line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
package LucyX::Suggester; |
2
|
2
|
|
|
2
|
|
50391
|
use warnings; |
|
2
|
|
|
|
|
6
|
|
|
2
|
|
|
|
|
63
|
|
3
|
2
|
|
|
2
|
|
10
|
use strict; |
|
2
|
|
|
|
|
3
|
|
|
2
|
|
|
|
|
70
|
|
4
|
2
|
|
|
2
|
|
10
|
use Carp; |
|
2
|
|
|
|
|
8
|
|
|
2
|
|
|
|
|
132
|
|
5
|
2
|
|
|
2
|
|
1813
|
use Data::Dump qw( dump ); |
|
2
|
|
|
|
|
18001
|
|
|
2
|
|
|
|
|
192
|
|
6
|
2
|
|
|
2
|
|
2101
|
use Search::Tools; |
|
2
|
|
|
|
|
3415
|
|
|
2
|
|
|
|
|
74
|
|
7
|
2
|
|
|
2
|
|
2247
|
use Lucy; |
|
0
|
|
|
|
|
|
|
|
0
|
|
|
|
|
|
|
8
|
|
|
|
|
|
|
|
9
|
|
|
|
|
|
|
our $VERSION = '0.005'; |
10
|
|
|
|
|
|
|
|
11
|
|
|
|
|
|
|
=head1 NAME |
12
|
|
|
|
|
|
|
|
13
|
|
|
|
|
|
|
LucyX::Suggester - suggest terms for Apache Lucy search |
14
|
|
|
|
|
|
|
|
15
|
|
|
|
|
|
|
=head1 SYNOPSIS |
16
|
|
|
|
|
|
|
|
17
|
|
|
|
|
|
|
use LucyX::Suggester; |
18
|
|
|
|
|
|
|
my $suggester = LucyX::Suggester->new( |
19
|
|
|
|
|
|
|
fields => [qw( foo bar )], |
20
|
|
|
|
|
|
|
indexes => $list_of_indexes, |
21
|
|
|
|
|
|
|
spellcheck => $search_tools_spellcheck, |
22
|
|
|
|
|
|
|
limit => 10, |
23
|
|
|
|
|
|
|
use_regex => 0, |
24
|
|
|
|
|
|
|
max_length => 64, |
25
|
|
|
|
|
|
|
); |
26
|
|
|
|
|
|
|
my $suggestions = $suggester->suggest('quiK brwn fox'); |
27
|
|
|
|
|
|
|
|
28
|
|
|
|
|
|
|
=head1 DESCRIPTION |
29
|
|
|
|
|
|
|
|
30
|
|
|
|
|
|
|
Inspired by the Solr Suggester feature, LucyX::Suggester |
31
|
|
|
|
|
|
|
will return a list of suggested terms based on |
32
|
|
|
|
|
|
|
actual terms in the specified index(es). Spellchecking |
33
|
|
|
|
|
|
|
on query terms is performed with Search::Tools::SpellCheck, |
34
|
|
|
|
|
|
|
which uses Text::Aspell. |
35
|
|
|
|
|
|
|
|
36
|
|
|
|
|
|
|
=head1 METHODS |
37
|
|
|
|
|
|
|
|
38
|
|
|
|
|
|
|
=head2 new( I ) |
39
|
|
|
|
|
|
|
|
40
|
|
|
|
|
|
|
Returns a new Suggester object. Supported I include: |
41
|
|
|
|
|
|
|
|
42
|
|
|
|
|
|
|
=over |
43
|
|
|
|
|
|
|
|
44
|
|
|
|
|
|
|
=item fields I |
45
|
|
|
|
|
|
|
|
46
|
|
|
|
|
|
|
List of fields to limit Lexicon scans to. |
47
|
|
|
|
|
|
|
|
48
|
|
|
|
|
|
|
=item indexes I |
49
|
|
|
|
|
|
|
|
50
|
|
|
|
|
|
|
List of indexes to search within. |
51
|
|
|
|
|
|
|
|
52
|
|
|
|
|
|
|
=item spellcheck I |
53
|
|
|
|
|
|
|
|
54
|
|
|
|
|
|
|
An instance of L. Set |
55
|
|
|
|
|
|
|
this to indicate custom values for language, dictionary, |
