all inner blocks. When the inner FOR EACH block starts, another timer is established for the first iteration of this block. When the first FOR EACH iteration completes, its timer is reset to 1 second for the next iteration. Meanwhile, the outer timer on the DO block continues without interruption.
The FOR EACH block completes and execution continues forward to the REPEAT block, which is an endless loop. The REPEAT block also has a 1 second timer for each iteration of the block. At some point, the outer 5 second timer elapses and the AVM raises the STOP condition. The STOP condition is raised on the statement the AVM was executing when the timer elapsed. Normal STOP handling proceeds from that point.
As the stack unwinds during STOP processing, the AVM encounters the ON STOP phrase on the DO block. The ON STOP phrase dismisses the STOP condition and resumes normal execution with the next statement following the DO block, as directed by the LEAVE option.
If you remove the comments from the IF statement in the REPEAT block, the block will complete within the outer time limit and the STOP condition is not raised.
If you want to experiment with elapsed timers on an inner block, insert a complex operation inside the FOR EACH block.
In the following example, the STOP-AFTER expression is modified during program execution:
DEFINE VARIABLE ix AS INTEGER NO-UNDO.
DEFINE VARIABLE stopTime AS INTEGER NO-UNDO INITIAL 30.
DO WHILE TRUE STOP-AFTER stopTime ON STOP UNDO, LEAVE:
RUN spinHere (10000).
stopTime = stopTime / 2.
END.
MESSAGE "program finished".
PROCEDURE spinHere:
DEFINE INPUT PARAMETER spinLimit AS INT64 NO-UNDO.
DEFINE VARIABLE endTime AS INT64 NO-UNDO.
DEFINE VARIABLE loopFlag AS LOGICAL NO-UNDO.
DO WHILE loopFlag:
IF (ETIME(FALSE) > endTime) THEN
loopFlag = FALSE.
END.
END PROCEDURE.
Because the STOP-AFTER expression is re-evaluated for each iteration of a looping block, any changes made to the expression during the iteration effect the timer for the block. In the example, the STOP-AFTER time limit is specified by the variable stopTime, which is initially set to 30 seconds. The procedure contains an iterating block which runs a procedure that executes for 10 seconds.
On the first iteration of the DO WHILE TRUE loop, stopTime is 30 seconds. The loop executes for 10 seconds, and then divides stopTime by 2. On the second iteration, the stopTime is 15 seconds; again the loop executes for 10 seconds, and then divides stopTime by 2. On the third iteration, the stopTime is 8 seconds. This time, the procedure spinHere runs for 8 seconds and then raises STOP. The STOP condition is handled by the DO block, and then the program displays the message program finished.
If a code block is called with a time limit of zero, the block is executed as if the STOP-AFTER phrase was omitted from the block declaration.
Consider the following example:
DEFINE VARIABLE barLimit AS INTEGER NO-UNDO.
DEFINE VARIABLE ix AS INTEGER NO-UNDO INITIAL 1.
DO STOP-AFTER 10 ON STOP UNDO, LEAVE:
RUN foo.
END.
PROCEDURE foo:
RUN bar.
END PROCEDURE.
PROCEDURE bar:
DO WHILE ix > 0 STOP-AFTER barLimit:
ix = ix + 1.
END.
END PROCEDURE.
In this example, procedure foo is run from within a timed block with a 10 second time limit; procedure bar is called from within the timed block, and contains an iterating block that specifies the STOP-AFTER phrase. Because the value of the STOP-AFTER expression evaluates to zero (that is, the current value of the barLimit variable), the block within bar is executed as an untimed block. However, the rules for execution of an untimed block within a timed block apply, so the untimed block in bar is executed with an implicit iteration time limit of 10 seconds.
Other points to consider are:
If the expression evaluates to zero or less, then this is the equivalent of not specifying a STOP-AFTER phrase.
STOP-AFTER phrases are not intended to interact with user interfaces.
Blocking calls to third party software components, where the AVM has transferred execution control, cannot be timed out. This category includes operating system calls, MS Windows system calls, and calls to any third party DLLs and Unix shared objects.
If the expression evaluates to zero or less, then this is the equivalent of not specifying a STOP-AFTER phrase.