1 |
package WebPAC::Validate; |
2 |
|
3 |
use warnings; |
4 |
use strict; |
5 |
|
6 |
use blib; |
7 |
|
8 |
use base 'WebPAC::Common'; |
9 |
use File::Slurp; |
10 |
use List::Util qw/first/; |
11 |
use Data::Dump qw/dump/; |
12 |
use WebPAC::Normalize qw/_pack_subfields_hash/; |
13 |
use Storable qw/dclone/; |
14 |
|
15 |
=head1 NAME |
16 |
|
17 |
WebPAC::Validate - provide simple validation for records |
18 |
|
19 |
=head1 VERSION |
20 |
|
21 |
Version 0.12 |
22 |
|
23 |
=cut |
24 |
|
25 |
our $VERSION = '0.12'; |
26 |
|
27 |
=head1 SYNOPSIS |
28 |
|
29 |
This module provide a simple way to validate your file against a simple |
30 |
configuration file in following format: |
31 |
|
32 |
# field 10 doesn't have any subfields |
33 |
10 |
34 |
# same with 101 |
35 |
101 |
36 |
# field 200 have valid subfields a-g |
37 |
# and field e is repeatable |
38 |
200 a b c d e* f g |
39 |
# field 205 can have only subfield a |
40 |
# and must exists |
41 |
205! a |
42 |
# while 210 can have a c or d |
43 |
210 a c d |
44 |
# field which is ignored in validation |
45 |
999- |
46 |
|
47 |
=head1 FUNCTIONS |
48 |
|
49 |
=head2 new |
50 |
|
51 |
Create new validation object |
52 |
|
53 |
my $validate = new WebPAC::Validate( |
54 |
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 |
C<path> and C<delimiters_path> can be specified by L<read_validate_file> and |
64 |
L<read_validate_delimiters> calls. |
65 |
|
66 |
=cut |
67 |
|
68 |
sub new { |
69 |
my $class = shift; |
70 |
my $self = {@_}; |
71 |
bless($self, $class); |
72 |
|
73 |
my $log = $self->_get_logger(); |
74 |
|
75 |
$self->read_validate_file( $self->{path} ) if ( $self->{path} ); |
76 |
|
77 |
if ( $self->{delimiters} ) { |
78 |
$self->{delimiters_regex} = '(\^[a-z0-9]|' . join('|', @{ $self->{delimiters} }) . ')'; |
79 |
$log->info("validation check delimiters with regex $self->{delimiters_regex}"); |
80 |
} |
81 |
|
82 |
$self->read_validate_delimiters_file( $self->{delimiters_path} ) if ( $self->{delimiters_path} ); |
83 |
|
84 |
return $self; |
85 |
} |
86 |
|
87 |
|
88 |
=head2 read_validate_file |
89 |
|
90 |
Specify validate rules file |
91 |
|
92 |
$validate->read_validate_file( 'conf/validate/file' ); |
93 |
|
94 |
Returns number of lines in file |
95 |
|
96 |
=cut |
97 |
|
98 |
sub read_validate_file { |
99 |
my $self = shift; |
100 |
|
101 |
my $path = shift || die "no path?"; |
102 |
|
103 |
my $log = $self->_get_logger(); |
104 |
|
105 |
my $v_file = read_file( $path ) || |
106 |
$log->logdie("can't open validate path $path: $!"); |
107 |
|
108 |
my $v; |
109 |
delete( $self->{must_exist} ); |
110 |
delete( $self->{must_exist_sf} ); |
111 |
delete( $self->{dont_validate} ); |
112 |
my $curr_line = 1; |
113 |
|
114 |
foreach my $l (split(/[\n\r]+/, $v_file)) { |
115 |
$curr_line++; |
116 |
|
117 |
# skip comments and whitespaces |
118 |
next if ($l =~ /^#/ || $l =~ /^\s*$/); |
119 |
|
120 |
$l =~ s/^\s+//; |
121 |
$l =~ s/\s+$//; |
122 |
|
123 |
my @d = split(/\s+/, $l); |
124 |
|
125 |
my $fld = shift @d; |
126 |
|
127 |
if ($fld =~ s/!$//) { |
128 |
$self->{must_exist}->{$fld}++; |
129 |
} elsif ($fld =~ s/-$//) { |
130 |
$self->{dont_validate}->{$fld}++; |
131 |
} |
132 |
|
133 |
$log->logdie("need field name in line $curr_line: $l") unless (defined($fld)); |
134 |
|
