--- trunk/lib/WebPAC/Validate.pm	2006/05/16 13:38:09	514
+++ trunk/lib/WebPAC/Validate.pm	2007/05/24 14:42:40	840
@@ -7,7 +7,10 @@
 
 use base 'WebPAC::Common';
 use File::Slurp;
-use Data::Dumper;
+use List::Util qw/first/;
+use Data::Dump qw/dump/;
+use WebPAC::Normalize qw/_pack_subfields_hash/;
+use Storable qw/dclone/;
 
 =head1 NAME
 
@@ -15,11 +18,11 @@
 
 =head1 VERSION
 
-Version 0.01
+Version 0.11
 
 =cut
 
-our $VERSION = '0.01';
+our $VERSION = '0.11';
 
 =head1 SYNOPSIS
 
@@ -31,11 +34,15 @@
   # same with 101
   101
   # field 200 have valid subfields a-g
-  200 a b c d e f g
+  # and field e is repeatable
+  200 a b c d e* f g
   # field 205 can have only subfield a
-  205 a
+  # and must exists
+  205! a
   # while 210 can have a c or d
   210 a c d
+  # field which is ignored in validation
+  999-
 
 =head1 FUNCTIONS
 
@@ -44,9 +51,15 @@
 Create new validation object
 
   my $validate = new WebPAC::Validate(
-  	path => '/path/to/input/validate_file',
+  	path => 'conf/validate/file',
+	delimiters => [ ' : ', ' / ', ' ; ', ' , ' ],
+	delimiters_path => 'conf/validate/delimiters/file',
   );
 
+Optional parametar C<delimiters> will turn on validating of delimiters. Be
+careful here, those delimiters are just stuck into regex, so they can
+contain L<perlre> regexpes.
+
 =cut
 
 sub new {
@@ -56,42 +69,449 @@
 
 	my $log = $self->_get_logger();
 
