--- trunk/all2xml.pl 2003/06/23 20:20:32 54 +++ trunk/all2xml.pl 2003/07/14 10:52:13 101 @@ -18,7 +18,8 @@ my $config; -use index_DBI; # there is no other, right now ;-) +#use index_DBI; # default DBI module for index +use index_DBI_cache; # faster DBI module using memory cache my $index; my %opts; @@ -33,7 +34,7 @@ my $path; # this is name of database -Text::Iconv->raise_error(1); # Conversion errors raise exceptions +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! @@ -55,7 +56,9 @@ # format in XML file my %type2tag = ( 'isis' => 'isis', - 'excel' => 'column' + 'excel' => 'column', + 'marc' => 'marc', + 'feed' => 'feed' ); sub data2xml { @@ -65,6 +68,9 @@ 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; @@ -76,17 +82,17 @@ # sort subrouting using order="" attribute sub by_order { - return 0 if (! $config->{indexer}->{$a}->{order}); - return 0 if (! $config->{indexer}->{$b}->{order}); + my $va = $config->{indexer}->{$a}->{order} || + $config->{indexer}->{$a}; + my $vb = $config->{indexer}->{$b}->{order} || + $config->{indexer}->{$b}; - return $config->{indexer}->{$a}->{order} <=> - $config->{indexer}->{$b}->{order} ; + return $va <=> $vb; } foreach my $field (sort by_order keys %{$config->{indexer}}) { $field=x($field); - $field_usage{$field}++; my $swish_data = ""; @@ -118,8 +124,18 @@ # init vars so that we go into while... ($swish,$display) = (1,1); - if ($swish || $display) { + # placeholder for all repeatable entries for index + my @index_data; + my $index_filter; + + # while because of repeatable fields + while ($swish || $display) { ($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; + } + # filter="name" ; filter this field through # filter/[name].pm my $filter = $x->{filter}; @@ -156,19 +172,45 @@ # type="index" ; insert into index if ($i && $display) { - my $index_data = $display; - if ($filter) { - no strict 'refs'; - foreach my $d (&$filter($index_data)) { - $index->insert($field, $d, $path); - } - } else { - $index->insert($field, $index_data, $path); + push @index_data, $display; + $index_filter = $filter if ($filter); + } + } + + # fill data in index + if (@index_data) { + if ($index_filter) { + no strict 'refs'; + foreach my $d (&$index_filter(@index_data)) { + $index->insert($field, $d, $path); + } + } else { + foreach my $d (@index_data) { + $index->insert($field, $d, $path); } } } } + # 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"); + ($s,$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 ($display_data) { @@ -206,7 +248,8 @@ # dump formatted output in if ($html) { - $xml .= xmlify("html",$html); + #$xml .= xmlify("html",$html); + $xml .= ""; } if ($xml) { @@ -233,6 +276,8 @@ $cfg_global->val('global', 'dbi_passwd') || '', ); +my $show_progress = $cfg_global->val('global', 'show_progress'); + foreach my $database ($cfg->Sections) { my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined"; @@ -240,14 +285,19 @@ print STDERR "reading ./import_xml/$type.xml\n"; - $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type} ], forcecontent => 1); + # extract just type basic + my $type_base = $type; + $type_base =~ s/_.+$//g; + + $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1); # output current progress indicator my $last_p = 0; +print STDERR "## show_progress: $show_progress ##\n"; # XXX sub progress { - #return if (! $opts{q}); # FIXME + return if (! $show_progress); my $current = shift; - my $total = shift; + my $total = shift || 1; my $p = int($current * 100 / $total); if ($p != $last_p) { printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p ); @@ -255,15 +305,42 @@ } } + my $fake_dir = 1; + sub fakeprogress { + my $current = shift @_; + + my @ind = ('-','\\','|','/','-','\\','|','/', '-'); + + $last_p += $fake_dir; + $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0); + if ($last_p % 10 == 0) { + printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]); + } + } + # now read database print STDERR "using: $type...