1 |
dpavlin |
1 |
package WebPAC::Input; |
2 |
|
|
|
3 |
|
|
use warnings; |
4 |
|
|
use strict; |
5 |
|
|
|
6 |
dpavlin |
1168 |
use lib 'lib'; |
7 |
dpavlin |
507 |
|
8 |
dpavlin |
487 |
use WebPAC::Common; |
9 |
dpavlin |
285 |
use base qw/WebPAC::Common/; |
10 |
dpavlin |
771 |
use Data::Dump qw/dump/; |
11 |
dpavlin |
1100 |
use Encode qw/decode from_to/; |
12 |
dpavlin |
1128 |
use YAML; |
13 |
dpavlin |
285 |
|
14 |
dpavlin |
1 |
=head1 NAME |
15 |
|
|
|
16 |
dpavlin |
286 |
WebPAC::Input - read different file formats into WebPAC |
17 |
dpavlin |
1 |
|
18 |
|
|
=cut |
19 |
|
|
|
20 |
dpavlin |
1100 |
our $VERSION = '0.19'; |
21 |
dpavlin |
1 |
|
22 |
|
|
=head1 SYNOPSIS |
23 |
|
|
|
24 |
dpavlin |
286 |
This module implements input as database which have fixed and known |
25 |
|
|
I<size> while indexing and single unique numeric identifier for database |
26 |
|
|
position ranging from 1 to I<size>. |
27 |
dpavlin |
1 |
|
28 |
dpavlin |
286 |
Simply, something that is indexed by unmber from 1 .. I<size>. |
29 |
|
|
|
30 |
|
|
Examples of such databases are CDS/ISIS files, MARC files, lines in |
31 |
|
|
text file, and so on. |
32 |
|
|
|
33 |
|
|
Specific file formats are implemented using low-level interface modules, |
34 |
|
|
located in C<WebPAC::Input::*> namespace which export C<open_db>, |
35 |
|
|
C<fetch_rec> and optional C<init> functions. |
36 |
|
|
|
37 |
dpavlin |
1 |
Perhaps a little code snippet. |
38 |
|
|
|
39 |
dpavlin |
597 |
use WebPAC::Input; |
40 |
dpavlin |
1 |
|
41 |
dpavlin |
597 |
my $db = WebPAC::Input->new( |
42 |
|
|
module => 'WebPAC::Input::ISIS', |
43 |
|
|
); |
44 |
dpavlin |
1 |
|
45 |
dpavlin |
597 |
$db->open( path => '/path/to/database' ); |
46 |
dpavlin |
416 |
print "database size: ",$db->size,"\n"; |
47 |
|
|
while (my $rec = $db->fetch) { |
48 |
|
|
# do something with $rec |
49 |
|
|
} |
50 |
dpavlin |
1 |
|
51 |
dpavlin |
286 |
|
52 |
|
|
|
53 |
dpavlin |
1 |
=head1 FUNCTIONS |
54 |
|
|
|
55 |
dpavlin |
3 |
=head2 new |
56 |
dpavlin |
1 |
|
57 |
dpavlin |
3 |
Create new input database object. |
58 |
|
|
|
59 |
dpavlin |
9 |
my $db = new WebPAC::Input( |
60 |
dpavlin |
286 |
module => 'WebPAC::Input::MARC', |
61 |
dpavlin |
416 |
recode => 'char pairs', |
62 |
dpavlin |
483 |
no_progress_bar => 1, |
63 |
dpavlin |
868 |
input_config => { |
64 |
|
|
mapping => [ 'foo', 'bar', 'baz' ], |
65 |
|
|
}, |
66 |
dpavlin |
9 |
); |
67 |
dpavlin |
3 |
|
68 |
dpavlin |
597 |
C<module> is low-level file format module. See L<WebPAC::Input::ISIS> and |
69 |
dpavlin |
286 |
L<WebPAC::Input::MARC>. |
70 |
|
|
|
71 |
dpavlin |
483 |
C<recode> is optional string constisting of character or words pairs that |
72 |
|
|
should be replaced in input stream. |
73 |
|
|
|
74 |
|
|
C<no_progress_bar> disables progress bar output on C<STDOUT> |
75 |
|
|
|
76 |
dpavlin |
285 |
This function will also call low-level C<init> if it exists with same |
77 |
|
|
parametars. |
78 |
|
|
|
79 |
dpavlin |
1 |
=cut |
80 |
|
|
|
81 |
dpavlin |
3 |
sub new { |
82 |
dpavlin |
285 |
my $class = shift; |
83 |
|
|
my $self = {@_}; |
84 |
dpavlin |
3 |
bless($self, $class); |
85 |
|
|
|
86 |
dpavlin |
285 |
my $log = $self->_get_logger; |
87 |
|
|
|
88 |
dpavlin |
1100 |
$log->logconfess("code_page argument is not suppored any more.") if $self->{code_page}; |
89 |
|
|
$log->logconfess("encoding argument is not suppored any more.") if $self->{encoding}; |
90 |
|
|
$log->logconfess("lookup argument is not suppored any more. rewrite call to lookup_ref") if $self->{lookup}; |
91 |
|
|
$log->logconfess("low_mem argument is not suppored any more. rewrite it to load_row and save_row") if $self->{low_mem}; |
