--- trunk/all2xml.pl 2004/03/05 23:33:19 233
+++ branches/humanistika/all2xml.pl 2005/01/16 18:57:15 633
@@ -1,11 +1,10 @@
#!/usr/bin/perl -w
use strict;
-use OpenIsis;
+use Biblio::Isis;
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;
@@ -17,6 +16,7 @@
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 $config;
@@ -139,18 +139,70 @@
} else {
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
}
+
if ($field_name) {
+ $field_name = x($field_name);
if (! $last_field_name) {
- $last_field_name = x($field_name);
+ $last_field_name = $field_name;
return $last_field_name;
} elsif ($field_name ne $last_field_name) {
- $last_field_name = x($field_name);
+ $last_field_name = $field_name;
return $last_field_name;
}
}
}
+ # init variables for different types
+ sub init_visible_type($) {
+ 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,$il) = (0,1,0,0,1);
+ }
+ return ($s,$se,$d,$i,$il);
+ }
+
+
+ # convert
+ #
+ #
+ # ,
+ # 200a
+ #
+ #
+ # to
+ #
+ # 200a
+ #
+ # 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) {
@@ -184,23 +236,18 @@
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) {
$line_delimiter = '
';
}
@@ -210,23 +257,10 @@
# placeholder for all repeatable entries for index
- sub chk_eval($) {
- 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 " is not defined!";
my $format_delimiter = x($x->{format_delimiter});
my @data;
@@ -239,10 +273,10 @@
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 ( tag)\n";
@@ -270,7 +304,7 @@
$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;
}
@@ -291,16 +325,14 @@
}
# 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
@@ -333,7 +365,7 @@
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
@@ -432,19 +464,17 @@
# 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);
}
@@ -497,7 +527,7 @@
$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];
@@ -507,7 +537,7 @@
# 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};
@@ -540,7 +570,7 @@
$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) {
@@ -549,7 +579,7 @@
# 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));
}
}
}
@@ -586,9 +616,18 @@
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) {
@@ -600,6 +639,10 @@
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
@@ -620,32 +663,64 @@
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 );
+
+ # helper for progress bar
+ sub fmt_time {
+ my $t = shift || 0;
+ my $out = "";
+
+ my ($ss,$mm,$hh) = gmtime($t);
+ $out .= "${hh}h" if ($hh);
+ $out .= sprintf("%02d:%02d", $mm,$ss);
+ $out .= " " if ($hh == 0);
+ return $out;
+ }
# output current progress indicator
my $last_p = 0;
+ my $start_t = time();
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).">", $p );
+ if ($p < $last_p || $current == 1) {
+ $start_t = time();
+ $last_p = 0;
+ } elsif ($p != $last_p) {
+ my $rate = ($current / (time() - $start_t || 1));
+ my $eta = ($total-$current) / ($rate || 1);
+ printf STDERR ("%5d [%-38s] %-5d %0.1f/s %s\r",$current,"=" x ($p/3)."$p%>", $total, $rate, fmt_time($eta));
$last_p = $p;
}
}
my $fake_dir = 1;
+ my $fake_pos = 0;
+ my $last_fake_t = time();
sub fakeprogress {
return if (! $show_progress);
my $current = shift @_;
- my @ind = ('-','\\','|','/','-','\\','|','/', '-');
+ 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) % $#ind]);
+ if ($current < $fake_pos) {
+ $start_t = time();
+ $last_fake_t = 0;
+ $fake_dir = 1;
+ $fake_pos = 0;
+ }
+
+ if (time()-$last_fake_t >= 1) {
+ $last_fake_t = time();
+ $fake_pos += $fake_dir;
+ $fake_dir = -$fake_dir if ($fake_pos > 38);
+ }
+
+ if ($current % 10 == 0) {
+ my $rate = ($current / (time() - $start_t || 1));
+ printf STDERR ("%5d [%-38s] %0.1f/s\r",$current, " " x $fake_pos .$ind[($current / 10) % 8], $rate);
}
}
@@ -660,66 +735,21 @@
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 );
+ my $db = new Biblio::Isis( isisdb => $isis_db );
- # check if .txt database for OpenIsis is zero length,
- # if so, erase it and re-open database
- sub check_txt_db {
- my $isis_db = shift || die "need isis database name";
- my $reopen = 0;
-
- if (-e $isis_db.".TXT") {
- print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
- unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
- $reopen++;
- }
- if (-e $isis_db.".PTR") {
- print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
- unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
- $reopen++;
- }
- return OpenIsis::open( $isis_db ) if ($reopen);
- }
-
- # EOF error
- if ($db == -1) {
- $db = check_txt_db($isis_db);
- if ($db == -1) {
- print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
- next;
- }
- }
-
- # OpenIsis::ERR_BADF
- if ($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( $db );
-
- # if 0 records, try to rease isis .txt database
- if ($max_rowid == 0) {
- # force removal of database
- $db = check_txt_db($isis_db);
- $max_rowid = OpenIsis::maxRowid( $db );
- }
+ my $max_rowid = $db->count || die "can't find maxmfn";
print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
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}) {
-
+ my $row = $db->to_hash( $row_id );
+ if ($row) {
+
+ $row->{mfn} = $row_id;
+ $row->{record} = $db->{record};
+
progress($row->{mfn}, $max_rowid);
my $swishpath = $path."#".int($row->{mfn});
@@ -733,15 +763,12 @@
}
}
}
- # for this to work with current version of OpenIsis (0.9.0)
- # you might need my patch from
- # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
- OpenIsis::close($db);
print STDERR "\n";
} elsif ($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!";
@@ -769,7 +796,11 @@
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) || $cell->Value;
}
}
@@ -795,36 +826,40 @@
}
} elsif ($type_base eq "marc") {
- use MARC;
+ require MARC::File::USMARC;
$import2cp = Text::Iconv->new($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->{marc_format}) || 'usmarc';
-
+ warn "marc_format is no longer used!" if ($config->{marc_format});
print STDERR "Reading MARC file '$marc_file'\n";
- my $marc = new MARC;
- my $nr = $marc->openmarc({
- file=>$marc_file, format=>$format
- }) || die "Can't open MARC file '$marc_file' with format '$format'";
+ my $marc = MARC::File::USMARC->in( $marc_file )
+ || die "Can't open MARC file '$marc_file': ".$MARC::File::ERROR;
- # read MARC file in memory
- $marc->nextmarc(-1);
+ # count records in MARC file
+ sub marc_count {
+ my $filename = shift || die;
+ my $file = MARC::File::USMARC->in($filename) || die $MARC::File::ERROR;
+ my $count = 0;
+ while ($file->skip()) {
+ $count++;
+ }
+ return $count;
+ }
- my $max_rec = $marc->marc_count();
+ my $count = marc_count($marc_file) || warn "no records in '$marc_file'?";
- for(my $i=1; $i<=$max_rec; $i++) {
+ my $i = 0;
- progress($i,$max_rec);
+ while( my $rec = $marc->next() ) {
- # store value for marc_sf.pm
- $main::cache->{marc_record} = $i;
+ progress($i++,$count);
my $swishpath = $database."#".$i;
- if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
+ if (my $xml = data2xml($type_base,$rec,$add_xml,$cfg,$database)) {
$xml = $cp2utf->convert($xml);
use bytes; # as opposed to chars
print "Path-Name: $swishpath\n";
@@ -889,14 +924,21 @@
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 command will read ISIS data file using IsisDB 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.
+If no configuration file is specified, it will use default one called
+C.
+
=head1 BUGS
Documentation is really lacking. However, in true Open Source spirit, source