Re: converting Time::HiRes

[tachyon]

in reply to converting Time::HiRes

use Time::HiRes qw(gettimeofday);

my $sec        = gettimeofday();   # scalar context
my ( $s, $ms ) = gettimeofday();   # list context

my $float_t = "$s.$ms";

printf "Seconds since epoch (scalar): %d\n", $sec;
printf "Seconds since epoch (list):   %d.%d\n", $s, $ms;
printf "Seconds since epoch (float):  %.6f\n", $float_t;

printf "Scalar: %s\nList:   %s\nFloat:  %s\n", 
    scalar(localtime($sec)),
    scalar(localtime("$s.$ms")),
    scalar(localtime($float_t));

__DATA__
Seconds since epoch (scalar): 1095126990
Seconds since epoch (list):   1095126990.878512
Seconds since epoch (float):  1095126990.878512
Scalar: Tue Sep 14 11:56:30 2004
List:   Tue Sep 14 11:56:30 2004
Float:  Tue Sep 14 11:56:30 2004

[Lhamo Latso]

Thanks, the scalar of localtime works, with a decimal correctly placed:

print scalar(localtime(1068837428.812901))."\n";

My real problem is with the Oracle trace data. It has a time value in microseconds since the epoch, but the conversion appears to be off by 9 months. The print statement above is from September 2004, but the conversion lands in November 2003. As in the following source data:

*** SESSION ID:(92.34308) 2004-09-06 11:52:07.103
APPNAME mod='sqlplus@prod90db (TNS V1-V3)' mh=0 act='' ah=0
=====================
PARSING IN CURSOR #1 len=69 dep=0 uid=0 oct=42 lid=0 tim=1068837428812
+901 hv=...

It's time to read the Oracle docs again!!

[tachyon]

The problem is that tim records the value in v$timer. Here is the man page that this was taken from:

V$TIMER

This view lists the elapsed time in hundredths of seconds. Time is measured since the beginning of the epoch, which is operating system specific, and wraps around to 0 again whenever the value overflows four bytes (roughly 497 days).

[tachyon]

Actually, scratch the V$Timer bit. There appears to be a plausible conversion factor.

my $time = convert_tim(1068837428812901);
print "Epoch $time\n", scalar gmtime $time, "\n";

# Seem to need a conversion factor of 976,562.5 is 1,000,000,000/1024
sub convert_tim {
    my $factor = 1_000_000_000/1024;
    return $_[0]/$factor;
}
__DATA__
Epoch 1094489527.10441
Mon Sep  6 16:52:07 2004

Now the plaintext date in your log was 2004-09-06 11:52:07.103 so we get the date right, and the minute and second (down to 5 sig figs). This would appear to be a little to improbable for it to be a coincidence. I expect your DB GMT offset is -5 which is why the hours are out.

[poulhs]

This is the wrong way to handle $ms !!

It will fail unless there are 6 significant digits!

$err = ( $ms < 100000 ? 'wrong' : '' );
printf "Seconds since epoch (float):  %.6f\n", $sec;
# $sec is a float, print it like that...
printf "Seconds since epoch (list):   %d.%d  $err\n", $s, $ms;
printf "Seconds since epoch (float):  %.6f  $err\n", $float_t;

Seconds since epocg (float):  1349360875.006262
Seconds since epoch (list):   1349360875.6267  wrong
Seconds since epoch (float):  1349360875.626700  wrong