92 |
dpavlin |
585 |
|
93 |
dpavlin |
286 |
$log->logconfess("specify low-level file format module") unless ($self->{module}); |
94 |
dpavlin |
726 |
my $module_path = $self->{module}; |
95 |
|
|
$module_path =~ s#::#/#g; |
96 |
|
|
$module_path .= '.pm'; |
97 |
|
|
$log->debug("require low-level module $self->{module} from $module_path"); |
98 |
dpavlin |
289 |
|
99 |
dpavlin |
726 |
require $module_path; |
100 |
dpavlin |
286 |
|
101 |
dpavlin |
3 |
$self ? return $self : return undef; |
102 |
dpavlin |
1 |
} |
103 |
|
|
|
104 |
dpavlin |
285 |
=head2 open |
105 |
|
|
|
106 |
|
|
This function will read whole database in memory and produce lookups. |
107 |
|
|
|
108 |
dpavlin |
761 |
my $store; # simple in-memory hash |
109 |
|
|
|
110 |
dpavlin |
286 |
$input->open( |
111 |
dpavlin |
285 |
path => '/path/to/database/file', |
112 |
dpavlin |
1100 |
input_encoding => 'cp852', |
113 |
|
|
strict_encoding => 0, |
114 |
dpavlin |
286 |
limit => 500, |
115 |
|
|
offset => 6000, |
116 |
dpavlin |
506 |
stats => 1, |
117 |
dpavlin |
707 |
lookup_coderef => sub { |
118 |
|
|
my $rec = shift; |
119 |
|
|
# store lookups |
120 |
dpavlin |
585 |
}, |
121 |
dpavlin |
597 |
modify_records => { |
122 |
|
|
900 => { '^a' => { ' : ' => '^b' } }, |
123 |
|
|
901 => { '*' => { '^b' => ' ; ' } }, |
124 |
|
|
}, |
125 |
dpavlin |
636 |
modify_file => 'conf/modify/mapping.map', |
126 |
dpavlin |
761 |
save_row => sub { |
127 |
|
|
my $a = shift; |
128 |
|
|
$store->{ $a->{id} } = $a->{row}; |
129 |
|
|
}, |
130 |
|
|
load_row => sub { |
131 |
|
|
my $a = shift; |
132 |
|
|
return defined($store->{ $a->{id} }) && |
133 |
|
|
$store->{ $a->{id} }; |
134 |
|
|
}, |
135 |
|
|
|
136 |
dpavlin |
285 |
); |
137 |
|
|
|
138 |
dpavlin |
1100 |
By default, C<input_encoding> is assumed to be C<cp852>. |
139 |
dpavlin |
285 |
|
140 |
dpavlin |
1304 |
C<offset> is optional parametar to skip records at beginning. |
141 |
dpavlin |
285 |
|
142 |
dpavlin |
286 |
C<limit> is optional parametar to read just C<limit> records from database |
143 |
dpavlin |
285 |
|
144 |
dpavlin |
506 |
C<stats> create optional report about usage of fields and subfields |
145 |
|
|
|
146 |
dpavlin |
707 |
C<lookup_coderef> is closure to called to save data into lookups |
147 |
dpavlin |
585 |
|
148 |
dpavlin |
597 |
C<modify_records> specify mapping from subfields to delimiters or from |
149 |
|
|
delimiters to subfields, as well as oprations on fields (if subfield is |
150 |
|
|
defined as C<*>. |
151 |
|
|
|
152 |
dpavlin |
636 |
C<modify_file> is alternative for C<modify_records> above which preserves order and offers |
153 |
|
|
(hopefully) simplier sintax than YAML or perl (see L</modify_file_regex>). This option |
154 |
|
|
overrides C<modify_records> if both exists for same input. |
155 |
|
|
|
156 |
dpavlin |
761 |
C<save_row> and C<load_row> are low-level implementation of store engine. Calling convention |
157 |
|
|
is documented in example above. |
158 |
|
|
|
159 |
dpavlin |
1100 |
C<strict_encoding> should really default to 1, but it doesn't for now. |
160 |
|
|
|
161 |
dpavlin |
286 |
Returns size of database, regardless of C<offset> and C<limit> |
162 |
|
|
parametars, see also C<size>. |
163 |
dpavlin |
285 |
|
164 |
|
|
=cut |
165 |
|
|
|
166 |
|
|
sub open { |
167 |
|
|
my $self = shift; |
168 |
|
|
my $arg = {@_}; |
169 |
|
|
|
170 |
|
|
my $log = $self->_get_logger(); |
171 |
dpavlin |
1100 |
$log->debug( "arguments: ",dump( $arg )); |
172 |
dpavlin |
285 |
|
173 |
dpavlin |
1100 |
$log->logconfess("encoding argument is not suppored any more.") if $self->{encoding}; |
174 |
|
|