135 |
if (@d) { |
136 |
$v->{$fld} = [ map { |
137 |
my $sf = $_; |
138 |
if ( $sf =~ s/!(\*)?$/$1/ ) { |
139 |
$self->{must_exist_sf}->{ $fld }->{ $sf }++; |
140 |
}; |
141 |
$sf; |
142 |
} @d ]; |
143 |
} else { |
144 |
$v->{$fld} = 1; |
145 |
} |
146 |
|
147 |
} |
148 |
|
149 |
$log->debug("current validation rules: ", dump($v)); |
150 |
|
151 |
$self->{rules} = $v; |
152 |
|
153 |
$log->info("validation uses rules from $path"); |
154 |
|
155 |
return $curr_line; |
156 |
} |
157 |
|
158 |
=head2 read_validate_delimiters_file |
159 |
|
160 |
$validate->read_validate_delimiters_file( 'conf/validate/delimiters/file' ); |
161 |
|
162 |
=cut |
163 |
|
164 |
sub read_validate_delimiters_file { |
165 |
my $self = shift; |
166 |
|
167 |
my $path = shift || die "no path?"; |
168 |
|
169 |
my $log = $self->_get_logger(); |
170 |
|
171 |
delete( $self->{_validate_delimiters_templates} ); |
172 |
delete( $self->{_delimiters_templates} ); |
173 |
|
174 |
if ( -e $path ) { |
175 |
$log->info("using delimiter validation rules from $path"); |
176 |
open(my $d, $path) || $log->fatal("can't open $path: $!"); |
177 |
while(<$d>) { |
178 |
chomp($d); |
179 |
if (/^\s*(#*)\s*(\d+)\t+(\d+)\t+(.*)$/) { |
180 |
my ($comment,$field,$count,$template) = ($1,$2,$3,$4); |
181 |
$self->{_validate_delimiters_templates}->{$field}->{$template} = $count unless ($comment); |
182 |
} else { |
183 |
warn "## ignored $d\n"; |
184 |
} |
185 |
} |
186 |
close($d); |
187 |
#warn "_validate_delimiters_templates = ",dump( $self->{_validate_delimiters_templates} ); |
188 |
} else { |
189 |
$log->warn("delimiters path $path doesn't exist, it will be created after this run"); |
190 |
} |
191 |
$self->{delimiters_path} = $path; |
192 |
} |
193 |
|
194 |
=head2 validate_rec |
195 |
|
196 |
Validate record and return errors |
197 |
|
198 |
my @errors = $validate->validate_rec( $rec, $rec_dump ); |
199 |
|
200 |
=cut |
201 |
|
202 |
sub validate_rec { |
203 |
my $self = shift; |
204 |
|
205 |
my $log = $self->_get_logger(); |
206 |
|
207 |
my $rec = shift || $log->logdie("validate_rec need record"); |
208 |
my $rec_dump = shift; |
209 |
|
210 |
$log->logdie("rec isn't HASH") unless (ref($rec) eq 'HASH'); |
211 |
# $log->logdie("can't find validation rules") unless (my $r = $self->{rules}); |
212 |
my $r = $self->{rules}; |
213 |
|
214 |
my $errors; |
215 |
|
216 |
$log->debug("rec = ", sub { dump($rec) }, "keys = ", keys %{ $rec }); |
217 |
|
218 |
my $fields; |
219 |
|
220 |
foreach my $f (keys %{ $rec }) { |
221 |
|
222 |
next if (!defined($f) || $f eq '' || $f eq '000'); |
223 |
|
224 |
# first check delimiters |
225 |
if ( my $regex = $self->{delimiters_regex} ) { |
226 |
|
227 |
foreach my $v (@{ $rec->{$f} }) { |
228 |
my $l = _pack_subfields_hash( $v, 1 ); |
229 |
my $subfield_dump = $l; |
230 |
my $template = ''; |
231 |
$l =~ s/$regex/$template.=$1/eg; |
232 |
#warn "## template: $template\n"; |
233 |
|
234 |
if ( $template ) { |
235 |
$self->{_delimiters_templates}->{$f}->{$template}++; |
236 |
|
237 |
if ( my $v = $self->{_validate_delimiters_templates} ) { |
238 |
if ( ! defined( $v->{$f}->{$template} ) ) { |
239 |
$errors->{$f}->{potentially_invalid_combination} = $template; |
240 |
$errors->{$f}->{dump} = $subfield_dump; |
241 |
#} else { |
242 |
