Couldn't one write ACME::Include based on source filters instead of the DB way you gave above?

Source filters are a bad idea in general, but note that this one does not attempt to parse perl code, so it can not be fooled by wierd-looking perl code like some source filters can be. I still don't say that soing such things would be a good idea, but it might be cleaner that the DB hack.

Here's my example, which just a quick draft, does not handle line numbers etc.

Filter/Include.pm is

package Filter::Include;
use warnings; use strict;
use Filter::Util::Call;
sub import
{
        @_==2 or die qq{usage: use Filter::Include "filename"};
        open my $f, "<", $_[1] or die qq{cannot open include file "$_[
+1]": $!};
        read $f, my $d, -s $f or die qq{cannot read file "$_[1]": $!};
        filter_add(sub {
                $_ = $d;
                filter_del();
                1;
                });
}
1;
__END__
[download]

#!perl -w
use warnings; use strict;
my(@a, @b);
@a = (
1, 2,
do{ use Filter::Include "./second"; },
7, 8);
print "a(@a) b(@b)\n";
__END__
[download]

3, 4); @b = (5, 6,
[download]