--- trunk/lib/WebPAC/Normalize.pm	2006/01/08 21:50:34	372
+++ trunk/lib/WebPAC/Normalize.pm	2006/10/08 18:21:26	752
@@ -1,752 +1,1252 @@
 package WebPAC::Normalize;
+use Exporter 'import';
+@EXPORT = qw/
+	_set_rec _set_lookup
+	_set_load_row
+	_get_ds _clean_ds
+	_debug
+	_pack_subfields_hash
+
+	tag search display
+	marc marc_indicators marc_repeatable_subfield
+	marc_compose marc_leader
+	marc_duplicate marc_remove
+	marc_original_order
+
+	rec1 rec2 rec
+	regex prefix suffix surround
+	first lookup join_with
+	save_into_lookup
+
+	split_rec_on
+/;
 
 use warnings;
 use strict;
-use blib;
-use WebPAC::Common;
-use base 'WebPAC::Common';
-use Data::Dumper;
+
+#use base qw/WebPAC::Common/;
+use Data::Dump qw/dump/;
+use Storable qw/dclone/;
+use Carp qw/confess/;
+
+# debugging warn(s)
+my $debug = 0;
+
 
 =head1 NAME
 
-WebPAC::Normalize - data mungling for normalisation
+WebPAC::Normalize - describe normalisaton rules using sets
 
 =head1 VERSION
 
-Version 0.08
+Version 0.23
 
 =cut
 
-our $VERSION = '0.08';
+our $VERSION = '0.23';
 
 =head1 SYNOPSIS
 
-This package contains code that mungle data to produce normalized format.
+This module uses C<conf/normalize/*.pl> files to perform normalisation
+from input records using perl functions which are specialized for set
+processing.
+
+Sets are implemented as arrays, and normalisation file is valid perl, which
+means that you check it's validity before running WebPAC using
+C<perl -c normalize.pl>.
+
+Normalisation can generate multiple output normalized data. For now, supported output
+types (on the left side of definition) are: C<tag>, C<display>, C<search> and
+C<marc>.
 
-It contains several assumptions:
+=head1 FUNCTIONS
 
-=over
+Functions which start with C<_> are private and used by WebPAC internally.
+All other functions are available for use within normalisation rules.
 
-=item *
+=head2 data_structure
+
+Return data structure
 
-format of fields is defined using C<v123^a> notation for repeatable fields
-or C<s123^a> for single (or first) value, where C<123> is field number and
-C<a> is subfield.
+  my $ds = WebPAC::Normalize::data_structure(
+	lookup => $lookup_hash,
+	row => $row,
+	rules => $normalize_pl_config,
+	marc_encoding => 'utf-8',
+	config => $config,
+	load_row_coderef => sub {
+		my ($database,$input,$mfn) = shift;
+		$store->load_row( database => $database, input => $input, id => $mfn );
+	},
+  );
 
-=item *
+Options C<row>, C<rules> and C<log> are mandatory while all
+other are optional.
 
-source data records (C<$rec>) have unique identifiers in field C<000>
+C<load_row_coderef> is closure only used when executing lookups, so they will
+die if it's not defined.
 
-=item *
+This function will B<die> if normalizastion can't be evaled.
 
-optional C<eval{length('v123^a') == 3}> tag at B<beginning of format> will be
-perl code that is evaluated before producing output (value of field will be
-interpolated before that)
+Since this function isn't exported you have to call it with 
+C<WebPAC::Normalize::data_structure>.
 
-=item *
+=cut
 
-optional C<filter{filter_name}> at B<begining of format> will apply perl
-code defined as code ref on format after field substitution to producing
-output
+my $load_row_coderef;
 
-There is one built-in filter called C<regex> which can be use like this:
+sub data_structure {
+	my $arg = {@_};
 
-  filter{regex(s/foo/bar/)}
+	die "need row argument" unless ($arg->{row});
+	die "need normalisation argument" unless ($arg->{rules});
 
-=item *
+	no strict 'subs';
+	_set_lookup( $arg->{lookup} ) if defined($arg->{lookup});
+	_set_rec( $arg->{row} );
+	_set_config( $arg->{config} ) if defined($arg->{config});
+	_clean_ds( %{ $arg } );
+	$load_row_coderef = $arg->{load_row_coderef};
 
-optional C<lookup{...}> will be then performed. See C<WebPAC::Lookups>.
+	eval "$arg->{rules}";
+	die "error evaling $arg->{rules}: $@\n" if ($@);
 
-=item *
+	return _get_ds();
+}
 
-at end, optional C<format>s rules are resolved. Format rules are similar to
-C<sprintf> and can also contain C<lookup{...}> which is performed after
-values are inserted in format.
+=head2 _set_rec
 
-=back
+Set current record hash
 
-This also describes order in which transformations are applied (eval,
-filter, lookup, format) which is important to undestand when deciding how to
-solve your data mungling and normalisation process.
+  _set_rec( $rec );
 
+=cut
 
+my $rec;
 
+sub _set_rec {
+	$rec = shift or die "no record hash";
+}
 
-=head1 FUNCTIONS
+=head2 _set_config
 
-=head2 new
+Set current config hash
 
-Create new normalisation object
+  _set_config( $config );
 
-  my $n = new WebPAC::Normalize::Something(
- 	filter => {
-		'filter_name_1' => sub {
-			# filter code
-			return length($_);
-		}, ...
-	},
-	db => $db_obj,
-	lookup_regex => $lookup->regex,
-	lookup => $lookup_obj,
-	prefix => 'foobar',
-  );
+Magic keys are:
 
-Parametar C<filter> defines user supplied snippets of perl code which can
-be use with C<filter{...}> notation.
+=over 4
 
-C<prefix> is used to form filename for database record (to support multiple
-source files which are joined in one database).
+=item _
 
-Recommended parametar C<lookup_regex> is used to enable parsing of lookups
-in structures. If you pass this parametar, you must also pass C<lookup>
-which is C<WebPAC::Lookup> object.
+Code of current database
+
+=item _mfn
+
+Current MFN
+
+=back
 
 =cut
 
-sub new {
-	my $class = shift;
-        my $self = {@_};
-        bless($self, $class);
+my $config;
 
-	my $r = $self->{'lookup_regex'} ? 1 : 0;
-	my $l = $self->{'lookup'} ? 1 : 0;
+sub _set_config {
+	$config = shift;
+}
 
-	my $log = $self->_get_logger();
+=head2 _get_ds
 
-	# those two must be in pair
-	if ( ($r & $l) != ($r || $l) ) {
-		my $log = $self->_get_logger();
-		$log->logdie("lookup_regex and lookup must be in pair");
-	}
+Return hash formatted as data structure
 
-	$log->logdie("lookup must be WebPAC::Lookup object") if ($self->{'lookup'} && ! $self->{'lookup'}->isa('WebPAC::Lookup'));
+  my $ds = _get_ds();
 
-	$log->warn("no prefix defined. please check that!") unless ($self->{'prefix'});
+=cut
 
-	$log->debug("using lookup regex: ", $self->{lookup_regex}) if ($r && $l);
+my ($out, $marc_record, $marc_encoding, $marc_repeatable_subfield, $marc_indicators, $leader);
+my ($marc_record_offset, $marc_fetch_offset) = (0, 0);
 