$log->logconfess("code_page argument is not suppored any more.") if $self->{code_page}; |
175 |
dpavlin |
585 |
$log->logconfess("lookup argument is not suppored any more. rewrite call to lookup_coderef") if ($arg->{lookup}); |
176 |
|
|
$log->logconfess("lookup_coderef must be CODE, not ",ref($arg->{lookup_coderef})) |
177 |
|
|
if ($arg->{lookup_coderef} && ref($arg->{lookup_coderef}) ne 'CODE'); |
178 |
|
|
|
179 |
dpavlin |
707 |
$log->debug( $arg->{lookup_coderef} ? '' : 'not ', "using lookup_coderef"); |
180 |
|
|
|
181 |
dpavlin |
285 |
$log->logcroak("need path") if (! $arg->{'path'}); |
182 |
dpavlin |
1100 |
my $input_encoding = $arg->{'input_encoding'} || $self->{'input_encoding'} || 'cp852'; |
183 |
dpavlin |
285 |
|
184 |
|
|
# store data in object |
185 |
dpavlin |
1307 |
$self->{$_} = $arg->{$_} foreach grep { defined $arg->{$_} } qw(path offset limit); |
186 |
dpavlin |
285 |
|
187 |
dpavlin |
761 |
if ($arg->{load_row} || $arg->{save_row}) { |
188 |
|
|
$log->logconfess("save_row and load_row must be defined in pair and be CODE") unless ( |
189 |
|
|
ref($arg->{load_row}) eq 'CODE' && |
190 |
|
|
ref($arg->{save_row}) eq 'CODE' |
191 |
|
|
); |
192 |
|
|
$self->{load_row} = $arg->{load_row}; |
193 |
|
|
$self->{save_row} = $arg->{save_row}; |
194 |
|
|
$log->debug("using load_row and save_row instead of in-memory hash"); |
195 |
|
|
} |
196 |
|
|
|
197 |
dpavlin |
416 |
my $filter_ref; |
198 |
dpavlin |
597 |
my $recode_regex; |
199 |
|
|
my $recode_map; |
200 |
dpavlin |
416 |
|
201 |
|
|
if ($self->{recode}) { |
202 |
|
|
my @r = split(/\s/, $self->{recode}); |
203 |
|
|
if ($#r % 2 != 1) { |
204 |
|
|
$log->logwarn("recode needs even number of elements (some number of valid pairs)"); |
205 |
|
|
} else { |
206 |
|
|
while (@r) { |
207 |
|
|
my $from = shift @r; |
208 |
|
|
my $to = shift @r; |
209 |
dpavlin |
597 |
$recode_map->{$from} = $to; |
210 |
dpavlin |
416 |
} |
211 |
|
|
|
212 |
dpavlin |
597 |
$recode_regex = join '|' => keys %{ $recode_map }; |
213 |
dpavlin |
416 |
|
214 |
dpavlin |
597 |
$log->debug("using recode regex: $recode_regex"); |
215 |
dpavlin |
416 |
} |
216 |
|
|
|
217 |
|
|
} |
218 |
|
|
|
219 |
dpavlin |
636 |
my $rec_regex; |
220 |
|
|
if (my $p = $arg->{modify_file}) { |
221 |
|
|
$log->debug("using modify_file $p"); |
222 |
|
|
$rec_regex = $self->modify_file_regexps( $p ); |
223 |
|
|
} elsif (my $h = $arg->{modify_records}) { |
224 |
dpavlin |
771 |
$log->debug("using modify_records ", sub { dump( $h ) }); |
225 |
dpavlin |
636 |
$rec_regex = $self->modify_record_regexps(%{ $h }); |
226 |
|
|
} |
227 |
dpavlin |
771 |
$log->debug("rec_regex: ", sub { dump($rec_regex) }) if ($rec_regex); |
228 |
dpavlin |
597 |
|
229 |
dpavlin |
726 |
my $class = $self->{module} || $log->logconfess("can't get low-level module name!"); |
230 |
|
|
|
231 |
dpavlin |
1286 |
$arg->{$_} = $self->{$_} foreach qw(offset limit); |
232 |
|
|
|
233 |
dpavlin |
726 |
my $ll_db = $class->new( |
234 |
dpavlin |
285 |
path => $arg->{path}, |
235 |
dpavlin |
873 |
input_config => $arg->{input_config} || $self->{input_config}, |
236 |
dpavlin |
624 |
# filter => sub { |
237 |
|
|
# my ($l,$f_nr) = @_; |
238 |
|
|
# return unless defined($l); |
239 |
dpavlin |
1100 |
# $l = decode($input_encoding, $l); |
240 |
dpavlin |
624 |
# $l =~ s/($recode_regex)/$recode_map->{$1}/g if ($recode_regex && $recode_map); |
241 |
|
|
# return $l; |
242 |
|
|
# }, |
243 |
dpavlin |
523 |
%{ $arg }, |
244 |
dpavlin |
285 |
); |
245 |
|
|
|
246 |
dpavlin |
1236 |
# save for dump and input_module |
247 |
|
|
$self->{ll_db} = $ll_db; |
248 |
|
|
|
249 |
dpavlin |
726 |
