isomorfeus-ferret 0.12.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +7 -0
- data/LICENSE +612 -0
- data/README.md +44 -0
- data/ext/isomorfeus_ferret_ext/benchmark.c +223 -0
- data/ext/isomorfeus_ferret_ext/benchmark.h +45 -0
- data/ext/isomorfeus_ferret_ext/benchmarks_all.h +25 -0
- data/ext/isomorfeus_ferret_ext/bm_bitvector.c +123 -0
- data/ext/isomorfeus_ferret_ext/bm_hash.c +118 -0
- data/ext/isomorfeus_ferret_ext/bm_micro_string.c +40 -0
- data/ext/isomorfeus_ferret_ext/bm_store.c +93 -0
- data/ext/isomorfeus_ferret_ext/email.rl +21 -0
- data/ext/isomorfeus_ferret_ext/extconf.rb +5 -0
- data/ext/isomorfeus_ferret_ext/fio_tmpfile.h +53 -0
- data/ext/isomorfeus_ferret_ext/frb_analysis.c +2577 -0
- data/ext/isomorfeus_ferret_ext/frb_index.c +3457 -0
- data/ext/isomorfeus_ferret_ext/frb_lang.c +9 -0
- data/ext/isomorfeus_ferret_ext/frb_lang.h +17 -0
- data/ext/isomorfeus_ferret_ext/frb_qparser.c +629 -0
- data/ext/isomorfeus_ferret_ext/frb_search.c +4460 -0
- data/ext/isomorfeus_ferret_ext/frb_store.c +515 -0
- data/ext/isomorfeus_ferret_ext/frb_threading.h +30 -0
- data/ext/isomorfeus_ferret_ext/frb_utils.c +1127 -0
- data/ext/isomorfeus_ferret_ext/frt_analysis.c +1644 -0
- data/ext/isomorfeus_ferret_ext/frt_analysis.h +247 -0
- data/ext/isomorfeus_ferret_ext/frt_array.c +124 -0
- data/ext/isomorfeus_ferret_ext/frt_array.h +54 -0
- data/ext/isomorfeus_ferret_ext/frt_bitvector.c +95 -0
- data/ext/isomorfeus_ferret_ext/frt_bitvector.h +586 -0
- data/ext/isomorfeus_ferret_ext/frt_compound_io.c +374 -0
- data/ext/isomorfeus_ferret_ext/frt_config.h +44 -0
- data/ext/isomorfeus_ferret_ext/frt_document.c +134 -0
- data/ext/isomorfeus_ferret_ext/frt_document.h +52 -0
- data/ext/isomorfeus_ferret_ext/frt_except.c +95 -0
- data/ext/isomorfeus_ferret_ext/frt_except.h +188 -0
- data/ext/isomorfeus_ferret_ext/frt_field_index.c +233 -0
- data/ext/isomorfeus_ferret_ext/frt_field_index.h +42 -0
- data/ext/isomorfeus_ferret_ext/frt_filter.c +157 -0
- data/ext/isomorfeus_ferret_ext/frt_fs_store.c +502 -0
- data/ext/isomorfeus_ferret_ext/frt_global.c +427 -0
- data/ext/isomorfeus_ferret_ext/frt_global.h +290 -0
- data/ext/isomorfeus_ferret_ext/frt_hash.c +518 -0
- data/ext/isomorfeus_ferret_ext/frt_hash.h +466 -0
- data/ext/isomorfeus_ferret_ext/frt_hashset.c +191 -0
- data/ext/isomorfeus_ferret_ext/frt_hashset.h +206 -0
- data/ext/isomorfeus_ferret_ext/frt_helper.c +62 -0
- data/ext/isomorfeus_ferret_ext/frt_helper.h +13 -0
- data/ext/isomorfeus_ferret_ext/frt_ind.c +353 -0
- data/ext/isomorfeus_ferret_ext/frt_ind.h +54 -0
- data/ext/isomorfeus_ferret_ext/frt_index.c +6377 -0
- data/ext/isomorfeus_ferret_ext/frt_index.h +880 -0
- data/ext/isomorfeus_ferret_ext/frt_lang.c +104 -0
- data/ext/isomorfeus_ferret_ext/frt_lang.h +44 -0
- data/ext/isomorfeus_ferret_ext/frt_mempool.c +87 -0
- data/ext/isomorfeus_ferret_ext/frt_mempool.h +33 -0
- data/ext/isomorfeus_ferret_ext/frt_multimapper.c +349 -0
- data/ext/isomorfeus_ferret_ext/frt_multimapper.h +52 -0
- data/ext/isomorfeus_ferret_ext/frt_posh.c +1006 -0
- data/ext/isomorfeus_ferret_ext/frt_posh.h +973 -0
- data/ext/isomorfeus_ferret_ext/frt_priorityqueue.c +147 -0
- data/ext/isomorfeus_ferret_ext/frt_priorityqueue.h +147 -0
- data/ext/isomorfeus_ferret_ext/frt_q_boolean.c +1612 -0
- data/ext/isomorfeus_ferret_ext/frt_q_const_score.c +157 -0
- data/ext/isomorfeus_ferret_ext/frt_q_filtered_query.c +209 -0
- data/ext/isomorfeus_ferret_ext/frt_q_fuzzy.c +281 -0
- data/ext/isomorfeus_ferret_ext/frt_q_match_all.c +147 -0
- data/ext/isomorfeus_ferret_ext/frt_q_multi_term.c +672 -0
- data/ext/isomorfeus_ferret_ext/frt_q_parser.c +3084 -0
- data/ext/isomorfeus_ferret_ext/frt_q_phrase.c +1182 -0
- data/ext/isomorfeus_ferret_ext/frt_q_prefix.c +98 -0
- data/ext/isomorfeus_ferret_ext/frt_q_range.c +665 -0
- data/ext/isomorfeus_ferret_ext/frt_q_span.c +2386 -0
- data/ext/isomorfeus_ferret_ext/frt_q_term.c +311 -0
- data/ext/isomorfeus_ferret_ext/frt_q_wildcard.c +166 -0
- data/ext/isomorfeus_ferret_ext/frt_ram_store.c +460 -0
- data/ext/isomorfeus_ferret_ext/frt_scanner.c +899 -0
- data/ext/isomorfeus_ferret_ext/frt_scanner.h +28 -0
- data/ext/isomorfeus_ferret_ext/frt_scanner_mb.c +6705 -0
- data/ext/isomorfeus_ferret_ext/frt_scanner_utf8.c +4419 -0
- data/ext/isomorfeus_ferret_ext/frt_search.c +1824 -0
- data/ext/isomorfeus_ferret_ext/frt_search.h +924 -0
- data/ext/isomorfeus_ferret_ext/frt_similarity.c +150 -0
- data/ext/isomorfeus_ferret_ext/frt_similarity.h +79 -0
- data/ext/isomorfeus_ferret_ext/frt_sort.c +796 -0
- data/ext/isomorfeus_ferret_ext/frt_stopwords.c +395 -0
- data/ext/isomorfeus_ferret_ext/frt_store.c +680 -0
- data/ext/isomorfeus_ferret_ext/frt_store.h +789 -0
- data/ext/isomorfeus_ferret_ext/frt_term_vectors.c +72 -0
- data/ext/isomorfeus_ferret_ext/frt_threading.h +23 -0
- data/ext/isomorfeus_ferret_ext/frt_win32.h +54 -0
- data/ext/isomorfeus_ferret_ext/isomorfeus_ferret.c +409 -0
- data/ext/isomorfeus_ferret_ext/isomorfeus_ferret.h +95 -0
- data/ext/isomorfeus_ferret_ext/libstemmer.c +93 -0
- data/ext/isomorfeus_ferret_ext/libstemmer.h +73 -0
- data/ext/isomorfeus_ferret_ext/q_parser.y +1366 -0
- data/ext/isomorfeus_ferret_ext/scanner.h +28 -0
- data/ext/isomorfeus_ferret_ext/scanner.in +43 -0
- data/ext/isomorfeus_ferret_ext/scanner.rl +84 -0
- data/ext/isomorfeus_ferret_ext/scanner_mb.rl +200 -0
- data/ext/isomorfeus_ferret_ext/scanner_utf8.rl +85 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_danish.c +324 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_danish.h +7 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_dutch.c +610 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_dutch.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_english.c +1104 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_english.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_finnish.c +749 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_finnish.h +7 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_french.c +1233 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_french.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_german.c +490 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_german.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_hungarian.c +1217 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_hungarian.h +7 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_italian.c +1052 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_italian.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_norwegian.c +283 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_norwegian.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_porter.c +735 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_porter.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_portuguese.c +1003 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_portuguese.h +7 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_spanish.c +1079 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_spanish.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_swedish.c +293 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_1_swedish.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_2_romanian.c +984 -0
- data/ext/isomorfeus_ferret_ext/stem_ISO_8859_2_romanian.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_KOI8_R_russian.c +686 -0