56
|
|
|
|
|
|
|
and other params to L. |
57
|
|
|
|
|
|
|
|
58
|
|
|
|
|
|
|
=item limit I |
59
|
|
|
|
|
|
|
|
60
|
|
|
|
|
|
|
Maximum number of suggestions to return. Defaults to 10. |
61
|
|
|
|
|
|
|
|
62
|
|
|
|
|
|
|
=item use_regex 1|0 |
63
|
|
|
|
|
|
|
|
64
|
|
|
|
|
|
|
Use a simple regex when comparing terms. Defaults to false (0), |
65
|
|
|
|
|
|
|
preferring index(). If your analyzer results in terms containing |
66
|
|
|
|
|
|
|
multiple words (e.g. phrases) then B is probably what you want. |
67
|
|
|
|
|
|
|
|
68
|
|
|
|
|
|
|
=item max_length I |
69
|
|
|
|
|
|
|
|
70
|
|
|
|
|
|
|
Set a max term length beyond which suggestions are trimmed |
71
|
|
|
|
|
|
|
with substr(). Default is 64 characters. Set to 0 to disable. |
72
|
|
|
|
|
|
|
|
73
|
|
|
|
|
|
|
=back |
74
|
|
|
|
|
|
|
|
75
|
|
|
|
|
|
|
=cut |
76
|
|
|
|
|
|
|
|
77
|
|
|
|
|
|
|
sub new { |
78
|
|
|
|
|
|
|
my $class = shift; |
79
|
|
|
|
|
|
|
my %args = @_; |
80
|
|
|
|
|
|
|
my $fields = delete $args{fields} || []; |
81
|
|
|
|
|
|
|
my $indexes = delete $args{indexes} or croak "indexes required"; |
82
|
|
|
|
|
|
|
my $spellcheck = delete $args{spellcheck}; |
83
|
|
|
|
|
|
|
my $limit = delete $args{limit} || 10; |
84
|
|
|
|
|
|
|
my $debug = delete $args{debug} || $ENV{LUCYX_DEBUG} || 0; |
85
|
|
|
|
|
|
|
my $use_regex = delete $args{use_regex} || 0; |
86
|
|
|
|
|
|
|
my $max_length = delete $args{max_length}; |
87
|
|
|
|
|
|
|
$max_length = 64 unless defined $max_length; |
88
|
|
|
|
|
|
|
|
89
|
|
|
|
|
|
|
if (%args) { |
90
|
|
|
|
|
|
|
croak "Too many arguments to new(): " . dump( \%args ); |
91
|
|
|
|
|
|
|
} |
92
|
|
|
|
|
|
|
if ( ref $indexes ne 'ARRAY' ) { |
93
|
|
|
|
|
|
|
croak "indexes must be an ARRAY ref"; |
94
|
|
|
|
|
|
|
} |
95
|
|
|
|
|
|
|
return bless( |
96
|
|
|
|
|
|
|
{ fields => $fields, |
97
|
|
|
|
|
|
|
indexes => $indexes, |
98
|
|
|
|
|
|
|
spellcheck => $spellcheck, |
99
|
|
|
|
|
|
|
limit => $limit, |
100
|
|
|
|
|
|
|
debug => $debug, |
101
|
|
|
|
|
|
|
use_regex => $use_regex, |
102
|
|
|
|
|
|
|
max_length => $max_length, |
103
|
|
|
|
|
|
|
}, |
104
|
|
|
|
|
|
|
$class |
105
|
|
|
|
|
|
|
); |
106
|
|
|
|
|
|
|
} |
107
|
|
|
|
|
|
|
|
108
|
|
|
|
|
|
|
=head2 suggest( I ) |
109
|
|
|
|
|
|
|
|
110
|
|
|
|
|
|
|
Returns arrayref of terms that match I. |
111
|
|
|
|
|
|
|
|
112
|
|
|
|
|
|
|
=cut |
113
|
|
|
|
|
|
|
|
114
|
