Datasets may be in the thousands, or millon rows...

The problem I have with delegating sorting and querying to providers is that paging gets really tricky.

my @rows;
for my $provider( qw/Provider1 Provider2 Provider3/ ) {
   push @rows, $provider->search( 
        query=>..., sort=>..., 
        page=>1, rows=>10 
    );
}
# now I may end up with 30 rows for page 1
# but the client grid expects 10, so let's cut it down...
@rows = @rows [ 0..9  ];
[download]

So, ideally, the "collection manager object" should work with this algorithm:

1. Ask each provider to search for a certain query and sort
2. Ask for a row from each provider starting at a given page
3. Is the page complete? Then call any external filters callbacks.
4. External filters dropped rows? Then keep fetching until page is complete again.
5. Sort the combined rows
6. Return resulting array, but keep track of the current paging state

--Miguel

Code tags added by GrandFather