- data/ext/isomorfeus_ferret_ext/stem_KOI8_R_russian.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_danish.c +325 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_danish.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_dutch.c +620 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_dutch.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_english.c +1111 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_english.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_finnish.c +754 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_finnish.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_french.c +1242 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_french.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_german.c +495 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_german.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_hungarian.c +1220 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_hungarian.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_italian.c +1059 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_italian.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_norwegian.c +285 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_norwegian.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_porter.c +741 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_porter.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_portuguese.c +1009 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_portuguese.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_romanian.c +990 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_romanian.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_russian.c +680 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_russian.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_spanish.c +1083 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_spanish.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_swedish.c +294 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_swedish.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_turkish.c +2191 -0
- data/ext/isomorfeus_ferret_ext/stem_UTF_8_turkish.h +6 -0
- data/ext/isomorfeus_ferret_ext/stem_api.c +66 -0
- data/ext/isomorfeus_ferret_ext/stem_api.h +26 -0
- data/ext/isomorfeus_ferret_ext/stem_header.h +57 -0
- data/ext/isomorfeus_ferret_ext/stem_modules.h +190 -0
- data/ext/isomorfeus_ferret_ext/stem_modules.txt +50 -0
- data/ext/isomorfeus_ferret_ext/stem_utilities.c +478 -0
- data/ext/isomorfeus_ferret_ext/test.c +850 -0
- data/ext/isomorfeus_ferret_ext/test.h +416 -0
- data/ext/isomorfeus_ferret_ext/test_1710.c +63 -0
- data/ext/isomorfeus_ferret_ext/test_analysis.c +1221 -0
- data/ext/isomorfeus_ferret_ext/test_array.c +272 -0
- data/ext/isomorfeus_ferret_ext/test_bitvector.c +600 -0
- data/ext/isomorfeus_ferret_ext/test_compound_io.c +170 -0
- data/ext/isomorfeus_ferret_ext/test_document.c +156 -0
- data/ext/isomorfeus_ferret_ext/test_except.c +244 -0
- data/ext/isomorfeus_ferret_ext/test_fields.c +522 -0
- data/ext/isomorfeus_ferret_ext/test_file_deleter.c +185 -0
- data/ext/isomorfeus_ferret_ext/test_filter.c +331 -0
- data/ext/isomorfeus_ferret_ext/test_fs_store.c +25 -0
- data/ext/isomorfeus_ferret_ext/test_global.c +299 -0
- data/ext/isomorfeus_ferret_ext/test_hash.c +485 -0
- data/ext/isomorfeus_ferret_ext/test_hashset.c +288 -0
- data/ext/isomorfeus_ferret_ext/test_helper.c +47 -0
- data/ext/isomorfeus_ferret_ext/test_highlighter.c +548 -0
- data/ext/isomorfeus_ferret_ext/test_index.c +2323 -0
- data/ext/isomorfeus_ferret_ext/test_lang.c +74 -0
- data/ext/isomorfeus_ferret_ext/test_mempool.c +102 -0
- data/ext/isomorfeus_ferret_ext/test_multimapper.c +64 -0
- data/ext/isomorfeus_ferret_ext/test_priorityqueue.c +213 -0
- data/ext/isomorfeus_ferret_ext/test_q_const_score.c +84 -0
- data/ext/isomorfeus_ferret_ext/test_q_filtered.c +61 -0
- data/ext/isomorfeus_ferret_ext/test_q_fuzzy.c +241 -0
- data/ext/isomorfeus_ferret_ext/test_q_parser.c +464 -0
- data/ext/isomorfeus_ferret_ext/test_q_span.c +575 -0
- data/ext/isomorfeus_ferret_ext/test_ram_store.c +77 -0
- data/ext/isomorfeus_ferret_ext/test_search.c +1874 -0
- data/ext/isomorfeus_ferret_ext/test_segments.c +167 -0
- data/ext/isomorfeus_ferret_ext/test_similarity.c +25 -0
- data/ext/isomorfeus_ferret_ext/test_sort.c +333 -0
- data/ext/isomorfeus_ferret_ext/test_store.c +591 -0
- data/ext/isomorfeus_ferret_ext/test_store.h +3 -0
- data/ext/isomorfeus_ferret_ext/test_term.c +351 -0
- data/ext/isomorfeus_ferret_ext/test_term_vectors.c +373 -0
- data/ext/isomorfeus_ferret_ext/test_test.c +83 -0
- data/ext/isomorfeus_ferret_ext/test_threading.c +188 -0
- data/ext/isomorfeus_ferret_ext/testhelper.c +561 -0
- data/ext/isomorfeus_ferret_ext/testhelper.h +25 -0
- data/ext/isomorfeus_ferret_ext/tests_all.h +87 -0
- data/ext/isomorfeus_ferret_ext/uchar-ucs4.rl +1854 -0
- data/ext/isomorfeus_ferret_ext/uchar-utf8.rl +1999 -0
- data/ext/isomorfeus_ferret_ext/url.rl +27 -0
- data/ext/isomorfeus_ferret_ext/word_list.h +15156 -0
- data/lib/isomorfeus/ferret/document.rb +132 -0
- data/lib/isomorfeus/ferret/field_symbol.rb +85 -0
- data/lib/isomorfeus/ferret/index/field_infos.rb +48 -0
- data/lib/isomorfeus/ferret/index/index.rb +970 -0
- data/lib/isomorfeus/ferret/monitor.rb +323 -0
- data/lib/isomorfeus/ferret/stdlib_patches.rb +151 -0
- data/lib/isomorfeus/ferret/version.rb +5 -0
- data/lib/isomorfeus-ferret.rb +8 -0
- metadata +307 -0
@@ -0,0 +1,1366 @@
|
|
1
|
+
/*****************************************************************************
|
2
|
+
* QueryParser
|
3
|
+
* ===========
|
4
|
+
*
|
5
|
+
* Brief Overview
|
6
|
+
* --------------
|
7
|
+
*
|
8
|
+
* === Creating a QueryParser
|
9
|
+
*
|
10
|
+
* +qp_new+ allocates a new QueryParser and assigns three very important
|
11
|
+
* HashSets; +qp->def_fields+, +qp->tkz_fields+ and +qp->all_fields+. The
|
12
|
+
* query language allows you to assign a field or a set of fields to each
|
13
|
+
* part of the query.
|
14
|
+
*
|
15
|
+
* - +qp->def_fields+ is the set of fields that a query is applied to by
|
16
|
+
* default when no fields are specified.
|
17
|
+
* - +qp->all_fields+ is the set of fields that gets searched when the user
|
18
|
+
* requests a search of all fields.
|
19
|
+
* - +qp->tkz_fields+ is the set of fields that gets tokenized before being
|
20
|
+
* added to the query parser.
|
21
|
+
*
|
22
|
+
* === qp_parse
|
23
|
+
*
|
24
|
+
* The main QueryParser method is +qp_parse+. It gets called with a the query
|
25
|
+
* string and returns a Query object which can then be passed to the
|
26
|
+
* IndexSearcher. The first thing it does is to clean the query string if
|
27
|
+
* +qp->clean_str+ is set to true. The cleaning is done with the
|
28
|
+
* +qp_clean_str+.
|
29
|
+
*
|
30
|
+
* It then calls the yacc parser which will set +qp->result+ to the parsed
|
31
|
+
* query. If parsing fails in any way, +qp->result+ should be set to NULL, in
|
32
|
+
* which case qp_parse does one of two things depending on the value of
|
33
|
+
* +qp->handle_parse_errors+;
|
34
|
+
*
|
35
|
+
* - If it is set to true, qp_parse attempts to do a very basic parsing of
|
36
|
+
* the query by ignoring all special characters and parsing the query as
|
37
|
+
* a plain boolean query.
|
38
|
+
* - If it is set to false, qp_parse will raise a PARSE_ERROR and hopefully
|
39
|
+
* free all allocated memory.
|
40
|
+
*
|
41
|
+
* === The Lexer
|
42
|
+
*
|
43
|
+
* +yylex+ is the lexing method called by the QueryParser. It breaks the
|
44
|
+
* query up into special characters;
|
45
|
+
*
|
46
|
+
* ( "&:()[]{}!\"~^|<>=*?+-" )
|
47
|
+
*
|
48
|
+
* and tokens;
|
49
|
+
*
|
50
|
+
* - QWRD
|
51
|
+
* - WILD_STR
|
52
|
+
* - AND['AND', '&&']
|
53
|
+
* - OR['OR', '||']
|
54
|
+
* - REQ['REQ', '+']
|
55
|
+
* - NOT['NOT', '-', '~']
|
56
|
+
*
|
57
|
+
* QWRD tokens are query word tokens which are made up of characters other
|
58
|
+
* than the special characters. They can also contain special characters when
|
59
|
+
* escaped with a backslash '\'. WILD_STR is the same as QWRD except that it
|
60
|
+
* may also contain '?' and '*' characters.