-	if (! $self->{filter} || ! $self->{filter}->{regex}) {
-		$log->debug("adding built-in filter regex");
-		$self->{filter}->{regex} = sub {
-			my ($val, $regex) = @_;
-			eval "\$val =~ $regex";
-			return $val;
-		};
-	}
+sub _get_ds {
+	return $out;
+}
+
+=head2 _clean_ds
+
+Clean data structure hash for next record
 
-	$self ? return $self : return undef;
+  _clean_ds();
+
+=cut
+
+sub _clean_ds {
+	my $a = {@_};
+	($out,$marc_record, $marc_encoding, $marc_repeatable_subfield, $marc_indicators, $leader) = ();
+	($marc_record_offset, $marc_fetch_offset) = (0,0);
+	$marc_encoding = $a->{marc_encoding};
 }
 
+=head2 _set_lookup
 
-=head2 data_structure
+Set current lookup hash
 
-Create in-memory data structure which represents normalized layout from
-C<conf/normalize/*.xml>.
+  _set_lookup( $lookup );
 
-This structures are used to produce output.
+=cut
 
- my $ds = $webpac->data_structure($rec);
+my $lookup;
+
+sub _set_lookup {
+	$lookup = shift;
+}
+
+=head2 _get_lookup
+
+Get current lookup hash
+
+  my $lookup = _get_lookup();
 
 =cut
 
-sub data_structure {
-	my $self = shift;
+sub _get_lookup {
+	return $lookup;
+}
 
-	my $log = $self->_get_logger();
+=head2 _set_load_row
 
-	my $rec = shift;
-	$log->logconfess("need HASH as first argument!") if ($rec !~ /HASH/o);
+Setup code reference which will return L<data_structure> from
+L<WebPAC::Store>
 
-	$log->debug("data_structure rec = ", sub { Dumper($rec) });
+  _set_load_row(sub {
+		my ($database,$input,$mfn) = @_;
+		$store->load_row( database => $database, input => $input, id => $mfn );
+  });
 
-	$log->logdie("need unique ID (mfn) in field 000 of record " . Dumper($rec) ) unless (defined($rec->{'000'}));
+=cut
 
-	my $id = $rec->{'000'}->[0] || $log->logdie("field 000 isn't array!");
+sub _set_load_row {
+	my $coderef = shift;
+	confess "argument isn't CODE" unless ref($coderef) eq 'CODE';
 
-	my $cache_file;
+	$load_row_coderef = $coderef;
+}
 
-	if ($self->{'db'}) {
-		my $ds = $self->{'db'}->load_ds( id => $id, prefix => $self->{prefix} );
-		$log->debug("load_ds( rec = ", sub { Dumper($rec) }, ") = ", sub { Dumper($ds) });
-		return $ds if ($ds);
-		$log->debug("cache miss, creating");
-	}
+=head2 _get_marc_fields
 
-	my @sorted_tags;
-	if ($self->{tags_by_order}) {
-		@sorted_tags = @{$self->{tags_by_order}};
-	} else {
-		@sorted_tags = sort { $self->_sort_by_order } keys %{$self->{'import_xml'}->{'indexer'}};
-		$self->{tags_by_order} = \@sorted_tags;
+Get all fields defined by calls to C<marc>
+
+	$marc->add_fields( WebPAC::Normalize:_get_marc_fields() );
+
+We are using I<magic> which detect repeatable fields only from
+sequence of field/subfield data generated by normalization.
+
+Repeatable field is created when there is second occurence of same subfield or
+if any of indicators are different.
+
+This is sane for most cases. Something like:
+
+  900a-1 900b-1 900c-1
+  900a-2 900b-2
+  900a-3
+
+will be created from any combination of:
+
+  900a-1 900a-2 900a-3 900b-1 900b-2 900c-1
+
+and following rules:
+
+  marc('900','a', rec('200','a') );
+  marc('900','b', rec('200','b') );
+  marc('900','c', rec('200','c') );
+
+which might not be what you have in mind. If you need repeatable subfield,
+define it using C<marc_repeatable_subfield> like this:
+
+  marc_repeatable_subfield('900','a');
+  marc('900','a', rec('200','a') );
+  marc('900','b', rec('200','b') );
+  marc('900','c', rec('200','c') );
+
+will create:
+
+  900a-1 900a-2 900a-3 900b-1 900c-1
+  900b-2
+
+There is also support for returning next or specific using:
+
+  while (my $mf = WebPAC::Normalize:_get_marc_fields( fetch_next => 1 ) ) {
+  	# do something with $mf
+  }
+
+will always return fields from next MARC record or
+
+  my $mf = WebPAC::Normalize::_get_marc_fields( offset => 42 );
+
+will return 42th copy record (if it exists).
+
+=cut
+
+sub _get_marc_fields {
+
+	my $arg = {@_};
+	warn "### _get_marc_fields arg: ", dump($arg), $/ if ($debug > 2);
+	my $offset = $marc_fetch_offset;
+	if ($arg->{offset}) {
+		$offset = $arg->{offset};
+	} elsif($arg->{fetch_next}) {
+		$marc_fetch_offset++;
 	}
 
-	my $ds;
+	return if (! $marc_record || ref($marc_record) ne 'ARRAY');
 
-	$log->debug("tags: ",sub { join(", ",@sorted_tags) });
+	warn "### full marc_record = ", dump( @{ $marc_record }), $/ if ($debug > 2);
 
-	foreach my $field (@sorted_tags) {
+	my $marc_rec = $marc_record->[ $offset ];
 
-		my $row;
+	warn "## _get_marc_fields (at offset: $offset) -- marc_record = ", dump( @$marc_rec ), $/ if ($debug > 1);
 
-#print "field $field [",$self->{'tag'},"] = ",Dumper($self->{'import_xml'}->{'indexer'}->{$field}->{$self->{'tag'}});
+	return if (! $marc_rec || ref($marc_rec) ne 'ARRAY' || $#{ $marc_rec } < 0);
 
-		foreach my $tag (@{$self->{'import_xml'}->{'indexer'}->{$field}->{$self->{'tag'}}}) {
-			my $format;
+	# first, sort all existing fields 
+	# XXX might not be needed, but modern perl might randomize elements in hash
+	my @sorted_marc_record = sort {
+		$a->[0] . ( $a->[3] || '' ) cmp $b->[0] . ( $b->[3] || '')
+	} @{ $marc_rec };
 
-			$log->logdie("expected tag HASH and got $tag") unless (ref($tag) eq 'HASH');
-			$format = $tag->{'value'} || $tag->{'content'};
+	@sorted_marc_record = @{ $marc_rec };	### FIXME disable sorting
+	
+	# output marc fields
+	my @m;
 
-			my @v;
-			if ($self->{'lookup_regex'} && $format =~ $self->{'lookup_regex'}) {
-				@v = $self->_rec_to_arr($rec,$format,'fill_in');
-			} else {
-				@v = $self->_rec_to_arr($rec,$format,'parse');
-			}
-			if (! @v) {
-				$log->debug("$field <",$self->{tag},"> format: $format no values");
-#				next;
-			} else {
-				$log->debug("$field <",$self->{tag},"> format: $format values: ", join(",", @v));
-			}
+	# count unique field-subfields (used for offset when walking to next subfield)
+	my $u;
+	map { $u->{ $_->[0] . ( $_->[3] || '')  }++ } @sorted_marc_record;
+
+	if ($debug) {
+		warn "## marc_repeatable_subfield = ", dump( $marc_repeatable_subfield ), $/ if ( $marc_repeatable_subfield );
+		warn "## marc_record[$offset] = ", dump( $marc_rec ), $/;
+		warn "## sorted_marc_record = ", dump( \@sorted_marc_record ), $/;
+		warn "## subfield count = ", dump( $u ), $/;
+	}
 
