Re: Cost of leaving debug statements in code

From: Chrysalis <cellis_at_iol.ie>
Date: 1997/06/06
Message-ID: <3397E857.578E@iol.ie>#1/1

I didn't know the answer, either, so I ran a test to try to determine the cost of evaluating a simple boolean test: results below (in seconds on a Mac PowerBook with Oracle V7.1.4).

To indulge in a "raging debate" seems unnecessary.

Chrysalis.

SQL> timing start BASE
SQL> declare
  2 bool boolean := TRUE;
  3 begin

  4  	for i in 0 .. 9999999	     -- 10^7
  5  	loop
  6  	   null;
  7  	end loop;

  8 end;
  9 /
SQL> timing stop;
timing for: BASE
 real: 56

SQL> 
SQL> timing start BOOL_TEST
SQL> declare
  2  	bool boolean := TRUE;
  3  begin
  4  	for i in 0 .. 9999999	     -- 10^7
  5  	loop
  6  	   if bool then
  7  	      null;
  8  	   end if;
  9  	end loop;

 10 end;
 11 /
SQL> timing stop;
timing for: BOOL_TEST
 real: 81

Allen Kirby wrote:
>
> Mark Everhart wrote:
> >
> > Debate raging in current organization about the cost vs benefit of leaving
> > debug output statements in production code. Currently, a private Boolean
> > variable can be set On or Off by using public procedures. This variable is
> > evaluated and if True then a debug package is called to output values of
> > variables or whatever.
> >
> > The argument rages over the cost of evaluating the Boolean variable at
> > numerous places in the code. Obviously, the poor souls manning the beepers
> > would like to have the option of re-running a production job that goes down
> > with the debug set on, so they can evaulate more accurately the cause of a
> > problem, fix it, and go back to sleep or whatever.
>
> The only way to resolve the debate is to run a load test, measuring
> system resources and response time, with the debug compiled in but
> turned off and with the debug not compiled in, and compare the results.
> Nobody can argue with that type of test. Personally, I would leave it
> in if at all possible.
> --
> ---
> Allen Kirby AT&T ITS Production Services
> akirby_at_att.com Alpharetta, GA.
Received on Fri Jun 06 1997 - 00:00:00 CDT