|
|
|
|
|
|
# |
115
|
|
|
|
|
|
|
# I tried re-writing this to use a PolySearcher |
116
|
|
|
|
|
|
|
# instead, in order to preserve phrases, etc |
117
|
|
|
|
|
|
|
# but it was up to 3x slower. |
118
|
|
|
|
|
|
|
# |
119
|
|
|
|
|
|
|
|
120
|
|
|
|
|
|
|
sub suggest { |
121
|
|
|
|
|
|
|
my $self = shift; |
122
|
|
|
|
|
|
|
my $query = shift; |
123
|
|
|
|
|
|
|
|
124
|
|
|
|
|
|
|
croak "query required" unless defined $query; |
125
|
|
|
|
|
|
|
|
126
|
|
|
|
|
|
|
my $debug = $self->{debug}; |
127
|
|
|
|
|
|
|
|
128
|
|
|
|
|
|
|
my $spellchecker = $self->{spellcheck}; |
129
|
|
|
|
|
|
|
if ( !$spellchecker ) { |
130
|
|
|
|
|
|
|
my $qparser = Search::Tools->parser( debug => $debug, ); |
131
|
|
|
|
|
|
|
$spellchecker = Search::Tools->spellcheck( |
132
|
|
|
|
|
|
|
debug => $debug, |
133
|
|
|
|
|
|
|
query_parser => $qparser, |
134
|
|
|
|
|
|
|
); |
135
|
|
|
|
|
|
|
} |
136
|
|
|
|
|
|
|
|
137
|
|
|
|
|
|
|
my $suggestions = $spellchecker->suggest($query); |
138
|
|
|
|
|
|
|
my %terms; |
139
|
|
|
|
|
|
|
for my $s (@$suggestions) { |
140
|
|
|
|
|
|
|
|
141
|
|
|
|
|
|
|
if ( !$s->{suggestions} ) { |
142
|
|
|
|
|
|
|
$terms{ $s->{word} }++; |
143
|
|
|
|
|
|
|
} |
144
|
|
|
|
|
|
|
else { |
145
|
|
|
|
|
|
|
for my $suggest ( @{ $s->{suggestions} } ) { |
146
|
|
|
|
|
|
|
$terms{$suggest}++; |
147
|
|
|
|
|
|
|
} |
148
|
|
|
|
|
|
|
} |
149
|
|
|
|
|
|
|
} |
150
|
|
|
|
|
|
|
|
151
|
|
|
|
|
|
|
# must analyze each query term and suggestion |
152
|
|
|
|
|
|
|
# per-field, which we cache for performance |
153
|
|
|
|
|
|
|
my %analyzed; |
154
|
|
|
|
|
|
|
|
155
|
|
|
|
|
|
|
# suggestions |
156
|
|
|
|
|
|
|
my %matches; |
157
|
|
|
|
|
|
|
|
158
|
|
|
|
|
|
|
my @my_fields = @{ $self->{fields} }; |
159
|
|
|
|
|
|
|
my $use_regex = $self->{use_regex}; |
160
|
|
|
|
|
|
|
my $maxl = $self->{max_length}; |
161
|
|
|
|
|
|
|
|
162
|
|
|
|
|
|
|
INDEX: for my $invindex ( @{ $self->{indexes} } ) { |
163
|
|
|
|
|
|
|
my $reader = Lucy::Index::IndexReader->open( index => $invindex ); |
164
|
|
|
|
|
|
|
my $schema = $reader->get_schema(); |
165
|
|
|
|
|
|
|
my $fields; |
166
|
|
|
|
|
|
|
if (@my_fields) { |
167
|
|
|
|
|
|
|
$fields = \@my_fields; |
168
|
|
|
|
|
|
|
} |
169
|
|
|
|
|
|
|
else { |
170
|
|
|
|
|
|
|
$fields = $schema->all_fields(); |
171
|
|
|
|
|
|
|
} |
172
|
|
|
|
|
|
|
my $seg_readers = $reader->seg_readers; |
173
|
|
|
|
|
|
|