-			if ($tag->{'sort'}) {
-				@v = $self->sort_arr(@v);
-			}
+	my $len = $#sorted_marc_record;
+	my $visited;
+	my $i = 0;
+	my $field;
 
-			# use format?
-			if ($tag->{'format_name'}) {
-				@v = map { $self->apply_format($tag->{'format_name'},$tag->{'format_delimiter'},$_) } @v;
-			}
+	foreach ( 0 .. $len ) {
 
-			# delimiter will join repeatable fields
-			if ($tag->{'delimiter'}) {
-				@v = ( join($tag->{'delimiter'}, @v) );
-			}
+		# find next element which isn't visited
+		while ($visited->{$i}) {
+			$i = ($i + 1) % ($len + 1);
+		}
 
-			# default types 
-			my @types = qw(display search);
-			# override by type attribute
-			@types = ( $tag->{'type'} ) if ($tag->{'type'});
-
-			foreach my $type (@types) {
-				# append to previous line?
-				$log->debug("tag $field / $type [",sub { join(",",@v) }, "] ", $row->{'append'} || 'no append');
-				if ($tag->{'append'}) {
-
-					# I will delimit appended part with
-					# delimiter (or ,)
-					my $d = $tag->{'delimiter'};
-					# default delimiter
-					$d ||= " ";
-
-					my $last = pop @{$row->{$type}};
-					$d = "" if (! $last);
-					$last .= $d . join($d, @v);
-					push @{$row->{$type}}, $last;
+		# mark it visited
+		$visited->{$i}++;
 
-				} else {
-					push @{$row->{$type}}, @v;
-				}
-			}
+		my $row = dclone( $sorted_marc_record[$i] );
 
+		# field and subfield which is key for
+		# marc_repeatable_subfield and u
+		my $fsf = $row->[0] . ( $row->[3] || '' );
+
+		if ($debug > 1) {
+
+			print "### field so far [", $#$field, "] : ", dump( $field ), " ", $field ? 'T' : 'F', $/;
+			print "### this [$i]: ", dump( $row ),$/;
+			print "### sf: ", $row->[3], " vs ", $field->[3],
+				$marc_repeatable_subfield->{ $row->[0] . $row->[3] } ? ' (repeatable)' : '', $/,
+				if ($#$field >= 0);
 
 		}
 
-		if ($row) {
-			$row->{'tag'} = $field;
+		# if field exists
+		if ( $#$field >= 0 ) {
+			if (
+				$row->[0] ne $field->[0] ||		# field
+				$row->[1] ne $field->[1] ||		# i1
+				$row->[2] ne $field->[2]		# i2
+			) {
+				push @m, $field;
+				warn "## saved/1 ", dump( $field ),$/ if ($debug);
+				$field = $row;
+
+			} elsif (
+				( $row->[3] lt $field->[-2] )		# subfield which is not next (e.g. a after c)
+				||
+				( $row->[3] eq $field->[-2] &&		# same subfield, but not repeatable
+					! $marc_repeatable_subfield->{ $fsf }
+				)
+			) {
+				push @m, $field;
+				warn "## saved/2 ", dump( $field ),$/ if ($debug);
+				$field = $row;
 
-			# TODO: name_sigular, name_plural
-			my $name = $self->{'import_xml'}->{'indexer'}->{$field}->{'name'};
-			my $row_name = $name ? $self->_x($name) : $field;
-
-			# post-sort all values in field
-			if ($self->{'import_xml'}->{'indexer'}->{$field}->{'sort'}) {
-				$log->warn("sort at field tag not implemented");
+			} else {
+				# append new subfields to existing field
+				push @$field, ( $row->[3], $row->[4] );
 			}
+		} else {
+			# insert first field
+			$field = $row;
+		}
 
-			$ds->{$row_name} = $row;
-
-			$log->debug("row $field: ",sub { Dumper($row) });
+		if (! $marc_repeatable_subfield->{ $fsf }) {
+			# make step to next subfield
+			$i = ($i + $u->{ $fsf } ) % ($len + 1);
 		}
+	}
 
+	if ($#$field >= 0) {
+		push @m, $field;
+		warn "## saved/3 ", dump( $field ),$/ if ($debug);
 	}
 
-	$self->{'db'}->save_ds(
-		id => $id,
-		ds => $ds,
-		prefix => $self->{prefix},
-	) if ($self->{'db'});
+	return \@m;
+}
 
-	$log->debug("ds: ", sub { Dumper($ds) });
+=head2 _debug
 
-	$log->logconfess("data structure returned is not array any more!") if wantarray;
+Change level of debug warnings
 
-	return $ds;
+  _debug( 2 );
 
-}
+=cut
 
-=head2 parse
+sub _debug {
+	my $l = shift;
+	return $debug unless defined($l);
+	warn "debug level $l",$/ if ($l > 0);
+	$debug = $l;
+}
 
-Perform smart parsing of string, skipping delimiters for fields which aren't
-defined. It can also eval code in format starting with C<eval{...}> and
-return output or nothing depending on eval code.
+=head1 Functions to create C<data_structure>
 
- my $text = $webpac->parse($rec,'eval{"v901^a" eq "Deskriptor"}descriptor: v250^a', $i);
+Those functions generally have to first in your normalization file.
 
-Filters are implemented here. While simple form of filters looks like this:
+=head2 tag
 
-  filter{name_of_filter}
+Define new tag for I<search> and I<display>.
 
-but, filters can also have variable number of parametars like this:
+  tag('Title', rec('200','a') );
 
-  filter{name_of_filter(param,param,param)}
 
 =cut
 
-my $warn_once;
+sub tag {
+	my $name = shift or die "tag needs name as first argument";
+	my @o = grep { defined($_) && $_ ne '' } @_;
+	return unless (@o);
+	$out->{$name}->{tag} = $name;
+	$out->{$name}->{search} = \@o;
+	$out->{$name}->{display} = \@o;
+}
 
-sub parse {
-	my $self = shift;
+=head2 display
 
-	my ($rec, $format_utf8, $i, $rec_size) = @_;
+Define tag just for I<display>
 
-	return if (! $format_utf8);
+  @v = display('Title', rec('200','a') );
 
-	my $log = $self->_get_logger();
+=cut
 
-	$log->logconfess("need HASH as first argument!") if ($rec !~ /HASH/o);
+sub display {
+	my $name = shift or die "display needs name as first argument";
+	my @o = grep { defined($_) && $_ ne '' } @_;
+	return unless (@o);
+	$out->{$name}->{tag} = $name;
+	$out->{$name}->{display} = \@o;
+}
 
-	$i = 0 if (! $i);
+=head2 search
 
-	my $format = $self->_x($format_utf8) || $log->logconfess("can't convert '$format_utf8' from UTF-8 to ",$self->{'code_page'});
+Prepare values just for I<search>
 
