If we don't want to bother with the problem of rowspanning for the moment, a simple solution is as follows:

#! /usr/bin/perl -w

use strict;
use CGI;
my $q = new CGI;
my %tab = ( abc => [4, 20], def => [9, 20000], ghi => [16, 200000]  );

print $q->table( {-border=>"1"},
    $q->TR([
        map {
            $q->td( $_ ) . $q->td( $tab{$_}->[0] )
        } sort keys %tab
    ])
);
[download]

But to introduce the second rowspanned cell into the picture had me scratching my head. How to you get the second TR element emitted? Looking at the HTML behind the target output, if we remove the table and TR elements, the contents of each row look something like:

<td rowspan="2">abc</td><td>4</td>
<td>20</td>
<td rowspan="2">def</td><td>9</td>
<td>20000</td>
<td rowspan="2">ghi</td><td>16</td>
<td>200000</td>
[download]

I see one possible solution, make the map emit the hash key twice, and use a flip-flop to emit upper row followed by the lower row

my $flipflop = 0;
print $q->table( {-border=>"1"},
    $q->TR([
        map {
            ++$flipflop % 2
                ? $q->td( {-rowspan=>"2"}, $_ ) . $q->td( $tab{$_}->[0
+] )
                : $q->td( $tab{$_}->[1] )
        } map { $_, $_ } sort keys %tab
    ])
);
[download]

That does the job, but it doesn't strike me as being particularly crystal clear for someone else to understand. Is there some other technique that monks have come up with for dealing with this problem?