Solution 3: Use native SEQUENCE database objects
Principle
Most recent database engines support SEQUENCE
database objects. If all database
server types you want to use support sequences, it is recommended that you use this solution.
Database Server Type | Supports sequences |
---|---|
IBM® Informix® | Yes |
Microsoft™ SQL Server | Yes |
Oracle® MySQL | No, see details. |
MariadDB | Yes |
Oracle Database Server | Yes |
PostgreSQL | Yes |
SAP HANA® | Yes |
SQLite | No, see details. |
Dameng® | Yes |
Implementation
- Create a
SEQUENCE
object for each table that previously used aSERIAL
column in the IBM Informix database. - In database creation scripts (
CREATE TABLE
), replace allSERIAL
types byINTEGER
(orBIGINT
if you need large integers). - Adapt your programs to retrieve a new sequence before inserting a new row. Consider writing a function to retrieve a new sequence number, using dynamic SQL to pass the name of the sequence as parameter, and adapt to the target database specifics to retrieve a single row (see example below).
For databases not supporting SEQUENCE
objects, the code uses SELECT
MAX(col)+1
to get a new number, assuming that it's a database like
SQLite used by a single process, where duplicates cannot occur. In a multi-user/process case, for
databases that do not support sequences, consider using the user-made sequence table solution.
Example
MAIN
DEFINE item_rec RECORD
item_num BIGINT,
item_name VARCHAR(40)
END RECORD
DEFINE x INT
DATABASE test1
CREATE TABLE item (
item_num BIGINT NOT NULL PRIMARY KEY,
item_name VARCHAR(50)
)
CALL sequence_create("item","item_num")
FOR x=1 TO 5
LET item_rec.item_num = sequence_next("item","item_num")
DISPLAY "New sequence: ", item_rec.item_num
LET item_rec.item_name = "Item#" || item_rec.item_num
INSERT INTO item VALUES ( item_rec.* )
END FOR
DROP TABLE item
CALL sequence_drop("item")
END MAIN
PRIVATE FUNCTION _supports_sequences() RETURNS BOOLEAN
DEFINE dt STRING
LET dt = fgl_db_driver_type()
RETURN ( dt!="sqt" AND dt!="mys" )
END FUNCTION
PRIVATE FUNCTION _get_next_pkey(tabname STRING, colname STRING) RETURNS BIGINT
DEFINE sql STRING
DEFINE val BIGINT
LET sql = SFMT("SELECT MAX(%1.%2)+1 FROM %1",tabname,colname)
PREPARE stmt1 FROM sql
EXECUTE stmt1 INTO val
RETURN NVL(val,1)
END FUNCTION
PUBLIC FUNCTION sequence_create(tabname STRING, colname STRING)
DEFINE val BIGINT
IF NOT _supports_sequences() THEN RETURN END IF
LET val = _get_next_pkey(tabname,colname)
EXECUTE IMMEDIATE SFMT("CREATE SEQUENCE %1_seq START WITH %2",tabname,val)
END FUNCTION
PUBLIC FUNCTION sequence_drop(tabname STRING)
IF NOT _supports_sequences() THEN RETURN END IF
EXECUTE IMMEDIATE SFMT("DROP SEQUENCE %1_seq",tabname)
END FUNCTION
PRIVATE FUNCTION _unique_row_condition() RETURNS STRING
CASE fgl_db_driver_type()
WHEN "ifx" RETURN " FROM systables WHERE tabid=1"
WHEN "db2" RETURN " FROM sysibm.systables WHERE name='SYSTABLES'"
WHEN "pgs" RETURN " FROM pg_class WHERE relname='pg_class'"
WHEN "ora" RETURN " FROM dual"
WHEN "hdb" RETURN " FROM dummy"
OTHERWISE RETURN " "
END CASE
END FUNCTION
PUBLIC FUNCTION sequence_next(tabname STRING, colname STRING) RETURNS BIGINT
DEFINE sql STRING
DEFINE seqname STRING
DEFINE val BIGINT
LET seqname = tabname||"_seq"
CASE
WHEN NOT _supports_sequences()
-- WARNING: Duplicates can be created in multi-user databases, must
-- be controlled with PRIMARY KEY or UNIQUE constraint!
RETURN _get_next_pkey(tabname,colname)
WHEN fgl_db_driver_type()=="pgs"
LET sql = SFMT("SELECT nextval('%1') %2",seqname,_unique_row_condition())
WHEN fgl_db_driver_type()=="esm"
OR fgl_db_driver_type()=="ftm"
OR fgl_db_driver_type()=="snc"
OR fgl_db_driver_type()=="ntz"
LET sql = SFMT("SELECT NEXT VALUE FOR %1",seqname)
OTHERWISE
LET sql = SFMT("SELECT %1.nextval %2",seqname,_unique_row_condition())
END CASE
TRY
PREPARE seq FROM sql
EXECUTE seq INTO val
CATCH
DISPLAY SFMT("ERROR: Could not produce new sequence for %1", tabname)
EXIT PROGRAM 1
RETURN NULL
END TRY
RETURN val
END FUNCTION