-	my @out;
+  @v = search('Title', rec('200','a') );
 
-	$log->debug("format: $format [$i]");
+=cut
 
-	my $eval_code;
-	# remove eval{...} from beginning
-	$eval_code = $1 if ($format =~ s/^eval{([^}]+)}//s);
+sub search {
+	my $name = shift or die "search needs name as first argument";
+	my @o = grep { defined($_) && $_ ne '' } @_;
+	return unless (@o);
+	$out->{$name}->{tag} = $name;
+	$out->{$name}->{search} = \@o;
+}
 
-	my $filter_name;
-	# remove filter{...} from beginning
-	$filter_name = $1 if ($format =~ s/^filter{([^}]+)}//s);
+=head2 marc_leader
 
-	# did we found any (att all) field from format in row?
-	my $found_any;
-	# prefix before first field which we preserve it $found_any
-	my $prefix;
+Setup fields within MARC leader or get leader
 
-	my $f_step = 1;
+  marc_leader('05','c');
+  my $leader = marc_leader();
 
-	while ($format =~ s/^(.*?)(v|s)(\d+)(?:\^(\w))?//s) {
+=cut
 
-		my $del = $1 || '';
-		$prefix = $del if ($f_step == 1);
+sub marc_leader {
+	my ($offset,$value) = @_;
 
-		my $fld_type = lc($2);
+	if ($offset) {
+		$leader->{ $offset } = $value;
+	} else {
+		return $leader;
+	}
+}
 
-		# repeatable index
-		my $r = $i;
-		if ($fld_type eq 's') {
-			if ($found_any->{'v'}) {
-				$r = 0;
-			} else {
-				return;
-			}
-		}
+=head2 marc
+
+Save value for MARC field
+
+  marc('900','a', rec('200','a') );
+  marc('001', rec('000') );
 
-		my $found = 0;
-		my $tmp = $self->get_data(\$rec,$3,$4,$r,\$found,$rec_size);
+=cut
 
-		if ($found) {
-			$found_any->{$fld_type} += $found;
+sub marc {
+	my $f = shift or die "marc needs field";
+	die "marc field must be numer" unless ($f =~ /^\d+$/);
+
+	my $sf;
+	if ($f >= 10) {
+		$sf = shift or die "marc needs subfield";
+	}
 
-			# we will skip delimiter before first occurence of field!
-			push @out, $del unless($found_any->{$fld_type} == 1);
-			push @out, $tmp;
+	foreach (@_) {
+		my $v = $_;		# make var read-write for Encode
+		next unless (defined($v) && $v !~ /^\s*$/);
+		my ($i1,$i2) = defined($marc_indicators->{$f}) ? @{ $marc_indicators->{$f} } : (' ',' ');
+		if (defined $sf) {
+			push @{ $marc_record->[ $marc_record_offset ] }, [ $f, $i1, $i2, $sf => $v ];
+		} else {
+			push @{ $marc_record->[ $marc_record_offset ] }, [ $f, $v ];
 		}
-		$f_step++;
 	}
+}
 
-	# test if any fields found?
-	return if (! $found_any->{'v'} && ! $found_any->{'s'});
+=head2 marc_repeatable_subfield
 
-	my $out = join('',@out);
+Save values for MARC repetable subfield
 
-	if ($out) {
-		# add rest of format (suffix)
-		$out .= $format;
+  marc_repeatable_subfield('910', 'z', rec('909') );
 
-		# add prefix if not there
-		$out = $prefix . $out if ($out !~ m/^\Q$prefix\E/);
+=cut
 
-		$log->debug("result: $out");
-	}
+sub marc_repeatable_subfield {
+	my ($f,$sf) = @_;
+	die "marc_repeatable_subfield need field and subfield!\n" unless ($f && $sf);
+	$marc_repeatable_subfield->{ $f . $sf }++;
+	marc(@_);
+}
+
+=head2 marc_indicators
+
+Set both indicators for MARC field
 
-	if ($eval_code) {
-		my $eval = $self->fill_in($rec,$eval_code,$i) || return;
-		$log->debug("about to eval{$eval} format: $out");
-		return if (! $self->_eval($eval));
+  marc_indicators('900', ' ', 1);
+
+Any indicator value other than C<0-9> will be treated as undefined.
+
+=cut
+
+sub marc_indicators {
+	my $f = shift || die "marc_indicators need field!\n";
+	my ($i1,$i2) = @_;
+	die "marc_indicators($f, ...) need i1!\n" unless(defined($i1));
+	die "marc_indicators($f, $i1, ...) need i2!\n" unless(defined($i2));
+
+	$i1 = ' ' if ($i1 !~ /^\d$/);
+	$i2 = ' ' if ($i2 !~ /^\d$/);
+	@{ $marc_indicators->{$f} } = ($i1,$i2);
+}
+
+=head2 marc_compose
+
+Save values for each MARC subfield explicitly
+
+  marc_compose('900',
+  	'a', rec('200','a')
+  	'b', rec('201','a')
+  	'a', rec('200','b')
+  	'c', rec('200','c')
+  );
+
+If you specify C<+> for subfield, value will be appended
+to previous defined subfield.
+
+=cut
+
+sub marc_compose {
+	my $f = shift or die "marc_compose needs field";
+	die "marc_compose field must be numer" unless ($f =~ /^\d+$/);
+
+	my ($i1,$i2) = defined($marc_indicators->{$f}) ? @{ $marc_indicators->{$f} } : (' ',' ');
+	my $m = [ $f, $i1, $i2 ];
+
+	warn "### marc_compose input subfields = ", dump(@_),$/ if ($debug > 2);
+
+	if ($#_ % 2 != 1) {
+		die "ERROR: marc_compose",dump($f,@_)," not valid (must be even).\nDo you need to add first() or join() around some argument?\n";
 	}
-	
-	if ($filter_name) {
-		my @filter_args;
-		if ($filter_name =~ s/(\w+)\((.*)\)/$1/) {
-			@filter_args = split(/,/, $2);
-		}
-		if ($self->{'filter'}->{$filter_name}) {
-			$log->debug("about to filter{$filter_name} format: $out with arguments: ", join(",", @filter_args));
-			unshift @filter_args, $out;
-			$out = $self->{'filter'}->{$filter_name}->(@filter_args);
-			return unless(defined($out));
-			$log->debug("filter result: $out");
-		} elsif (! $warn_once->{$filter_name}) {
-			$log->warn("trying to use undefined filter $filter_name");
-			$warn_once->{$filter_name}++;
+
+	while (@_) {
+		my $sf = shift;
+		my $v = shift;
+
+		next unless (defined($v) && $v !~ /^\s*$/);
+		warn "## ++ marc_compose($f,$sf,$v) ", dump( $m ),$/ if ($debug > 1);
+		if ($sf ne '+') {
+			push @$m, ( $sf, $v );
+		} else {
+			$m->[ $#$m ] .= $v;
 		}
 	}
 
-	return $out;
+	warn "## marc_compose current marc = ", dump( $m ),$/ if ($debug > 1);
+
+	push @{ $marc_record->[ $marc_record_offset ] }, $m if ($#{$m} > 2);
 }
 
-=head2 fill_in
+=head2 marc_duplicate
 
-Workhourse of all: takes record from in-memory structure of database and
-strings with placeholders and returns string or array of with substituted
-values from record.
+Generate copy of current MARC record and continue working on copy
 
- my $text = $webpac->fill_in($rec,'v250^a');
+  marc_duplicate();
 
-Optional argument is ordinal number for repeatable fields. By default,
-it's assume to be first repeatable field (fields are perl array, so first
-element is 0).
-Following example will read second value from repeatable field.
+Copies can be accessed using C<< _get_marc_fields( fetch_next => 1 ) >> or
+C<< _get_marc_fields( offset => 42 ) >>.
 
