--- trunk/all2xml.pl 2003/01/11 19:55:30 9
+++ trunk/all2xml.pl 2003/11/29 15:59:19 182
@@ -6,16 +6,24 @@
use Data::Dumper;
use XML::Simple;
use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
-require Unicode::Map8;
-use DBI;
-
-my $config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1);
-my $dbh = DBI->connect("DBI:Pg:dbname=webpac","","") || die $DBI::errstr; # FIX
-# FIX; select relname from pg_class where relname like 'index_%' ;
-$dbh->begin_work || die $dbh->errstr();
-
-$dbh->do("delete from index_author") || die $dbh->errstr();
-$dbh->do("delete from index_title") || die $dbh->errstr();
+use Text::Iconv;
+use Config::IniFiles;
+use Encode;
+#use GDBM_File;
+use Fcntl; # for O_RDWR
+use TDB_File;
+
+$|=1;
+
+my $config_file = $0;
+$config_file =~ s/\.pl$/.conf/;
+die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
+
+my $config;
+
+#use index_DBI; # default DBI module for index
+use index_DBI_cache; # faster DBI module using memory cache
+my $index;
my %opts;
@@ -27,117 +35,473 @@
getopts('d:m:qs', \%opts);
-my $db_dir = $opts{d} || "ps"; # FIX
+my $path; # this is name of database
+
+Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
+
+# this is encoding of all files on disk, including import_xml/*.xml file and
+# filter/*.pm files! It will be used to store strings in perl internally!
+my $codepage = 'ISO-8859-2';
+
+my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
+# this function will convert data from XML files to local encoding
+sub x {
+ return $utf2cp->convert($_[0]);
+}
-#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts);
+# decode isis/excel or other import codepage
+my $import2cp;
-#print Dumper($config->{indexer});
-#print "-" x 70,"\n";
+# outgoing xml must be in UTF-8
+my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
-# how to convert isis code page to UTF8?
-my $isis_map = Unicode::Map8->new($config->{isis_codepage}) || die;
+# mapping between data type and tag which specify
+# format in XML file
+my %type2tag = (
+ 'isis' => 'isis',
+ 'excel' => 'column',
+ 'marc' => 'marc',
+ 'feed' => 'feed'
+);
+
+my $cache; # for cacheing
+
+# lookup hash (tied to file)
+my %lhash;
+# this option will cache all lookup entries in memory.
+# if you are tight on memory, turn this off
+my $use_lhash_cache = 1;
-sub isis2xml {
+sub data2xml {
+ use xmlify;
+
+ my $type = shift @_;
my $row = shift @_;
+ my $add_xml = shift @_;
+ # needed to read values from configuration file
+ my $cfg = shift @_;
+ my $database = shift @_;
my $xml;
- $xml->{db_dir} = [ $db_dir ]; # FIX remove?
- sub isis_sf {
- my $row = shift @_;
- my $isis_id = shift @_;
- my $subfield = shift @_;
- if ($row->{$isis_id}->[0]) {
- my $sf = OpenIsis::subfields($row->{$isis_id}->[0]);
- if (! defined $subfield || length($subfield) == 0) {
- # subfield list undef, empty or no defined subfields for this record
- my $all_sf = $row->{$isis_id}->[0];
- $all_sf =~ s/\^./ /g; nuke definirions
- return $all_sf;
- } elsif ($sf->{$subfield}) {
- return $sf->{$subfield};
- }
+ use parse_format;
+
+ my $html = ""; # html formatted display output
+
+ my %field_usage; # counter for usage of each field
+
+ # sort subrouting using order="" attribute
+ sub by_order {
+ my $va = $config->{indexer}->{$a}->{order} ||
+ $config->{indexer}->{$a};
+ my $vb = $config->{indexer}->{$b}->{order} ||
+ $config->{indexer}->{$b};
+
+ return $va <=> $vb;
+ }
+
+ my @sorted_tags;
+ if ($cache->{tags_by_order}) {
+ @sorted_tags = @{$cache->{tags_by_order}};
+ } else {
+ @sorted_tags = sort by_order keys %{$config->{indexer}};
+ $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};
+ 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) {
+ return x($field_name);
}
}
- foreach my $field (keys %{$config->{indexer}}) {
- my $display_data = "";
+ # begin real work: go field by field
+ foreach my $field (@sorted_tags) {
+
+ $field=x($field);
+ $field_usage{$field}++;
+
my $swish_data = "";
- my $index_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} || '
';
+
+ foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
+
+ my $format = x($x->{content});
+ my $delimiter = x($x->{delimiter}) || ' ';
- foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
+ my $repeat_off = 0; # init repeatable offset
- my $display_tmp = "";
- my $swish_tmp = "";
- my $index_tmp = "";
-
- my $format = $x->{content};
- my $s = 1; # swish only
- my $d = 1; # display only
- my $i = 0; # index only
+ # 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");
- ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
-#print STDERR "## s: $s d: $d i: $i ## $format ##\n";
- # parse format
- my $prefix = "";
- if ($format =~ s/^([^\d]+)//) {
- $prefix = $1;
- }
- while ($format) {
- if ($format =~ s/^(\d\d\d)(\w?)//) {
- my $isis_tmp = isis_sf($row,$1,$2);
- if ($isis_tmp) {
-# $display_tmp .= $prefix . "/$1/$2/".$isis_tmp if ($d);
- $display_tmp .= $prefix . $isis_tmp if ($d);
- $swish_tmp .= $isis_tmp." " if ($s);
- $index_tmp .= $prefix . $isis_tmp if ($i);
-#print STDERR " $isis_tmp <--\n";
- }
- $prefix = "";
- } elsif ($format =~ s/^([^\d]+)//) {
- $prefix = $1;
+ ($s,$se,$d,$i) = (0,0,0,1) 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}) {
+ $line_delimiter = ' ';
+ } elsif ($display_data) {
+ $line_delimiter = '
';
+ }
+
+ # init vars so that we go into while...