SEG: for my $seg_reader (@$seg_readers) { |
174
|
|
|
|
|
|
|
my $lex_reader |
175
|
|
|
|
|
|
|
= $seg_reader->obtain('Lucy::Index::LexiconReader'); |
176
|
|
|
|
|
|
|
FIELD: for my $field (@$fields) { |
177
|
|
|
|
|
|
|
$self->_analyze_terms( $schema, $field, \%analyzed, \%terms ); |
178
|
|
|
|
|
|
|
|
179
|
|
|
|
|
|
|
# sort in order to seek() below. |
180
|
|
|
|
|
|
|
my @to_check = sort keys %{ $analyzed{$field} }; |
181
|
|
|
|
|
|
|
|
182
|
|
|
|
|
|
|
$debug and warn "$field=" . dump \@to_check; |
183
|
|
|
|
|
|
|
|
184
|
|
|
|
|
|
|
my $lexicon = $lex_reader->lexicon( field => $field ); |
185
|
|
|
|
|
|
|
next FIELD unless $lexicon; |
186
|
|
|
|
|
|
|
|
187
|
|
|
|
|
|
|
CHECK: for my $check_term (@to_check) { |
188
|
|
|
|
|
|
|
|
189
|
|
|
|
|
|
|
my $check_initial = substr( $check_term, 0, 1 ); |
190
|
|
|
|
|
|
|
|
191
|
|
|
|
|
|
|
$debug and warn " seek($check_term)"; |
192
|
|
|
|
|
|
|
$lexicon->seek($check_term); |
193
|
|
|
|
|
|
|
|
194
|
|
|
|
|
|
|
TERM: while ( defined( my $term = $lexicon->get_term ) ) { |
195
|
|
|
|
|
|
|
|
196
|
|
|
|
|
|
|
$debug and warn " $check_term -> $term"; |
197
|
|
|
|
|
|
|
|
198
|
|
|
|
|
|
|
my $initial = substr( $term, 0, 1 ); |
199
|
|
|
|
|
|
|
if ( $initial and $initial gt $check_initial ) { |
200
|
|
|
|
|
|
|
$debug |
201
|
|
|
|
|
|
|
and warn |
202
|
|
|
|
|
|
|
" reset: initial=$initial > check_initial=$check_initial"; |
203
|
|
|
|
|
|
|
next CHECK; |
204
|
|
|
|
|
|
|
} |
205
|
|
|
|
|
|
|
|
206
|
|
|
|
|
|
|
# TODO better weighting than simple freq? |
207
|
|
|
|
|
|
|
|
208
|
|
|
|
|
|
|
if ( !$use_regex |
209
|
|
|
|
|
|
|
&& index( $term, $check_term, 0 ) == 0 ) |
210
|
|
|
|
|
|
|
{ |
211
|
|
|
|
|
|
|
my $freq = $lex_reader->doc_freq( |
212
|
|
|
|
|
|
|
field => $field, |
213
|
|
|
|
|
|
|
term => $term, |
214
|
|
|
|
|
|
|
); |
215
|
|
|
|
|
|
|
$debug and warn "ok term=$term [$freq]"; |
216
|
|
|
|
|
|
|
$matches{ |
217
|
|
|
|
|
|
|
$maxl |
218
|
|
|
|
|
|
|
? substr( $term, 0, $maxl ) |
219
|
|
|
|
|
|
|
: $term |
220
|
|
|
|
|
|
|
} += $freq; |
221
|
|
|
|
|
|
|
|
222
|
|
|
|
|
|
|
# abort everything if we've hit our limit |
223
|
|
|
|
|
|
|
if ( scalar( keys %matches ) >= $self->{limit} ) { |
224
|
|
|
|
|
|
|
last INDEX; |
225
|
|
|
|
|
|
|
} |
226
|
|
|
|
|
|
|
|
227
|
|
|
|
|
|
|
} |
228
|
|
|
|
|
|
|
elsif ( $use_regex && $term =~ m/\b\Q$check_term/ ) { |
229
|
|
|
|
|
|
|