- my $text = $webpac->fill_in($rec,'Title: v250^a',1);
+=cut
 
-This function B<does not> perform parsing of format to inteligenty skip
-delimiters before fields which aren't used.
+sub marc_duplicate {
+	 my $m = $marc_record->[ -1 ];
+	 die "can't duplicate record which isn't defined" unless ($m);
+	 push @{ $marc_record }, dclone( $m );
+	 warn "## marc_duplicate = ", dump(@$marc_record), $/ if ($debug > 1);
+	 $marc_record_offset = $#{ $marc_record };
+	 warn "## marc_record_offset = $marc_record_offset", $/ if ($debug > 1);
+}
 
-This method will automatically decode UTF-8 string to local code page
-if needed.
+=head2 marc_remove
 
-There is optional parametar C<$record_size> which can be used to get sizes of
-all C<field^subfield> combinations in this format.
+Remove some field or subfield from MARC record.
 
- my $text = $webpac->fill_in($rec,'got: v900^a v900^x',0,\$rec_size); 
+  marc_remove('200');
+  marc_remove('200','a');
 
-=cut
+This will erase field C<200> or C<200^a> from current MARC record.
 
-sub fill_in {
-	my $self = shift;
+This is useful after calling C<marc_duplicate> or on it's own (but, you
+should probably just remove that subfield definition if you are not
+using C<marc_duplicate>).
 
-	my $log = $self->_get_logger();
+FIXME: support fields < 10.
 
-	my ($rec,$format,$i,$rec_size) = @_;
+=cut
 
-	$log->logconfess("need data record") unless ($rec);
-	$log->logconfess("need format to parse") unless($format);
+sub marc_remove {
+	my ($f, $sf) = @_;
 
-	# iteration (for repeatable fields)
-	$i ||= 0;
+	die "marc_remove needs record number" unless defined($f);
 
-	$log->logdie("infitite loop in format $format") if ($i > ($self->{'max_mfn'} || 9999));
+	my $marc = $marc_record->[ $marc_record_offset ];
 
-	# FIXME remove for speedup?
-	$log->logconfess("need HASH as first argument!") if ($rec !~ /HASH/o);
+	warn "### marc_remove before = ", dump( $marc ), $/ if ($debug > 2);
 
-	if (utf8::is_utf8($format)) {
-		$format = $self->_x($format);
+	my $i = 0;
+	foreach ( 0 .. $#{ $marc } ) {
+		last unless (defined $marc->[$i]);
+		warn "#### working on ",dump( @{ $marc->[$i] }), $/ if ($debug > 3);
+		if ($marc->[$i]->[0] eq $f) {
+			if (! defined $sf) {
+				# remove whole field
+				splice @$marc, $i, 1;
+				warn "#### slice \@\$marc, $i, 1 = ",dump( @{ $marc }), $/ if ($debug > 3);
+				$i--;
+			} else {
+				foreach my $j ( 0 .. (( $#{ $marc->[$i] } - 3 ) / 2) ) {
+					my $o = ($j * 2) + 3;
+					if ($marc->[$i]->[$o] eq $sf) {
+						# remove subfield
+						splice @{$marc->[$i]}, $o, 2;
+						warn "#### slice \@{\$marc->[$i]}, $o, 2 = ", dump( @{ $marc }), $/ if ($debug > 3);
+						# is record now empty?
+						if ($#{ $marc->[$i] } == 2) {
+							splice @$marc, $i, 1;
+							warn "#### slice \@\$marc, $i, 1 = ", dump( @{ $marc }), $/ if ($debug > 3);
+							$i--;
+						};
+					}
+				}
+			}
+		}
+		$i++;
 	}
 
-	my $found = 0;
-	my $just_single = 1;
+	warn "### marc_remove($f", $sf ? ",$sf" : "", ") after = ", dump( $marc ), $/ if ($debug > 2);
 
-	my $eval_code;
-	# remove eval{...} from beginning
-	$eval_code = $1 if ($format =~ s/^eval{([^}]+)}//s);
+	$marc_record->[ $marc_record_offset ] = $marc;
 
-	my $filter_name;
-	# remove filter{...} from beginning
-	$filter_name = $1 if ($format =~ s/^filter{([^}]+)}//s);
+	warn "## full marc_record = ", dump( @{ $marc_record }), $/ if ($debug > 1);
+}
 
-	# do actual replacement of placeholders
-	# repeatable fields
-	if ($format =~ s/v(\d+)(?:\^(\w))?/$self->get_data(\$rec,$1,$2,$i,\$found,$rec_size)/ges) {
-		$just_single = 0;
-	}
+=head2 marc_original_order
 
-	# non-repeatable fields
-	if ($format =~ s/s(\d+)(?:\^(\w))?/$self->get_data(\$rec,$1,$2,0,\$found,$rec_size)/ges) {
-		return if ($i > 0 && $just_single);
-	}
+Copy all subfields preserving original order to marc field.
 
-	if ($found) {
-		$log->debug("format: $format");
-		if ($eval_code) {
-			my $eval = $self->fill_in($rec,$eval_code,$i);
-			return if (! $self->_eval($eval));
-		}
-		if ($filter_name && $self->{'filter'}->{$filter_name}) {
-			$log->debug("filter '$filter_name' for $format");
-			$format = $self->{'filter'}->{$filter_name}->($format);
-			return unless(defined($format));
-			$log->debug("filter result: $format");
+  marc_original_order( marc_field_number, original_input_field_number );
+
+Please note that field numbers are consistent with other commands (marc
+field number first), but somewhat counter-intuitive (destination and then
+source).
+
+You might want to use this command if you are just renaming subfields or
+using pre-processing modify_record in C<config.yml> and don't need any
+post-processing or want to preserve order of original subfields.
+
+
+=cut
+
+sub marc_original_order {
+
+	my ($to, $from) = @_;
+	die "marc_original_order needs from and to fields\n" unless ($from && $to);
+
+	return unless defined($rec->{$from});
+
+	my $r = $rec->{$from};
+	die "record field $from isn't array\n" unless (ref($r) eq 'ARRAY');
+
+	my ($i1,$i2) = defined($marc_indicators->{$to}) ? @{ $marc_indicators->{$to} } : (' ',' ');
+	warn "## marc_original_order($to,$from) source = ", dump( $r ),$/ if ($debug > 1);
+
+	foreach my $d (@$r) {
+
+		if (! defined($d->{subfields}) && ref($d->{subfields}) ne 'ARRAY') {
+			warn "# marc_original_order($to,$from): field $from doesn't have subfields specification\n";
+			next;
 		}
-		# do we have lookups?
-		if ($self->{'lookup'}) {
-			if ($self->{'lookup'}->can('lookup')) {
-				my @lookup = $self->{lookup}->lookup($format);
-				$log->debug("lookup $format", join(", ", @lookup));
-				return @lookup;
+	
+		my @sfs = @{ $d->{subfields} };
+
+		die "field $from doesn't have even number of subfields specifications\n" unless($#sfs % 2 == 1);
+
+		warn "#--> d: ",dump($d), "\n#--> sfs: ",dump(@sfs),$/ if ($debug > 2);
+
+		my $m = [ $to, $i1, $i2 ];
+
+		while (my $sf = shift @sfs) {
+
+			warn "#--> sf: ",dump($sf), $/ if ($debug > 2);
+			my $offset = shift @sfs;
+			die "corrupted sufields specification for field $from\n" unless defined($offset);
+
+			my $v;
+			if (ref($d->{$sf}) eq 'ARRAY') {
+				$v = $d->{$sf}->[$offset] if (defined($d->{$sf}->[$offset]));
+			} elsif ($offset == 0) {
+				$v = $d->{$sf};
 			} else {
-				$log->warn("Have lookup object but can't invoke lookup method");
+				die "field $from subfield '$sf' need occurence $offset which doesn't exist", dump($d->{$sf});
 			}
-		} else {
-			return $format;
+			push @$m, ( $sf, $v ) if (defined($v));
+		}
+
+		if ($#{$m} > 2) {
+			push @{ $marc_record->[ $marc_record_offset ] }, $m;
 		}
-	} else {
-		return;
 	}
+
+	warn "## marc_record = ", dump( $marc_record ),$/ if ($debug > 1);
 }
 
