--- trunk/all2xml.pl 2003/11/24 01:19:15 177
+++ trunk/all2xml.pl 2003/12/15 00:12:16 196
@@ -72,6 +72,8 @@
# if you are tight on memory, turn this off
my $use_lhash_cache = 1;
+my $last_field_name; # cache to prevent repeated fields
+
sub data2xml {
use xmlify;
@@ -102,16 +104,54 @@
}
my @sorted_tags;
- if ($cache->{tags_by_order}->{$type}) {
- @sorted_tags = @{$cache->{tags_by_order}->{$type}};
+ if ($cache->{tags_by_order}) {
+ @sorted_tags = @{$cache->{tags_by_order}};
} else {
@sorted_tags = sort by_order keys %{$config->{indexer}};
- $cache->{tags_by_order}->{$type} = \@sorted_tags;
+ $cache->{tags_by_order} = \@sorted_tags;
}
# lookup key
my $lookup_key;
+ # cache for field in pages
+ delete $cache->{display_data};
+ delete $cache->{swish_data};
+ delete $cache->{swish_exact_data};
+ delete $cache->{index_data};
+ delete $cache->{index_delimiter};
+ my @page_fields; # names of fields
+
+
+ # subs used to produce output
+
+ sub get_field_name($$$) {
+ my ($config,$field,$field_usage) = @_;
+
+ # find field name (signular, plural)
+ my $field_name = "";
+ if ($config->{indexer}->{$field}->{name_singular} && $field_usage == 1) {
+ $field_name = $config->{indexer}->{$field}->{name_singular};
+ } elsif ($config->{indexer}->{$field}->{name_plural}) {
+ $field_name = $config->{indexer}->{$field}->{name_plural};
+ } elsif ($config->{indexer}->{$field}->{name}) {
+ $field_name = $config->{indexer}->{$field}->{name};
+ } else {
+ print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
+ }
+ if ($field_name) {
+ if (! $last_field_name) {
+ $last_field_name = x($field_name);
+ return $last_field_name;
+ } elsif ($field_name ne $last_field_name) {
+ $last_field_name = x($field_name);
+ return $last_field_name;
+ }
+ }
+ }
+
+
+ # begin real work: go field by field
foreach my $field (@sorted_tags) {
$field=x($field);
@@ -120,30 +160,46 @@
my $swish_data = "";
my $swish_exact_data = "";
my $display_data = "";
+ my @index_data;
my $line_delimiter;
my ($swish,$display);
my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
+
+ # is this field page-by-page?
+ my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page};
+ push @page_fields,$field if ($iterate_by_page);
+ my %page_max = ();
+ # 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 $format_name = $config->{indexer}->{$field}->{format_name};
+ my $format_delimiter = $config->{indexer}->{$field}->{format_delimiter};
+ if ($format_name && $format_delimiter) {
+ $cache->{format}->{$field}->{format_name} = $format_name;
+ $cache->{format}->{$field}->{format_delimiter} = $format_delimiter;
+ }
+
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
my $format = x($x->{content});
my $delimiter = x($x->{delimiter}) || ' ';
- my $repeat_off = 0; # repeatable offset
+ my $repeat_off = 0; # init repeatable offset
# swish, swish_exact, display, index, index_lookup
# swish and display defaults
my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
$s = 0 if (lc($x->{type}) eq "display");
$d = 0 if (lc($x->{type}) eq "swish");
- $se = 1 if (lc($x->{type}) eq "swish_exact");
- ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
- $il = 1 if (lc($x->{type}) =~ /^lookup/);
-
+ ($s,$se,$d,$i) = (0,1,0,1) if (lc($x->{type}) eq "index");
+ ($s,$se,$d,$i) = (0,1,0,0) if (lc($x->{type}) eq "swish_exact");
+ ($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/);