unless (defined($ll_db)) { |
250 |
dpavlin |
285 |
$log->logwarn("can't open database $arg->{path}, skipping..."); |
251 |
|
|
return; |
252 |
|
|
} |
253 |
|
|
|
254 |
dpavlin |
726 |
my $size = $ll_db->size; |
255 |
|
|
|
256 |
dpavlin |
285 |
unless ($size) { |
257 |
|
|
$log->logwarn("no records in database $arg->{path}, skipping..."); |
258 |
|
|
return; |
259 |
|
|
} |
260 |
|
|
|
261 |
dpavlin |
339 |
my $from_rec = 1; |
262 |
|
|
my $to_rec = $size; |
263 |
dpavlin |
285 |
|
264 |
dpavlin |
286 |
if (my $s = $self->{offset}) { |
265 |
dpavlin |
1304 |
$log->debug("offset $s records"); |
266 |
|
|
$from_rec = $s + 1; |
267 |
dpavlin |
285 |
} else { |
268 |
dpavlin |
1304 |
$self->{offset} = $from_rec - 1; |
269 |
dpavlin |
285 |
} |
270 |
|
|
|
271 |
dpavlin |
286 |
if ($self->{limit}) { |
272 |
dpavlin |
301 |
$log->debug("limiting to ",$self->{limit}," records"); |
273 |
dpavlin |
339 |
$to_rec = $from_rec + $self->{limit} - 1; |
274 |
|
|
$to_rec = $size if ($to_rec > $size); |
275 |
dpavlin |
285 |
} |
276 |
|
|
|
277 |
dpavlin |
1100 |
my $strict_encoding = $arg->{strict_encoding} || $self->{strict_encoding}; ## FIXME should be 1 really |
278 |
dpavlin |
285 |
|
279 |
dpavlin |
1100 |
$log->info("processing $self->{size}/$size records [$from_rec-$to_rec]", |
280 |
|
|
" encoding $input_encoding ", $strict_encoding ? ' [strict]' : '', |
281 |
|
|
$self->{stats} ? ' [stats]' : '', |
282 |
|
|
); |
283 |
|
|
|
284 |
dpavlin |
1307 |
$self->{size} = 0; |
285 |
|
|
|
286 |
dpavlin |
285 |
# read database |
287 |
dpavlin |
339 |
for (my $pos = $from_rec; $pos <= $to_rec; $pos++) { |
288 |
dpavlin |
285 |
|
289 |
dpavlin |
286 |
$log->debug("position: $pos\n"); |
290 |
dpavlin |
285 |
|
291 |
dpavlin |
1307 |
$self->{size}++; # XXX I could move this more down if I didn't want empty records... |
292 |
|
|
|
293 |
dpavlin |
726 |
my $rec = $ll_db->fetch_rec($pos, sub { |
294 |
dpavlin |
799 |
my ($l,$f_nr,$debug) = @_; |
295 |
dpavlin |
624 |
# return unless defined($l); |
296 |
|
|
# return $l unless ($rec_regex && $f_nr); |
297 |
dpavlin |
285 |
|
298 |
dpavlin |
818 |
return unless ( defined($l) && defined($f_nr) ); |
299 |
|
|
|
300 |
dpavlin |
800 |
warn "-=> $f_nr ## |$l|\n" if ($debug); |
301 |
dpavlin |
625 |
$log->debug("-=> $f_nr ## $l"); |
302 |
|
|
|
303 |
dpavlin |
624 |
# codepage conversion and recode_regex |
304 |
dpavlin |
1107 |
$l = decode($input_encoding, $l, 1); |
305 |
dpavlin |
624 |
$l =~ s/($recode_regex)/$recode_map->{$1}/g if ($recode_regex && $recode_map); |
306 |
|
|
|
307 |
dpavlin |
619 |
# apply regexps |
308 |
|
|
if ($rec_regex && defined($rec_regex->{$f_nr})) { |
309 |
|
|
$log->logconfess("regexps->{$f_nr} must be ARRAY") if (ref($rec_regex->{$f_nr}) ne 'ARRAY'); |
310 |
|
|
my $c = 0; |
311 |
|
|
foreach my $r (@{ $rec_regex->{$f_nr} }) { |
312 |
dpavlin |
626 |
my $old_l = $l; |
313 |
dpavlin |
800 |
$log->logconfess("expected regex in ", dump( $r )) unless defined($r->{regex}); |
314 |
|
|
eval '$l =~ ' . $r->{regex}; |
315 |
dpavlin |
626 |
if ($old_l ne $l) { |
316 |
dpavlin |
800 |
my $d = "|$old_l| -> |$l| "; # . $r->{regex}; |
317 |
|
|
$d .= ' +' . $r->{line} . ' ' . $r->{file} if defined($r->{line}); |
318 |
|
|
$d .= ' ' . $r->{debug} if defined($r->{debug}); |
319 |
|
|
$log->debug("MODIFY $d"); |
320 |
|
|
warn "*** $d\n" if ($debug); |
321 |
|
|
|
322 |
dpavlin |
626 |
} |
323 |
dpavlin |
1122 |
$log->error("error applying regex: ",dump($r), $@) if $@; |
324 |
dpavlin |
619 |
} |
325 |
|
|
} |
326 |
|
|
|
327 |
dpavlin |
800 |
$log->debug("<=- $f_nr ## |$l|"); |
328 |
dpavlin |
799 |
warn "<=- $f_nr ## $l\n" if ($debug); |
329 |