|
61
|
+
*
|
62
|
+
* === The Parser
|
63
|
+
*
|
64
|
+
* For a better understanding of the how the query parser works, it is a good
|
65
|
+
* idea to study the Ferret Query Language (FQL) described below. Once you
|
66
|
+
* understand FQL the one tricky part that needs to be mentioned is how
|
67
|
+
* fields are handled. This is where +qp->def_fields+ and +qp->all_fields
|
68
|
+
* come into play. When no fields are specified then the default fields are
|
69
|
+
* used. The '*:' field specifier will search all fields contained in the
|
70
|
+
* all_fields set. Otherwise all fields specified in the field descripter
|
71
|
+
* separated by '|' will be searched. For example 'title|content:' will
|
72
|
+
* search the title and content fields. When fields are specified like this,
|
73
|
+
* the parser will push the fields onto a stack and all queries modified by
|
74
|
+
* the field specifier will be applied to the fields on top of the stack.
|
75
|
+
* The parser uses the FLDS macro to handle the current fields. It takes the
|
76
|
+
* current query building function in the parser and calls it for all the
|
77
|
+
* current search fields (on top of the stack).
|
78
|
+
*
|
79
|
+
* Ferret Query Language (FQL)
|
80
|
+
* ---------------------------
|
81
|
+
*
|
82
|
+
* FIXME to be continued...
|
83
|
+
*****************************************************************************/
|
84
|
+
%{
|
85
|
+
#include <string.h>
|
86
|
+
#include <ctype.h>
|
87
|
+
#include <wctype.h>
|
88
|
+
#include <assert.h>
|
89
|
+
#include "frt_global.h"
|
90
|
+
#include "frt_except.h"
|
91
|
+
#include "frt_search.h"
|
92
|
+
#include "frt_array.h"
|
93
|
+
#include "frt_internal.h"
|
94
|
+
|
95
|
+
typedef struct Phrase {
|
96
|
+
int size;
|
97
|
+
int capa;
|
98
|
+
int pos_inc;
|
99
|
+
FrtPhrasePosition *positions;
|
100
|
+
} Phrase;
|
101
|
+
|
102
|
+
#define BCA_INIT_CAPA 4
|
103
|
+
typedef struct BCArray {
|
104
|
+
int size;
|
105
|
+
int capa;
|
106
|
+
BooleanClause **clauses;
|
107
|
+
} BCArray;
|
108
|
+
|
109
|
+
float qp_default_fuzzy_min_sim = 0.5;
|
110
|
+
int qp_default_fuzzy_pre_len = 0;
|
111
|
+
|
112
|
+
%}
|
113
|
+
%union {
|
114
|
+
FrtQuery *query;
|
115
|
+
BooleanClause *bcls;
|
116
|
+
BCArray *bclss;
|
117
|
+
FrtHashSet *hashset;
|
118
|
+
Phrase *phrase;
|
119
|
+
char *str;
|
120
|
+
}
|
121
|
+
%{
|
122
|
+
static int yylex(YYSTYPE *lvalp, FrtQParser *qp);
|
123
|
+
static int yyerror(QParser *qp, char const *msg);
|
124
|
+
|
125
|
+
#define PHRASE_INIT_CAPA 4
|
126
|
+
static FrtQuery *get_bool_q(BCArray *bca);
|
127
|
+
|
128
|
+
static BCArray *first_cls(BooleanClause *boolean_clause);
|
129
|
+
static BCArray *add_and_cls(BCArray *bca, BooleanClause *clause);
|
130
|
+
static BCArray *add_or_cls(BCArray *bca, BooleanClause *clause);
|
131
|
+
static BCArray *add_default_cls(QParser *qp, BCArray *bca,
|
132
|
+
BooleanClause *clause);
|
133
|
+
static void bca_destroy(BCArray *bca);
|
134
|
+
|
135
|
+
static BooleanClause *get_bool_cls(FrtQuery *q, BCType occur);
|
136
|
+
|
137
|
+
static FrtQuery *get_term_q(QParser *qp, FrtSymbol field, char *word);
|
138
|
+
static FrtQuery *get_fuzzy_q(QParser *qp, FrtSymbol field, char *word,
|
139
|
+
char *slop);
|
140
|
+
static FrtQuery *get_wild_q(QParser *qp, FrtSymbol field, char *pattern);
|
141
|
+
|
142
|
+
static FrtHashSet *first_field(QParser *qp, const char *field);
|
143
|
+
static FrtHashSet *add_field(QParser *qp, const char *field);
|
144
|
+
|
145
|
+
static FrtQuery *get_phrase_q(QParser *qp, Phrase *phrase, char *slop);
|
146
|
+
|
147
|
+
static Phrase *ph_first_word(char *word);
|
148
|
+
static Phrase *ph_add_word(Phrase *self, char *word);
|
149
|
+
static Phrase *ph_add_multi_word(Phrase *self, char *word);
|
150
|
+
static void ph_destroy(Phrase *self);
|
151
|
+
|
152
|
+
static FrtQuery *get_r_q(QParser *qp, FrtSymbol field, char *from, char *to,
|
153
|
+
bool inc_lower, bool inc_upper);
|
154
|
+
|
155
|
+
static void qp_push_fields(QParser *self, FrtHashSet *fields, bool destroy);
|
156
|
+
static void qp_pop_fields(QParser *self);
|
157
|
+
|
158
|
+
/**
|
159
|
+
* +FLDS+ calls +func+ for all fields on top of the field stack. +func+
|
160
|
+
* must return a query. If there is more than one field on top of FieldStack
|
161
|
+
* then +FLDS+ will combing all the queries returned by +func+ into a single
|
162
|
+
* BooleanQuery which it than assigns to +q+. If there is only one field, the
|
163
|
+
* return value of +func+ is assigned to +q+ directly.
