18 |
|
|
19 |
=head1 VERSION |
=head1 VERSION |
20 |
|
|
21 |
Version 0.07 |
Version 0.12 |
22 |
|
|
23 |
=cut |
=cut |
24 |
|
|
25 |
our $VERSION = '0.07'; |
our $VERSION = '0.12'; |
26 |
|
|
27 |
=head1 SYNOPSIS |
=head1 SYNOPSIS |
28 |
|
|
41 |
205! a |
205! a |
42 |
# while 210 can have a c or d |
# while 210 can have a c or d |
43 |
210 a c d |
210 a c d |
44 |
|
# field which is ignored in validation |
45 |
|
999- |
46 |
|
|
47 |
=head1 FUNCTIONS |
=head1 FUNCTIONS |
48 |
|
|
52 |
|
|
53 |
my $validate = new WebPAC::Validate( |
my $validate = new WebPAC::Validate( |
54 |
path => 'conf/validate/file', |
path => 'conf/validate/file', |
55 |
|
delimiters => [ ' : ', ' / ', ' ; ', ' , ' ], |
56 |
|
delimiters_path => 'conf/validate/delimiters/file', |
57 |
); |
); |
58 |
|
|
59 |
|
Optional parametar C<delimiters> will turn on validating of delimiters. Be |
60 |
|
careful here, those delimiters are just stuck into regex, so they can |
61 |
|
contain L<perlre> regexpes. |
62 |
|
|
63 |
=cut |
=cut |
64 |
|
|
65 |
sub new { |
sub new { |
69 |
|
|
70 |
my $log = $self->_get_logger(); |
my $log = $self->_get_logger(); |
71 |
|
|
72 |
foreach my $p (qw/path/) { |
$log->logdie("need path or delimiters_path") unless ( $self->{path} || $self->{delimiters_path} ); |
|
$log->logconfess("need $p") unless ($self->{$p}); |
|
|
} |
|
73 |
|
|
74 |
my $v_file = read_file( $self->{path} ) || |
if ( $self->{path} ) { |
|
$log->logdie("can't open validate path $self->{path}: $!"); |
|
75 |
|
|
76 |
my $v; |
my $v_file = read_file( $self->{path} ) || |
77 |
my $curr_line = 1; |
$log->logdie("can't open validate path $self->{path}: $!"); |
78 |
|
|
79 |
foreach my $l (split(/[\n\r]+/, $v_file)) { |
my $v; |
80 |
$curr_line++; |
my $curr_line = 1; |
81 |
|
|
82 |
# skip comments and whitespaces |
foreach my $l (split(/[\n\r]+/, $v_file)) { |
83 |
next if ($l =~ /^#/ || $l =~ /^\s*$/); |
$curr_line++; |
84 |
|
|
85 |
$l =~ s/^\s+//; |
# skip comments and whitespaces |
86 |
$l =~ s/\s+$//; |
next if ($l =~ /^#/ || $l =~ /^\s*$/); |
87 |
|
|
88 |
my @d = split(/\s+/, $l); |
$l =~ s/^\s+//; |
89 |
|
$l =~ s/\s+$//; |
90 |
|
|
91 |
my $fld = shift @d; |
my @d = split(/\s+/, $l); |
92 |
|
|
93 |
if ($fld =~ s/!$//) { |
my $fld = shift @d; |
|
$self->{must_exist}->{$fld}++; |
|
|
} |
|
94 |
|
|
95 |
$log->logdie("need field name in line $curr_line: $l") unless (defined($fld)); |
if ($fld =~ s/!$//) { |
96 |
|
$self->{must_exist}->{$fld}++; |
97 |
|
} elsif ($fld =~ s/-$//) { |
98 |
|
$self->{dont_validate}->{$fld}++; |
99 |
|
} |
100 |
|
|
101 |
|
$log->logdie("need field name in line $curr_line: $l") unless (defined($fld)); |
102 |
|
|
103 |
|
if (@d) { |
104 |
|
$v->{$fld} = [ map { |
105 |
|
my $sf = $_; |
106 |
|
if ( $sf =~ s/!(\*)?$/$1/ ) { |
107 |
|
$self->{must_exist_sf}->{ $fld }->{ $sf }++; |
108 |
|
}; |
109 |
|
$sf; |
110 |
|
} @d ]; |
111 |
|
} else { |
112 |
|
$v->{$fld} = 1; |
113 |
|
} |
114 |
|
|
|
if (@d) { |
|
|