dpavlin |
619 |
return $l; |
330 |
|
|
}); |
331 |
|
|
|
332 |
dpavlin |
771 |
$log->debug(sub { dump($rec) }); |
333 |
dpavlin |
308 |
|
334 |
dpavlin |
285 |
if (! $rec) { |
335 |
dpavlin |
286 |
$log->warn("record $pos empty? skipping..."); |
336 |
dpavlin |
285 |
next; |
337 |
|
|
} |
338 |
|
|
|
339 |
|
|
# store |
340 |
dpavlin |
761 |
if ($self->{save_row}) { |
341 |
|
|
$self->{save_row}->({ |
342 |
|
|
id => $pos, |
343 |
|
|
row => $rec, |
344 |
|
|
}); |
345 |
dpavlin |
285 |
} else { |
346 |
dpavlin |
286 |
$self->{data}->{$pos} = $rec; |
347 |
dpavlin |
285 |
} |
348 |
|
|
|
349 |
|
|
# create lookup |
350 |
dpavlin |
585 |
$arg->{'lookup_coderef'}->( $rec ) if ($rec && $arg->{'lookup_coderef'}); |
351 |
dpavlin |
285 |
|
352 |
dpavlin |
506 |
# update counters for statistics |
353 |
|
|
if ($self->{stats}) { |
354 |
dpavlin |
593 |
|
355 |
dpavlin |
619 |
# fetch clean record with regexpes applied for statistics |
356 |
dpavlin |
726 |
my $rec = $ll_db->fetch_rec($pos); |
357 |
dpavlin |
619 |
|
358 |
dpavlin |
593 |
foreach my $fld (keys %{ $rec }) { |
359 |
dpavlin |
506 |
$self->{_stats}->{fld}->{ $fld }++; |
360 |
dpavlin |
593 |
|
361 |
dpavlin |
1076 |
#$log->logdie("invalid record fild $fld, not ARRAY") |
362 |
|
|
next unless (ref($rec->{ $fld }) eq 'ARRAY'); |
363 |
dpavlin |
593 |
|
364 |
|
|
foreach my $row (@{ $rec->{$fld} }) { |
365 |
|
|
|
366 |
|
|
if (ref($row) eq 'HASH') { |
367 |
|
|
|
368 |
|
|
foreach my $sf (keys %{ $row }) { |
369 |
dpavlin |
619 |
next if ($sf eq 'subfields'); |
370 |
dpavlin |
593 |
$self->{_stats}->{sf}->{ $fld }->{ $sf }->{count}++; |
371 |
|
|
$self->{_stats}->{sf}->{ $fld }->{ $sf }->{repeatable}++ |
372 |
|
|
if (ref($row->{$sf}) eq 'ARRAY'); |
373 |
dpavlin |
506 |
} |
374 |
dpavlin |
593 |
|
375 |
|
|
} else { |
376 |
|
|
$self->{_stats}->{repeatable}->{ $fld }++; |
377 |
|
|
} |
378 |
dpavlin |
506 |
} |
379 |
dpavlin |
593 |
} |
380 |
dpavlin |
506 |
} |
381 |
|
|
|
382 |
dpavlin |
483 |
$self->progress_bar($pos,$to_rec) unless ($self->{no_progress_bar}); |
383 |
dpavlin |
285 |
|
384 |
|
|
} |
385 |
|
|
|
386 |
dpavlin |
286 |
$self->{pos} = -1; |
387 |
|
|
$self->{last_pcnt} = 0; |
388 |
dpavlin |
285 |
|
389 |
|
|
# store max mfn and return it. |
390 |
dpavlin |
339 |
$self->{max_pos} = $to_rec; |
391 |
|
|
$log->debug("max_pos: $to_rec"); |
392 |
dpavlin |
285 |
|
393 |
|
|
return $size; |
394 |
|
|
} |
395 |
|
|
|
396 |
dpavlin |
1221 |
sub input_module { $_[0]->{ll_db} } |
397 |
|
|
|
398 |
dpavlin |
285 |
=head2 fetch |
399 |
|
|
|
400 |
|
|
Fetch next record from database. It will also displays progress bar. |
401 |
|
|
|
402 |
|
|
my $rec = $isis->fetch; |
403 |
|
|
|
404 |
|
|
Record from this function should probably go to C<data_structure> for |
405 |
|
|
normalisation. |
406 |
|
|
|
407 |
|
|
=cut |
408 |
|
|
|
409 |
|
|
sub fetch { |
410 |
|
|
my $self = shift; |
411 |
|
|
|
412 |
|
|
my $log = $self->_get_logger(); |
413 |
|
|
|
414 |
dpavlin |
286 |
$log->logconfess("it seems that you didn't load database!") unless ($self->{pos}); |
415 |
dpavlin |
285 |
|
416 |
dpavlin |
286 |
if ($self->{pos} == -1) { |
417 |
dpavlin |
1304 |
$self->{pos} = $self->{offset} + 1; |
418 |
dpavlin |
285 |
} else { |
419 |
dpavlin |
286 |
$self->{pos}++; |
420 |
dpavlin |
285 |
} |
421 |
|
|
|
422 |
dpavlin |
286 |
my $mfn = $self->{pos}; |
423 |
dpavlin |
285 |
|
424 |
dpavlin |
286 |
if ($mfn > $self->{max_pos}) { |
425 |
|
|
$self->{pos} = $self->{max_pos}; |
426 |
dpavlin |
285 |
$log->debug("at EOF"); |
427 |
|
|
return; |
428 |
|
|
} |
429 |
|
|
|
430 |
dpavlin |
483 |
$self->progress_bar($mfn,$self->{max_pos}) unless ($self->{no_progress_bar}); |