+ ($swish,$display) = (1,1);
+
+ # placeholder for all repeatable entries for index
+
+ sub mkformat {
+ my $x = shift || die "mkformat needs tag reference";
+ my $data = shift || return;
+ my $format_name = x($x->{format_name}) || return $data;
+ my $fmt = x($config->{format}->{$format_name}->{content}) || die " is not defined!";
+ my $format_delimiter = x($x->{format_delimiter});
+ my @data;
+ if ($format_delimiter) {
+ @data = split(/$format_delimiter/,$data);
+ } else {
+ push @data,$data;
+ }
+
+ if ($fmt) {
+ my $nr = scalar $fmt =~ s/%s/%s/g;
+ if (($#data+1) == $nr) {
+ return sprintf($fmt,@data);
+ } else {
+ print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
+ return $data;
+ }
} else {
- print STDERR "WARNING: unparsed format '$format'\n";
+ print STDERR "usage of link '$format_name' without defined format ( tag)\n";
+ }
+ }
+
+ # 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";
last;
- };
+ }
+
+ # 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 (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} = $null;
+ }
+ } else {
+ $display = $cache->{lhash}->{$display};
+ }
+ } else {
+ $display = $lhash{$display} || $null;
+ }
+ }
+
+ # filter="name" ; filter this field through
+ # filter/[name].pm
+ my $filter = $x->{filter};
+ if ($filter && !$cache->{filter_loaded}->{$filter}) {
+ require "filter/".$filter.".pm";
+ $cache->{filter_loaded}->{$filter}++;
+ }
+ # type="swish" ; field for swish
+ if ($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);
+ }
+ }
+
+ # type="display" ; field for display
+ if ($d && $display) {
+ if ($line_delimiter && $display_data) {
+ $display_data .= $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)));
+ }
+ } else {
+ if ($display_data) {
+ $display_data .= $delimiter.mkformat($x,$display);
+ } else {
+ $display_data = mkformat($x,$display);
+ }
+ }
+ }
+
+ # type="index" ; insert into index
+ my $idisplay;
+ if ($i && $display) {
+ $idisplay = $display;
+ if ($filter) {
+ no strict 'refs';
+ $idisplay = &$filter($idisplay);
+ }
+ push @index_data, $idisplay if (! $iterate_by_page);
+ }
+
+ # store fields in lookup
+ if ($il && $display) {
+ if (lc($x->{type}) eq "lookup_key") {
+ if ($lookup_key) {
+ print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
+ } else {
+ $lookup_key = $display;
+ }
+ } elsif (lc($x->{type}) eq "lookup_val") {
+ if ($lookup_key) {
+ $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}) {
+ $index_data[$#index_data] .= $idisplay;
+ } else {
+ push @index_data, $idisplay;
+ }
+ $idisplay = "";
+ @{$cache->{index_data}->{$field}->[$page]} = @index_data;
+ }
+ }
+ }
+
+ if (! $iterate_by_page) {
+ # fill data in index
+ foreach my $d (@index_data) {
+ $index->insert($field, $d, $path);
+ }
+ @index_data = ();
+ }
+ }
+
+ # now try to parse variables from configuration file
+ foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
+
+ 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");
+
+ if ($val) {
+ $display_data .= $delimiter.$val if ($d);
+ $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 = "";
+ }
}
- # add suffix
- $display_tmp .= $prefix if ($display_tmp);
- $index_tmp .= $prefix if ($index_tmp);
-
-# $display_data .= $display_tmp if ($display_tmp ne "");
-# $swish_data .= $swish_tmp if ($swish_tmp ne "");
- $display_data .= $display_tmp;
- $swish_data .= $swish_tmp;
- $index_data .= $index_tmp;
-
- }
-#print "--display:$display_data\n--swish:$swish_data\n";
- #$xml->{$field."_display"} = $isis_map->tou($display_data)->utf8 if ($display_data);
- #$xml->{$field."_swish"} = unac_string($config->{isis_codepage},$swish_data) if ($swish_data);
- $xml->{$field."_display" } = [ $isis_map->tou($display_data)->utf8 ] if ($display_data);
- $xml->{$field."_swish"} = [ unac_string($config->{isis_codepage},$swish_data) ] if ($swish_data);
-
- # index
- if ($index_data && $index_data ne "") {
- my $sql = "select $field from index_$field where upper($field)=upper(?)";
- my $sth = $dbh->prepare($sql) || die $dbh->errstr();
- $sth->execute($index_data) || die "SQL: $sql; ".$dbh->errstr();
-#print STDERR "--->$index_data<---\n";
- if (! $sth->fetchrow_hashref) {
- my $sql = "insert into index_$field values (?)";
- my $sth = $dbh->prepare($sql) || die $dbh->errstr();
-#print STDERR "$sql: $index_data