$v->{$fld} = [ map { |
|
|
my $sf = $_; |
|
|
if ( $sf =~ s/!(\*)?$/$1/ ) { |
|
|
$self->{must_exist_sf}->{ $fld }->{ $sf }++; |
|
|
}; |
|
|
$sf; |
|
|
} @d ]; |
|
|
} else { |
|
|
$v->{$fld} = 1; |
|
115 |
} |
} |
116 |
|
|
117 |
} |
$log->debug("current validation rules: ", dump($v)); |
118 |
|
|
119 |
$log->debug("current validation rules: ", dump($v)); |
$self->{rules} = $v; |
120 |
|
|
121 |
$self->{rules} = $v; |
$log->info("validation uses rules from $self->{path}"); |
122 |
|
} |
123 |
|
|
124 |
$log->info("validation uses rules from $self->{path}"); |
if ( $self->{delimiters} ) { |
125 |
|
$self->{delimiters_regex} = '(\^[a-z0-9]|' . join('|', @{ $self->{delimiters} }) . ')'; |
126 |
|
$log->info("validation check delimiters with regex $self->{delimiters_regex}"); |
127 |
|
} |
128 |
|
|
129 |
|
if ( my $path = $self->{delimiters_path} ) { |
130 |
|
if ( -e $path ) { |
131 |
|
$log->info("using delimiter validation rules from $path"); |
132 |
|
open(my $d, $path) || $log->fatal("can't open $path: $!"); |
133 |
|
while(<$d>) { |
134 |
|
chomp($d); |
135 |
|
if (/^\s*(#*)\s*(\d+)\t+(\d+)\t+(.*)$/) { |
136 |
|
my ($comment,$field,$count,$template) = ($1,$2,$3,$4); |
137 |
|
$self->{_validate_delimiters_templates}->{$field}->{$template} = $count unless ($comment); |
138 |
|
} else { |
139 |
|
warn "## ignored $d\n"; |
140 |
|
} |
141 |
|
} |
142 |
|
close($d); |
143 |
|
#warn "_validate_delimiters_templates = ",dump( $self->{_validate_delimiters_templates} ); |
144 |
|
} else { |
145 |
|
$log->warn("delimiters path $path doesn't exist, it will be created after this run"); |
146 |
|
} |
147 |
|
} |
148 |
|
|
149 |
$self ? return $self : return undef; |
$self ? return $self : return undef; |
150 |
} |
} |
151 |
|
|
152 |
=head2 validate_errors |
=head2 validate_rec |
153 |
|
|
154 |
Validate record and return errors |
Validate record and return errors |
155 |
|
|
156 |
my @errors = $validate->validate_errors( $rec, $rec_dump ); |
my @errors = $validate->validate_rec( $rec, $rec_dump ); |
157 |
|
|
158 |
=cut |
=cut |
159 |
|
|
160 |
sub validate_errors { |
sub validate_rec { |
161 |
my $self = shift; |
my $self = shift; |
162 |
|
|
163 |
my $log = $self->_get_logger(); |
my $log = $self->_get_logger(); |
164 |
|
|
165 |
my $rec = shift || $log->logdie("validate_errors need record"); |
my $rec = shift || $log->logdie("validate_rec need record"); |
166 |
my $rec_dump = shift; |
my $rec_dump = shift; |
167 |
|
|
168 |
$log->logdie("rec isn't HASH") unless (ref($rec) eq 'HASH'); |
$log->logdie("rec isn't HASH") unless (ref($rec) eq 'HASH'); |
169 |
$log->logdie("can't find validation rules") unless (my $r = $self->{rules}); |
# $log->logdie("can't find validation rules") unless (my $r = $self->{rules}); |
170 |
|
my $r = $self->{rules}; |
171 |
|
|
172 |
my $errors; |
my $errors; |
173 |
|
|
179 |
|
|
180 |
next if (!defined($f) || $f eq '' || $f eq '000'); |
next if (!defined($f) || $f eq '' || $f eq '000'); |
