use base 'WebPAC::Common';
use File::Slurp;
use List::Util qw/first/;
-use Data::Dumper;
+use Data::Dump qw/dump/;
use WebPAC::Normalize qw/_pack_subfields_hash/;
use Storable qw/dclone/;
=head1 VERSION
-Version 0.03
+Version 0.12
=cut
-our $VERSION = '0.03';
+our $VERSION = '0.12';
=head1 SYNOPSIS
205! a
# while 210 can have a c or d
210 a c d
+ # field which is ignored in validation
+ 999-
=head1 FUNCTIONS
my $validate = new WebPAC::Validate(
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.
+
+C<path> and C<delimiters_path> can be specified by L<read_validate_file> and
+L<read_validate_delimiters> calls.
+
=cut
sub new {
my $log = $self->_get_logger();
- foreach my $p (qw/path/) {
- $log->logconfess("need $p") unless ($self->{$p});
+ $self->read_validate_file( $self->{path} ) if ( $self->{path} );
+
+ if ( $self->{delimiters} ) {
+ $self->{delimiters_regex} = '(\^[a-z0-9]|' . join('|', @{ $self->{delimiters} }) . ')';
+ $log->info("validation check delimiters with regex $self->{delimiters_regex}");
}
- my $v_file = read_file( $self->{path} ) ||
- $log->logdie("can't open validate path $self->{path}: $!");
+ $self->read_validate_delimiters_file( $self->{delimiters_path} ) if ( $self->{delimiters_path} );
+
+ return $self;
+}
+
+
+=head2 read_validate_file
+
+Specify validate rules file
+
+ $validate->read_validate_file( 'conf/validate/file' );
+
+Returns number of lines in file
+
+=cut
+
+sub read_validate_file {
+ my $self = shift;
+
+ my $path = shift || die "no path?";
+
+ my $log = $self->_get_logger();
+
+ my $v_file = read_file( $path ) ||
+ $log->logdie("can't open validate path $path: $!");
my $v;
+ delete( $self->{must_exist} );
+ delete( $self->{must_exist_sf} );
+ delete( $self->{dont_validate} );
my $curr_line = 1;
foreach my $l (split(/[\n\r]+/, $v_file)) {
if ($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} = \@d;
+ $v->{$fld} = [ map {
+ my $sf = $_;
+ if ( $sf =~ s/!// ) {
+ $self->{must_exist_sf}->{ $fld }->{ $sf }++;
+ };
+ $sf;
+ } @d ];
} else {
$v->{$fld} = 1;
}
}
- $log->debug("current validation rules: ", Dumper($v));
+ $log->debug("current validation rules: ", dump($v));
$self->{rules} = $v;
- $log->info("validation uses rules from $self->{path}");
+ $log->info("validation uses rules from $path");
+
+ return $curr_line;
+}
+
+=head2 read_validate_delimiters_file
+
+ $validate->read_validate_delimiters_file( 'conf/validate/delimiters/file' );
- $self ? return $self : return undef;
+=cut
+
+sub read_validate_delimiters_file {
+ my $self = shift;
+
+ my $path = shift || die "no path?";
+
+ my $log = $self->_get_logger();
+
+ delete( $self->{_validate_delimiters_templates} );
+ delete( $self->{_delimiters_templates} );
+
+ 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->{delimiters_path} = $path;
}
-=head2 validate_errors
+=head2 validate_rec
Validate record and return errors
- my @errors = $validate->validate_errors( $rec );
+ my @errors = $validate->validate_rec( $rec, $rec_dump );
=cut
-sub validate_errors {
+sub validate_rec {
my $self = shift;
my $log = $self->_get_logger();
- my $rec = shift || $log->logdie("validate_errors need record");
+ 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});
+# $log->logdie("can't find validation rules") unless (my $r = $self->{rules});
+ my $r = $self->{rules};
- my @errors;
+ my $errors;
- $log->debug("rec = ", sub { Dumper($rec) }, "keys = ", keys %{ $rec });
+ $log->debug("rec = ", sub { dump($rec) }, "keys = ", keys %{ $rec });
my $fields;
next if (!defined($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}->{potentially_invalid_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}) ) {
- push @errors, "field '$f' shouldn't exists";
+ $errors->{ $f }->{unexpected} = "this field is not expected";
next;
}
if (ref($rec->{$f}) ne 'ARRAY') {
- push @errors, "field '$f' isn't repetable, probably bug in parsing input data";
+ $errors->{ $f }->{not_repeatable} = "probably bug in parsing input data";
next;
}
# can we have subfields?
if (ref($r->{$f}) eq 'ARRAY') {
# are values hashes? (has subfields)
- if (ref($v) ne 'HASH') {
- push @errors, "$f has value without subfields: $v";
+ if (! defined($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";
+ $errors->{$f}->{dump} = $v;
next;
} else {
delete($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} }) {
- push @errors, "$f has unknown subfield: $sf";
+ if (! defined first { $_ eq $sf } @{ $r->{$f} }) {
+ $errors->{ $f }->{subfield}->{extra}->{$sf}++;
}
}
}
if (my @r_sf = sort keys( %$sf_repeatable )) {
- my $plural = $#r_sf > 0 ? 1 : 0;
-
- push @errors, "$f subfield" .
