I've conditioned myself at this point to drop in a 'local $_;' whenever I use the <> operator in a function. But, today I was bitten by an oddity of $.. It was in very non-production code, but it's still worth knowing what's going on.

A simple example (what to look for: the line numbers in the warnings):

#!/usr/bin/perl
use strict;
use warnings;

{
    open my $fh, '<', 'somefile' or die "Opening somefile for reading:
+ $!";
    sub seeky { seek $fh, 0, 0 }
}

while (<DATA>) {
    chomp;
    my $line = $.;
    warn "Pre match: Line $. == $line\n";
    next if /match/;
    seeky();
    warn "Postmatch: Line $. != $line\n";
}
__END__
foo match
bar
baz match
bot
[download]

From perldoc perlvar, I think the following bit me:

When a line is read from a filehandle (via readline() or <> ), or when tell() or seek() is called on it, $. becomes an alias to the line counter for that filehandle.

I have two questions:

• Why should I have expected this behavior? $fh isn't even visible from the scope in which I use $.
• What other examples of these kinds of "gotchas" when using predefined variables can people provide?

(UPDATE): Also, I noticed the section from perldoc perlvar about localizing $.:

Localizing $. will not localize the filehandle's line count. Instead, it will localize perl's notion of which filehandle $. is currently aliased to.

... but I couldn't seem to elicit any helpful effects by doing so. (figured a local $.; after the while(<DATA>) might keep it associated with *DATA.)