RE: What stupid mistake am I making with this onlogin trigger?

From: Powell, Mark D <>
Date: Tue, 26 Feb 2008 09:09:10 -0500
Message-ID: <>


I have not tested it recently but the sys_context call used to be significantly slower than just issuing
select sid from v$mystat where rownum = 1

This form of the query will eliminate the need to read all 370 or so rows and then through away the duplicates.

  • Mark D Powell -- Phone (313) 592-5148

-----Original Message-----

[] On Behalf Of Nigel Thomas Sent: Tuesday, February 26, 2008 7:14 AM To:; Subject: Re: What stupid mistake am I making with this onlogin trigger?

Small extra point:

IMHO it's not best practice to use v$ views when you could use SYS_CONTEXT and DBMS_APPLICATION_INFO instead - so you don't have to have too many users with unrestricted read access to some or all of the v$ views. It's always a good idea to minimise the scope of any grants.

(select unique(sid) from v$mystat)

can be replaced with:

which has the added benefit of avoiding a hash(unique) on v$mystat (370 rows in my case).

To demonstrate:


 l_module varchar2(48);
 l_action varchar2(32);
 l_username varchar2(30);
 l_sid integer;

 dbms_application_info.read_module_info(l_module, l_action); l_username := sys_context('userenv','session_user');  l_sid := sys_context('userenv','sessionid');

 dbms_output.put_line('User:'||l_user||', SID:'||l_sid||', Module:'||l_module);


Shame that DBMS_APPLICATION_INFO doesn't have usable GET_xxx functions, but forces the use of a procedure with output parameters. How 1960s is that? Also, in some circumstances it might be appropriate to cover DBMS_APPLICATION_INFO with a package that only allows gets, not sets.

NB I haven't compared the cost of these with accessing the views.

Regards Nigel


-- Received on Tue Feb 26 2008 - 08:09:10 CST

Original text of this message