-	foreach my $p (qw/path/) {
-		$log->logconfess("need $p") unless ($self->{$p});
+	$log->logdie("need path or delimiters_path") unless ( $self->{path} || $self->{delimiters_path} );
+
+	if ( $self->{path} ) {
+
+		my $v_file = read_file( $self->{path} ) ||
+			$log->logdie("can't open validate path $self->{path}: $!");
+
+		my $v;
+		my $curr_line = 1;
+
+		foreach my $l (split(/[\n\r]+/, $v_file)) {
+			$curr_line++;
+
+			# skip comments and whitespaces
+			next if ($l =~ /^#/ || $l =~ /^\s*$/);
+
+			$l =~ s/^\s+//;
+			$l =~ s/\s+$//;
+
+			my @d = split(/\s+/, $l);
+
+			my $fld = shift @d;
+
+			if ($fld =~ s/!$//) {
+				$self->{must_exist}->{$fld}++;
+			} elsif ($fld =~ s/-$//) {
+				$self->{dont_validate}->{$fld}++;
+			}
+
+			$log->logdie("need field name in line $curr_line: $l") unless (defined($fld));
+
+			if (@d) {
+				$v->{$fld} = [ map {
+					my $sf = $_;
+					if ( $sf =~ s/!(\*)?$/$1/ ) {
+						$self->{must_exist_sf}->{ $fld }->{ $sf }++;
+					};
+					$sf;
+				} @d ];
+			} else {
+				$v->{$fld} = 1;
+			}
+
+		}
+
+		$log->debug("current validation rules: ", dump($v));
+
+		$self->{rules} = $v;
+
+		$log->info("validation uses rules from $self->{path}");
+	}
+
+	if ( $self->{delimiters} ) {
+		$self->{delimiters_regex} = '(\^[a-z0-9]|' . join('|', @{ $self->{delimiters} }) . ')';
+		$log->info("validation check delimiters with regex $self->{delimiters_regex}");
+	}
+
+	if ( my $path = $self->{delimiters_path} ) {
+		if ( -e $path ) {
+			$log->info("using delimiter validation rules from $path");
+			open(my $d, $path) || $log->fatal("can't open $path: $!");
+			while(<$d>) {
+				chomp($d);
+				if (/^\s*(#*)\s*(\d+)\t+(\d+)\t+(.*)$/) {
+					my ($comment,$field,$count,$template) = ($1,$2,$3,$4);
+					$self->{_validate_delimiters_templates}->{$field}->{$template} = $count unless ($comment);
+				} else {
+					warn "## ignored $d\n";
+				}
+			}
+			close($d);
+			warn "_validate_delimiters_templates = ",dump( $self->{_validate_delimiters_templates} );
+		} else {
+			$log->warn("delimiters path $path doesn't exist, it will be created after this run");
+		}
+	}
+
+	$self ? return $self : return undef;
+}
+
+=head2 validate_rec
+
+Validate record and return errors
+
+  my @errors = $validate->validate_rec( $rec, $rec_dump );
+
+=cut
+
+sub validate_rec {
+	my $self = shift;
+
+	my $log = $self->_get_logger();
+
+	my $rec = shift || $log->logdie("validate_rec need record");
+	my $rec_dump = shift;
+
+	$log->logdie("rec isn't HASH") unless (ref($rec) eq 'HASH');
+#	$log->logdie("can't find validation rules") unless (my $r = $self->{rules});
+	my $r = $self->{rules};
+
+	my $errors;
+
+	$log->debug("rec = ", sub { dump($rec) }, "keys = ", keys %{ $rec });
+
+	my $fields;
+
+	foreach my $f (keys %{ $rec }) {
+
+		next if (!defined($f) || $f eq '' || $f eq '000');
+
+		# first check delimiters
+		if ( my $regex = $self->{delimiters_regex} ) {
+
+			foreach my $v (@{ $rec->{$f} }) {
+				my $l = _pack_subfields_hash( $v, 1 );
+				my $subfield_dump = $l;
+				my $template = '';
+				$l =~ s/$regex/$template.=$1/eg;
+				#warn "## template: $template\n";
+
+				if ( $template ) {
+					$self->{_delimiters_templates}->{$f}->{$template}++;
+
+					if ( my $v = $self->{_validate_delimiters_templates} ) {
+						if ( ! defined( $v->{$f}->{$template} ) ) {
+							$errors->{$f}->{invalid_delimiters_combination} = $template;
+							$errors->{$f}->{dump} = $subfield_dump;
+						#} else {
+						#	warn "## $f $template ok\n";
+						}
+					}
+				}
+			}
+		}
+
+		next unless ( $r );	# skip validation of no rules are specified
+
+		next if (defined( $self->{dont_validate}->{$f} ));
+
+		# track field usage
+		$fields->{$f}++;
+
+		if ( ! defined($r->{$f}) ) {
+			$errors->{ $f }->{unexpected} = "this field is not expected";
+			next;
+		}
+
+
+		if (ref($rec->{$f}) ne 'ARRAY') {
+			$errors->{ $f }->{not_repeatable} = "probably bug in parsing input data";
+			next;
+		}
+
+		foreach my $v (@{ $rec->{$f} }) {
+			# can we have subfields?
+			if (ref($r->{$f}) eq 'ARRAY') {
+				# are values hashes? (has subfields)
+				if (! defined($v)) {
+#					$errors->{$f}->{empty} = undef;
+#					$errors->{dump} = $rec_dump if ($rec_dump);
+				} elsif (ref($v) ne 'HASH') {
+					$errors->{$f}->{missing_subfield} = join(",", @{ $r->{$f} }) . " required";
+					next;
+				} else {
+
+					my $h = dclone( $v );
+
+					my $sf_repeatable;
+
+					delete($v->{subfields}) if (defined($v->{subfields}));
+
+					my $subfields;
+
+					foreach my $sf (keys %{ $v }) {
+
+						$subfields->{ $sf }++;
+
+						# is non-repeatable but with multiple values?
+						if ( ! first { $_ eq $sf.'*' } @{$r->{$f}} ) {
+							if ( ref($v->{$sf}) eq 'ARRAY' ) {
+								$sf_repeatable->{$sf}++;
+							};
+							if (! first { $_ eq $sf } @{ $r->{$f} }) {
+								$errors->{ $f }->{subfield}->{extra}->{$sf}++;
+							}
+						}
+
+					}
+					if (my @r_sf = sort keys( %$sf_repeatable )) {
+
+						foreach my $sf (@r_sf) {
+							$errors->{$f}->{subfield}->{extra_repeatable}->{$sf}++;
+							$errors->{$f}->{dump} = _pack_subfields_hash( $h, 1 );
+						}
+
+					}
+
+					if ( defined( $self->{must_exist_sf}->{$f} ) ) {
+						foreach my $sf (sort keys %{ $self->{must_exist_sf}->{$f} }) {
+#warn "====> $f $sf must exist\n";
+							$errors->{$f}->{subfield}->{missing}->{$sf}++
+								unless defined( $subfields->{$sf} );
+						}
+					}
+
+				}
+			} elsif (ref($v) eq 'HASH') {
+				$errors->{$f}->{unexpected_subfields}++;
+				$errors->{$f}->{dump} = _pack_subfields_hash( $v, 1 );
+			}
+		}
 	}
 