181 |
|
|
182 |
|
# first check delimiters |
183 |
|
if ( my $regex = $self->{delimiters_regex} ) { |
184 |
|
|
185 |
|
foreach my $v (@{ $rec->{$f} }) { |
186 |
|
my $l = _pack_subfields_hash( $v, 1 ); |
187 |
|
my $subfield_dump = $l; |
188 |
|
my $template = ''; |
189 |
|
$l =~ s/$regex/$template.=$1/eg; |
190 |
|
#warn "## template: $template\n"; |
191 |
|
|
192 |
|
if ( $template ) { |
193 |
|
$self->{_delimiters_templates}->{$f}->{$template}++; |
194 |
|
|
195 |
|
if ( my $v = $self->{_validate_delimiters_templates} ) { |
196 |
|
if ( ! defined( $v->{$f}->{$template} ) ) { |
197 |
|
$errors->{$f}->{potentially_invalid_combination} = $template; |
198 |
|
$errors->{$f}->{dump} = $subfield_dump; |
199 |
|
#} else { |
200 |
|
# warn "## $f $template ok\n"; |
201 |
|
} |
202 |
|
} |
203 |
|
} |
204 |
|
} |
205 |
|
} |
206 |
|
|
207 |
|
next unless ( $r ); # skip validation of no rules are specified |
208 |
|
|
209 |
|
next if (defined( $self->{dont_validate}->{$f} )); |
210 |
|
|
211 |
|
# track field usage |
212 |
$fields->{$f}++; |
$fields->{$f}++; |
213 |
|
|
214 |
if ( ! defined($r->{$f}) ) { |
if ( ! defined($r->{$f}) ) { |
215 |
$errors->{field}->{ $f }->{unexpected} = "this field is not expected"; |
$errors->{ $f }->{unexpected} = "this field is not expected"; |
216 |
next; |
next; |
217 |
} |
} |
218 |
|
|
219 |
|
|
220 |
if (ref($rec->{$f}) ne 'ARRAY') { |
if (ref($rec->{$f}) ne 'ARRAY') { |
221 |
$errors->{field}->{ $f }->{not_repeatable} = "probably bug in parsing input data"; |
$errors->{ $f }->{not_repeatable} = "probably bug in parsing input data"; |
222 |
next; |
next; |
223 |
} |
} |
224 |
|
|
227 |
if (ref($r->{$f}) eq 'ARRAY') { |
if (ref($r->{$f}) eq 'ARRAY') { |
228 |
# are values hashes? (has subfields) |
# are values hashes? (has subfields) |
229 |
if (! defined($v)) { |
if (! defined($v)) { |
230 |
# $errors->{field}->{$f}->{empty} = undef; |
# $errors->{$f}->{empty} = undef; |
231 |
# $errors->{dump} = $rec_dump if ($rec_dump); |
# $errors->{dump} = $rec_dump if ($rec_dump); |
232 |
} elsif (ref($v) ne 'HASH') { |
} elsif (ref($v) ne 'HASH') { |
233 |
$errors->{field}->{$f}->{missing_subfield} = join(",", @{ $r->{$f} }) . " required"; |
$errors->{$f}->{missing_subfield} = join(",", @{ $r->{$f} }) . " required"; |
234 |
next; |
next; |
235 |
} else { |
} else { |
236 |
|
|
252 |
$sf_repeatable->{$sf}++; |
$sf_repeatable->{$sf}++; |
253 |
}; |
}; |
254 |
if (! first { $_ eq $sf } @{ $r->{$f} }) { |
if (! first { $_ eq $sf } @{ $r->{$f} }) { |
255 |
$errors->{field}->{ $f }->{subfield}->{extra}->{$sf}++; |
$errors->{ $f }->{subfield}->{extra}->{$sf}++; |
256 |
} |
} |
257 |
} |
} |
258 |
|
|
260 |
if (my @r_sf = sort keys( %$sf_repeatable )) { |
if (my @r_sf = sort keys( %$sf_repeatable )) { |
261 |
|
|
262 |
foreach my $sf (@r_sf) { |
foreach my $sf (@r_sf) { |
263 |
$errors->{field}->{$f}->{subfield}->{extra_repeatable}->{$sf}++; |
$errors->{$f}->{subfield}->{extra_repeatable}->{$sf}++; |
