Since people have shown how the use of a regex approach tends to be slower, here's a way to observe word boundaries (well, spaces between words, anyway) without using a regex:

my $maxlen = 20;
my $longtext = "This is some very long string that needs to be truncat
+ed to $maxlen characters...";

my $trunctext = substr( $longtext, 0, rindex( $longtext, " ", $maxlen 
+));

print "$longtext\n$trunctext\n";
[download]

The rindex function, like substr, is faster than a regex match.