-	my $v_file = read_file( $self->{path} ) ||
-		$log->logdie("can't open validate path $self->{path}: $!");
+	$log->debug("_delimiters_templates = ", dump( $self->{_delimiters_templates} ) );
 
-	my $v;
-	my $curr_line = 1;
+	foreach my $must (sort keys %{ $self->{must_exist} }) {
+		next if ($fields->{$must});
+		$errors->{$must}->{missing}++;
+		$errors->{dump} = $rec_dump if ($rec_dump);
+	}
 
-	foreach my $l (split(/[\n\r]+/, $v_file)) {
-		$curr_line++;
-		# skip comments
-		next if ($l =~ m/^#/);
+	if ($errors) {
+		$log->debug("errors: ", $self->report_error( $errors ) );
 
-		$l =~ s/^\s+//;
-		$l =~ s/\s+$//;
+		my $mfn = $rec->{'000'}->[0] || $log->logconfess("record ", dump( $rec ), " doesn't have MFN");
+		$self->{errors}->{$mfn} = $errors;
+	}
+
+	#$log->logcluck("return from this function is ARRAY") unless wantarray;
+
+	return $errors;
+}
 
-		my @d = split(/\s+/, $l);
+=head2 reset_errors
 
-		my $fld = shift @d || $log->logdie("need field name in line $curr_line: $l");
+Clean all accumulated errors for this input
+
+  $validate->reset_errors;
+
+=cut
+
+sub reset_errors {
+	my $self = shift;
+	delete ($self->{errors});
+}
+
+=head2 all_errors
+
+Return hash with all errors
+
+  print dump( $validate->all_errors );
+
+=cut
+
+sub all_errors {
+	my $self = shift;
+	return $self->{errors};
+}
+
+=head2 report_error
+
+Produce nice humanly readable report of single error
+
+  print $validate->report_error( $error_hash );
+
+=cut
 