my $freq = $lex_reader->doc_freq( |
230
|
|
|
|
|
|
|
field => $field, |
231
|
|
|
|
|
|
|
term => $term, |
232
|
|
|
|
|
|
|
); |
233
|
|
|
|
|
|
|
$debug and warn "ok term=$term [$freq]"; |
234
|
|
|
|
|
|
|
$matches{ |
235
|
|
|
|
|
|
|
$maxl |
236
|
|
|
|
|
|
|
? substr( |
237
|
|
|
|
|
|
|
$term, index( $term, $check_term, 0 ), |
238
|
|
|
|
|
|
|
$maxl |
239
|
|
|
|
|
|
|
) |
240
|
|
|
|
|
|
|
: $term |
241
|
|
|
|
|
|
|
} += $freq; |
242
|
|
|
|
|
|
|
|
243
|
|
|
|
|
|
|
# abort everything if we've hit our limit |
244
|
|
|
|
|
|
|
if ( scalar( keys %matches ) >= $self->{limit} ) { |
245
|
|
|
|
|
|
|
last INDEX; |
246
|
|
|
|
|
|
|
} |
247
|
|
|
|
|
|
|
} |
248
|
|
|
|
|
|
|
else { |
249
|
|
|
|
|
|
|
$debug |
250
|
|
|
|
|
|
|
and warn |
251
|
|
|
|
|
|
|
" No match - skipping to next CHECK term"; |
252
|
|
|
|
|
|
|
next CHECK; |
253
|
|
|
|
|
|
|
} |
254
|
|
|
|
|
|
|
|
255
|
|
|
|
|
|
|
last TERM unless $lexicon->next; |
256
|
|
|
|
|
|
|
|
257
|
|
|
|
|
|
|
} |
258
|
|
|
|
|
|
|
} |
259
|
|
|
|
|
|
|
} |
260
|
|
|
|
|
|
|
} |
261
|
|
|
|
|
|
|
} |
262
|
|
|
|
|
|
|
|
263
|
|
|
|
|
|
|
# boost phrase matches |
264
|
|
|
|
|
|
|
for my $m ( keys %matches ) { |
265
|
|
|
|
|
|
|
next unless $m =~ m/ /; |
266
|
|
|
|
|
|
|
for my $m2 ( keys %matches ) { |
267
|
|
|
|
|
|
|
if ( $m =~ m/\b\Q$m2\E\b/ ) { |
268
|
|
|
|
|
|
|
$matches{$m} += $matches{$m2}; |
269
|
|
|
|
|
|
|
} |
270
|
|
|
|
|
|
|
} |
271
|
|
|
|
|
|
|
} |
272
|
|
|
|
|
|
|
|
273
|
|
|
|
|
|
|
$debug and warn "matches=" . dump( \%matches ); |
274
|
|
|
|
|
|
|
|
275
|
|
|
|
|
|
|
return [ |
276
|
|
|
|
|
|
|
sort { $matches{$b} <=> $matches{$a} || $a cmp $b } |
277
|
|
|
|
|
|
|
keys %matches |
278
|
|
|
|
|
|
|
]; |
279
|
|
|
|
|
|
|
} |
280
|
|
|
|
|
|
|
|
281
|
|
|
|
|
|
|
sub _analyze_terms { |
282
|
|
|
|
|
|
|
my ( $self, $schema, $field, $analyzed, $terms ) = @_; |
283
|
|
|
|
|
|
|
|
284
|
|
|
|
|
|
|
return if exists $analyzed->{$field}; |
285
|
|
|
|
|
|
|
|
286
|
|
|
|
|
|
|
my $analyzer = $schema->fetch_analyzer($field); |
287
|
|
|
|
|
|
|
for my $t ( keys %$terms ) { |
288
|
|
|
|
|
|
|
my $baked = $analyzer ? $analyzer->split($t)->[0] : $t; |
289
|
|
|
|
|
|
|
next if length $baked == 1; # too much noise |
290
|
|
|
|
|
|
|
$analyzed->{$field}->{$baked} = $t; |
291
|
|
|
|
|
|
|
} |
292
|
|
|
|
|
|
|
} |
293
|
|
|
|
|
|
|
|
294
|
|
|
|
|
|
|
1; |
295
|
|
|
|
|
|
|
|
296
|
|
|
|
|
|
|
__END__ |