# what will separate last line from this one?
- if ($display_data && $x->{append} && $x->{append} eq "1") {
+ if ($display_data && $x->{append}) {
$line_delimiter = ' ';
} elsif ($display_data) {
$line_delimiter = '
';
@@ -153,9 +209,8 @@
($swish,$display) = (1,1);
# placeholder for all repeatable entries for index
- my @index_data;
- sub mkformat {
+ sub mkformat($$) {
my $x = shift || die "mkformat needs tag reference";
my $data = shift || return;
my $format_name = x($x->{format_name}) || return $data;
@@ -183,6 +238,8 @@
# while because of repeatable fields
while ($swish || $display) {
+ my $page = $repeat_off;
+ $page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0));
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
if ($repeat_off > 1000) {
print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
@@ -191,23 +248,24 @@
# is this field is lookup?
if ($display && $x->{lookup}) {
+ my $null = "";
if ($use_lhash_cache) {
if (!defined($cache->{lhash}->{$display})) {
my $new_display = $lhash{$display};
- if ($new_display) {
+ if (defined($new_display)) {
#print STDERR "lookup cache store '$display' = '$new_display'\n";
$display = $new_display;
$cache->{lhash}->{$display} = $new_display;
} else {
print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
$display = "";
- $cache->{lhash}->{$display} = "";
+ $cache->{lhash}->{$display} = $null;
}
} else {
$display = $cache->{lhash}->{$display};
}
} else {
- $display = $lhash{$display} || "";
+ $display = $lhash{$display} || $null;
}
}
@@ -224,11 +282,23 @@
no strict 'refs';
my $tmp = join(" ",&$filter($swish)) if ($s || $se);
$swish_data .= $tmp if ($s);
- $swish_exact_data .= $tmp if ($se);
+ if ($se) {
+ if ($swish_exact_data) {
+ $swish_exact_data .= "xxexx xxbxx ".$tmp;
+ } else {
+ $swish_exact_data .= $tmp;
+ }
+ }
} else {
$swish_data .= $swish if ($s);
- $swish_exact_data .= $swish if ($se);
+ if ($se) {
+ if ($swish_exact_data) {
+ $swish_exact_data .= "xxexx xxbxx ".$swish;
+ } else {
+ $swish_exact_data .= $swish;
+ }
+ }
}
}
@@ -236,7 +306,6 @@
if ($d && $display) {
if ($line_delimiter && $display_data) {
$display_data .= $line_delimiter;
- undef $line_delimiter;
}
if ($filter) {
no strict 'refs';
@@ -255,16 +324,14 @@
}
# type="index" ; insert into index
+ my $idisplay;
if ($i && $display) {
+ $idisplay = $display;
if ($filter) {
no strict 'refs';
- $display = &$filter($display);
- }
- if ($x->{append} && @index_data) {
- $index_data[$#index_data].=$display;
- } else {
- push @index_data, $display;
+ $idisplay = &$filter($idisplay);
}
+ push @index_data, $idisplay if (! $iterate_by_page);
}
# store fields in lookup
@@ -282,12 +349,71 @@
print STDERR "WARNING: no lookup_key defined for '$display'?";
}
}
+
+ }
+
+ # store data for page-by-page repeatable fields
+ if ($iterate_by_page) {
+ sub iterate_fld($$$$$$) {
+ my ($cache,$what,$field,$page,$data,$append) = @_;
+ return if (!$data);
+
+ my $ldel = $page_line_delimiter;
+ $ldel = " " if ($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;
+ }
+ }
+
+ if ($display_data) {
+ iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append});
+ }
+ $display_data = "";
+ if ($swish_data) {
+ iterate_fld($cache,'swish_data',$field,$page,$swish_data,$x->{append});
+ $swish_data = "";
+ }
+ if ($swish_exact_data) {
+ iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append});
+ $swish_exact_data = "";
+ }
+
+ if ($idisplay) {
+ my $ldel=$page_line_delimiter;
+ my @index_data;
+ if ($cache->{index_data}->{$field}->[$page]) {
+
+ @index_data = @{$cache->{index_data}->{$field}->[$page]};
+ }
+ if ($x->{append}) {
+ if (@index_data) {
+ $index_data[$#index_data] .= $idisplay;
+ } else {
+ push @index_data, $idisplay;
+ }
+ } else {
+ push @index_data, $idisplay;
+ }
+ $idisplay = "";
+ @{$cache->{index_data}->{$field}->[$page]} = @index_data;
+ }
}
}
- # fill data in index
- foreach my $d (@index_data) {
- $index->insert($field, $d, $path);
+ if (! $iterate_by_page) {
+ my $idel = $x->{index_delimiter};
+ # fill data in index
+ foreach my $tmp (@index_data) {
+ my $i = $d = $tmp;
+ if ($idel && $tmp =~ m/$idel/) {
+ ($i,$d) = split(/$idel/,$tmp);
+ }
+ $index->insert($field, $i, $d, $path);
+ }
+ @index_data = ();
}
}
@@ -310,50 +436,109 @@
$index->insert($field, $val, $path) if ($i);
}
+ if ($iterate_by_page) {
+ # FIXME data from config tag will appear just
+ # on first page!!!
