--- trunk/all2xml.pl 2003/11/24 01:19:15 177
+++ trunk/all2xml.pl 2004/03/14 20:19:42 290
@@ -72,6 +72,8 @@
# if you are tight on memory, turn this off
my $use_lhash_cache = 1;
+my $last_field_name; # cache to prevent repeated fields
+
sub data2xml {
use xmlify;
@@ -102,16 +104,104 @@
}
my @sorted_tags;
- if ($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;
+ }
+ }
+ }
+
+
+ # init variables for different types
+ sub init_visible_type($) {
+ my $type = shift;
+
+ # swish, swish_exact, display, index, index_lookup
+ # swish and display defaults
+ my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
+ if (lc($type) eq "display") {
+ $s = 0;
+ } elsif (lc($type) eq "swish") {
+ $d = 0;
+ } elsif (lc($type) eq "index") {
+ ($s,$se,$d,$i) = (0,1,0,1);
+ } elsif (lc($type) eq "swish_exact") {
+ ($s,$se,$d,$i) = (0,1,0,0);
+ } elsif (lc($type) =~ /^lookup/) {
+ ($s,$se,$d,$i,$il) = (0,1,0,0,1);
+ }
+ return ($s,$se,$d,$i,$il);
+ }
+
+
+ # convert
+ #
+ #
+ # ,
+ # 200a
+ #
+ #
+ # to
+ #
+ # 200a
+ #
+ # but without loosing spaces in delimiter (becasue
+ # new XML::Simple strips spaces in attribute values
+ # as defined in XML specification)
+ #
+ sub unroll_x($) {
+ my $x = shift;
+
+ if (defined $x->{value}) {
+ my ($v,$d) = ($x->{value}->{content}, $x->{delimiter}->{content});
+ delete $x->{value};
+ delete $x->{delimiter};
+ $x->{content} = $v;
+ $x->{delimiter} = $d;
+ }
+ return $x;
+ }
+
+ # begin real work: go field by field
foreach my $field (@sorted_tags) {
$field=x($field);
@@ -120,31 +210,42 @@
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}}) {
+ $x = unroll_x($x);
+
my $format = x($x->{content});
my $delimiter = x($x->{delimiter}) || ' ';
- my $repeat_off = 0; # 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/);
+ my $repeat_off = 0; # init repeatable offset
+ my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
# what will separate last line from this one?
- if ($display_data && $x->{append} && $x->{append} eq "1") {
- $line_delimiter = ' ';
+ if ($display_data && $x->{append}) {
+ $line_delimiter = $delimiter;
} elsif ($display_data) {
$line_delimiter = '
';
}
@@ -153,9 +254,8 @@
($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 $data;
@@ -173,7 +273,7 @@
if (($#data+1) == $nr) {
return sprintf($fmt,@data);
} else {
- print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
+ #print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
return $data;
}
} else {
@@ -183,6 +283,8 @@
# while because of repeatable fields
while ($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, $format\n";
@@ -191,23 +293,24 @@
# is this field is lookup?
if ($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";
+# 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} || $null;
}
}
@@ -220,51 +323,47 @@
}
# type="swish" ; field for swish
if ($swish) {
+ my $tmp = $swish;
if ($filter && ($s || $se)) {
no strict 'refs';
- my $tmp = join(" ",&$filter($swish)) if ($s || $se);
- $swish_data .= $tmp if ($s);
- $swish_exact_data .= $tmp if ($se);
-
- } else {
- $swish_data .= $swish if ($s);
- $swish_exact_data .= $swish if ($se);
+ $tmp = join(" ",&$filter($tmp)) if ($s || $se);
}
+
+ $swish_data .= $tmp if ($s && $tmp);
+ $swish_exact_data .= "xxbxx $tmp xxexx " if ($tmp && $tmp ne "" && $se);
}
# type="display" ; field for display
if ($d && $display) {
+ my $ldel = $delimiter;
if ($line_delimiter && $display_data) {
- $display_data .= $line_delimiter;
- undef $line_delimiter;
+ $ldel = $line_delimiter;
}
if ($filter) {
no strict 'refs';
- if ($display_data) {
- $display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display)));
- } else {
- $display_data = join($delimiter,mkformat($x,&$filter($display)));
+ my @arr;
+ foreach my $tmp (&$filter($display)) {
+ my $tmp2 = mkformat($x,$tmp);
+ push @arr,$tmp2 if ($tmp2);
}
+ $display_data .= $ldel if ($display_data && @arr);
+ $display_data .= join($delimiter,@arr);
} else {
- if ($display_data) {
- $display_data .= $delimiter.mkformat($x,$display);
- } else {
- $display_data = mkformat($x,$display);
- }
+ $display_data .= $ldel if ($display_data);
+ my $tmp = mkformat($x,$display);