# warn "## $f $template ok\n"; |
243 |
} |
244 |
} |
245 |
} |
246 |
} |
247 |
} |
248 |
|
249 |
next unless ( $r ); # skip validation of no rules are specified |
250 |
|
251 |
next if (defined( $self->{dont_validate}->{$f} )); |
252 |
|
253 |
# track field usage |
254 |
$fields->{$f}++; |
255 |
|
256 |
if ( ! defined($r->{$f}) ) { |
257 |
$errors->{ $f }->{unexpected} = "this field is not expected"; |
258 |
next; |
259 |
} |
260 |
|
261 |
|
262 |
if (ref($rec->{$f}) ne 'ARRAY') { |
263 |
$errors->{ $f }->{not_repeatable} = "probably bug in parsing input data"; |
264 |
next; |
265 |
} |
266 |
|
267 |
foreach my $v (@{ $rec->{$f} }) { |
268 |
# can we have subfields? |
269 |
if (ref($r->{$f}) eq 'ARRAY') { |
270 |
# are values hashes? (has subfields) |
271 |
if (! defined($v)) { |
272 |
# $errors->{$f}->{empty} = undef; |
273 |
# $errors->{dump} = $rec_dump if ($rec_dump); |
274 |
} elsif (ref($v) ne 'HASH') { |
275 |
$errors->{$f}->{missing_subfield} = join(",", @{ $r->{$f} }) . " required"; |
276 |
$errors->{$f}->{dump} = $v; |
277 |
next; |
278 |
} else { |
279 |
|
280 |
my $h = dclone( $v ); |
281 |
|
282 |
my $sf_repeatable; |
283 |
|
284 |
delete($v->{subfields}) if (defined($v->{subfields})); |
285 |
|
286 |
my $subfields; |
287 |
|
288 |
foreach my $sf (keys %{ $v }) { |
289 |
|
290 |
$subfields->{ $sf }++; |
291 |
|
292 |
# is non-repeatable but with multiple values? |
293 |
if ( ! first { $_ eq $sf.'*' } @{$r->{$f}} ) { |
294 |
if ( ref($v->{$sf}) eq 'ARRAY' ) { |
295 |
$sf_repeatable->{$sf}++; |
296 |
}; |
297 |
if (! first { $_ eq $sf } @{ $r->{$f} }) { |
298 |
$errors->{ $f }->{subfield}->{extra}->{$sf}++; |
299 |
} |
300 |
} |
301 |
|
302 |
} |
303 |
if (my @r_sf = sort keys( %$sf_repeatable )) { |
304 |
|
305 |
foreach my $sf (@r_sf) { |
306 |
$errors->{$f}->{subfield}->{extra_repeatable}->{$sf}++; |
307 |
$errors->{$f}->{dump} = _pack_subfields_hash( $h, 1 ); |
308 |
} |
309 |
|
310 |
} |
311 |
|
312 |
if ( defined( $self->{must_exist_sf}->{$f} ) ) { |
313 |
foreach my $sf (sort keys %{ $self->{must_exist_sf}->{$f} }) { |
314 |
#warn "====> $f $sf must exist\n"; |
315 |
$errors->{$f}->{subfield}->{missing}->{$sf}++ |
316 |
unless defined( $subfields->{$sf} ); |
317 |
} |
318 |
} |
319 |
|
320 |
} |
321 |
} elsif (ref($v) eq 'HASH') { |
322 |
$errors->{$f}->{unexpected_subfields}++; |
323 |
$errors->{$f}->{dump} = _pack_subfields_hash( $v, 1 ); |
324 |
} |
325 |
} |
326 |
} |
327 |
|
328 |
$log->debug("_delimiters_templates = ", sub { dump( $self->{_delimiters_templates} ) } ); |
329 |
|
330 |
foreach my $must (sort keys %{ $self->{must_exist} }) { |
331 |
next if ($fields->{$must}); |
332 |
$errors->{$must}->{missing}++; |
333 |
$errors->{dump} = $rec_dump if ($rec_dump); |
334 |
} |
335 |
|
336 |
if ($errors) { |
337 |
$log->debug("errors: ", $self->report_error( $errors ) ); |
338 |
|
339 |
my $mfn = $rec->{'000'}->[0] || $log->logconfess("record ", sub { dump( $rec ) }, " doesn't have MFN"); |
340 |
$self->{errors}->{$mfn} = $errors; |
341 |
} |
342 |
|
343 |
#$log->logcluck("return from this function is ARRAY") unless wantarray; |
344 |
|
345 |
return $errors; |
346 |
} |
347 |
|
348 |
=head2 reset |
349 |
|
350 |
Clean all accumulated errors for this input and remember delimiter templates |
351 |
for L<save_delimiters_templates> |
