--- trunk/all2xml.pl 2003/01/16 17:35:54 10 +++ trunk2/all2all.pl 2004/09/11 08:36:38 422 @@ -1,223 +1,233 @@ #!/usr/bin/perl -w -use strict; -use OpenIsis; -use Getopt::Std; -use Data::Dumper; -use XML::Simple; -use Text::Unaccent 1.02; # 1.01 won't compile on my platform, -use Text::Iconv; - - -$|=1; - -my $config; - -$config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1); - -use index_DBI; # there is no other, right now ;-) +=head1 NAME -my $index = new index_DBI(); # open index +all2all.pl - basic script for all WebPAC needs -my %opts; +=cut -# usage: -# -d directory name -# -m multiple directories -# -q quiet -# -s run swish +use strict; +use Data::Dumper; +use Carp; -getopts('d:m:qs', \%opts); +use lib './lib'; +use WebPAC; +use WebPAC::jsFind; +use WebPAC::Index; + +my $webpac = new WebPAC( + code_page => 'ISO-8859-2', +# limit_mfn => 100, +# debug => 1, + low_mem => 1, +) || die; + +my $log = $webpac->_get_logger() || die "can't get logger"; + +$log->debug("creating WebPAC::jsFind object"); + +my $index = new WebPAC::jsFind( + index_path => './out/index', + keys => 10, +) || die; -my $db_dir = $opts{d} || "ps"; # FIX +my $thes; -#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); +$|=1; -#print Dumper($config->{indexer}); -#print "-" x 70,"\n"; +my $maxmfn = $webpac->open_isis( + filename => shift @ARGV || '/data/hidra/THS/THS', + lookup => [ + { 'key' => 'd:v900', 'val' => 'v250^a' }, +# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, +# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, +# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, + { 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, + { 'key' => '900_mfn:v900', 'val' => 'v000' }, + # tree structure + { 'eval' => 'length("v251") == 2', 'key' => 'root:v251', 'val' => 'v900' }, + { 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' }, + ], +); -Text::Iconv->raise_error(1); # Conversion errors raise exceptions +$log->debug("isis file ",$webpac->{'isis_filename'}," opened"); -my $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); -my $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); -my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); +$log->info("rows: $maxmfn"); -sub isis2xml { +$webpac->open_import_xml(type => 'isis_hidra_ths'); - use xmlify; +$log->debug("data: ",Dumper($webpac->{'data'})); - my $row = shift @_; +while (my $rec = $webpac->fetch_rec) { - my $xml; - $xml .= xmlify('db_dir',$db_dir); # FIX remove? + my @ds = $webpac->data_structure($rec); - use parse_format; + if (0 && $log->is_debug) { + $log->debug("rec = ",Dumper($rec)); + $log->debug("ds = ",Dumper(\@ds)); + } + next if (! @ds); - foreach my $field (keys %{$config->{indexer}}) { + my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object"); - my $swish_data = ""; - my $display_data = ""; - my $index_data = ""; + if ($filename) { + $webpac->output_file( + file => $filename, + template => 'html.tt', + data => \@ds, + headline => $webpac->{'headline'}, + ); + } else { + print $webpac->output( + template => 'text.tt', + data => \@ds, + headline => $webpac->{'headline'}, + ); + } - foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { + my $headline = $webpac->{'headline'}; - my $format = $x->{content}; - my ($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"); - ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); -#print STDERR "## s: $s d: $d i: $i ## $format ##\n"; + my $f = $filename; + $f =~ s!out/!!; - $format = $cludge_codepage->convert($format); - my ($swish,$display) = parse_format($format,$row); -#print STDERR "s: $swish\nd: $display\n" if ($swish); + # save into index + foreach my $ds (@ds) { + next if (! $ds->{'swish'}); -#print STDERR "swish: $swish<-- display: $display<--\n"; - # FIX: this is ugly, UGLY, cludge: OpenIsis return - # UTF8 encoding of strings, but as if source charset - # is ISO8859-1 and not some other. This breaks our - # isis character encoding, so we convert it first - # back to ISO8859-1 (which can actually be different - # encoding in isis) + $index->insert( + index_name => $ds->{'tag'}, + path => $f, + headline => $headline, + words => join(" ",@{$ds->{'swish'}}) + ); + } - $swish_data .= $swish if ($s && $swish); - $display_data .= $display if ($d && $display); - $index_data .= $display if ($i && $display); - } + # save into sorted index (thesaurus) + foreach my $ds (@ds) { + next if (! $ds->{'index'}); -#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); - if ($display_data) { - $display_data = $isis_codepage->convert($display_data)."##" || $display_data; - $xml .= xmlify($field."_display", $display_data); - } - if ($swish_data) { - my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); - $swish_data = $i->convert($swish_data); - $xml .= xmlify($field."_swish",unac_string('ISO8859-2',$swish_data)); - #$swish_data = $isis_codepage->convert($swish_data)."##" || $swish_data; - #$xml .= xmlify($field."_swish",unac_string($config->{isis_codepage},$swish_data)); - } + $thes->{$ds->{'tag'}} ||= new WebPAC::Index; - # index - if ($index_data && $index_data ne "") { - $index_data = $index_codepage->convert($index_data) || $index_data; - $index->insert($field, $index_data, $db_dir); + foreach my $h (@{$ds->{'index'}}) { + $thes->{$ds->{'tag'}}->insert( + path => $f, + headline => $h, + ); } - - } - if ($xml) { -#print STDERR "x: $xml\n"; - return "\n$xml\n"; - } else { - return; } -} -########################################################################## +# print Dumper(\@ds); -my $last_tell=0; +} -my @isis_dirs = ( '.' ); # use dirname as database name +foreach my $t (keys %{$thes}) { -if ($opts{m}) { - @isis_dirs = split(/,/,$opts{m}); + my @e = $thes->{$t}->elements; + if (! @e) { + $log->logwarn("no elements in sorted index $t?"); + next; + } + + my $file = "./out/bfilter/$t.txt"; + $log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]"); + + $webpac->output_file( + file => $file, + template => 'index.tt', + data => \@e, + index_name => $t, + ); } -my @isis_dbs; - -foreach (@isis_dirs) { - if (-e $config->{isis_data}."/$db_dir/$_/LIBRI") { - push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI"; - } - if (-e $config->{isis_data}."/$db_dir/$_/PERI") { - push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI"; - } - if (-e $config->{isis_data}."/$db_dir/$_/AMS") { - push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS"; - } - if (-e $config->{isis_data}."/$db_dir/$_/ARTI") { -# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI"; +if (0 && $log->is_debug) { + $log->debug("lookup hash: ",Dumper($webpac->{'lookup'})); + $log->debug("data hash: ",Dumper($webpac->{'data'})); + foreach my $t (keys %{$thes}) { + $log->debug("thesaurus $t hash: ",Dumper($thes->{$t})); } } -print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs); +$log->info("lookup hash: ",Dumper($webpac->{'lookup'})); -my $db; +# +# tree dump implementation which shouldn't be here :-) +# -foreach my $isis_db (@isis_dbs) { +my $file = 'out/browse.html'; +open(HTML, "> $file") || $log->logdie("can't open '$file'"); +print HTML qq{ + + + Browse + + + + + + +}; -SWISH-E web site at http://www.swish-e.org +$index->close; -=cut +$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'}));