 
-=head2 _rec_to_arr
+=head1 Functions to extract data from input
+
+This function should be used inside functions to create C<data_structure> described
+above.
 
-Similar to C<parse> and C<fill_in>, but returns array of all repeatable fields. Usable
-for fields which have lookups, so they shouldn't be parsed but rather
-C<paste>d or C<fill_id>ed. Last argument is name of operation: C<paste> or C<fill_in>.
+=head2 _pack_subfields_hash
 
- my @arr = $webpac->fill_in_to_arr($rec,'[v900];;[v250^a]','paste');
+ @subfields = _pack_subfields_hash( $h );
+ $subfields = _pack_subfields_hash( $h, 1 );
+
+Return each subfield value in array or pack them all together and return scalar
+with subfields (denoted by C<^>) and values.
 
 =cut
 
-sub _rec_to_arr {
-	my $self = shift;
+sub _pack_subfields_hash {
 
-	my ($rec, $format_utf8, $code) = @_;
+	warn "## _pack_subfields_hash( ",dump(@_), " )\n" if ($debug > 1);
 
-	my $log = $self->_get_logger();
+	my ($h,$include_subfields) = @_;
 
-	$log->logconfess("need HASH as first argument!") if ($rec !~ /HASH/o);
-	return if (! $format_utf8);
+	if ( defined($h->{subfields}) ) {
+		my $sfs = delete $h->{subfields} || die "no subfields?";
+		my @out;
+		while (@$sfs) {
+			my $sf = shift @$sfs;
+			push @out, '^' . $sf if ($include_subfields);
+			my $o = shift @$sfs;
+			if ($o == 0 && ref( $h->{$sf} ) ne 'ARRAY' ) {
+				# single element subfields are not arrays
+#warn "====> $sf $o / $#$sfs ", dump( $sfs, $h->{$sf} ), "\n";
 
-	my $i = 0;
-	my $max = 0;
-	my @arr;
-	my $rec_size = {};
-
-	while ($i <= $max) {
-		my $v = $self->$code($rec,$format_utf8,$i++,\$rec_size) || next;
-		push @arr, $v;
-		if ($rec_size) {
-			foreach my $f (keys %{ $rec_size }) {
-				$max = $rec_size->{$f} if ($rec_size->{$f} > $max);
+				push @out, $h->{$sf};
+			} else {
+#warn "====> $sf $o / $#$sfs ", dump( $sfs, $h->{$sf} ), "\n";
+				push @out, $h->{$sf}->[$o];
 			}
-			warn "max set to $max, rec_size = ", Dumper($rec_size);
-			undef $rec_size;
+		}
+		if ($include_subfields) {
+			return join('', @out);
+		} else {
+			return @out;
+		}
+	} else {
+		if ($include_subfields) {
+			my $out = '';
+			foreach my $sf (sort keys %$h) {
+				if (ref($h->{$sf}) eq 'ARRAY') {
+					$out .= '^' . $sf . join('^' . $sf, @{ $h->{$sf} });
+				} else {
+					$out .= '^' . $sf . $h->{$sf};
+				}
+			}
+			return $out;
+		} else {
+			# FIXME this should probably be in alphabetical order instead of hash order
+			values %{$h};
 		}
 	}
+}
+
+=head2 rec1
+
+Return all values in some field
 
-	$log->debug("format '$format_utf8' returned ",--$i," elements: ", sub { join(" | ",@arr) }) if (@arr);
+  @v = rec1('200')
 
-	return @arr;
+TODO: order of values is probably same as in source data, need to investigate that
+
+=cut
+
+sub rec1 {
+	my $f = shift;
+	warn "rec1($f) = ", dump( $rec->{$f} ), $/ if ($debug > 1);
+	return unless (defined($rec) && defined($rec->{$f}));
+	warn "rec1($f) = ", dump( $rec->{$f} ), $/ if ($debug > 1);
+	if (ref($rec->{$f}) eq 'ARRAY') {
+		my @out;
+		foreach my $h ( @{ $rec->{$f} } ) {
+			if (ref($h) eq 'HASH') {
+				push @out, ( _pack_subfields_hash( $h ) );
+			} else {
+				push @out, $h;
+			}
+		}
+		return @out;
+	} elsif( defined($rec->{$f}) ) {
+		return $rec->{$f};
+	}
 }
 
+=head2 rec2
 
-=head2 get_data
+Return all values in specific field and subfield
 
-Returns value from record.
+  @v = rec2('200','a')
 
- my $text = $self->get_data(\$rec,$f,$sf,$i,\$found,\$rec_size);
+=cut
 
-Required arguments are:
+sub rec2 {
+	my $f = shift;
+	return unless (defined($rec && $rec->{$f}));
+	my $sf = shift;
+	warn "rec2($f,$sf) = ", dump( $rec->{$f} ), $/ if ($debug > 1);
+	return map {
+		if (ref($_->{$sf}) eq 'ARRAY') {
+			@{ $_->{$sf} };
+		} else {
+			$_->{$sf};
+		}
+	} grep { ref($_) eq 'HASH' && $_->{$sf} } @{ $rec->{$f} };
+}
 
-=over 8
+=head2 rec
 
-=item C<$rec>
+syntaxtic sugar for
 
-record reference
+  @v = rec('200')
+  @v = rec('200','a')
 
-=item C<$f>
+If rec() returns just single value, it will
+return scalar, not array.
 
