--- trunk/all2xml.pl 2003/02/23 15:47:40 35
+++ trunk/all2xml.pl 2003/07/14 10:54:34 102
@@ -8,6 +8,7 @@
use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
use Text::Iconv;
use Config::IniFiles;
+use Encode;
$|=1;
@@ -17,9 +18,9 @@
my $config;
-use index_DBI; # there is no other, right now ;-)
-
-my $index = new index_DBI(); # open index
+#use index_DBI; # default DBI module for index
+use index_DBI_cache; # faster DBI module using memory cache
+my $index;
my %opts;
@@ -31,29 +32,45 @@
getopts('d:m:qs', \%opts);
-my $db_dir;
-
-#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts);
+my $path; # this is name of database
-#print Dumper($config->{indexer});
-#print "-" x 70,"\n";
+Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
-Text::Iconv->raise_error(1); # Conversion errors 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]);
+}
-my $isis_codepage;
-my $index_codepage;
-my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1');
-my $xml_codepage;
+# decode isis/excel or other import codepage
+my $import2cp;
-my $XML_CHARSET = 'UTF8';
+# outgoing xml must be in UTF-8
+my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
+# mapping between data type and tag which specify
+# format in XML file
+my %type2tag = (
+ 'isis' => 'isis',
+ 'excel' => 'column',
+ 'marc' => 'marc',
+ 'feed' => 'feed'
+);
-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;
@@ -65,15 +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 $swish_data = "";
@@ -82,26 +101,18 @@
my ($swish,$display);
- foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
+ my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
+ foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
- my $format = $x->{content};
- my $delimiter = $x->{delimiter} || ' ';
+ my $format = x($x->{content});
+ my $delimiter = x($x->{delimiter}) || ' ';
- # FIX: this is ugly, UGLY, cludge string is returned
- # in UTF8 encoding , but as if source charset
- # is ISO8859-1 and not some other. This break other
- # character encodings, so we convert it first
- # back to ISO8859-1
- $format = $xml_codepage->convert($format);
- $delimiter = $xml_codepage->convert($delimiter) if ($delimiter);
-
- my $isis_i = 0; # isis repeatable offset
+ my $repeat_off = 0; # repeatable offset
my ($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 ## $format ##\n";
# what will separate last line from this one?
if ($display_data && $x->{append} && $x->{append} eq "1") {
@@ -113,12 +124,18 @@
# init vars so that we go into while...
($swish,$display) = (1,1);
- while ($swish || $display) {
- ($swish,$display) = parse_format($format,$row,$isis_i++);
-#print STDERR "s: $swish\nd: $display\n" if ($swish);
-
-#print STDERR "swish: $swish<-- display: $display<--\n";
+ # 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 = $x->{filter};
@@ -128,7 +145,6 @@
# type="swish" ; field for swish
if ($s && $swish) {
if ($filter) {
-#print STDERR "using filter '$filter'\n";
no strict 'refs';
$swish_data .= join(" ",&$filter($swish));
} else {
@@ -156,26 +172,47 @@
# type="index" ; insert into index
if ($i && $display) {
- my $index_data = $index_codepage->convert($display) || $display;
- if ($filter) {
- no strict 'refs';
- foreach my $d (&$filter($index_data)) {
- $index->insert($field, $d, $db_dir);
- }
- } else {
- $index->insert($field, $index_data, $db_dir);
+ 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'}}) {
-#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data);
- if ($display_data) {
+ my $delimiter = x($x->{delimiter}) || ' ';
+ my $val = $cfg->val($database, x($x->{content}));
-# $display_data = $isis_codepage->convert($display_data) || die "Can't convert '$display_data' !";
- # FIX: this is removed and replaced by html tag.