- ( $plural ? 's ' : ' ' ) .
- join(', ', @r_sf) .
- ( $plural ? ' are ' : ' is ' ) .
- 'repeatable in: ' .
- join('', _pack_subfields_hash( $h, 1) );
+
+ 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($v) eq 'HASH') {
- push @errors, "$f has subfields which is not valid";
+ $errors->{$f}->{unexpected_subfields}++;
+ $errors->{$f}->{dump} = _pack_subfields_hash( $v, 1 );
}
}
}
+ $log->debug("_delimiters_templates = ", sub { dump( $self->{_delimiters_templates} ) } );
+
foreach my $must (sort keys %{ $self->{must_exist} }) {
next if ($fields->{$must});
- push @errors,
- "field $must should exist, but it doesn't";
+ $errors->{$must}->{missing}++;
+ $errors->{dump} = $rec_dump if ($rec_dump);
+ }
+
+ if ($errors) {
+ $log->debug("errors: ", $self->report_error( $errors ) );
+
+ my $mfn = $rec->{'000'}->[0] || $log->logconfess("record ", sub { dump( $rec ) }, " doesn't have MFN");
+ $self->{errors}->{$mfn} = $errors;
}
#$log->logcluck("return from this function is ARRAY") unless wantarray;
- $log->debug("errors: ", join(", ", @errors)) if (@errors);
+ return $errors;
+}
+
+=head2 reset
+
+Clean all accumulated errors for this input and remember delimiter templates
+for L<save_delimiters_templates>
+
+ $validate->reset;
+
+This function B<must> be called after each input to provide accurate statistics.
+
+=cut
+
+sub reset {
+ my $self = shift;
+
+ my $log = $self->_get_logger;
+
+ delete ($self->{errors});
+
+ if ( ! $self->{_delimiters_templates} ) {
+ $log->debug("called without _delimiters_templates?");
+ return;
+ }
+
+ foreach my $f ( keys %{ $self->{_delimiters_templates} } ) {
+ foreach my $t ( keys %{ $self->{_delimiters_templates}->{$f} } ) {
+ $self->{_accumulated_delimiters_templates}->{$f}->{$t} +=
+ $self->{_delimiters_templates}->{$f}->{$t};
+ }
+ }
+ $log->debug("_accumulated_delimiters_templates = ", sub { dump( $self->{_accumulated_delimiters_templates} ) } );
+ delete ($self->{_delimiters_templates});
+}
+
+=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
+
+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
+ );
+
+ $log->debug( "new_results: ", sub { dump($new_results) } ) if ( $new_results );
+
+ push @$results, $new_results if ($new_results);
+ $dump = $new_dump if ($new_dump);
+
+ }
+
+ $log->debug( "results: ", sub { dump($results) } ) if ( $results );
+
+ if ($#$results == 0) {
+ return ($results->[0], $dump);
+ } else {
+ return ($results, $dump);
+ }
+ }
+
+
+ sub _reformat {
+ my $l = shift;
+ $l =~ s/\t/ /g;
+ $l =~ s/_/ /g;
+ return $l;
+ }
+
+ my $out = '';
+
+ 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);
+
+ $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,
+ current_input => 1,
+ );
+
+Options:
+
+=over 4
+
+=item report
+
+Generate humanly readable report with single fields
+
+=item current_input
+
+Report just current_input and not accumulated data
+
+=back
+
+=cut
+
+sub delimiters_templates {
+ my $self = shift;
+
+ my $args = {@_};
+
+ my $t = $self->{_accumulated_delimiters_templates};
+ $t = $self->{_delimiters_templates} if ( $args->{current_input} );
+
+ 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
+
+Save accumulated delimiter templates
+
+ $validator->save_delimiters_template( '/path/to/validate/delimiters' );
+
+=cut
+
+sub save_delimiters_templates {
+ my $self = shift;
+
+ my $path = shift;
+ $path ||= $self->{delimiters_path};
+
+ my $log = $self->_get_logger;
+
+ $log->logdie("need path") unless ( $path );
+
+
+ if ( ! $self->{_accumulated_delimiters_templates} ) {
+ $log->error('no _accumulated_delimiters_templates found, reset');
+ $self->reset;
+ }
+
+ if ( $self->{_delimiters_templates} ) {
+ $log->error('found _delimiters_templates, calling reset');
+ $self->reset;
+ }
+
+ $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");
- return @errors;
+ return 1;
}
=head1 AUTHOR
projects / webpac2 / blobdiff