264 |
$errors->{field}->{$f}->{dump} = |
$errors->{$f}->{dump} = _pack_subfields_hash( $h, 1 ); |
|
join('', _pack_subfields_hash( $h, 1 ) ); |
|
265 |
} |
} |
266 |
|
|
267 |
} |
} |
269 |
if ( defined( $self->{must_exist_sf}->{$f} ) ) { |
if ( defined( $self->{must_exist_sf}->{$f} ) ) { |
270 |
foreach my $sf (sort keys %{ $self->{must_exist_sf}->{$f} }) { |
foreach my $sf (sort keys %{ $self->{must_exist_sf}->{$f} }) { |
271 |
#warn "====> $f $sf must exist\n"; |
#warn "====> $f $sf must exist\n"; |
272 |
$errors->{field}->{$f}->{subfield}->{missing}->{$sf}++ |
$errors->{$f}->{subfield}->{missing}->{$sf}++ |
273 |
unless defined( $subfields->{$sf} ); |
unless defined( $subfields->{$sf} ); |
274 |
} |
} |
275 |
} |
} |
276 |
|
|
277 |
} |
} |
278 |
} elsif (ref($v) eq 'HASH') { |
} elsif (ref($v) eq 'HASH') { |
279 |
$errors->{field}->{$f}->{unexpected_subfields}++; |
$errors->{$f}->{unexpected_subfields}++; |
280 |
$errors->{field}->{$f}->{dump} = |
$errors->{$f}->{dump} = _pack_subfields_hash( $v, 1 ); |
|
join('', _pack_subfields_hash( $v, 1 ) ); |
|
281 |
} |
} |
282 |
} |
} |
283 |
} |
} |
284 |
|
|
285 |
|
$log->debug("_delimiters_templates = ", dump( $self->{_delimiters_templates} ) ); |
286 |
|
|
287 |
foreach my $must (sort keys %{ $self->{must_exist} }) { |
foreach my $must (sort keys %{ $self->{must_exist} }) { |
288 |
next if ($fields->{$must}); |
next if ($fields->{$must}); |
289 |
$errors->{field}->{$must}->{missing}++; |
$errors->{$must}->{missing}++; |
290 |
$errors->{dump} = $rec_dump if ($rec_dump); |
$errors->{dump} = $rec_dump if ($rec_dump); |
291 |
} |
} |
292 |
|
|
293 |
if ($errors) { |
if ($errors) { |
294 |
$log->debug("errors: ", sub { dump( $errors ) } ); |
$log->debug("errors: ", $self->report_error( $errors ) ); |
295 |
|
|
296 |
my $mfn = $rec->{'000'}->[0] || $log->logconfess("record ", dump( $rec ), " doesn't have MFN"); |
my $mfn = $rec->{'000'}->[0] || $log->logconfess("record ", dump( $rec ), " doesn't have MFN"); |
297 |
$self->{errors}->{$mfn} = $errors; |
$self->{errors}->{$mfn} = $errors; |
302 |
return $errors; |
return $errors; |
303 |
} |
} |
304 |
|
|
305 |
=head2 reset_errors |
=head2 reset |
306 |
|
|
307 |
Clean all accumulated errors for this input |
Clean all accumulated errors for this input and remember delimiter templates |
308 |
|
for L<save_delimiters_templates> |
309 |
|
|
310 |
$validate->reset_errors; |
$validate->reset; |
311 |
|
|
312 |
|
This function B<must> be called after each input to provide accurate statistics. |
313 |
|
|
314 |
=cut |
=cut |
315 |
|
|
316 |
sub reset_errors { |
sub reset { |
317 |
my $self = shift; |
my $self = shift; |
318 |
|
|
319 |
|
my $log = $self->_get_logger; |
320 |
|
|
321 |
delete ($self->{errors}); |
delete ($self->{errors}); |
322 |
|
|
323 |
|
if ( ! $self->{_delimiters_templates} ) { |
324 |
|
$log->debug("called without _delimiters_templates?"); |
325 |
|
return; |
326 |
|
} |
327 |
|
|
328 |
|
