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;
use Config::IniFiles;
use Encode;
my $config_file = $0;
$config_file =~ s/\.pl$/.conf/;
+$config_file = $ARGV[0] if ($ARGV[0] && -f $ARGV[0]);
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
my $config;
}
+ # init variables for different types
+ sub init_visible_type($) {
+ 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,$il) = (0,1,0,0,1);
+ } elsif ($type) {
+ print STDERR "WARNING: unknown type: $type\n";
+ }
+ return ($s,$se,$d,$i,$il);
+ }
+
+
+ # convert
+ #
+ # <tag>
+ # <delimiter>, </delimiter>
+ # <value>200a</value>
+ # </tag>
+ #
+ # to
+ #
+ # <tag delimiter=", ">200a</tag>
+ #
+ # 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) {
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
+ $x = unroll_x($x);
+
my $format = x($x->{content});
my $delimiter = x($x->{delimiter}) || ' ';
my $repeat_off = 0; # init repeatable offset
- # swish, swish_exact, display, index, index_lookup
- # swish and display defaults
- my ($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");
- ($s,$se,$d,$i) = (0,1,0,1) 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/);
+ my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
# what will separate last line from this one?
if ($display_data && $x->{append}) {
- $line_delimiter = ' ';
+ $line_delimiter = $delimiter;
} elsif ($display_data) {
$line_delimiter = '<br/>';
}
# placeholder for all repeatable entries for index
- sub chk_eval($) {
- my $data = shift;
- if ($data =~ s/eval{([^}]+)}//) {
- if (eval "$1") {
- return $data;
- } else {
- return undef;
- }
- } else {
- return $data;
- }
- }
-
sub mkformat($$) {
my $x = shift || die "mkformat needs tag reference";
my $data = shift || return;
- my $format_name = x($x->{format_name}) || return chk_eval($data);
+ my $format_name = x($x->{format_name}) || return $data;
my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!";
my $format_delimiter = x($x->{format_delimiter});
my @data;
if ($fmt) {
my $nr = scalar $fmt =~ s/%s/%s/g;
if (($#data+1) == $nr) {
- return chk_eval(sprintf($fmt,@data));
+ return sprintf($fmt,@data);
} else {
#print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
- return chk_eval($data);
+ return $data;
}
} else {
print STDERR "usage of link '$format_name' without defined format (<link> tag)\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} = $null;
}
}
# 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 .= "xxbxx $tmp xxexx " if ($se && $tmp ne "");
-
- } else {
- $swish_data .= $swish if ($s);
- $swish_exact_data .= "xxbxx $swish xxexx " if ($se && $swish ne "");
+ $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
no strict 'refs';
$idisplay = &$filter($idisplay);
}
- push @index_data, $idisplay if (! $iterate_by_page);
+ push @index_data, $idisplay if ($idisplay && !$iterate_by_page);
}
# store fields in lookup
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'?";
}
# 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);
}
$swish_data =~ s/ +/ /g;
$swish_data =~ s/ +$//g;
- $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
+ $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
}
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
# add delimiters before and after word.
# That is required to produce exact match
- $xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data));
+ $xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data));
}
my $idel = $cache->{index_delimiter}->{$field};
$swish_data =~ s/ +/ /g;
$swish_data =~ s/ +$//g;
- $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
+ $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
}
if ($swish_exact_data) {
# add delimiters before and after word.
# That is required to produce exact match
- $xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data));
+ $xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data));
}
}
}
my $show_progress = $cfg_global->val('global', 'show_progress');
-my $unac_filter = $cfg_global->val('global', 'unac_filter');
-if ($unac_filter) {
- require $unac_filter;
+my $my_unac_filter = $cfg_global->val('global', 'my_unac_filter');
+if ($my_unac_filter) {
+ print STDERR "using $my_unac_filter to filter characters for search\n";
+ require $my_unac_filter;
+} else {
+ print STDERR "### fallback to default my_unac_string!\n";
+ eval q{
+ sub main::my_unac_string($$) {
+ my ($charset, $string) = (@_);
+ return $string;
+ }
+ };
}
foreach my $database ($cfg->Sections) {
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional
if ($lookup_file) {
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
+ if (! -e $lookup_file) {
+ open(LOOKUP, "> $lookup_file") || die "can't create $lookup_file': $!";
+ close(LOOKUP);
+ }
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
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 $last_p = 0;
print STDERR "\n";
} elsif ($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!";
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;
+ # this conversion is a cludge.
+ # Files from Excell could have
+ # characters which don't fit into
+ # destination encoding.
+ $row->{int2col($iC)} = $utf2cp->convert($cell->Value) || $cell->Value;
}
}
}
} elsif ($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!";
all2xml.pl - read various file formats and dump XML for SWISH-E
+=head1 SYNOPSYS
+
+ $ all2xml.pl [test.conf]
+
=head1 DESCRIPTION
This command will read ISIS data file using OpenIsis perl module, MARC
this script B<isn't general xml generator> from isis files (isis allready
has something like that). Output of this script is tailor-made for SWISH-E.
+If no configuration file is specified, it will use default one called
+C<all2xml.conf>.
+
=head1 BUGS
Documentation is really lacking. However, in true Open Source spirit, source
projects / webpac / blobdiff