431 |
dpavlin |
285 |
|
432 |
|
|
my $rec; |
433 |
|
|
|
434 |
dpavlin |
761 |
if ($self->{load_row}) { |
435 |
|
|
$rec = $self->{load_row}->({ id => $mfn }); |
436 |
dpavlin |
285 |
} else { |
437 |
dpavlin |
286 |
$rec = $self->{data}->{$mfn}; |
438 |
dpavlin |
285 |
} |
439 |
|
|
|
440 |
|
|
$rec ||= 0E0; |
441 |
|
|
} |
442 |
|
|
|
443 |
|
|
=head2 pos |
444 |
|
|
|
445 |
|
|
Returns current record number (MFN). |
446 |
|
|
|
447 |
|
|
print $isis->pos; |
448 |
|
|
|
449 |
|
|
First record in database has position 1. |
450 |
|
|
|
451 |
|
|
=cut |
452 |
|
|
|
453 |
|
|
sub pos { |
454 |
|
|
my $self = shift; |
455 |
dpavlin |
286 |
return $self->{pos}; |
456 |
dpavlin |
285 |
} |
457 |
|
|
|
458 |
|
|
|
459 |
|
|
=head2 size |
460 |
|
|
|
461 |
|
|
Returns number of records in database |
462 |
|
|
|
463 |
|
|
print $isis->size; |
464 |
|
|
|
465 |
|
|
Result from this function can be used to loop through all records |
466 |
|
|
|
467 |
|
|
foreach my $mfn ( 1 ... $isis->size ) { ... } |
468 |
|
|
|
469 |
dpavlin |
286 |
because it takes into account C<offset> and C<limit>. |
470 |
dpavlin |
285 |
|
471 |
|
|
=cut |
472 |
|
|
|
473 |
|
|
sub size { |
474 |
|
|
my $self = shift; |
475 |
dpavlin |
1306 |
return $self->{size}; # FIXME this is buggy if open is called multiple times! |
476 |
dpavlin |
285 |
} |
477 |
|
|
|
478 |
|
|
=head2 seek |
479 |
|
|
|
480 |
|
|
Seek to specified MFN in file. |
481 |
|
|
|
482 |
|
|
$isis->seek(42); |
483 |
|
|
|
484 |
|
|
First record in database has position 1. |
485 |
|
|
|
486 |
|
|
=cut |
487 |
|
|
|
488 |
|
|
sub seek { |
489 |
|
|
my $self = shift; |
490 |
dpavlin |
793 |
my $pos = shift; |
491 |
dpavlin |
285 |
|
492 |
|
|
my $log = $self->_get_logger(); |
493 |
|
|
|
494 |
dpavlin |
797 |
$log->logconfess("called without pos") unless defined($pos); |
495 |
dpavlin |
793 |
|
496 |
dpavlin |
285 |
if ($pos < 1) { |
497 |
|
|
$log->warn("seek before first record"); |
498 |
|
|
$pos = 1; |
499 |
dpavlin |
286 |
} elsif ($pos > $self->{max_pos}) { |
500 |
dpavlin |
285 |
$log->warn("seek beyond last record"); |
501 |
dpavlin |
286 |
$pos = $self->{max_pos}; |
502 |
dpavlin |
285 |
} |
503 |
|
|
|
504 |
dpavlin |
286 |
return $self->{pos} = (($pos - 1) || -1); |
505 |
dpavlin |
285 |
} |
506 |
|
|
|
507 |
dpavlin |
506 |
=head2 stats |
508 |
dpavlin |
285 |
|
509 |
dpavlin |
506 |
Dump statistics about field and subfield usage |
510 |
|
|
|
511 |
dpavlin |
507 |
print $input->stats; |
512 |
dpavlin |
506 |
|
513 |
|
|
=cut |
514 |
|
|
|
515 |
|
|
sub stats { |
516 |
|
|
my $self = shift; |
517 |
dpavlin |
507 |
|
518 |
|
|
my $log = $self->_get_logger(); |
519 |
|
|
|
520 |
|
|
my $s = $self->{_stats}; |
521 |
|
|
if (! $s) { |
522 |
|
|
$log->warn("called stats, but there is no statistics collected"); |
523 |
|
|
return; |
524 |
|
|
} |
525 |
|
|
|
526 |
|
|
my $max_fld = 0; |
527 |
|
|
|
528 |
|
|
my $out = join("\n", |
529 |
|
|
map { |
530 |
dpavlin |
844 |
my $f = $_; |
531 |
|
|
die "no field in ", dump( $s->{fld} ) unless defined( $f ); |
532 |
dpavlin |
507 |
my $v = $s->{fld}->{$f} || die "no s->{fld}->{$f}"; |
533 |
|
|
$max_fld = $v if ($v > $max_fld); |
534 |
|
|
|
535 |
dpavlin |
519 |
my $o = sprintf("%4s %d ~", $f, $v); |
536 |
dpavlin |
507 |
|
537 |
|
|
if (defined($s->{sf}->{$f})) { |
538 |
dpavlin |
909 |
my @subfields = keys %{ $s->{sf}->{$f} }; |
539 |
dpavlin |
507 |
map { |
540 |
dpavlin |
593 |
$o .= sprintf(" %s:%d%s", $_, |
541 |
|
|
$s->{sf}->{$f}->{$_}->{count}, |
542 |
|
|
$s->{sf}->{$f}->{$_}->{repeatable} ? '*' : '', |
543 |
|
|
); |
544 |
dpavlin |
909 |
} ( |
545 |