foreach my $f ( keys %{ $self->{_delimiters_templates} } ) { |
329 |
|
foreach my $t ( keys %{ $self->{_delimiters_templates}->{$f} } ) { |
330 |
|
$self->{_accumulated_delimiters_templates}->{$f}->{$t} += |
331 |
|
$self->{_delimiters_templates}->{$f}->{$t}; |
332 |
|
} |
333 |
|
} |
334 |
|
$log->debug("_accumulated_delimiters_templates = ", dump( $self->{_accumulated_delimiter_templates} )); |
335 |
|
delete ($self->{_delimiters_templates}); |
336 |
} |
} |
337 |
|
|
338 |
=head2 all_errors |
=head2 all_errors |
348 |
return $self->{errors}; |
return $self->{errors}; |
349 |
} |
} |
350 |
|
|
351 |
=head2 report |
=head2 report_error |
352 |
|
|
353 |
Produce nice humanly readable report of errors |
Produce nice humanly readable report of single error |
354 |
|
|
355 |
print $validate->report; |
print $validate->report_error( $error_hash ); |
356 |
|
|
357 |
=cut |
=cut |
358 |
|
|
359 |
sub report { |
sub report_error { |
360 |
my $self = shift; |
my $self = shift; |
361 |
|
|
362 |
my $log = $self->_get_logger(); |
my $h = shift || die "no hash?"; |
363 |
|
|
364 |
|
sub _unroll { |
365 |
|
my ($self, $tree, $accumulated) = @_; |
366 |
|
|
367 |
sub unroll { |
my $log = $self->_get_logger(); |
|
my ($tree, $accumulated) = @_; |
|
368 |
|
|
369 |
$log->debug("# ", |
$log->debug("# ", |
370 |
( $tree ? "tree: $tree " : '' ), |
( $tree ? "tree: $tree " : '' ), |
383 |
|
|
384 |
if ($k eq 'dump') { |
if ($k eq 'dump') { |
385 |
$dump = $tree->{dump}; |
$dump = $tree->{dump}; |
386 |
warn "## dump: $dump\n"; |
#warn "## dump ",dump($dump),"\n"; |
387 |
next; |
next; |
388 |
} |
} |
389 |
|
|
390 |
$log->debug("current: $k"); |
$log->debug("current: $k"); |
391 |
|
|
392 |
my ($new_results, $new_dump) = unroll($tree->{$k}, |
my ($new_results, $new_dump) = $self->_unroll($tree->{$k}, |
393 |
$accumulated ? "$accumulated\t$k" : $k |
$accumulated ? "$accumulated\t$k" : $k |
394 |
); |
); |
395 |
|
|
413 |
} |
} |
414 |
} |
} |
415 |
|
|
416 |
|
|
417 |
|
sub _reformat { |
418 |
|
my $l = shift; |
419 |
|
$l =~ s/\t/ /g; |
420 |
|
$l =~ s/_/ /g; |
421 |
|
return $l; |
422 |
|
} |
423 |
|
|
424 |
my $out = ''; |
my $out = ''; |
425 |
|
|
426 |
|
for my $f (sort keys %{ $h }) { |
427 |
|
$out .= "$f: "; |
428 |
|
|
429 |
|
my ($r, $d) = $self->_unroll( $h->{$f} ); |
430 |
|
my $e; |
431 |
|
if (ref($r) eq 'ARRAY') { |
432 |
|
$e .= join(", ", map { _reformat( $_ ) } @$r); |
433 |
|
} else { |
434 |
|
$e .= _reformat( $r ); |
435 |
|
} |
436 |
|
$e .= "\n\t$d" if ($d); |
437 |
|
|
438 |
|
$out .= $e . "\n"; |
439 |
|
} |
440 |
|
return $out; |
441 |
|
} |
442 |
|
|
443 |
|
|
444 |
|
=head2 report |
445 |
|
|
446 |
|
Produce nice humanly readable report of errors |
447 |
|
|
448 |
|
print $validate->report; |
449 |
|
|
450 |
|
=cut |
451 |
|
|
452 |
|
sub report { |
453 |
|
my $self = shift; |
454 |
my $e = $self->{errors} || return; |
my $e = $self->{errors} || return; |
455 |
|
|
456 |
foreach my $mfn (sort keys %$e) { |
my $out; |
