Different plsql for different Oracle versions

Question

Hi everyone,
I am working on a plsql script which may be run on multiple databases. Unfortunately, I have a need to run one query type if the version is 10g, and another if it is 9i. The 10g query is preferable for the type & amount of information it returns, but 9i does not have the same level of support. I wrote the below (dumbed down for posting purposes) to do this:
spool sqlLog.txt append
set heading off;
SET NEWPAGE 0;
SET SPACE 0;
SET LINESIZE 32767;
SET PAGESIZE 0;
--SET ECHO OFF;
SET FEEDBACK OFF;
SET VERIFY OFF;
SET HEADING OFF;
SET MARKUP HTML OFF;
SET TERMOUT OFF;
SET TRIMOUT ON;
SET TRIMSPOOL ON;
SET WRAP OFF ;
SET LONG 4000;
SET LONGCHUNKSIZE 500;
set serverout on;
set serveroutput on;
declare     
ver VARCHAR2(64);
BEGIN
select v.version into ver from product_component_version v where product like '%Oracle%';
ver:=substr(ver,0,2); --remove anything but the highest version number, a.k.a. 9 or 10.
dbms_output.put_line('ver: '||ver);
--10g
IF to_number(ver)>=10
THEN
     for csr_10g in (SELECT sql_fulltext, a.parsing_schema_name, b.name, b.value_string
                         FROM V$SQL a left outer join v$sql_bind_capture b ON a.SQL_ID = b.SQL_ID)
          loop
            dbms_output.put_line(csr_10g.sql_fulltext||';'|| csr_10g.parsing_schema_name||';'|| csr_10g.name||';'|| csr_10g.value_string);
         end loop;
ELSE
--9i
     for csr_9i in (SELECT a.sql_Text, b.Schemaname
                         FROM v$sqlarea a left outer join v$session b ON a.parsing_schema_id=b.schema#
          loop
            dbms_output.put_line(csr_9i.sql_text||';'|| csr_9i.schemaname||';');
         end loop;
END IF;
END;
--run;
spool off;
exit;If I run this on a 10g database, it works correctly, spooling the query return to a file.
If I run this on a 9i database, it will not run, and i get these errors:
                         FROM V$SQL a left outer join v$sql_bind_capture b ON a.SQL_ID = b.SQL_ID --combine to also find the bound variables used
ERROR at line 13:
ORA-06550: line 13, column 35:
PL/SQL: ORA-00942: table or view does not exist
ORA-06550: line 12, column 17:
PL/SQL: SQL Statement ignored
ORA-06550: line 27, column 30:
PLS-00364: loop index variable 'CSR_10G' use is invalid
ORA-06550: line 27, column 9:
PL/SQL: Statement ignored
Am I doing something wrong with my syntax, or is there another way to go about this?
Thanks!

Answer 1

Hi,
Dynamic SQL, like John suggested, might be the most elegant solution, but here are some others to consider:
(1) Create dummy objects. Your code guarantees that you'll never get a run-time error trying to query v$sql in Oracle 9, but the problem is that you're getting a compile-time error just by referencing it. So create a v$sql: a dummy table in your own schema, a synonym to the Oracle 10 data dictionary via a database link, anything just so it compiles.
(2) Create real objects. For example, in your Oracle 9 database, create a view called v$sql, based on v$sqlarea and v$session, that has all the same columns you use in Oracle 10. If you can do this, you won't even need an IF statement in your PL/SQL; the same code that runs against the data dictionary v$sql in Oracle 10 will run against your own schema's v$sql in Oracle 9.
(3) Isolate the problems. Write two different versions of a package, one for each version, and put all version-dependent code in the package. In most of your code, call the procedure to do the version-dependent stuff. For example, don't open a cursor in your program: instead, call the package to open a cursor.
(4) Comment Out. Klugy but cute. Write your PL/SQL without IF statements, like this:
BEGIN
&v10     for csr_10g in (SELECT sql_fulltext, a.parsing_schema_name, b.name, b.value_string
&v10                         FROM V$SQL a left outer
&v9     for csr_9i in (SELECT a.sql_Text, b.Schemaname
&v9                         FROM v$sqlarea a left outer join v$session b ON a.parsing_schema_id=b.schema#
&v9                                                                                  join v$sql_bind_capture b ON a.SQL_ID = b.SQL_ID)
          loop
&v10            dbms_output.put_line(csr_10g.sql_fulltext||';'|| csr_10g.parsing_schema_name||';'|| csr_10g.name||';'|| csr_10g.value_string);
&v9            dbms_output.put_line(csr_9i.sql_text||';'|| csr_9i.schemaname||';');
         end loop;
END;Compile your PL/SQL in SQL*Plus.
Define two substitution variables: v10 and v9.
In Oracle 10, v10 make " " and make v9 "--".
In Oracle 9, v10 make "--" and make v9 " ".
You can do the assignments on the fly with the SQL*Plus "COLUMN ... NEW_VALUE" command.
In either database, code that only works in the other version will be commented out. Your code will look like hell (good spacing and comments will make it less hard to read), but it will compile and work.
None of these ideas are mutually exclusive. You might want to do dynamic SQL in some places, and use on dummy objects in others, for example.