-field
+=cut
 
-=item C<$sf>
+sub rec {
+	my @out;
+	if ($#_ == 0) {
+		@out = rec1(@_);
+	} elsif ($#_ == 1) {
+		@out = rec2(@_);
+	}
+	if ($#out == 0 && ! wantarray) {
+		return $out[0];
+	} elsif (@out) {
+		return @out;
+	} else {
+		return '';
+	}
+}
 
-optional subfield
+=head2 regex
 
-=item C<$i>
+Apply regex to some or all values
 
-index offset for repeatable values ( 0 ... $rec_size->{'400^a'} )
+  @v = regex( 's/foo/bar/g', @v );
+
+=cut
 
-=item C<$found>
+sub regex {
+	my $r = shift;
+	my @out;
+	#warn "r: $r\n", dump(\@_);
+	foreach my $t (@_) {
+		next unless ($t);
+		eval "\$t =~ $r";
+		push @out, $t if ($t && $t ne '');
+	}
+	return @out;
+}
 
-optional variable that will be incremeted if preset
+=head2 prefix
 
-=item C<$rec_size>
+Prefix all values with a string
 
-hash to hold maximum occurances of C<field^subfield> combinations
-(which can be accessed using keys in same format)
+  @v = prefix( 'my_', @v );
 
-=back
+=cut
+
+sub prefix {
+	my $p = shift or return;
+	return map { $p . $_ } grep { defined($_) } @_;
+}
+
+=head2 suffix
+
+suffix all values with a string
 
-Returns value or empty string, updates C<$found> and C<rec_size>
-if present.
+  @v = suffix( '_my', @v );
 
 =cut
 
-sub get_data {
-	my $self = shift;
+sub suffix {
+	my $s = shift or die "suffix needs string as first argument";
+	return map { $_ . $s } grep { defined($_) } @_;
+}
 
-	my ($rec,$f,$sf,$i,$found,$cache) = @_;
+=head2 surround
 
-	return '' unless ($$rec->{$f} && ref($$rec->{$f}) eq 'ARRAY');
+surround all values with a two strings
 
-	if (defined($$cache)) {
-		$$cache->{ $f . ( $sf ? '^' . $sf : '' ) } ||= scalar @{ $$rec->{$f} };
-	}
+  @v = surround( 'prefix_', '_suffix', @v );
 
-	return '' unless ($$rec->{$f}->[$i]);
+=cut
 
-	{
-		no strict 'refs';
-		if (defined($sf)) {
-			$$found++ if (defined($$found) && $$rec->{$f}->[$i]->{$sf});
-			return $$rec->{$f}->[$i]->{$sf};
-		} else {
-			$$found++ if (defined($$found));
-			# it still might have subfields, just
-			# not specified, so we'll dump some debug info
-			if ($$rec->{$f}->[$i] =~ /HASH/o) {
-				my $out;
-				foreach my $k (keys %{$$rec->{$f}->[$i]}) {
-					$out .= '$' . $k .':' . $$rec->{$f}->[$i]->{$k}." ";
-				}
-				return $out;
-			} else {
-				return $$rec->{$f}->[$i];
-			}
-		}
-	}
+sub surround {
+	my $p = shift or die "surround need prefix as first argument";
+	my $s = shift or die "surround needs suffix as second argument";
+	return map { $p . $_ . $s } grep { defined($_) } @_;
 }
 
+=head2 first
 
-=head2 apply_format
+Return first element
 
-Apply format specified in tag with C<format_name="name"> and
-C<format_delimiter=";;">.
+  $v = first( @v );
+
+=cut
+
+sub first {
+	my $r = shift;
+	return $r;
+}
 
- my $text = $webpac->apply_format($format_name,$format_delimiter,$data);
+=head2 lookup
 
-Formats can contain C<lookup{...}> if you need them.
+Consult lookup hashes for some value
+
+  @v = lookup(
+  	sub {
+		'ffkk/peri/mfn'.rec('000')
+	},
+	'ffkk','peri','200-a-200-e',
+	sub {
+		first(rec(200,'a')).' '.first(rec('200','e'))
+	}
+  );
+
+Code like above will be B<automatically generated> using L<WebPAC::Parse> from
+normal lookup definition in C<conf/lookup/something.pl> which looks like:
+
+  lookup(
+	# which results to return from record recorded in lookup
+	sub { 'ffkk/peri/mfn' . rec('000') },
+	# from which database and input
+	'ffkk','peri',
+	# such that following values match
+	sub { first(rec(200,'a')) . ' ' . first(rec('200','e')) },
+	# if this part is missing, we will try to match same fields
+	# from lookup record and current one, or you can override
+	# which records to use from current record using
+	sub { rec('900','x') . ' ' . rec('900','y') },
+  )
+
+You can think about this lookup as SQL (if that helps):
+
+  select
+  	sub { what }
+  from
+  	database, input
+  where
+    sub { filter from lookuped record }
+  having
+    sub { optional filter on current record }
+
+Easy as pie, right?
 
 =cut
 
-sub apply_format {
-	my $self = shift;
+sub lookup {
+	my ($what, $database, $input, $key, $having) = @_;
+
+	confess "lookup needs 5 arguments: what, database, input, key, having" unless ($#_ == 4);
+
+	warn "## lookup ($database, $input, $key)", $/ if ($debug > 1);
+	return unless (defined($lookup->{$database}->{$input}->{$key}));
 
-	my ($name,$delimiter,$data) = @_;
+	confess "lookup really need load_row_coderef added to data_structure\n" unless ($load_row_coderef);
 
-	my $log = $self->_get_logger();
+	my $mfns;
+	my @having = $having->();
 
-	if (! $self->{'import_xml'}->{'format'}->{$name}) {
-		$log->warn("<format name=\"$name\"> is not defined in ",$self->{'import_xml_file'});
-		return $data;
+	warn "## having = ", dump( @having ) if ($debug > 2);
+
+	foreach my $h ( @having ) {
+		if (defined($lookup->{$database}->{$input}->{$key}->{$h})) {
+			warn "lookup for $database/$input/$key/$h return ",dump($lookup->{$database}->{$input}->{$key}->{$h}),"\n" if ($debug);
+			$mfns->{$_}++ foreach keys %{ $lookup->{$database}->{$input}->{$key}->{$h} };
+		}
 	}
 
-	$log->warn("no delimiter for format $name") if (! $delimiter);
+	return unless ($mfns);
 
-	my $format = $self->_x($self->{'import_xml'}->{'format'}->{$name}->{'content'}) || $log->logdie("can't find format '$name'");
+	my @mfns = sort keys %$mfns;
 
-	my @data = split(/\Q$delimiter\E/, $data);
+	warn "# lookup loading $database/$input/$key mfn ", join(",",@mfns)," having ",dump(@having),"\n" if ($debug);
 
-	my $out = sprintf($format, @data);
-	$log->debug("using format $name [$format] on $data to produce: $out");
+	my $old_rec = $rec;
+	my @out;
 
-	if ($self->{'lookup_regex'} && $out =~ $self->{'lookup_regex'}) {
-		return $self->{'lookup'}->lookup($out);
-	} else {
-		return $out;
+	foreach my $mfn (@mfns) {
+		$rec = $load_row_coderef->( $database, $input, $mfn );
+
+		warn "got $database/$input/$mfn = ", dump($rec), $/ if ($debug);
+
+		my @vals = $what->();
+
+		push @out, ( @vals );
+
+		warn "lookup for mfn $mfn returned ", dump(@vals), $/ if ($debug);
 	}
 
+#	if (ref($lookup->{$k}) eq 'ARRAY') {
+#		return @{ $lookup->{$k} };
+#	} else {
+#		return $lookup->{$k};
+#	}
+
+	$rec = $old_rec;
+
+	warn "## lookup returns = ", dump(@out), $/ if ($debug);
+
+	if ($#out == 0) {
+		return $out[0];
+	} else {
+		return @out;
+	}
 }
 
-=head2 sort_arr
+=head2 save_into_lookup
 
-Sort array ignoring case and html in data
+Save value into lookup. It associates current database, input
+and specific keys with one or more values which will be
+associated over MFN.
 
