--- trunk/all2xml.pl 2006/04/13 19:35:19 726 +++ trunk/all2xml.pl 2006/04/13 19:44:51 730 @@ -126,6 +126,7 @@ delete $cache->{swish_exact_data}; delete $cache->{index_data}; delete $cache->{index_delimiter}; + delete $cache->{distinct}; my @page_fields; # names of fields @@ -234,6 +235,11 @@ # default line_delimiter if using my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '
'; $cache->{index_delimiter}->{$field} = $config->{indexer}->{$field}->{index_delimiter}; + my $distinct = $config->{indexer}->{$field}->{distinct}; + if ($distinct && !$iterate_by_page) { + warn "WARNING: distinct is currently not supported without iterate_by_page!\n"; + $distinct = 0; + } my $format_name = $config->{indexer}->{$field}->{format_name}; my $format_delimiter = $config->{indexer}->{$field}->{format_delimiter}; @@ -412,9 +418,10 @@ $ldel = " " if ($append); #print STDERR "line delimiter: ",Dumper($ldel) if ($ldel); if (! $cache->{$what}->{$field}->[$page]) { - $cache->{$what}->{$field}->[$page] = $data; - } else { - $cache->{$what}->{$field}->[$page] .= $ldel.$data; + push @{$cache->{$what}->{$field}->[$page]}, { + data => $data, + delimiter => $ldel, + }; } } @@ -489,15 +496,15 @@ # on first page!!! my $page = 0; if ($display_data) { - $cache->{display_data}->{$field}->[$page] = $display_data; + push @{$cache->{display_data}->{$field}->[$page]}, { data => $display_data }; $display_data = ""; } if ($swish_data) { - $cache->{swish_data}->{$field}->[$page] = $swish_data; + push @{$cache->{swish_data}->{$field}->[$page]}, { data => $swish_data }; $swish_data = ""; } if ($swish_exact_data) { - $cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data; + push @{$cache->{swish_exact_data}->{$field}->[$page]}, { data => $swish_exact_data }; $swish_exact_data = ""; } } @@ -508,14 +515,28 @@ my $nr_pages = $page_max{$field} || next; #print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; #print STDERR Dumper($cache->{display_data}); + my $seen; # used for distinct for (my $page=0; $page <= $nr_pages; $page++) { my $display_data; - if ($cache->{format}->{$field}) { - my $tmp = mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]); - $display_data=$tmp if ($tmp); - } else { - $display_data = $cache->{display_data}->{$field}->[$page]; + my $delimiter = ''; + foreach my $element (@{ $cache->{display_data}->{$field}->[$page] }) { + my $data = $element->{data}; + die "BUG! no data in element?" unless ($data); + + if ($distinct) { + next if ($cache->{distinct}->{$field}->{ $data }); + $cache->{distinct}->{$field}->{ $data } = 1; + } + + if ($cache->{format}->{$field}) { + my $tmp = mkformat($cache->{format}->{$field},$data); + $display_data .= $delimiter . $tmp if ($tmp); + } else { + $display_data .= $delimiter . $data; + } + $delimiter = $element->{delimiter} if ($element->{delimiter}); } + if ($display_data) { # default if ($field eq "headline") { $xml .= xmlify("headline", $display_data); @@ -527,7 +548,7 @@ } } - my $swish_data = $cache->{swish_data}->{$field}->[$page]; + my $swish_data = join(" ",map { $_->{data} } @{ $cache->{swish_data}->{$field}->[$page] }); if ($swish_data) { # remove extra spaces $swish_data =~ s/ +/ /g; @@ -536,7 +557,7 @@ $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data)); } - my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page]; + my $swish_exact_data = join(" ", map { $_->{data} } @{ $cache->{swish_exact_data}->{$field}->[$page] }); if ($swish_exact_data) { $swish_exact_data =~ s/ +/ /g; $swish_exact_data =~ s/ +$//g;