|
164
|
+
*/
|
165
|
+
#define FLDS(q, func) do {\
|
166
|
+
FRT_TRY {\
|
167
|
+
FrtSymbol field;\
|
168
|
+
if (qp->fields->size == 0) {\
|
169
|
+
q = NULL;\
|
170
|
+
} else if (qp->fields->size == 1) {\
|
171
|
+
field = (Symbol)qp->fields->first->elem;\
|
172
|
+
q = func;\
|
173
|
+
} else {\
|
174
|
+
FrtQuery *volatile sq; FrtHashSetEntry *volatile hse;\
|
175
|
+
q = bq_new_max(false, qp->max_clauses);\
|
176
|
+
for (hse = qp->fields->first; hse; hse = hse->next) {\
|
177
|
+
field = (Symbol)hse->elem;\
|
178
|
+
sq = func;\
|
179
|
+
FRT_TRY\
|
180
|
+
if (sq) frt_bq_add_query_nr(q, sq, FRT_BC_SHOULD);\
|
181
|
+
FRT_XCATCHALL\
|
182
|
+
if (sq) frt_q_deref(sq);\
|
183
|
+
FRT_XENDTRY\
|
184
|
+
}\
|
185
|
+
if (((FrtBooleanQuery *)q)->clause_cnt == 0) {\
|
186
|
+
frt_q_deref(q);\
|
187
|
+
q = NULL;\
|
188
|
+
}\
|
189
|
+
}\
|
190
|
+
} FRT_XCATCHALL\
|
191
|
+
qp->destruct = true;\
|
192
|
+
FRT_HANDLED();\
|
193
|
+
FRT_XENDTRY\
|
194
|
+
if (qp->destruct && !qp->recovering && q) {q_deref(q); q = NULL;}\
|
195
|
+
} while (0)
|
196
|
+
|
197
|
+
#define Y if (qp->destruct) goto yyerrorlab;
|
198
|
+
#define T FRT_TRY
|
199
|
+
#define E\
|
200
|
+
FRT_XCATCHALL\
|
201
|
+
qp->destruct = true;\
|
202
|
+
FRT_HANDLED();\
|
203
|
+
FRT_XENDTRY\
|
204
|
+
if (qp->destruct) Y;
|
205
|
+
%}
|
206
|
+
%expect 1
|
207
|
+
%pure-parser
|
208
|
+
%parse-param { FrtQParser *qp }
|
209
|
+
%lex-param { FrtQParser *qp }
|
210
|
+
%token <str> QWRD WILD_STR
|
211
|
+
%type <query> q bool_q boosted_q term_q wild_q field_q phrase_q range_q
|
212
|
+
%type <bcls> bool_cls
|
213
|
+
%type <bclss> bool_clss
|
214
|
+
%type <hashset> field
|
215
|
+
%type <phrase> ph_words
|
216
|
+
%nonassoc LOW
|
217
|
+
%left AND OR
|
218
|
+
%nonassoc REQ NOT
|
219
|
+
%left ':'
|
220
|
+
%nonassoc HIGH
|
221
|
+
%destructor { if ($$ && qp->destruct) frt_q_deref($$); } q bool_q boosted_q term_q wild_q field_q phrase_q range_q
|
222
|
+
%destructor { if ($$ && qp->destruct) bc_deref($$); } bool_cls
|
223
|
+
%destructor { if ($$ && qp->destruct) bca_destroy($$); } bool_clss
|
224
|
+
%destructor { if ($$ && qp->destruct) ph_destroy($$); } ph_words
|
225
|
+
%%
|
226
|
+
bool_q : /* Nothing */ { qp->result = $$ = NULL; }
|
227
|
+
| bool_clss { T qp->result = $$ = get_bool_q($1); E }
|
228
|
+
;
|
229
|
+
bool_clss : bool_cls { T $$ = first_cls($1); E }
|
230
|
+
| bool_clss AND bool_cls { T $$ = add_and_cls($1, $3); E }
|
231
|
+
| bool_clss OR bool_cls { T $$ = add_or_cls($1, $3); E }
|
232
|
+
| bool_clss bool_cls { T $$ = add_default_cls(qp, $1, $2); E }
|
233
|
+
;
|
234
|
+
bool_cls : REQ boosted_q { T $$ = get_bool_cls($2, FRT_BC_MUST); E }
|
235
|
+
| NOT boosted_q { T $$ = get_bool_cls($2, FRT_BC_MUST_NOT); E }
|
236
|
+
| boosted_q { T $$ = get_bool_cls($1, FRT_BC_SHOULD); E }
|
237
|
+
;
|
238
|
+
boosted_q : q
|
239
|
+
| q '^' QWRD { T if ($1) sscanf($3,"%f",&($1->boost)); $$=$1; E }
|
240
|
+
;
|
241
|
+
q : term_q
|
242
|
+
| '(' ')' { T $$ = bq_new_max(true, qp->max_clauses); E }
|
243
|
+
| '(' bool_clss ')' { T $$ = get_bool_q($2); E }
|
244
|
+
| field_q
|
245
|
+
| phrase_q
|
246
|
+
| range_q
|
247
|
+
| wild_q
|
248
|
+
;
|
249
|
+
term_q : QWRD { FLDS($$, get_term_q(qp, field, $1)); Y}
|
250
|
+
| QWRD '~' QWRD %prec HIGH { FLDS($$, get_fuzzy_q(qp, field, $1, $3)); Y}
|
251
|
+
| QWRD '~' %prec LOW { FLDS($$, get_fuzzy_q(qp, field, $1, NULL)); Y}
|
252
|
+
;
|
253
|
+
wild_q : WILD_STR { FLDS($$, get_wild_q(qp, field, $1)); Y}
|
254
|
+
;
|
255
|
+
field_q : field ':' q { qp_pop_fields(qp); }
|
256
|
+
{ $$ = $3; }
|
257
|
+
| '*' { qp_push_fields(qp, qp->all_fields, false); } ':' q { qp_pop_fields(qp); }
|
258
|
+
{ $$ = $4; }
|
259
|
+
;
|
260
|
+
field : QWRD { $$ = first_field(qp, $1); }
|
261
|
+
| field '|' QWRD { $$ = add_field(qp, $3);}
|
262
|
+
;
|
263
|
+
phrase_q : '"' ph_words '"' { $$ = get_phrase_q(qp, $2, NULL); }
|
264
|
+
| '"' ph_words '"' '~' QWRD { $$ = get_phrase_q(qp, $2, $5); }
|
265
|
+
| '"' '"' { $$ = NULL; }
|
266
|
+
| '"' '"' '~' QWRD { $$ = NULL; (void)$4;}
|
267
|
+
;
|
268
|
+
ph_words : QWRD { $$ = ph_first_word($1); }
|
269
|
+
| '<' '>' { $$ = ph_first_word(NULL); }
|
270
|
+
| ph_words QWRD { $$ = ph_add_word($1, $2); }
|
271
|
+
| ph_words '<' '>' { $$ = ph_add_word($1, NULL); }
|
272
|
+
| ph_words '|' QWRD { $$ = ph_add_multi_word($1, $3); }
|
273
|
+
;
|
274
|
+
range_q : '[' QWRD QWRD ']' { FLDS($$, get_r_q(qp, field, $2, $3, true, true)); Y}
|
275
|
+
| '[' QWRD QWRD '}' { FLDS($$, get_r_q(qp, field, $2, $3, true, false)); Y}
|
276
|
+
| '{' QWRD QWRD ']' { FLDS($$, get_r_q(qp, field, $2, $3, false, true)); Y}
|
277
|
+
| '{' QWRD QWRD '}' { FLDS($$, get_r_q(qp, field, $2, $3, false, false)); Y}
|
278
|
+
| '<' QWRD '}' { FLDS($$, get_r_q(qp, field, NULL,$2, false, false)); Y}
|
279
|
+
| '<' QWRD ']' { FLDS($$, get_r_q(qp, field, NULL,$2, false, true)); Y}
|
280
|
+
| '[' QWRD '>' { FLDS($$, get_r_q(qp, field, $2, NULL,true, false)); Y}
|
281
|
+
| '{' QWRD '>' { FLDS($$, get_r_q(qp, field, $2, NULL,false, false)); Y}
|
282
|
+
| '<' QWRD { FLDS($$, get_r_q(qp, field, NULL,$2, false, false)); Y}
|
283
|
+
| '<' '=' QWRD { FLDS($$, get_r_q(qp, field, NULL,$3, false, true)); Y}
|
284
|
+
| '>' '=' QWRD { FLDS($$, get_r_q(qp, field, $3, NULL,true, false)); Y}
|
285
|
+
| '>' QWRD { FLDS($$, get_r_q(qp, field, $2, NULL,false, false)); Y}
|
286
|
+
;
|
287
|
+
%%
|
288
|
+
|
289
|
+
static const char *special_char = "&:()[]{}!\"~^|<>=*?+-";
|
290
|
+
static const char *not_word = " \t()[]{}!\"~^|<>=";
|
291
|
+
|
292
|
+
/**
|
293
|
+
* +get_word+ gets the next query-word from the query string. A query-word is
|
294
|
+
* basically a string of non-special or escaped special characters. It is
|
295
|
+
* FrtAnalyzer agnostic. It is up to the get_*_q methods to tokenize the word and
|
296
|
+
* turn it into a +Query+. See the documentation for each get_*_q method to
|
297
|
+
* see how it handles tokenization.
|
298
|
+
*
|
299
|
+
* Note that +get_word+ is also responsible for returning field names and
|
300
|
+
* matching the special tokens 'AND', 'NOT', 'REQ' and 'OR'.