dpavlin |
910 |
# first indicators and other special subfields |
546 |
|
|
sort( grep { length($_) > 1 } @subfields ), |
547 |
|
|
# then subfileds (single char) |
548 |
|
|
sort( grep { length($_) == 1 } @subfields ), |
549 |
dpavlin |
909 |
); |
550 |
dpavlin |
507 |
} |
551 |
|
|
|
552 |
|
|
if (my $v_r = $s->{repeatable}->{$f}) { |
553 |
|
|
$o .= " ($v_r)" if ($v_r != $v); |
554 |
|
|
} |
555 |
|
|
|
556 |
|
|
$o; |
557 |
dpavlin |
1076 |
} sort { |
558 |
|
|
if ( $a =~ m/^\d+$/ && $b =~ m/^\d+$/ ) { |
559 |
|
|
$a <=> $b |
560 |
|
|
} else { |
561 |
|
|
$a cmp $b |
562 |
|
|
} |
563 |
|
|
} keys %{ $s->{fld} } |
564 |
dpavlin |
507 |
); |
565 |
|
|
|
566 |
dpavlin |
771 |
$log->debug( sub { dump($s) } ); |
567 |
dpavlin |
507 |
|
568 |
dpavlin |
1128 |
my $path = 'var/stats.yml'; |
569 |
|
|
YAML::DumpFile( $path, $s ); |
570 |
|
|
$log->info( 'created ', $path, ' with ', -s $path, ' bytes' ); |
571 |
|
|
|
572 |
dpavlin |
507 |
return $out; |
573 |
dpavlin |
506 |
} |
574 |
|
|
|
575 |
dpavlin |
774 |
=head2 dump_ascii |
576 |
dpavlin |
652 |
|
577 |
|
|
Display humanly readable dump of record |
578 |
|
|
|
579 |
|
|
=cut |
580 |
|
|
|
581 |
dpavlin |
774 |
sub dump_ascii { |
582 |
dpavlin |
652 |
my $self = shift; |
583 |
|
|
|
584 |
dpavlin |
771 |
return unless $self->{ll_db}; |
585 |
dpavlin |
652 |
|
586 |
dpavlin |
860 |
if ($self->{ll_db}->can('dump_ascii')) { |
587 |
dpavlin |
774 |
return $self->{ll_db}->dump_ascii( $self->{pos} ); |
588 |
dpavlin |
771 |
} else { |
589 |
|
|
return dump( $self->{ll_db}->fetch_rec( $self->{pos} ) ); |
590 |
|
|
} |
591 |
dpavlin |
652 |
} |
592 |
|
|
|
593 |
dpavlin |
823 |
=head2 _get_regex |
594 |
dpavlin |
597 |
|
595 |
dpavlin |
823 |
Helper function called which create regexps to be execute on code. |
596 |
dpavlin |
597 |
|
597 |
dpavlin |
823 |
_get_regex( 900, 'regex:[0-9]+' ,'numbers' ); |
598 |
|
|
_get_regex( 900, '^b', ' : ^b' ); |
599 |
dpavlin |
597 |
|
600 |
dpavlin |
823 |
It supports perl regexps with C<regex:> prefix to from value and has |
601 |
|
|
additional logic to skip empty subfields. |
602 |
|
|
|
603 |
dpavlin |
597 |
=cut |
604 |
|
|
|
605 |
dpavlin |
636 |
sub _get_regex { |
606 |
|
|
my ($sf,$from,$to) = @_; |
607 |
dpavlin |
797 |
|
608 |
|
|
# protect / |
609 |
|
|
$from =~ s!/!\\/!gs; |
610 |
|
|
$to =~ s!/!\\/!gs; |
611 |
|
|
|
612 |
dpavlin |
784 |
if ($from =~ m/^regex:(.+)$/) { |
613 |
|
|
$from = $1; |
614 |
|
|
} else { |
615 |
|
|
$from = '\Q' . $from . '\E'; |
616 |
|
|
} |
617 |
dpavlin |
636 |
if ($sf =~ /^\^/) { |
618 |
dpavlin |
823 |
my $need_subfield_data = '*'; # no |
619 |
|
|
# if from is also subfield, require some data in between |
620 |
|
|
# to correctly skip empty subfields |
621 |
|
|
$need_subfield_data = '+' if ($from =~ m/^\\Q\^/); |
622 |
dpavlin |
636 |
return |
623 |
dpavlin |
823 |
's/\Q'. $sf .'\E([^\^]' . $need_subfield_data . '?)'. $from .'([^\^]*?)/'. $sf .'$1'. $to .'$2/'; |
624 |
dpavlin |
636 |
} else { |
625 |
|
|
return |
626 |
dpavlin |
784 |
's/'. $from .'/'. $to .'/g'; |
627 |
dpavlin |
636 |
} |
628 |
|
|
} |
629 |
|
|
|
630 |
dpavlin |
823 |
|
631 |
|
|
=head2 modify_record_regexps |
632 |
|
|
|
633 |
|
|
Generate hash with regexpes to be applied using L<filter>. |
634 |
|
|
|
635 |
|
|
my $regexpes = $input->modify_record_regexps( |
636 |
|
|
900 => { '^a' => { ' : ' => '^b' } }, |
637 |
|
|
901 => { '*' => { '^b' => ' ; ' } }, |
638 |
|
|
); |
639 |
|
|
|
640 |
|
|
=cut |
641 |
|
|
|
642 |
dpavlin |
597 |
sub modify_record_regexps { |
643 |
|
|
my $self = shift; |
644 |
|
|
my $modify_record = {@_}; |
645 |
|
|
|
646 |
|
|
my $regexpes; |
647 |
|