352 |
|
353 |
$validate->reset; |
354 |
|
355 |
This function B<must> be called after each input to provide accurate statistics. |
356 |
|
357 |
=cut |
358 |
|
359 |
sub reset { |
360 |
my $self = shift; |
361 |
|
362 |
my $log = $self->_get_logger; |
363 |
|
364 |
delete ($self->{errors}); |
365 |
|
366 |
if ( ! $self->{_delimiters_templates} ) { |
367 |
$log->debug("called without _delimiters_templates?"); |
368 |
return; |
369 |
} |
370 |
|
371 |
foreach my $f ( keys %{ $self->{_delimiters_templates} } ) { |
372 |
foreach my $t ( keys %{ $self->{_delimiters_templates}->{$f} } ) { |
373 |
$self->{_accumulated_delimiters_templates}->{$f}->{$t} += |
374 |
$self->{_delimiters_templates}->{$f}->{$t}; |
375 |
} |
376 |
} |
377 |
$log->debug("_accumulated_delimiters_templates = ", sub { dump( $self->{_accumulated_delimiters_templates} ) } ); |
378 |
delete ($self->{_delimiters_templates}); |
379 |
} |
380 |
|
381 |
=head2 all_errors |
382 |
|
383 |
Return hash with all errors |
384 |
|
385 |
print dump( $validate->all_errors ); |
386 |
|
387 |
=cut |
388 |
|
389 |
sub all_errors { |
390 |
my $self = shift; |
391 |
return $self->{errors}; |
392 |
} |
393 |
|
394 |
=head2 report_error |
395 |
|
396 |
Produce nice humanly readable report of single error |
397 |
|
398 |
print $validate->report_error( $error_hash ); |
399 |
|
400 |
=cut |
401 |
|
402 |
sub report_error { |
403 |
my $self = shift; |
404 |
|
405 |
my $h = shift || die "no hash?"; |
406 |
|
407 |
sub _unroll { |
408 |
my ($self, $tree, $accumulated) = @_; |
409 |
|
410 |
my $log = $self->_get_logger(); |
411 |
|
412 |
$log->debug("# ", |
413 |
( $tree ? "tree: $tree " : '' ), |
414 |
( $accumulated ? "accumulated: $accumulated " : '' ), |
415 |
); |
416 |
|
417 |
my $results; |
418 |
|
419 |
if (ref($tree) ne 'HASH') { |
420 |
return ("$accumulated\t($tree)", undef); |
421 |
} |
422 |
|
423 |
my $dump; |
424 |
|
425 |
foreach my $k (sort keys %{ $tree }) { |
426 |
|
427 |
if ($k eq 'dump') { |
428 |
$dump = $tree->{dump}; |
429 |
#warn "## dump ",dump($dump),"\n"; |
430 |
next; |
431 |
} |
432 |
|
433 |
$log->debug("current: $k"); |
434 |
|
435 |
my ($new_results, $new_dump) = $self->_unroll($tree->{$k}, |
436 |
$accumulated ? "$accumulated\t$k" : $k |
437 |
); |
438 |
|
439 |
$log->debug( "new_results: ", sub { dump($new_results) } ) if ( $new_results ); |
440 |
|
441 |
push @$results, $new_results if ($new_results); |
442 |
$dump = $new_dump if ($new_dump); |
443 |
|
444 |
} |
445 |
|
446 |
$log->debug( "results: ", sub { dump($results) } ) if ( $results ); |
447 |
|
448 |
if ($#$results == 0) { |
449 |
return ($results->[0], $dump); |
450 |
} else { |
451 |
return ($results, $dump); |
452 |
} |
453 |
} |
454 |
|
455 |
|
456 |
sub _reformat { |
457 |
my $l = shift; |
458 |
$l =~ s/\t/ /g; |
459 |
$l =~ s/_/ /g; |
460 |
return $l; |
461 |
} |
462 |
|
463 |
my $out = ''; |
464 |
|
465 |
for my $f (sort keys %{ $h }) { |
466 |
$out .= "$f: "; |
467 |
|
468 |
my ($r, $d) = $self->_unroll( $h->{$f} ); |
469 |
my $e; |
470 |
if (ref($r) eq 'ARRAY') { |
471 |
$e .= join(", ", map { _reformat( $_ ) } @$r); |
472 |
} else { |
473 |
$e .= _reformat( $r ); |
474 |
} |
475 |
$e .= "\n\t$d" if ($d); |
476 |
|
477 |
$out .= $e . "\n"; |
478 |
} |
479 |
return $out; |
480 |
} |
481 |
|