-		if (@d) {
-			$v->{$fld}->{ref} = 'ARRAY';
-			$v->{$fld}->{sf} = \@d;
+sub report_error {
+	my $self = shift;
+
+	my $h = shift || die "no hash?";
+
+	sub _unroll {
+		my ($self, $tree, $accumulated) = @_;
+
+		my $log = $self->_get_logger();
+
+		$log->debug("# ",
+			( $tree			? "tree: $tree "					: '' ),
+			( $accumulated	? "accumulated: $accumulated "		: '' ),
+		);
+
+		my $results;
+
+		if (ref($tree) ne 'HASH') {
+			return ("$accumulated\t($tree)", undef);
+		}
+
+		my $dump;
+
+		foreach my $k (sort keys %{ $tree }) {
+
+			if ($k eq 'dump') {
+				$dump = $tree->{dump};
+				#warn "## dump ",dump($dump),"\n";
+				next;
+			}
+
+			$log->debug("current: $k");
+
+			my ($new_results, $new_dump) = $self->_unroll($tree->{$k},
+				$accumulated ? "$accumulated\t$k" : $k
+			);
+
+			$log->debug(
+				( $new_results		? "new_results: " . dump($new_results) ." "	: '' ),
+			);
+
+			push @$results, $new_results if ($new_results);
+			$dump = $new_dump if ($new_dump);
+
+		}
+
+		$log->debug(
+			( $results		? "results: " . dump($results) ." "	: '' ),
+		);
+
+		if ($#$results == 0) {
+			return ($results->[0], $dump);
 		} else {
-			$v->{$fld}->{ref} = '';
+			return ($results, $dump);
 		}
+	}
+
 
+	sub _reformat {
+		my $l = shift;
+		$l =~ s/\t/ /g;
+		$l =~ s/_/ /g;
+		return $l;
 	}
 
-	$log->debug("current validation rules: ", Dumper($v));
+	my $out = '';
 
-	$self->{rules} = $v;
+	for my $f (sort keys %{ $h }) {
+		$out .= "$f: ";
+		
+		my ($r, $d) = $self->_unroll( $h->{$f} );
+		my $e;
+		if (ref($r) eq 'ARRAY') {
+			$e .= join(", ", map { _reformat( $_ ) } @$r);
+		} else {
+			$e .= _reformat( $r );
+		}
+		$e .= "\n\t$d" if ($d);
 
-	$self ? return $self : return undef;
+		$out .= $e . "\n";
+	}
+	return $out;
+}
+
+
+=head2 report
+
+Produce nice humanly readable report of errors
+
+  print $validate->report;
+
+=cut
+
+sub report {
+	my $self = shift;
+	my $e = $self->{errors} || return;
+
+	my $out;
+	foreach my $mfn (sort { $a <=> $b } keys %$e) {
+		$out .= "MFN $mfn\n" . $self->report_error( $e->{$mfn} ) . "\n";
+	}
+
+	return $out;
+
+}
+
+=head2 delimiters_templates
+
+Generate report of delimiter tamplates 
+
+  my $report = $validate->delimiter_teplates(
+	report => 1,
+  );
+
+Options:
+
+=over 4
+
+=item report
+
+Generate humanly readable report with single fields
+
+=back
+
+=cut
+
+sub delimiters_templates {
+	my $self = shift;
+
+	my $args = {@_};
+
+	my $t = $self->{_delimiters_templates};
+
+	my $log = $self->_get_logger;
+
+	unless ($t) {
+		$log->error("called without delimiters");
+		return;
+	}
+
+	my $out;
+
+	foreach my $f (sort { $a <=> $b } keys %$t) {
+		$out .= "$f\n" if ( $args->{report} );
+		foreach my $template (sort { $a cmp $b } keys %{ $t->{$f} }) {
+			my $count = $t->{$f}->{$template};
+			$out .= 
+				( $count ? "" : "# " ) .
+				( $args->{report} ? "" : "$f" ) .
+				"\t$count\t$template\n";
+		}
+	}
+
+	return $out;
+}
+
+=head2 save_delimiters_templates
+
+=cut
+
+sub save_delimiters_templates {
+	my $self = shift;
+
+	my $path = $self->{delimiters_path};
+
+	return unless ( $path );
+
+	my $log = $self->_get_logger;
+
+	$path .= '.new' if ( -e $path );
+
+	open(my $d, '>', $path) || $log->fatal("can't open $path: $!");
+	print $d $self->delimiters_templates;
+	close($d);
+
+	$log->info("new delimiters templates saved to $path");
 }
 
 =head1 AUTHOR