|
301
|
+
*/
|
302
|
+
static int get_word(YYSTYPE *lvalp, FrtQParser *qp)
|
303
|
+
{
|
304
|
+
bool is_wild = false;
|
305
|
+
int len;
|
306
|
+
char c;
|
307
|
+
char *buf = qp->buf[qp->buf_index];
|
308
|
+
char *bufp = buf;
|
309
|
+
qp->buf_index = (qp->buf_index + 1) % QP_CONC_WORDS;
|
310
|
+
|
311
|
+
if (qp->dynbuf) {
|
312
|
+
free(qp->dynbuf);
|
313
|
+
qp->dynbuf = NULL;
|
314
|
+
}
|
315
|
+
|
316
|
+
qp->qstrp--; /* need to back up one character */
|
317
|
+
|
318
|
+
while (!strchr(not_word, (c = *qp->qstrp++))) {
|
319
|
+
switch (c) {
|
320
|
+
case '\\':
|
321
|
+
if ((c = *qp->qstrp) == '\0') {
|
322
|
+
*bufp++ = '\\';
|
323
|
+
}
|
324
|
+
else {
|
325
|
+
*bufp++ = c;
|
326
|
+
qp->qstrp++;
|
327
|
+
}
|
328
|
+
break;
|
329
|
+
case ':':
|
330
|
+
if ((*qp->qstrp) == ':') {
|
331
|
+
qp->qstrp++;
|
332
|
+
*bufp++ = ':';
|
333
|
+
*bufp++ = ':';
|
334
|
+
}
|
335
|
+
else {
|
336
|
+
goto get_word_done;
|
337
|
+
}
|
338
|
+
break;
|
339
|
+
case '*': case '?':
|
340
|
+
is_wild = true;
|
341
|
+
/* fall through */
|
342
|
+
default:
|
343
|
+
*bufp++ = c;
|
344
|
+
}
|
345
|
+
/* we've exceeded the static buffer. switch to the dynamic one. The
|
346
|
+
* dynamic buffer is allocated enough space to hold the whole query
|
347
|
+
* string so it's capacity doesn't need to be checked again once
|
348
|
+
* allocated. */
|
349
|
+
if (!qp->dynbuf && ((bufp - buf) == MAX_WORD_SIZE)) {
|
350
|
+
qp->dynbuf = FRT_ALLOC_AND_ZERO_N(char, strlen(qp->qstr) + 1);
|
351
|
+
strncpy(qp->dynbuf, buf, MAX_WORD_SIZE);
|
352
|
+
buf = qp->dynbuf;
|
353
|
+
bufp = buf + MAX_WORD_SIZE;
|
354
|
+
}
|
355
|
+
}
|
356
|
+
get_word_done:
|
357
|
+
qp->qstrp--;
|
358
|
+
/* check for keywords. There are only four so we have a bit of a hack
|
359
|
+
* which just checks for all of them. */
|
360
|
+
*bufp = '\0';
|
361
|
+
len = (int)(bufp - buf);
|
362
|
+
if (qp->use_keywords) {
|
363
|
+
if (len == 3) {
|
364
|
+
if (buf[0] == 'A' && buf[1] == 'N' && buf[2] == 'D') return AND;
|
365
|
+
if (buf[0] == 'N' && buf[1] == 'O' && buf[2] == 'T') return NOT;
|
366
|
+
if (buf[0] == 'R' && buf[1] == 'E' && buf[2] == 'Q') return REQ;
|
367
|
+
}
|
368
|
+
if (len == 2 && buf[0] == 'O' && buf[1] == 'R') return OR;
|
369
|
+
}
|
370
|
+
|
371
|
+
/* found a word so return it. */
|
372
|
+
lvalp->str = buf;
|
373
|
+
if (is_wild) {
|
374
|
+
return WILD_STR;
|
375
|
+
}
|
376
|
+
return QWRD;
|
377
|
+
}
|
378
|
+
|
379
|
+
/**
|
380
|
+
* +yylex+ is the lexing method called by the QueryParser. It breaks the
|
381
|
+
* query up into special characters;
|
382
|
+
*
|
383
|
+
* ( "&:()[]{}!\"~^|<>=*?+-" )
|
384
|
+
*
|
385
|
+
* and tokens;
|
386
|
+
*
|
387
|
+
* - QWRD
|
388
|
+
* - WILD_STR
|
389
|
+
* - AND['AND', '&&']
|
390
|
+
* - OR['OR', '||']
|
391
|
+
* - REQ['REQ', '+']
|
392
|
+
* - NOT['NOT', '-', '~']
|
393
|
+
*
|
394
|
+
* QWRD tokens are query word tokens which are made up of characters other
|
395
|
+
* than the special characters. They can also contain special characters when
|
396
|
+
* escaped with a backslash '\'. WILD_STR is the same as QWRD except that it
|
397
|
+
* may also contain '?' and '*' characters.
|
398
|
+
*
|
399
|
+
* If any of the special chars are seen they will usually be returned straight
|
400
|
+
* away. The exceptions are the wild chars '*' and '?', and '&' which will be
|
401
|
+
* treated as a plain old word character unless followed by another '&'.
|
402
|
+
*
|
403
|
+
* If no special characters or tokens are found then yylex delegates to
|
404
|
+
* +get_word+ which will fetch the next query-word.
|
405
|
+
*/
|
406
|
+
static int yylex(YYSTYPE *lvalp, FrtQParser *qp)
|
407
|
+
{
|
408
|
+
char c, nc;
|
409
|
+
|
410
|
+
while ((c=*qp->qstrp++) == ' ' || c == '\t') {
|
411
|
+
}
|
412
|
+
|
413
|
+
if (c == '\0') return 0;
|
414
|
+
|
415
|
+
if (strchr(special_char, c)) { /* comment */
|
416
|
+
nc = *qp->qstrp;
|
417
|
+
switch (c) {
|
418
|
+
case '-': case '!': return NOT;
|
419
|
+
case '+': return REQ;
|
420
|
+
case '*':
|
421
|
+
if (nc == ':') return c;
|
422
|
+
break;
|
423
|
+
case '?':
|
424
|
+
break;
|
425
|
+
case '&':
|
426
|
+
if (nc == '&') {
|
427
|
+
qp->qstrp++;
|
428
|
+
return AND;
|
429
|
+
}
|
430
|
+
break; /* Don't return single & character. Use in word. */
|
431
|
+
case '|':
|
432
|
+
if (nc == '|') {
|
433
|
+
qp->qstrp++;
|
434
|
+
return OR;
|
435
|
+
}
|
436
|
+
default:
|
437
|
+
return c;
|
438
|
+
}
|
439
|
+
}
|
440
|
+
|
441
|
+
return get_word(lvalp, qp);
|
442
|
+
}
|
443
|
+
|
444
|
+
/**
|
445
|
+
* yyerror gets called if there is an parse error with the yacc parser.
|
446
|
+
* It is responsible for clearing any memory that was allocated during the
|
447
|
+
* parsing process.
|
448
|
+
*/
|
449
|
+
static int yyerror(QParser *qp, char const *msg)
|
450
|
+
{
|
451
|
+
qp->destruct = true;
|
452
|
+
if (!qp->handle_parse_errors) {
|
453
|
+
char buf[1024];
|
454
|
+
buf[1023] = '\0';
|
455
|
+
strncpy(buf, qp->qstr, 1023);
|
456
|
+
if (qp->clean_str) {
|
457
|
+
free(qp->qstr);
|
458
|
+
}
|
459
|
+
frt_mutex_unlock(&qp->mutex);
|
460
|
+
snprintf(xmsg_buffer, XMSG_BUFFER_SIZE,
|
461
|
+
"couldn't parse query ``%s''. Error message "
|
462
|
+
" was %s", buf, (char *)msg);
|
463
|
+
}
|
464
|
+
while (qp->fields_top->next != NULL) {
|
465
|
+
qp_pop_fields(qp);
|
466
|
+
}
|
467
|
+
return 0;
|
468
|
+
}
|
469
|
+
|
470
|
+
#define BQ(query) ((FrtBooleanQuery *)(query))
|
471
|
+
|
472
|
+
/**
|
473
|
+
* The QueryParser caches a tokenizer for each field so that it doesn't need
|
474
|
+
* to allocate a new tokenizer for each term in the query. This would be quite
|
475
|
+
* expensive as tokenizers use quite a large hunk of memory.
|
476
|
+
*
|
477
|
+
* This method returns the query parser for a particular field and sets it up
|
478
|
+
* with the text to be tokenized.
|
479
|
+
*/
|
480
|
+
static FrtTokenStream *get_cached_ts(QParser *qp, FrtSymbol field, char *text)
|
481
|
+
{
|
482
|
+
FrtTokenStream *ts;
|
483
|
+
if (frt_hs_exists(qp->tokenized_fields, field)) {
|
484
|
+
ts = (FrtTokenStream *)frt_h_get(qp->ts_cache, field);
|
485
|
+
if (!ts) {
|
486
|
+
ts = frt_a_get_ts(qp->analyzer, field, text);
|
487
|
+
frt_h_set(qp->ts_cache, field, ts);
|
488
|
+
}
|
489
|
+
else {
|
490
|
+
ts->reset(ts, text);
|
491
|
+
}
|
492
|
+
}
|
493
|
+
else {
|
494
|
+
ts = qp->non_tokenizer;
|
495
|
+
ts->reset(ts, text);
|
496
|
+
}
|
497
|
+
return ts;
|
498
|
+
}
|
499
|
+
|
500
|
+
/**
|
501
|
+
* Turns a BooleanClause array into a BooleanQuery. It will optimize the query
|
502
|
+
* if 0 or 1 clauses are present to NULL or the actual query in the clause
|
503
|
+
* respectively.