457 |
my ($r, $d) = unroll( $e->{$mfn} ); |
foreach my $mfn (sort { $a <=> $b } keys %$e) { |
458 |
$out .= "MFN $mfn\n", dump($r), "\t$d\n\n"; |
$out .= "MFN $mfn\n" . $self->report_error( $e->{$mfn} ) . "\n"; |
459 |
} |
} |
460 |
|
|
461 |
return $out; |
return $out; |
462 |
|
|
463 |
|
} |
464 |
|
|
465 |
|
=head2 delimiters_templates |
466 |
|
|
467 |
|
Generate report of delimiter tamplates |
468 |
|
|
469 |
|
my $report = $validate->delimiter_teplates( |
470 |
|
report => 1, |
471 |
|
accumulated => 1, |
472 |
|
); |
473 |
|
|
474 |
|
Options: |
475 |
|
|
476 |
|
=over 4 |
477 |
|
|
478 |
|
=item report |
479 |
|
|
480 |
|
Generate humanly readable report with single fields |
481 |
|
|
482 |
|
=item accumulated |
483 |
|
|
484 |
|
Use accumulated data from all inputs |
485 |
|
|
486 |
|
=back |
487 |
|
|
488 |
|
=cut |
489 |
|
|
490 |
|
sub delimiters_templates { |
491 |
|
my $self = shift; |
492 |
|
|
493 |
|
my $args = {@_}; |
494 |
|
|
495 |
|
my $t; |
496 |
|
if ( $args->{accumulated} ) { |
497 |
|
$t = $self->{_accumulated_delimiters_templates}; |
498 |
|
} else { |
499 |
|
$t = $self->{_delimiters_templates}; |
500 |
|
} |
501 |
|
|
502 |
|
my $log = $self->_get_logger; |
503 |
|
|
504 |
|
unless ($t) { |
505 |
|
$log->error("called without delimiters"); |
506 |
|
return; |
507 |
|
} |
508 |
|
|
509 |
|
my $out; |
510 |
|
|
511 |
|
foreach my $f (sort { $a <=> $b } keys %$t) { |
512 |
|
$out .= "$f\n" if ( $args->{report} ); |
513 |
|
foreach my $template (sort { $a cmp $b } keys %{ $t->{$f} }) { |
514 |
|
my $count = $t->{$f}->{$template}; |
515 |
|
$out .= |
516 |
|
( $count ? "" : "# " ) . |
517 |
|
( $args->{report} ? "" : "$f" ) . |
518 |
|
"\t$count\t$template\n"; |
519 |
|
} |
520 |
|
} |
521 |
|
|
522 |
|
return $out; |
523 |
|
} |
524 |
|
|
525 |
|
=head2 save_delimiters_templates |
526 |
|
|
527 |
|
Save accumulated delimiter templates |
528 |
|
|
529 |
|
$validator->save_delimiters_template( '/path/to/validate/delimiters' ); |
530 |
|
|
531 |
|
=cut |
532 |
|
|
533 |
|
sub save_delimiters_templates { |
534 |
|
my $self = shift; |
535 |
|
|
536 |
|
my $path = $self->{delimiters_path}; |
537 |
|
|
538 |
|
return unless ( $path ); |
539 |
|
|
540 |
|
my $log = $self->_get_logger; |
541 |
|
|
542 |
|
if ( ! $self->{_accumulated_delimiters_templates} ) { |
543 |
|
$log->error('no _accumulated_delimiters_templates found, reset'); |
544 |
|
$self->reset; |
545 |
|
} |
546 |
|
|
547 |
|
if ( ! $self->{_delimiters_templates} ) { |
548 |
|
$log->error('found _delimiters_templates, calling reset'); |
549 |
|
$self->reset; |
550 |
|
} |
551 |
|
|
552 |
|
$path .= '.new' if ( -e $path ); |
553 |
|
|
554 |
|
open(my $d, '>', $path) || $log->fatal("can't open $path: $!"); |
555 |
|
print $d $self->delimiters_templates( accumulated => 1 ); |
556 |
|
close($d); |
557 |
|
|
558 |
|
$log->info("new delimiters templates saved to $path"); |
559 |
} |
} |
560 |
|
|
561 |
=head1 AUTHOR |
=head1 AUTHOR |