Procedure scope

Stored procedures can include handlers that are called when certain conditions occur within the procedure. The applicability of each handler depends on its position in the procedure definition and the one or more conditions it handles.

Scope and rules for handlers

The scope and rules for handlers in OceanBase Database are as follows:

• Handlers declared within a BEGIN ... END block are only applicable to SQL statements following the handler declaration within that block. If a handler itself raises a condition, it cannot handle that condition and no other handlers can be declared within the same block. For example, in the following code, handlers h1 and h2 are applicable to conditions raised by statements st1 and st2, respectively, but not to conditions raised within the bodies of h1 or h2.

  BEGIN -- outer block
    DECLARE EXIT HANDLER FOR ...;  -- handler h1
    DECLARE EXIT HANDLER FOR ...;  -- handler h2
    st1;
    st2;
  END;
  

• A handler is only effective within the block in which it is declared and cannot be activated by conditions outside that block. For instance, in the following example, handler h1 is effective within the st1 scope of the inner block but not within the st2 scope of the outer block:

  BEGIN -- outer block
    BEGIN -- inner block
      DECLARE EXIT HANDLER FOR ...;  -- handler h1
      st1;
    END;
    st2;
  END;
  

• Handlers can be specific or general. Specific handlers are used for error codes, SQLSTATE values, or condition names. General handlers are used for conditions in the SQLWARNING, SQLEXCEPTION, or NOT FOUND categories. Condition specificity is related to condition priority.

  Multiple handlers with different characteristics can be declared in different scopes. For example, an outer block might have a specific error code handler, while an inner block might have a general SQLWARNING handler. Alternatively, a single block might contain both specific error code handlers and general SQLWARNING handlers.

The activation of a handler depends not only on its scope and condition value but also on other handlers. When a condition occurs in a stored program, the server searches for applicable handlers within the current scope (the current BEGIN ... END block). If no applicable handler is found, it continues searching in the next outer scope (block). When the server finds one or more applicable handlers in the specified scope, it selects based on condition priority. The condition priorities are as follows:

• Error code handlers have higher priority than SQLSTATE value handlers.

• SQLSTATE value handlers have higher priority than general SQLWARNING, SQLEXCEPTION, or NOT FOUND handlers.

• SQLEXCEPTION handlers have higher priority than SQLWARNING handlers.

• Multiple applicable handlers with the same priority can exist. For example, a statement can generate multiple warnings with different error codes, each having a specific error handler. In such cases, the server's choice of which handler to activate is indeterminate and may vary based on the situation.

If multiple applicable handlers exist in different scopes, the handler in the innermost scope takes precedence over those in outer scopes, even if the outer scope handlers have higher priority for the specific condition.

If no applicable handler is found when a condition occurs, the following actions are taken based on the condition's category:

• For SQLEXCEPTION conditions, the stored program terminates at the statement that raised the condition, similar to using an EXIT handler. If the program is called by another stored program, the calling program will apply its own selection rules to handle the condition.

• For SQLWARNING conditions, the program continues execution, similar to using a CONTINUE handler.

• For NOT FOUND conditions, if the condition is raised normally, the action is CONTINUE. If the condition is raised by SIGNAL or RESIGNAL, the action is EXIT.

Examples

Examples 1, 2, 3, and 4 show how the handler selection rules apply.

The following example 1 shows a procedure that contains two handlers: one for the specified SQLSTATE value ('42S02') that occurs when an attempt is made to drop a nonexistent table, and one for the general SQLEXCEPTION category.

-- Example 1
CREATE PROCEDURE proc1()
BEGIN
  DECLARE CONTINUE HANDLER FOR SQLSTATE '42S02'
    SELECT 'SQLSTATE handler was activated' AS msg;
  DECLARE CONTINUE HANDLER FOR SQLEXCEPTION
    SELECT 'SQLEXCEPTION handler was activated' AS msg;

  DROP TABLE test.tbl1;
END;

In the case of example 1, both handlers are declared in the same block and have the same scope. However, the SQLSTATE handler has priority over the SQLEXCEPTION handler, so if the table tbl1 does not exist, the DROP TABLE statement will trigger the condition and activate the SQLSTATE handler. As shown below:

obclient> CALL proc1();
+--------------------------------+
| msg                            |
+--------------------------------+
| SQLSTATE handler was activated |
+--------------------------------+
1 row in set

In the case of example 2, the procedure contains the same two handlers as in example 1. However, the DROP TABLE statement and the SQLEXCEPTION handler are located in an inner block.

-- Example 2
CREATE PROCEDURE proc2()
BEGIN -- outer block
    DECLARE CONTINUE HANDLER FOR SQLSTATE '42S02'
      SELECT 'SQLSTATE handler was activated' AS msg;
  BEGIN -- inner block
    DECLARE CONTINUE HANDLER FOR SQLEXCEPTION
      SELECT 'SQLEXCEPTION handler was activated' AS msg;

    DROP TABLE test.tbl1; -- occurs within the inner block
  END;
END;

In the case of example 2, the handler that is located closer to the point where the condition occurs is activated first, so the SQLEXCEPTION handler is activated, even though it is more general than the SQLSTATE handler. As shown below:

obclient> CALL proc2();
+------------------------------------+
| msg                                |
+------------------------------------+
| SQLEXCEPTION handler was activated |
+------------------------------------+
1 row in set

In the case of example 3, one handler is declared in a block within the scope of the DROP TABLE statement:

-- Example 3
CREATE PROCEDURE proc3()
BEGIN -- outer block
  DECLARE CONTINUE HANDLER FOR SQLEXCEPTION
    SELECT 'SQLEXCEPTION handler was activated' AS msg;
  BEGIN -- inner block
    DECLARE CONTINUE HANDLER FOR SQLSTATE '42S02'
      SELECT 'SQLSTATE handler was activated' AS msg;
  END;

  DROP TABLE test.tbl1; -- occurs within the outer block
END;

In the case of example 3, the SQLEXCEPTION handler is activated because the SQLSTATE condition is not within the scope of the DROP TABLE statement. As shown below:

obclient> CALL proc3();
+------------------------------------+
| msg                                |
+------------------------------------+
| SQLEXCEPTION handler was activated |
+------------------------------------+
1 row in set

In the case of example 4, both handlers are declared in a block within the scope of the DROP TABLE statement:

-- Example 4
CREATE PROCEDURE proc4()
BEGIN -- outer block
  BEGIN -- inner block
    DECLARE CONTINUE HANDLER FOR SQLEXCEPTION
      SELECT 'SQLEXCEPTION handler was activated' AS msg;
    DECLARE CONTINUE HANDLER FOR SQLSTATE '42S02'
      SELECT 'SQLSTATE handler was activated' AS msg;
  END;

  DROP TABLE test.tbl1; -- occurs within the outer block
END;

In the case of example 4, neither handler is applicable because they are not within the scope of the DROP TABLE statement. The condition triggered by the statement is not handled and results in an error that terminates the procedure. As shown below:

obclient> CALL proc4();

ERROR 1051 (42S02): Unknown table 'test.tbl1'