|
504
|
+
*/
|
505
|
+
static FrtQuery *get_bool_q(BCArray *bca)
|
506
|
+
{
|
507
|
+
FrtQuery *q;
|
508
|
+
const int clause_count = bca->size;
|
509
|
+
|
510
|
+
if (clause_count == 0) {
|
511
|
+
q = NULL;
|
512
|
+
free(bca->clauses);
|
513
|
+
}
|
514
|
+
else if (clause_count == 1) {
|
515
|
+
BooleanClause *bc = bca->clauses[0];
|
516
|
+
if (bc->is_prohibited) {
|
517
|
+
q = frt_bq_new(false);
|
518
|
+
frt_bq_add_query_nr(q, bc->query, FRT_BC_MUST_NOT);
|
519
|
+
frt_bq_add_query_nr(q, frt_maq_new(), FRT_BC_MUST);
|
520
|
+
}
|
521
|
+
else {
|
522
|
+
q = bc->query;
|
523
|
+
}
|
524
|
+
free(bc);
|
525
|
+
free(bca->clauses);
|
526
|
+
}
|
527
|
+
else {
|
528
|
+
q = frt_bq_new(false);
|
529
|
+
/* copy clauses into query */
|
530
|
+
|
531
|
+
BQ(q)->clause_cnt = clause_count;
|
532
|
+
BQ(q)->clause_capa = bca->capa;
|
533
|
+
free(BQ(q)->clauses);
|
534
|
+
BQ(q)->clauses = bca->clauses;
|
535
|
+
}
|
536
|
+
free(bca);
|
537
|
+
return q;
|
538
|
+
}
|
539
|
+
|
540
|
+
/**
|
541
|
+
* Base method for appending BooleanClauses to a BooleanClause array. This
|
542
|
+
* method doesn't care about the type of clause (MUST, SHOULD, MUST_NOT).
|
543
|
+
*/
|
544
|
+
static void bca_add_clause(BCArray *bca, BooleanClause *clause)
|
545
|
+
{
|
546
|
+
if (bca->size >= bca->capa) {
|
547
|
+
bca->capa <<= 1;
|
548
|
+
FRT_REALLOC_N(bca->clauses, BooleanClause *, bca->capa);
|
549
|
+
}
|
550
|
+
bca->clauses[bca->size] = clause;
|
551
|
+
bca->size++;
|
552
|
+
}
|
553
|
+
|
554
|
+
/**
|
555
|
+
* Add the first clause to a BooleanClause array. This method is also
|
556
|
+
* responsible for allocating a new BooleanClause array.
|
557
|
+
*/
|
558
|
+
static BCArray *first_cls(BooleanClause *clause)
|
559
|
+
{
|
560
|
+
BCArray *bca = FRT_ALLOC_AND_ZERO(BCArray);
|
561
|
+
bca->capa = BCA_INIT_CAPA;
|
562
|
+
bca->clauses = FRT_ALLOC_N(BooleanClause *, BCA_INIT_CAPA);
|
563
|
+
if (clause) {
|
564
|
+
bca_add_clause(bca, clause);
|
565
|
+
}
|
566
|
+
return bca;
|
567
|
+
}
|
568
|
+
|
569
|
+
/**
|
570
|
+
* Add AND clause to the BooleanClause array. The means that it will set the
|
571
|
+
* clause being added and the previously added clause from SHOULD clauses to
|
572
|
+
* MUST clauses. (If they are currently MUST_NOT clauses they stay as they
|
573
|
+
* are.)
|
574
|
+
*/
|
575
|
+
static BCArray *add_and_cls(BCArray *bca, BooleanClause *clause)
|
576
|
+
{
|
577
|
+
if (clause) {
|
578
|
+
if (bca->size == 1) {
|
579
|
+
if (!bca->clauses[0]->is_prohibited) {
|
580
|
+
bc_set_occur(bca->clauses[0], FRT_BC_MUST);
|
581
|
+
}
|
582
|
+
}
|
583
|
+
if (!clause->is_prohibited) {
|
584
|
+
bc_set_occur(clause, FRT_BC_MUST);
|
585
|
+
}
|
586
|
+
bca_add_clause(bca, clause);
|
587
|
+
}
|
588
|
+
return bca;
|
589
|
+
}
|
590
|
+
|
591
|
+
/**
|
592
|
+
* Add SHOULD clause to the BooleanClause array.
|
593
|
+
*/
|
594
|
+
static BCArray *add_or_cls(BCArray *bca, BooleanClause *clause)
|
595
|
+
{
|
596
|
+
if (clause) {
|
597
|
+
bca_add_clause(bca, clause);
|
598
|
+
}
|
599
|
+
return bca;
|
600
|
+
}
|
601
|
+
|
602
|
+
/**
|
603
|
+
* Add AND or OR clause to the BooleanClause array, depending on the default
|
604
|
+
* clause type.
|
605
|
+
*/
|
606
|
+
static BCArray *add_default_cls(QParser *qp, BCArray *bca,
|
607
|
+
BooleanClause *clause)
|
608
|
+
{
|
609
|
+
if (qp->or_default) {
|
610
|
+
add_or_cls(bca, clause);
|
611
|
+
}
|
612
|
+
else {
|
613
|
+
add_and_cls(bca, clause);
|
614
|
+
}
|
615
|
+
return bca;
|
616
|
+
}
|
617
|
+
|
618
|
+
/**
|
619
|
+
* destroy array of BooleanClauses
|
620
|
+
*/
|
621
|
+
static void bca_destroy(BCArray *bca)
|
622
|
+
{
|
623
|
+
int i;
|
624
|
+
for (i = 0; i < bca->size; i++) {
|
625
|
+
bc_deref(bca->clauses[i]);
|
626
|
+
}
|
627
|
+
free(bca->clauses);
|
628
|
+
free(bca);
|
629
|
+
}
|
630
|
+
|
631
|
+
/**
|
632
|
+
* Turn a query into a BooleanClause for addition to a BooleanQuery.
|
633
|
+
*/
|
634
|
+
static BooleanClause *get_bool_cls(FrtQuery *q, BCType occur)
|
635
|
+
{
|
636
|
+
if (q) {
|
637
|
+
return bc_new(q, occur);
|
638
|
+
}
|
639
|
+
else {
|
640
|
+
return NULL;
|
641
|
+
}
|
642
|
+
}
|
643
|
+
|
644
|
+
/**
|
645
|
+
* Create a TermQuery. The word will be tokenized and if the tokenization
|
646
|
+
* produces more than one token, a PhraseQuery will be returned. For example,
|
647
|
+
* if the word is dbalmain@gmail.com and a LetterTokenizer is used then a
|
648
|
+
* PhraseQuery "dbalmain gmail com" will be returned which is actually exactly
|
649
|
+
* what we want as it will match any documents containing the same email
|
650
|
+
* address and tokenized with the same tokenizer.
|
651
|
+
*/
|
652
|
+
static FrtQuery *get_term_q(QParser *qp, FrtSymbol field, char *word)
|
653
|
+
{
|
654
|
+
FrtQuery *q;
|
655
|
+
FrtToken *token;
|
656
|
+
FrtTokenStream *stream = get_cached_ts(qp, field, word);
|
657
|
+
|
658
|
+
if ((token = frt_ts_next(stream)) == NULL) {
|
659
|
+
q = NULL;
|
660
|
+
}
|
661
|
+
else {
|
662
|
+
q = frt_tq_new(field, token->text);
|
663
|
+
if ((token = frt_ts_next(stream)) != NULL) {
|
664
|
+
/* Less likely case, destroy the term query and create a
|
665
|
+
* phrase query instead */
|
666
|
+
FrtQuery *phq = frt_phq_new(field);
|
667
|
+
frt_phq_add_term(phq, ((TermQuery *)q)->term, 0);
|
668
|
+
q->destroy_i(q);
|
669
|
+
q = phq;
|
670
|
+
do {
|
671
|
+
if (token->pos_inc) {
|
672
|
+
frt_phq_add_term(q, token->text, token->pos_inc);
|
673
|
+
/* add some slop since single term was expected */
|
674
|
+
((FrtPhraseQuery *)q)->slop++;
|
675
|
+
}
|
676
|
+
else {
|
677
|
+
frt_phq_append_multi_term(q, token->text);
|
678
|
+
}
|
679
|
+
} while ((token = frt_ts_next(stream)) != NULL);
|
680
|
+
}
|
681
|
+
}
|
682
|
+
return q;
|
683
|
+
}
|
684
|
+
|
685
|
+
/**
|
686
|
+
* Create a FuzzyQuery. The word will be tokenized and only the first token
|
687
|
+
* will be used. If there are any more tokens after tokenization, they will be
|
688
|
+
* ignored.