- #$xml .= xmlify($field."_display", $display_data);
+ my ($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 ($display_data) {
if ($field eq "headline") {
$xml .= xmlify("headline", $display_data);
@@ -193,21 +230,17 @@
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
}
if ($field_name) {
- $html .= $xml_codepage->convert($field_name);
+ $html .= x($field_name);
}
$html .= $display_data."###\n";
}
}
if ($swish_data) {
- my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2');
# remove extra spaces
$swish_data =~ s/ +/ /g;
$swish_data =~ s/ +$//g;
- $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));
+ $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
}
@@ -215,11 +248,11 @@
# dump formatted output in
if ($html) {
- $xml .= xmlify("html",$html);
+ #$xml .= xmlify("html",$html);
+ $xml .= "";
}
if ($xml) {
-#print STDERR "x: $xml\n";
$xml .= $add_xml if ($add_xml);
return "\n$xml\n";
} else {
@@ -229,59 +262,238 @@
##########################################################################
+# read configuration for this script
my $cfg = new Config::IniFiles( -file => $config_file );
+# read global.conf configuration
+my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
+
+# open index
+$index = new index_DBI(
+ $cfg_global->val('global', 'dbi_dbd'),
+ $cfg_global->val('global', 'dbi_dsn'),
+ $cfg_global->val('global', 'dbi_user'),
+ $cfg_global->val('global', 'dbi_passwd') || '',
+ );
+
+my $show_progress = $cfg_global->val('global', 'show_progress');
+
foreach my $database ($cfg->Sections) {
- my $isis_db = $cfg -> val($database, 'isis_db');
- my $type = $cfg -> val($database, 'type');
- my $add_xml = $cfg -> val($database, 'xml');
-
- # read configuration for this type
- $config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1);
- $isis_codepage = Text::Iconv->new($config->{isis_codepage},$XML_CHARSET);
- $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage});
- $xml_codepage = Text::Iconv->new($cfg->val($database,'xml_codepage'),$XML_CHARSET);
-
- my $db = OpenIsis::open( $isis_db );
- if (0) {
-# # FIX
-# if (! $db ) {
- print STDERR "WARNING: can't open '$isis_db'\n";
- next ;
- }
+ my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
+ my $add_xml = $cfg -> val($database, 'xml'); # optional
- my $max_rowid = OpenIsis::maxRowid( $db );
+print STDERR "reading ./import_xml/$type.xml\n";
- print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
+ # extract just type basic
+ my $type_base = $type;
+ $type_base =~ s/_.+$//g;
- my $path = $database; # was $isis_db
+ $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1);
+ # output current progress indicator
my $last_p = 0;
+ sub progress {
+ return if (! $show_progress);
+ my $current = 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 );
+ $last_p = $p;
+ }
+ }
+
+ 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_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";
-# { my $row_id = 4514;
-# FIX
- for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
- my $row = OpenIsis::read( $db, $row_id );
- if ($row && $row->{mfn}) {
-#print STDERR "mfn: ",$row->{mfn},"\n";
- # output current process indicator
- my $p = int($row->{mfn} * 100 / $max_rowid);
- if ($p != $last_p) {
- printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q});
- $last_p = $p;
+ my $path = $database;
+
+ for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
+ my $row = OpenIsis::read( $db, $row_id );
+ if ($row && $row->{mfn}) {
+
+ progress($row->{mfn}, $max_rowid);
+
+ my $swishpath = $path."#".int($row->{mfn});
+
+ if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
+ $xml = $cp2utf->convert($xml);
+ use bytes; # as opposed to chars
+ print "Path-Name: $swishpath\n";
+ print "Content-Length: ".(length($xml)+1)."\n";
+ print "Document-Type: XML\n\n$xml\n";
+ }
+ }
+ }
+ print STDERR "\n";
+
+ } elsif ($type_base eq "excel") {
+ use Spreadsheet::ParseExcel;
+ use 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!";
+
+ my $sheet = x($config->{sheet}) || die "no sheet 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'";
+
+ my $sheet_nr = 0;
+ foreach my $oWks (@{$oBook->{Worksheet}}) {
+ #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
+ last if ($oWks->{Name} eq $sheet);
+ $sheet_nr++;
+ }
+
+ my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
+
+ print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
+ defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
+ my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
+
+ 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][$iC];
+ if ($cell) {
+ $row->{int2col($iC)} = $cell->Value;
+ }
}
- if (my $xml = isis2xml($row,$add_xml)) {
-#print STDERR "--ret-->$xml\n";
+ progress($iR, $end_row);
+
+# print "row[$iR/$end_row] ";
+# foreach (keys %{$row}) {
+# print "$_: ",$row->{$_},"\t";
+# }
+# print "\n";
+
+ my $swishpath = $database."#".$iR;
+
+ next if (! $row);
+
+ if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
+ $xml = $cp2utf->convert($xml);
use bytes; # as opposed to chars
- print "Path-Name: $path#".int($row->{mfn})."\n";
+ print "Path-Name: $swishpath\n";
print "Content-Length: ".(length($xml)+1)."\n";
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);
+
+ }
}
- print STDERR "\n";
}
# call this to commit index
@@ -293,15 +505,21 @@
=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