Re^4: "require Carp" may be hazardous to your code (assume)

[ikegami]

Note that I kick quite a kick out of the last half of that statement. I'm sure the very first thing that require does is exactly that anyway, so I'd never add such trivial code in an obvious attempt at micro-optimization.

Did you read the comment next to "require Carp::Heavy unless $INC{"Carp/Heavy.pm"};" in the source? It's not an optimization. It's done that way so the module runs when Safe disallows require.

Update:

I don't recall you having posted data to support your claim either.

( On average, it takes me 4.24 ns to load Carp. Oops, Time::HiRes loads Carp. New benchmarks in progress. )

>perl -MTime::HiRes=time -le "$t1=time; require Carp; $t2=time; print 
+$t2-$t1"
4.05311584472656e-006

>perl -MTime::HiRes=time -le "$t1=time; require Carp; $t2=time; print 
+$t2-$t1"
4.05311584472656e-006

>perl -MTime::HiRes=time -le "$t1=time; require Carp; $t2=time; print 
+$t2-$t1"
4.05311584472656e-006

>perl -MTime::HiRes=time -le "$t1=time; require Carp; $t2=time; print 
+$t2-$t1"
5.00679016113281e-006

>perl -MTime::HiRes=time -le "$t1=time; require Carp; $t2=time; print 
+$t2-$t1"
4.05311584472656e-006

[tye]

Congratulations on your broken benchmark:

 > perl -MTime::HiRes -le"print $INC{'Carp.pm'}"
.../Carp.pm

Note that you also didn't address that this practice predates the factorying out of the "heavy" parts.

[ikegami]

Nice catch. I thought it was a bit fast...

Without heavy bits: 1.34 ms

>perl -le "$t1=time; for (1..$ARGV[0]) { delete $INC{'Carp.pm'}; requi
+re Carp; } $t2=time; print(($t2-$t1)/$ARGV[0])" 100000
0.00134

>perl -le "$t1=time; for (1..$ARGV[0]) { delete $INC{'Carp.pm'}; requi
+re Carp; } $t2=time; print(($t2-$t1)/$ARGV[0])" 100000
0.00134

>perl -le "$t1=time; for (1..$ARGV[0]) { delete $INC{'Carp.pm'}; requi
+re Carp; } $t2=time; print(($t2-$t1)/$ARGV[0])" 100000
0.00134

With heavy bits: 4.75 ms

>perl -le "$t1=time; for (1..$ARGV[0]) { delete $INC{'Carp.pm'}; delet
+e $INC{'Carp/Heavy.pm'}; require Carp; require Carp::Heavy; } $t2=tim
+e; print(($t2-$t1)/$ARGV[0])" 100000
0.00475

>perl -le "$t1=time; for (1..$ARGV[0]) { delete $INC{'Carp.pm'}; delet
+e $INC{'Carp/Heavy.pm'}; require Carp; require Carp::Heavy; } $t2=tim
+e; print(($t2-$t1)/$ARGV[0])" 100000
0.00474

[grinder]

Using much more lightweight code (for a FreeBSD system), I get around 30 milliseconds when not loading Carp, and 90 msecs when I do. Here is the code I used:

use Carp; # comment this line out, or not
use BSD::Process;

my $t = BSD::Process->new->elapsed_time;

print $t, $/; # time in milliseconds

The times are very erratic. It seems to boil down to whether the process executes within one timeslice of the processor or not: there are definite quanta to the amount of time taken. Nevertheless, sometimes people may care about 0.06 seconds of startup time (certainly less non-negligible than 4 usec).