\n"; - if ($type eq "isis") { + if ($type_base eq "isis") { + my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage); my $db = OpenIsis::open( $isis_db ); + # OpenIsis::ERR_BADF + if ($db == -4) { + print STDERR "FATAL: OpenIsis can't find file $isis_db\n"; + next; + # OpenIsis::ERR_IO + } elsif ($db == -5) { + print STDERR "FATAL: OpenIsis can't access file $isis_db\n"; + next; + } elsif ($db < 0) { + print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n"; + next; + } + my $max_rowid = OpenIsis::maxRowid( $db ); print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; @@ -278,7 +355,7 @@ my $swishpath = $path."#".int($row->{mfn}); - if (my $xml = data2xml($type,$row,$add_xml)) { + if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { $xml = $cp2utf->convert($xml); use bytes; # as opposed to chars print "Path-Name: $swishpath\n"; @@ -289,7 +366,7 @@ } print STDERR "\n"; - } elsif ($type eq "excel") { + } elsif ($type_base eq "excel") { use Spreadsheet::ParseExcel; use Spreadsheet::ParseExcel::Utility qw(int2col); @@ -297,7 +374,7 @@ my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; my $sheet = x($config->{sheet}) || die "no sheet in $type.xml"; - my $start_row = x($config->{start_row}) || die "no start_row in $type.xml"; + my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml"; my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'"; @@ -314,7 +391,7 @@ defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file"; my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow}; - for(my $iR = $oWorksheet->{MinRow} ; defined $end_row && $iR <= $end_row ; $iR++) { + for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) { my $row; for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { my $cell = $oWorksheet->{Cells}[$iR][$iC]; @@ -335,7 +412,7 @@ next if (! $row); - if (my $xml = data2xml($type,$row,$add_xml)) { + if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { $xml = $cp2utf->convert($xml); use bytes; # as opposed to chars print "Path-Name: $swishpath\n"; @@ -343,6 +420,80 @@ print "Document-Type: XML\n\n$xml\n"; } } + } elsif ($type_base eq "marc") { + + use 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'; + + 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'"; + + my $i=0; # record nr. + + my $rec; + + while ($marc->nextmarc(1)) { + + # XXX + fakeprogress($i++); + + my $swishpath = $database."#".$i; + + if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) { + $xml = $cp2utf->convert($xml); + use bytes; # as opposed to chars + print "Path-Name: $swishpath\n"; + print "Content-Length: ".(length($xml)+1)."\n"; + print "Document-Type: XML\n\n$xml\n"; + } + } + } elsif ($type_base eq "feed") { + + $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); + my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!"; + + print STDERR "Reading feed from program '$prog'\n"; + + open(FEED,"feeds/$prog |") || die "can't start $prog: $!"; + + my $i=1; # record nr. + + my $data; + my $line=1; + + while () { + chomp; + + if (/^$/) { + my $swishpath = $database."#".$i++; + + if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) { + $xml = $cp2utf->convert($xml); + use bytes; # as opposed to chars + print "Path-Name: $swishpath\n"; + print "Content-Length: ".(length($xml)+1)."\n"; + print "Document-Type: XML\n\n$xml\n"; + } + $line = 1; + $data = {}; + next; + } + + $line = $1 if (s/^(\d+):\s*//); + $data->{$line++} = $_; + + fakeprogress($i); + + } } } @@ -355,15 +506,21 @@ =head1 NAME -isis2xml.pl - read isis file and dump XML +all2xml.pl - read various file formats and dump XML for SWISH-E =head1 DESCRIPTION -This command will read ISIS data file using OpenIsis perl module and -create XML file for usage with I -indexer. Dispite it's name, this script B -from isis files (isis allready has something like that). Output of this -script is tailor-made for SWISH-E. +This command will read ISIS data file using OpenIsis perl module, MARC +records using MARC module and optionally Micro$oft Excel files to +create one XML file for usage with I indexer. Dispite it's name, +this script B from isis files (isis allready +has something like that). Output of this script is tailor-made for SWISH-E. + +=head1 BUGS + +Documentation is really lacking. However, in true Open Source spirit, source +is best documentation. I even made considerable effort to comment parts +which are not intuitively clear, so... =head1 AUTHOR