Anyway, as it is, you would want to get the ids from all rows of "STATION2", and the ids from all rows of STATION3, and do something for the ones that match. Does it matter whether or not the $row values match too? If not, then you will probably be better off re-ordering the hash layers as suggested in the previous paragraph.

And what sort of "efficiency" are you looking for: run-time speed, or memory footprint? If the structure you showed is essential for some other aspect of your program, but you also want to do this thing with matching ID's, you can either stick with just the one copy of the data in memory and do a lot of extra processing (to navigate through all the rows), or make additional copies of the data (making it easier to work with all the ID's) and do less processing.

Let's suppose there's a good reason to keep the existing structure as-is, and that it's okay to just make a copy of the data so that you can work with the ids. Here's a way to "transpose" the hash layers:

my %rehash;
for my $station ( keys %hash ) {
    for my $row ( keys %{$hash{$station}} ) {
        for my $id ( keys %{$hash{$station}{$row}} ) {
            $rehash{$station}{$id}{$row} = $hash{$station}{$row}{$id};
        }
    }
}

# now check for matching ids among station2 and station3

for my $s2_key ( keys %{$rehash{"STATION2"}} ) {
    if ( exists( $rehash{"STATION3"}{$s2_key} )) {
        # do something...
    }
}
[download]

Often if you can't figure out how to drive a data structure in a clean way to achieve some end, it is because the data structure is not appropriate.

$row implies a contiguously numbered sequence starting at 0 or 1 to me. If that is the case then an array is much more appropriate than a hash.

In any case a code fragment with no data nor any explanation of what you are actually trying to achieve won't get you very good answers. Try presenting a small sample script with some sample data and telling us about the larger problem you are trying to solve.

Yeah I suppose it's easier when there is a visual So I have a spreadsheet that has 4 tabs. STATION1 thru STATION4. This first column in each tab is the rowId , just incremental row counter. data on STATION1 would look like this:

1,,business,first,last,email,,,,,
[download]

STATION2(Column2 is a reference to Column1 in STATION1):

1,1,,,123,Main,St,City,State,Zip
[download]

STATION3 (Column2 references Column1 in STATION2):

1,1,1,P-550,11223344
2,1,2,P-330,22334455
[download]

I am inserting data into two hashes like so:

$excel{$tab}{$row}{$col} = $cell->{Val};
$rehash{$tab}{$id}{$row} = $excel{$tab}{$row}{$col};
[download]

This attempt is finding no results:

for my $s2_key (keys %{$rehash{"1"}}) {
                        if ( exists( $rehash{"1"}{$s2_key} )) {
                                print SOMETHING;
                        }
[download]

Update: Durrr! That's what I get for operating on no sleep. Put "1" instead of "STATION2". I will continue to fiddle.

I want to compare every ID value at STATION3 to see if it matches the ID value at STATION2, if it matches, do something

Some other alternatives for visualizing complex data structures are described here: Re: How can I visualize my complex data structure?, including some graphical options for the more visually inclined. :-)