482 |
|
483 |
=head2 report |
484 |
|
485 |
Produce nice humanly readable report of errors |
486 |
|
487 |
print $validate->report; |
488 |
|
489 |
=cut |
490 |
|
491 |
sub report { |
492 |
my $self = shift; |
493 |
my $e = $self->{errors} || return; |
494 |
|
495 |
my $out; |
496 |
foreach my $mfn (sort { $a <=> $b } keys %$e) { |
497 |
$out .= "MFN $mfn\n" . $self->report_error( $e->{$mfn} ) . "\n"; |
498 |
} |
499 |
|
500 |
return $out; |
501 |
|
502 |
} |
503 |
|
504 |
=head2 delimiters_templates |
505 |
|
506 |
Generate report of delimiter tamplates |
507 |
|
508 |
my $report = $validate->delimiter_teplates( |
509 |
report => 1, |
510 |
current_input => 1, |
511 |
); |
512 |
|
513 |
Options: |
514 |
|
515 |
=over 4 |
516 |
|
517 |
=item report |
518 |
|
519 |
Generate humanly readable report with single fields |
520 |
|
521 |
=item current_input |
522 |
|
523 |
Report just current_input and not accumulated data |
524 |
|
525 |
=back |
526 |
|
527 |
=cut |
528 |
|
529 |
sub delimiters_templates { |
530 |
my $self = shift; |
531 |
|
532 |
my $args = {@_}; |
533 |
|
534 |
my $t = $self->{_accumulated_delimiters_templates}; |
535 |
$t = $self->{_delimiters_templates} if ( $args->{current_input} ); |
536 |
|
537 |
my $log = $self->_get_logger; |
538 |
|
539 |
unless ($t) { |
540 |
$log->error("called without delimiters"); |
541 |
return; |
542 |
} |
543 |
|
544 |
my $out; |
545 |
|
546 |
foreach my $f (sort { $a <=> $b } keys %$t) { |
547 |
$out .= "$f\n" if ( $args->{report} ); |
548 |
foreach my $template (sort { $a cmp $b } keys %{ $t->{$f} }) { |
549 |
my $count = $t->{$f}->{$template}; |
550 |
$out .= |
551 |
( $count ? "" : "# " ) . |
552 |
( $args->{report} ? "" : "$f" ) . |
553 |
"\t$count\t$template\n"; |
554 |
} |
555 |
} |
556 |
|
557 |
return $out; |
558 |
} |
559 |
|
560 |
=head2 save_delimiters_templates |
561 |
|
562 |
Save accumulated delimiter templates |
563 |
|
564 |
$validator->save_delimiters_template( '/path/to/validate/delimiters' ); |
565 |
|
566 |
=cut |
567 |
|
568 |
sub save_delimiters_templates { |
569 |
my $self = shift; |
570 |
|
571 |
my $path = shift; |
572 |
$path ||= $self->{delimiters_path}; |
573 |
|
574 |
my $log = $self->_get_logger; |
575 |
|
576 |
$log->logdie("need path") unless ( $path ); |
577 |
|
578 |
|
579 |
if ( ! $self->{_accumulated_delimiters_templates} ) { |
580 |
$log->error('no _accumulated_delimiters_templates found, reset'); |
581 |
$self->reset; |
582 |
} |
583 |
|
584 |
if ( $self->{_delimiters_templates} ) { |
585 |
$log->error('found _delimiters_templates, calling reset'); |
586 |
$self->reset; |
587 |
} |
588 |
|
589 |
$path .= '.new' if ( -e $path ); |
590 |
|
591 |
open(my $d, '>', $path) || $log->fatal("can't open $path: $!"); |
592 |
print $d $self->delimiters_templates; |
593 |
close($d); |
594 |
|
595 |
$log->info("new delimiters templates saved to $path"); |
596 |
} |
597 |
|
598 |
=head1 AUTHOR |
599 |
|
600 |
Dobrica Pavlinusic, C<< <dpavlin@rot13.org> >> |
601 |
|
602 |
=head1 COPYRIGHT & LICENSE |
603 |
|
604 |
Copyright 2006 Dobrica Pavlinusic, All Rights Reserved. |
605 |
|
606 |
This program is free software; you can redistribute it and/or modify it |
607 |
under the same terms as Perl itself. |
608 |
|
609 |
=cut |
610 |
|
611 |
1; # End of WebPAC::Validate |