- my @sorted = $webpac->sort_arr(@unsorted);
+MFN will be extracted from first occurence current of field 000
+in current record, or if it doesn't exist from L<_set_config> C<_mfn>.
 
-=cut
+  my $nr = save_into_lookup($database,$input,$key,sub {
+ 	# code which produce one or more values 
+  });
 
-sub sort_arr {
-	my $self = shift;
+It returns number of items saved.
 
-	my $log = $self->_get_logger();
+This function shouldn't be called directly, it's called from code created by
+L<WebPAC::Parser>. 
 
-	# FIXME add Schwartzian Transformation?
+=cut
 
-	my @sorted = sort {
-		$a =~ s#<[^>]+/*>##;
-		$b =~ s#<[^>]+/*>##;
-		lc($b) cmp lc($a)
-	} @_;
-	$log->debug("sorted values: ",sub { join(", ",@sorted) });
+sub save_into_lookup {
+	my ($database,$input,$key,$coderef) = @_;
+	die "save_into_lookup needs database" unless defined($database);
+	die "save_into_lookup needs input" unless defined($input);
+	die "save_into_lookup needs key" unless defined($key);
+	die "save_into_lookup needs CODE" unless ( defined($coderef) && ref($coderef) eq 'CODE' );
+
+	warn "## save_into_lookup rec = ", dump($rec), " config = ", dump($config), $/ if ($debug > 2);
+
+	my $mfn = 
+		defined($rec->{'000'}->[0])	?	$rec->{'000'}->[0]	:
+		defined($config->{_mfn})	?	$config->{_mfn}		:
+										die "mfn not defined or zero";
+
+	my $nr = 0;
+
+	foreach my $v ( $coderef->() ) {
+		$lookup->{$database}->{$input}->{$key}->{$v}->{$mfn}++;
+		warn "# saved lookup $database/$input/$key [$v] $mfn\n" if ($debug > 1);
+		$nr++;
+	}
 
-	return @sorted;
+	return $nr;
 }
 
+=head2 config
 
-=head1 INTERNAL METHODS
+Consult config values stored in C<config.yml>
 
-=head2 _sort_by_order
+  # return database code (key under databases in yaml)
+  $database_code = config();	# use _ from hash
+  $database_name = config('name');
+  $database_input_name = config('input name');
+  $tag = config('input normalize tag');
 
-Sort xml tags data structure accoding to C<order=""> attribute.
+Up to three levels are supported.
 
 =cut
 
-sub _sort_by_order {
-	my $self = shift;
+sub config {
+	return unless ($config);
+
+	my $p = shift;
 
-	my $va = $self->{'import_xml'}->{'indexer'}->{$a}->{'order'} ||
-		$self->{'import_xml'}->{'indexer'}->{$a};
-	my $vb = $self->{'import_xml'}->{'indexer'}->{$b}->{'order'} ||
-		$self->{'import_xml'}->{'indexer'}->{$b};
+	$p ||= '';
 
-	return $va <=> $vb;
+	my $v;
+
+	warn "### getting config($p)\n" if ($debug > 1);
+
+	my @p = split(/\s+/,$p);
+	if ($#p < 0) {
+		$v = $config->{ '_' };	# special, database code
+	} else {
+
+		my $c = dclone( $config );
+
+		foreach my $k (@p) {
+			warn "### k: $k c = ",dump($c),$/ if ($debug > 1);
+			if (ref($c) eq 'ARRAY') {
+				$c = shift @$c;
+				warn "config($p) taking first occurence of '$k', probably not what you wanted!\n";
+				last;
+			}
+
+			if (! defined($c->{$k}) ) {
+				$c = undef;
+				last;
+			} else {
+				$c = $c->{$k};
+			}
+		}
+		$v = $c if ($c);
+
+	}
+
+	warn "## config( '$p' ) = ",dump( $v ),$/ if ($v && $debug);
+	warn "config( '$p' ) is empty\n" if (! $v);
+
+	return $v;
 }
 
-=head2 _x
+=head2 id
 
-Convert strings from C<conf/normalize/*.xml> encoding into application
-specific encoding (optinally specified using C<code_page> to C<new>
-constructor).
+Returns unique id of this record
 
- my $text = $n->_x('normalize text string');
+  $id = id();
 
-This is a stub so that other modules doesn't have to implement it.
+Returns C<42/2> for 2nd occurence of MFN 42.
 
 =cut
 
-sub _x {
-	my $self = shift;
-	return shift;
+sub id {
+	my $mfn = $config->{_mfn} || die "no _mfn in config data";
+	return $mfn . $#{$marc_record} ? $#{$marc_record} + 1 : '';
 }
 
+=head2 join_with
 
-=head1 AUTHOR
+Joins walues with some delimiter
 
-Dobrica Pavlinusic, C<< <dpavlin@rot13.org> >>
+  $v = join_with(", ", @v);
+
+=cut
+
+sub join_with {
+	my $d = shift;
+	warn "### join_with('$d',",dump(@_),")\n" if ($debug > 2);
+	my $v = join($d, grep { defined($_) && $_ ne '' } @_);
+	return '' unless defined($v);
+	return $v;
+}
 
-=head1 COPYRIGHT & LICENSE
+=head2 split_rec_on
 
-Copyright 2005 Dobrica Pavlinusic, All Rights Reserved.
+Split record subfield on some regex and take one of parts out
 
-This program is free software; you can redistribute it and/or modify it
-under the same terms as Perl itself.
+  $a_before_semi_column =
+  	split_rec_on('200','a', /\s*;\s*/, $part);
+
+C<$part> is optional number of element. First element is
+B<1>, not 0!
+
+If there is no C<$part> parameter or C<$part> is 0, this function will
+return all values produced by splitting.
 
 =cut
 
-1; # End of WebPAC::Normalize
+sub split_rec_on {
+	die "split_rec_on need (fld,sf,regex[,part]" if ($#_ < 2);
+
+	my ($fld, $sf, $regex, $part) = @_;
+	warn "### regex ", ref($regex), $regex, $/ if ($debug > 2);
+
+	my @r = rec( $fld, $sf );
+	my $v = shift @r;
+	warn "### first rec($fld,$sf) = ",dump($v),$/ if ($debug > 2);
+
+	return '' if ( ! defined($v) || $v =~ /^\s*$/);
+
+	my @s = split( $regex, $v );
+	warn "## split_rec_on($fld,$sf,$regex,$part) = ",dump(@s),$/ if ($debug > 1);
+	if ($part && $part > 0) {
+		return $s[ $part - 1 ];
+	} else {
+		return @s;
+	}
+}
+
+# END
+1;