Trapping Perl divide-by-zero exceptions using eval( )

[Griff]

Why is it that this code produces a trappable exception -
#--------------------------------

$b = 1;
$c = 0;

eval { $b / $c ; } ;

if ($@)
{
print "I caught an exception\n";
print $@;
}

#--------------------------------

but this code seems to crash without catching the exception -

eval { 1 / 0 ; } ;

if ($@)
{
print "I caught an exception\n";
print $@;
}

I know that the second piece of code is even dumber than the first,
but just wondered what the difference was between the errors produced.

Thanks - Griff

 

[Scott W Gifford]

(e-mail address removed) (Griff) writes:

[...]

but this code seems to crash without catching the exception -

eval { 1 / 0 ; } ;

if ($@)
{
print "I caught an exception\n";
print $@;
}

I know that the second piece of code is even dumber than the first,
but just wondered what the difference was between the errors
produced.

My guess is that this error is produced at compile-time, when the Perl
compiler is calculating constant expressions and your code isn't yet
running. If you eval 1/0 in a string you get a runtime exception, and
if there are any variables at all in the expression (even $a/0) you
get a runtime exception, which seems to support this theory.

For a more definitive answer somebody will probably have to delve into
Perl's guts.

----ScottG.

 

[Uri Guttman]

G> Why is it that this code produces a trappable exception -
G> #--------------------------------

G> $b = 1;
G> $c = 0;

G> eval { $b / $c ; } ;

G> but this code seems to crash without catching the exception -

G> eval { 1 / 0 ; } ;

G> I know that the second piece of code is even dumber than the first,
G> but just wondered what the difference was between the errors produced.

the latter is executed at compile time as perl does constant folding. so
the eval hasn't been compiled yet and can't be run.

uri