|
689
|
+
*/
|
690
|
+
static FrtQuery *get_fuzzy_q(QParser *qp, FrtSymbol field, char *word,
|
691
|
+
char *slop_str)
|
692
|
+
{
|
693
|
+
FrtQuery *q;
|
694
|
+
FrtToken *token;
|
695
|
+
FrtTokenStream *stream = get_cached_ts(qp, field, word);
|
696
|
+
|
697
|
+
if ((token = frt_ts_next(stream)) == NULL) {
|
698
|
+
q = NULL;
|
699
|
+
}
|
700
|
+
else {
|
701
|
+
/* it only makes sense to find one term in a fuzzy query */
|
702
|
+
float slop = qp_default_fuzzy_min_sim;
|
703
|
+
if (slop_str) {
|
704
|
+
sscanf(slop_str, "%f", &slop);
|
705
|
+
}
|
706
|
+
q = frt_fuzq_new_conf(field, token->text, slop, qp_default_fuzzy_pre_len,
|
707
|
+
qp->max_clauses);
|
708
|
+
}
|
709
|
+
return q;
|
710
|
+
}
|
711
|
+
|
712
|
+
/**
|
713
|
+
* Downcase a string taking locale into account and works for multibyte
|
714
|
+
* character sets.
|
715
|
+
*/
|
716
|
+
static char *lower_str(char *str)
|
717
|
+
{
|
718
|
+
const int max_len = (int)strlen(str) + 1;
|
719
|
+
int cnt;
|
720
|
+
wchar_t *wstr = FRT_ALLOC_N(wchar_t, max_len);
|
721
|
+
if ((cnt = mbstowcs(wstr, str, max_len)) > 0) {
|
722
|
+
wchar_t *w = wstr;
|
723
|
+
while (*w) {
|
724
|
+
*w = towlower(*w);
|
725
|
+
w++;
|
726
|
+
}
|
727
|
+
wcstombs(str, wstr, max_len);
|
728
|
+
}
|
729
|
+
else {
|
730
|
+
char *s = str;
|
731
|
+
while (*s) {
|
732
|
+
*s = tolower(*s);
|
733
|
+
s++;
|
734
|
+
}
|
735
|
+
}
|
736
|
+
free(wstr);
|
737
|
+
str[max_len] = '\0';
|
738
|
+
return str;
|
739
|
+
}
|
740
|
+
|
741
|
+
/**
|
742
|
+
* Create a WildCardQuery. No tokenization will be performed on the pattern
|
743
|
+
* but the pattern will be downcased if +qp->wild_lower+ is set to true and
|
744
|
+
* the field in question is a tokenized field.
|
745
|
+
*
|
746
|
+
* Note: this method will not always return a WildCardQuery. It could be
|
747
|
+
* optimized to a MatchAllQuery if the pattern is '*' or a PrefixQuery if the
|
748
|
+
* only wild char (*, ?) in the pattern is a '*' at the end of the pattern.
|
749
|
+
*/
|
750
|
+
static FrtQuery *get_wild_q(QParser *qp, FrtSymbol field, char *pattern)
|
751
|
+
{
|
752
|
+
FrtQuery *q;
|
753
|
+
bool is_prefix = false;
|
754
|
+
char *p;
|
755
|
+
int len = (int)strlen(pattern);
|
756
|
+
|
757
|
+
if (qp->wild_lower
|
758
|
+
&& (!qp->tokenized_fields || frt_hs_exists(qp->tokenized_fields, field))) {
|
759
|
+
lower_str(pattern);
|
760
|
+
}
|
761
|
+
|
762
|
+
/* simplify the wildcard query to a prefix query if possible. Basically a
|
763
|
+
* prefix query is any wildcard query that has a '*' as the last character
|
764
|
+
* and no other wildcard characters before it. "*" by itself will expand
|
765
|
+
* to a MatchAllQuery */
|
766
|
+
if (strcmp(pattern, "*") == 0) {
|
767
|
+
return frt_maq_new();
|
768
|
+
}
|
769
|
+
if (pattern[len - 1] == '*') {
|
770
|
+
is_prefix = true;
|
771
|
+
for (p = &pattern[len - 2]; p >= pattern; p--) {
|
772
|
+
if (*p == '*' || *p == '?') {
|
773
|
+
is_prefix = false;
|
774
|
+
break;
|
775
|
+
}
|
776
|
+
}
|
777
|
+
}
|
778
|
+
if (is_prefix) {
|
779
|
+
/* chop off the '*' temporarily to create the query */
|
780
|
+
pattern[len - 1] = 0;
|
781
|
+
q = frt_prefixq_new(field, pattern);
|
782
|
+
pattern[len - 1] = '*';
|
783
|
+
}
|
784
|
+
else {
|
785
|
+
q = frt_wcq_new(field, pattern);
|
786
|
+
}
|
787
|
+
MTQMaxTerms(q) = qp->max_clauses;
|
788
|
+
return q;
|
789
|
+
}
|
790
|
+
|
791
|
+
/**
|
792
|
+
* Adds another field to the top of the FieldStack.
|
793
|
+
*/
|
794
|
+
static FrtHashSet *add_field(QParser *qp, const char *field_name)
|
795
|
+
{
|
796
|
+
FrtSymbol field = field_name;
|
797
|
+
if (qp->allow_any_fields || frt_hs_exists(qp->all_fields, field)) {
|
798
|
+
frt_hs_add(qp->fields, field);
|
799
|
+
}
|
800
|
+
return qp->fields;
|
801
|
+
}
|
802
|
+
|
803
|
+
/**
|
804
|
+
* The method gets called when a field modifier ("field1|field2:") is seen. It
|
805
|
+
* will push a new FieldStack object onto the stack and add +field+ to its
|
806
|
+
* fields set.
|
807
|
+
*/
|
808
|
+
static FrtHashSet *first_field(QParser *qp, const char *field)
|
809
|
+
{
|
810
|
+
qp_push_fields(qp, frt_hs_new_ptr(NULL), true);
|
811
|
+
return add_field(qp, field);
|
812
|
+
}
|
813
|
+
|
814
|
+
/**
|
815
|
+
* Destroy a phrase object freeing all allocated memory.
|
816
|
+
*/
|
817
|
+
static void ph_destroy(Phrase *self)
|
818
|
+
{
|
819
|
+
int i;
|
820
|
+
for (i = 0; i < self->size; i++) {
|
821
|
+
frt_ary_destroy(self->positions[i].terms, &free);
|
822
|
+
}
|
823
|
+
free(self->positions);
|
824
|
+
free(self);
|
825
|
+
}
|
826
|
+
|
827
|
+
|
828
|
+
/**
|
829
|
+
* Allocate a new Phrase object
|
830
|
+
*/
|
831
|
+
static Phrase *ph_new()
|
832
|
+
{
|
833
|
+
Phrase *self = FRT_ALLOC_AND_ZERO(Phrase);
|
834
|
+
self->capa = PHRASE_INIT_CAPA;
|
835
|
+
self->positions = FRT_ALLOC_AND_ZERO_N(PhrasePosition, PHRASE_INIT_CAPA);
|
836
|
+
return self;
|
837
|
+
}
|
838
|
+
|
839
|
+
/**
|
840
|
+
* Add the first word to the phrase. This method is also in charge of
|
841
|
+
* allocating a new Phrase object.
|
842
|
+
*/
|
843
|
+
static Phrase *ph_first_word(char *word)
|
844
|
+
{
|
845
|
+
Phrase *self = ph_new();
|
846
|
+
if (word) { /* no point in adding NULL in start */
|
847
|
+
self->positions[0].terms = frt_ary_new_type_capa(char *, 1);
|
848
|
+
frt_ary_push(self->positions[0].terms, frt_estrdup(word));
|
849
|
+
self->size = 1;
|
850
|
+
}
|
851
|
+
return self;
|
852
|
+
}
|
853
|
+
|
854
|
+
/**
|
855
|
+
* Add a new word to the Phrase
|
856
|
+
*/
|
857
|
+
static Phrase *ph_add_word(Phrase *self, char *word)
|
858
|
+
{
|
859
|
+
if (word) {
|
860
|
+
const int index = self->size;
|
861
|
+
FrtPhrasePosition *pp = self->positions;
|
862
|
+
if (index >= self->capa) {
|
863
|
+
self->capa <<= 1;
|
864
|
+
FRT_REALLOC_N(pp, PhrasePosition, self->capa);
|
865
|
+
self->positions = pp;
|
866
|
+
}
|
867
|
+
pp[index].pos = self->pos_inc;
|
868
|
+
pp[index].terms = frt_ary_new_type_capa(char *, 1);
|
869
|
+
frt_ary_push(pp[index].terms, frt_estrdup(word));
|
870
|
+
self->size++;
|
871
|
+
self->pos_inc = 0;
|
872
|
+
}
|
873
|
+
else {
|
874
|
+
self->pos_inc++;
|
875
|
+
}
|
876
|
+
return self;
|
877
|
+
}
|
878
|
+
|
879
|
+
/**
|
880
|
+
* Adds a word to the Phrase object in the same position as the previous word
|
881
|
+
* added to the Phrase. This will later be turned into a multi-PhraseQuery.