+ $display_data .= $tmp if ($tmp);
}
}
# 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, $idisplay if ($idisplay && !$iterate_by_page);
}
# store fields in lookup
@@ -273,87 +372,214 @@
if ($lookup_key) {
print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
} else {
- $lookup_key = $display;
+ if ($filter) {
+ no strict 'refs';
+ $lookup_key = &$filter($display);
+ } else {
+ $lookup_key = $display;
+ }
}
} elsif (lc($x->{type}) eq "lookup_val") {
if ($lookup_key) {
- $lhash{$lookup_key} = $display;
+ if ($filter) {
+ no strict 'refs';
+ $lhash{$lookup_key} = &$filter($display);
+ } else {
+ $lhash{$lookup_key} = $display;
+ }
} else {
print STDERR "WARNING: no lookup_key defined for '$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 = ();
}
}
# now try to parse variables from configuration file
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
+ $x = unroll_x($x);
+
my $delimiter = x($x->{delimiter}) || ' ';
my $val = $cfg->val($database, x($x->{content}));
- my ($s,$d,$i) = (1,1,0); # swish, display default
- $s = 0 if (lc($x->{type}) eq "display");
- $d = 0 if (lc($x->{type}) eq "swish");
- # no support for swish exact in config.
- # IMHO, it's useless
- ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
+ # FIXME index_lookup is not supported!
+ my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
if ($val) {
$display_data .= $delimiter.$val if ($d);
- $swish_data .= $val if ($s);
+ $swish_data .= " ".$val if ($s);
$index->insert($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}) {
+ my $tmp = mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]);
+ $display_data=$tmp if ($tmp);
} else {
- print STDERR "WARNING: field '$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($codepage,$swish_exact_data));
+ }
+
+ 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;
+ # 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;
- # 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'));
- }
+ $xml .= xmlify($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($codepage,$swish_exact_data));
+ }
+ }
}
# dump formatted output in
@@ -404,6 +630,8 @@
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
print STDERR "creating lookup file '$lookup_file'\n";
+ # delete memory cache for lookup file
+ delete $cache->{lhash};
}
# open existing lookup file
@@ -420,7 +648,7 @@
my $type_base = $type;
$type_base =~ s/_.+$//g;
- $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1);
+ $config=XMLin("./import_xml/$type.xml", ForceArray => [ $type2tag{$type_base}, 'config', 'format' ], ForceContent => 1 );
# output current progress indicator
my $last_p = 0;
@@ -452,6 +680,9 @@
# 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 "$database doesn't have 'isis_db' defined!";
@@ -537,8 +768,9 @@
print STDERR "\n";
} elsif ($type_base eq "excel") {
- use Spreadsheet::ParseExcel;
- use Spreadsheet::ParseExcel::Utility qw(int2col);
+ require Spreadsheet::ParseExcel;
+ require Spreadsheet::ParseExcel::Utility;
+ import Spreadsheet::ParseExcel::Utility qw(int2col);
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
@@ -592,29 +824,32 @@
}
} elsif ($type_base eq "marc") {
- use MARC;
+ require MARC;
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
# optional argument is format
- my $format = x($config->{format}) || 'usmarc';
+ my $format = x($config->{marc_format}) || 'usmarc';
print STDERR "Reading MARC file '$marc_file'\n";
my $marc = new MARC;
my $nr = $marc->openmarc({
file=>$marc_file, format=>$format
- }) || die "Can't open MARC file '$marc_file'";
+ }) || die "Can't open MARC file '$marc_file' with format '$format'";
- my $i=0; # record nr.
+ # read MARC file in memory
+ $marc->nextmarc(-1);
- my $rec;
+ my $max_rec = $marc->marc_count();
- while ($marc->nextmarc(1)) {
+ for(my $i=1; $i<=$max_rec; $i++) {
- # XXX
- fakeprogress($i++);
+ progress($i,$max_rec);
+
+ # store value for marc_sf.pm
+ $main::cache->{marc_record} = $i;
my $swishpath = $database."#".$i;
@@ -626,6 +861,9 @@
print "Document-Type: XML\n\n$xml\n";
}
}
+
+ print STDERR "\n";
+
} elsif ($type_base eq "feed") {
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);