|
|
648 |
dpavlin |
634 |
my $log = $self->_get_logger(); |
649 |
|
|
|
650 |
dpavlin |
597 |
foreach my $f (keys %$modify_record) { |
651 |
dpavlin |
634 |
$log->debug("field: $f"); |
652 |
|
|
|
653 |
dpavlin |
597 |
foreach my $sf (keys %{ $modify_record->{$f} }) { |
654 |
dpavlin |
634 |
$log->debug("subfield: $sf"); |
655 |
|
|
|
656 |
dpavlin |
597 |
foreach my $from (keys %{ $modify_record->{$f}->{$sf} }) { |
657 |
|
|
my $to = $modify_record->{$f}->{$sf}->{$from}; |
658 |
|
|
#die "no field?" unless defined($to); |
659 |
dpavlin |
800 |
my $d = "|$from| -> |$to|"; |
660 |
|
|
$log->debug("transform: $d"); |
661 |
dpavlin |
597 |
|
662 |
dpavlin |
636 |
my $regex = _get_regex($sf,$from,$to); |
663 |
dpavlin |
800 |
push @{ $regexpes->{$f} }, { regex => $regex, debug => $d }; |
664 |
dpavlin |
636 |
$log->debug("regex: $regex"); |
665 |
dpavlin |
597 |
} |
666 |
|
|
} |
667 |
|
|
} |
668 |
|
|
|
669 |
|
|
return $regexpes; |
670 |
|
|
} |
671 |
|
|
|
672 |
dpavlin |
636 |
=head2 modify_file_regexps |
673 |
|
|
|
674 |
dpavlin |
799 |
Generate hash with regexpes to be applied using L<filter> from |
675 |
dpavlin |
636 |
pseudo hash/yaml format for regex mappings. |
676 |
|
|
|
677 |
|
|
It should be obvious: |
678 |
|
|
|
679 |
|
|
200 |
680 |
|
|
'^a' |
681 |
|
|
' : ' => '^e' |
682 |
|
|
' = ' => '^d' |
683 |
|
|
|
684 |
|
|
In field I<200> find C<'^a'> and then C<' : '>, and replace it with C<'^e'>. |
685 |
|
|
In field I<200> find C<'^a'> and then C<' = '>, and replace it with C<'^d'>. |
686 |
|
|
|
687 |
|
|
my $regexpes = $input->modify_file_regexps( 'conf/modify/common.pl' ); |
688 |
|
|
|
689 |
|
|
On undef path it will just return. |
690 |
|
|
|
691 |
|
|
=cut |
692 |
|
|
|
693 |
|
|
sub modify_file_regexps { |
694 |
|
|
my $self = shift; |
695 |
|
|
|
696 |
|
|
my $modify_path = shift || return; |
697 |
|
|
|
698 |
|
|
my $log = $self->_get_logger(); |
699 |
|
|
|
700 |
|
|
my $regexpes; |
701 |
|
|
|
702 |
dpavlin |
697 |
CORE::open(my $fh, $modify_path) || $log->logdie("can't open modify file $modify_path: $!"); |
703 |
dpavlin |
636 |
|
704 |
|
|
my ($f,$sf); |
705 |
|
|
|
706 |
|
|
while(<$fh>) { |
707 |
|
|
chomp; |
708 |
|
|
next if (/^#/ || /^\s*$/); |
709 |
|
|
|
710 |
|
|
if (/^\s*(\d+)\s*$/) { |
711 |
|
|
$f = $1; |
712 |
|
|
$log->debug("field: $f"); |
713 |
|
|
next; |
714 |
|
|
} elsif (/^\s*'([^']*)'\s*$/) { |
715 |
|
|
$sf = $1; |
716 |
|
|
$log->die("can't define subfiled before field in: $_") unless ($f); |
717 |
|
|
$log->debug("subfield: $sf"); |
718 |
|
|
} elsif (/^\s*'([^']*)'\s*=>\s*'([^']*)'\s*$/) { |
719 |
|
|
my ($from,$to) = ($1, $2); |
720 |
|
|
|
721 |
|
|
$log->debug("transform: |$from| -> |$to|"); |
722 |
|
|
|
723 |
|
|
my $regex = _get_regex($sf,$from,$to); |
724 |
dpavlin |
800 |
push @{ $regexpes->{$f} }, { |
725 |
|
|
regex => $regex, |
726 |
|
|
file => $modify_path, |
727 |
|
|
line => $., |
728 |
|
|
}; |
729 |
dpavlin |
636 |
$log->debug("regex: $regex"); |
730 |
dpavlin |
1121 |
} else { |
731 |
|
|
die "can't parse: $_"; |
732 |
dpavlin |
636 |
} |
733 |
|
|
} |
734 |
|
|
|
735 |
|
|
return $regexpes; |
736 |
|
|
} |
737 |
|
|
|
738 |
dpavlin |
3 |
=head1 AUTHOR |
739 |
|
|
|
740 |
|
|
Dobrica Pavlinusic, C<< <dpavlin@rot13.org> >> |
741 |
|
|
|
742 |
dpavlin |
1 |
=head1 COPYRIGHT & LICENSE |
743 |
|
|
|
744 |
dpavlin |
599 |
Copyright 2005-2006 Dobrica Pavlinusic, All Rights Reserved. |
745 |
dpavlin |
1 |
|
746 |
|
|
This program is free software; you can redistribute it and/or modify it |
747 |
|
|
under the same terms as Perl itself. |
748 |
|
|
|
749 |
|
|
=cut |
750 |
|
|
|
751 |
|
|
1; # End of WebPAC::Input |