+ my $page = 0;
+ if ($display_data) {
+ $cache->{display_data}->{$field}->[$page] = $display_data;
+ $display_data = "";
+ }
+ if ($swish_data) {
+ $cache->{swish_data}->{$field}->[$page] = $swish_data;
+ $swish_data = "";
+ }
+ if ($swish_exact_data) {
+ $cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data;
+ $swish_exact_data = "";
+ }
+ }
}
-
- if ($display_data) {
-
- if ($field eq "headline") {
- $xml .= xmlify("headline", $display_data);
- } else {
-
- # find field name (signular, plural)
- my $field_name = "";
- if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
- $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
- } elsif ($config->{indexer}->{$field}->{name_plural}) {
- $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
- } elsif ($config->{indexer}->{$field}->{name}) {
- $field_name = $config->{indexer}->{$field}->{name}."#-#";
+ # save data page-by-page
+ foreach my $field (@page_fields) {
+ my $nr_pages = $page_max{$field} || next;
+#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n";
+#print STDERR Dumper($cache->{display_data});
+ for (my $page=0; $page <= $nr_pages; $page++) {
+ my $display_data;
+ if ($cache->{format}->{$field}) {
+ $display_data=mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]);
} else {
- print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
+ $display_data = $cache->{display_data}->{$field}->[$page];
+ }
+ if ($display_data) { # default
+ if ($field eq "headline") {
+ $xml .= xmlify("headline", $display_data);
+ } else {
+
+ # fallback to empty field name if needed
+ $html .= get_field_name($config,$field,$field_usage{$field}) || '';
+ $html .= "#-#".$display_data."###\n";
+ }
}
- if ($field_name) {
- $html .= x($field_name);
+
+ my $swish_data = $cache->{swish_data}->{$field}->[$page];
+ if ($swish_data) {
+ # remove extra spaces
+ $swish_data =~ s/ +/ /g;
+ $swish_data =~ s/ +$//g;
+
+ $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
+ }
+
+ my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
+ if ($swish_exact_data) {
+ $swish_exact_data =~ s/ +/ /g;
+ $swish_exact_data =~ s/ +$//g;
+
+ # add delimiters before and after word.
+ # That is required to produce exact match
+ $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
+ }
+
+ my $idel = $cache->{index_delimiter}->{$field};
+ foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) {
+ my $i = $tmp;
+ my $d = $tmp;
+ if ($idel && $tmp =~ m/$idel/) {
+ ($i,$d) = split(/$idel/,$tmp);
+ }
+ $index->insert($field, $i, $d, $path);
+#print STDERR "index [$idel] $field: $i --> $d [$path]\n";
}
- $html .= $display_data."###\n";
}
+
}
- if ($swish_data) {
- # remove extra spaces
- $swish_data =~ s/ +/ /g;
- $swish_data =~ s/ +$//g;
-
- $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
- }
+
+ if (! $iterate_by_page) {
+ if ($display_data) {
+ if ($field eq "headline") {
+ $xml .= xmlify("headline", $display_data);
+ } else {
- if ($swish_exact_data) {
- $swish_exact_data =~ s/ +/ /g;
- $swish_exact_data =~ s/ +$//g;
-
- # add delimiters before and after word.
- # That is required to produce exact match
- $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
- }
+ # fallback to empty field name if needed
+ $html .= get_field_name($config,$field,$field_usage{$field}) || '';
+ $html .= "#-#".$display_data."###\n";
+ }
+ }
+ if ($swish_data) {
+ # remove extra spaces
+ $swish_data =~ s/ +/ /g;
+ $swish_data =~ s/ +$//g;
+ $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
+ }
+ if ($swish_exact_data) {
+ $swish_exact_data =~ s/ +/ /g;
+ $swish_exact_data =~ s/ +$//g;
+
+ # add delimiters before and after word.
+ # That is required to produce exact match
+ $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
+ }
+ }
}
# dump formatted output in
@@ -452,6 +637,9 @@
# now read database
print STDERR "using: $type...\n";
+ # erase cache for tags by order in this database
+ delete $cache->{tags_by_order};
+
if ($type_base eq "isis") {
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";