CGIs derive their parameters from either $ENV{'QUERY_STRING'} or <STDIN> (or both!), while Getopt::Long expects everything to be in @ARGV.

Converting one to the other is trivial ...

my $extract;
my $file;
{
    local @ARGV = split /&/, $ENV{'QUERY_STRING'};

    GetOptions(
        'file=s'   => \$file,
        'extract!' => \$extract,
    );
}
[download]

That'll break if you have URI-escaped characters (ie, %7E instead of ~), so a more robust solution would be to iteratively build @ARGV using CGI's param funcs.

my $extract;
my $file;
{
    local @ARGV;

    foreach my $key ( param ) {
        if ( param($key) ) {
            foreach my $value ( param($key) ) {
                push @ARGV, $key, $value;
            }
        } else {
              # arg-less booleans
            push @ARGV, $key;
        }
    }

    GetOptions(
        'file=s'   => \$file,
        'extract!' => \$extract,
    );
}
[download]

Your answer raises two further questions if you'll indulge me:

-why the different handling of arguments re: command line vs CGI -- and where, if anyplace, is that documented? I read and looked up references for 'parameters' and 'arguments' for a full day before I posted the question, and didn't see anything that hinted at the issue;

-could you possibly explain the REALLY bizarre CGI response when I test-run your logic on Win32 and my ISPs SUN server?

use CGI;
my $qry = new CGI;

my $argCt = @ARGV;
for ($i=0; $i < $argCt; $i++)
{
    print "Arg $i = [$ARGV[$i]]\n";
}
print "No args found\n" if $argCt == 0;

foreach my $key ( $qry->param ) 
{
    if ( $qry->param($key) )
    {
        foreach my $value ( $qry->param($key) ) 
        {
            print "key, value are $key, $value\n";
        }
    } else
    {
        # arg-less booleans
        print "key-only is $key\n";
    }
}

print "Arg(0) is [$ARGV[0]], param list is [", $qry->param(),"]\n";
[download]

perl stuff.pl a b c
Arg[0] = a
Arg[1] = b
Arg[2] = c
key, value are keywords, a
key, value are keywords, b
key, value are keywords, c
Arg(0) is [a], param list is [keywords]
[download]

perl stuff.pl a b=
Arg[0] = a
Arg[1] = b=
key-only is a
key-only is b
Arg(0) is [a], param list is [ab]
[download]

http:...stuff.cgi?a&b
[download]

Thanks!

why the different handling of arguments re: command line vs CGI -- and where, if anyplace, is that documented?

A good place to start is the CGI spec, but to answer your Perl isn't a CGI-centric language, so there's no reason why command-line and CGI arguments to be handle the same way.

Heck, if you happened to write a Perl based webserver, handling them the same way could be detrimental; after all, how would you pass arguments to the webserver and process CGIs?

could you possibly explain the REALLY bizarre CGI response

An oversight on my part:

If CGI.pm thinks it's being called ISINDEX-style (see the spec again and the CGI.pm's keyword section) it behaves a little differently, making arguments available via keywords or param('keywords').

Something like the below may get things working as you expect, but my testing was inconsistent across different versions of CGI.pm.

if ( param($key) ) {
    foreach my $value ( param($key) ) {
        if ( $key eq 'keywords' && keywords ) {
              # We ONLY have booleans
            push @ARGV, param('keywords');
            next;
        }
        push @ARGV, $key, $value;
    }
} else {
    # ...
[download]

    --k.