|
882
|
+
*/
|
883
|
+
static Phrase *ph_add_multi_word(Phrase *self, char *word)
|
884
|
+
{
|
885
|
+
const int index = self->size - 1;
|
886
|
+
FrtPhrasePosition *pp = self->positions;
|
887
|
+
|
888
|
+
if (word) {
|
889
|
+
frt_ary_push(pp[index].terms, frt_estrdup(word));
|
890
|
+
}
|
891
|
+
return self;
|
892
|
+
}
|
893
|
+
|
894
|
+
/**
|
895
|
+
* Build a phrase query for a single field. It might seem like a better idea
|
896
|
+
* to build the PhraseQuery once and duplicate it for each field but this
|
897
|
+
* would be buggy in the case of PerFieldAnalyzers in which case a different
|
898
|
+
* tokenizer could be used for each field.
|
899
|
+
*
|
900
|
+
* Note that the query object returned by this method is not always a
|
901
|
+
* PhraseQuery. If there is only one term in the query then the query is
|
902
|
+
* simplified to a TermQuery. If there are multiple terms but only a single
|
903
|
+
* position, then a MultiTermQuery is retured.
|
904
|
+
*
|
905
|
+
* Note that each word in the query gets tokenized. Unlike get_term_q, if the
|
906
|
+
* word gets tokenized into more than one token, the rest of the tokens are
|
907
|
+
* ignored. For example, if you have the phrase;
|
908
|
+
*
|
909
|
+
* "email: dbalmain@gmail.com"
|
910
|
+
*
|
911
|
+
* the Phrase object will contain to positions with the words 'email:' and
|
912
|
+
* 'dbalmain@gmail.com'. Now, if you are using a LetterTokenizer then the
|
913
|
+
* second word will be tokenized into the tokens ['dbalmain', 'gmail', 'com']
|
914
|
+
* and only the first token will be used, so the resulting phrase query will
|
915
|
+
* actually look like this;
|
916
|
+
*
|
917
|
+
* "email dbalmain"
|
918
|
+
*
|
919
|
+
* This problem can easily be solved by using the StandardTokenizer or any
|
920
|
+
* custom tokenizer which will leave dbalmain@gmail.com as a single token.
|
921
|
+
*/
|
922
|
+
static FrtQuery *get_phrase_query(QParser *qp, FrtSymbol field,
|
923
|
+
Phrase *phrase, char *slop_str)
|
924
|
+
{
|
925
|
+
const int pos_cnt = phrase->size;
|
926
|
+
FrtQuery *q = NULL;
|
927
|
+
|
928
|
+
if (pos_cnt == 1) {
|
929
|
+
char **words = phrase->positions[0].terms;
|
930
|
+
const int word_count = frt_ary_size(words);
|
931
|
+
if (word_count == 1) {
|
932
|
+
q = get_term_q(qp, field, words[0]);
|
933
|
+
}
|
934
|
+
else {
|
935
|
+
int i;
|
936
|
+
int term_cnt = 0;
|
937
|
+
FrtToken *token;
|
938
|
+
char *last_word = NULL;
|
939
|
+
|
940
|
+
for (i = 0; i < word_count; i++) {
|
941
|
+
token = frt_ts_next(get_cached_ts(qp, field, words[i]));
|
942
|
+
if (token) {
|
943
|
+
free(words[i]);
|
944
|
+
last_word = words[i] = frt_estrdup(token->text);
|
945
|
+
++term_cnt;
|
946
|
+
}
|
947
|
+
else {
|
948
|
+
/* empty words will later be ignored */
|
949
|
+
words[i][0] = '\0';
|
950
|
+
}
|
951
|
+
}
|
952
|
+
|
953
|
+
switch (term_cnt) {
|
954
|
+
case 0:
|
955
|
+
q = frt_bq_new(false);
|
956
|
+
break;
|
957
|
+
case 1:
|
958
|
+
q = frt_tq_new(field, last_word);
|
959
|
+
break;
|
960
|
+
default:
|
961
|
+
q = frt_multi_tq_new_conf(field, term_cnt, 0.0);
|
962
|
+
for (i = 0; i < word_count; i++) {
|
963
|
+
/* ignore empty words */
|
964
|
+
if (words[i][0]) {
|
965
|
+
frt_multi_tq_add_term(q, words[i]);
|
966
|
+
}
|
967
|
+
}
|
968
|
+
break;
|
969
|
+
}
|
970
|
+
}
|
971
|
+
}
|
972
|
+
else if (pos_cnt > 1) {
|
973
|
+
FrtToken *token;
|
974
|
+
FrtTokenStream *stream;
|
975
|
+
int i, j;
|
976
|
+
int pos_inc = 0;
|
977
|
+
q = frt_phq_new(field);
|
978
|
+
if (slop_str) {
|
979
|
+
int slop;
|
980
|
+
sscanf(slop_str,"%d",&slop);
|
981
|
+
((FrtPhraseQuery *)q)->slop = slop;
|
982
|
+
}
|
983
|
+
|
984
|
+
for (i = 0; i < pos_cnt; i++) {
|
985
|
+
char **words = phrase->positions[i].terms;
|
986
|
+
const int word_count = frt_ary_size(words);
|
987
|
+
if (pos_inc) {
|
988
|
+
((FrtPhraseQuery *)q)->slop++;
|
989
|
+
}
|
990
|
+
pos_inc += phrase->positions[i].pos + 1; /* Actually holds pos_inc*/
|
991
|
+
|
992
|
+
if (word_count == 1) {
|
993
|
+
stream = get_cached_ts(qp, field, words[0]);
|
994
|
+
while ((token = frt_ts_next(stream))) {
|
995
|
+
if (token->pos_inc) {
|
996
|
+
frt_phq_add_term(q, token->text,
|
997
|
+
pos_inc ? pos_inc : token->pos_inc);
|
998
|
+
}
|
999
|
+
else {
|
1000
|
+
frt_phq_append_multi_term(q, token->text);
|
1001
|
+
((FrtPhraseQuery *)q)->slop++;
|
1002
|
+
}
|
1003
|
+
pos_inc = 0;
|
1004
|
+
}
|
1005
|
+
}
|
1006
|
+
else {
|
1007
|
+
bool added_position = false;
|
1008
|
+
|
1009
|
+
for (j = 0; j < word_count; j++) {
|
1010
|
+
stream = get_cached_ts(qp, field, words[j]);
|
1011
|
+
if ((token = frt_ts_next(stream))) {
|
1012
|
+
if (!added_position) {
|
1013
|
+
frt_phq_add_term(q, token->text,
|
1014
|
+
pos_inc ? pos_inc : token->pos_inc);
|
1015
|
+
added_position = true;
|
1016
|
+
pos_inc = 0;
|
1017
|
+
}
|
1018
|
+
else {
|
1019
|
+
frt_phq_append_multi_term(q, token->text);
|
1020
|
+
}
|
1021
|
+
}
|
1022
|
+
}
|
1023
|
+
}
|
1024
|
+
}
|
1025
|
+
}
|
1026
|
+
return q;
|
1027
|
+
}
|
1028
|
+
|
1029
|
+
/**
|
1030
|
+
* Get a phrase query from the Phrase object. The Phrase object is built up by
|
1031
|
+
* the query parser as the all PhraseQuery didn't work well for this. Once the
|
1032
|
+
* PhraseQuery has been built the Phrase object needs to be destroyed.
|
1033
|
+
*/
|
1034
|
+
static FrtQuery *get_phrase_q(QParser *qp, Phrase *phrase, char *slop_str)
|
1035
|
+
{
|
1036
|
+
FrtQuery *volatile q = NULL;
|
1037
|
+
FLDS(q, get_phrase_query(qp, field, phrase, slop_str));
|
1038
|
+
ph_destroy(phrase);
|
1039
|
+
return q;
|
1040
|
+
}
|
1041
|
+
|
1042
|
+
/**
|
1043
|
+
* Gets a RangeQuery object.
|
1044
|
+
*
|
1045
|
+
* Just like with WildCardQuery, RangeQuery needs to downcase its terms if the
|
1046
|
+
* tokenizer also downcased its terms.
|
1047
|
+
*/
|
1048
|
+
static FrtQuery *get_r_q(QParser *qp, FrtSymbol field, char *from, char *to,
|
1049
|
+
bool inc_lower, bool inc_upper)
|
1050
|
+
{
|
1051
|
+
FrtQuery *rq;
|
1052
|
+
if (qp->wild_lower
|
1053
|
+
&& (!qp->tokenized_fields || frt_hs_exists(qp->tokenized_fields, field))) {
|
1054
|
+
if (from) {
|
1055
|
+
lower_str(from);
|
1056
|
+
}
|
1057
|
+
if (to) {
|
1058
|
+
lower_str(to);
|
1059
|
+
}
|
1060
|
+
}
|
1061
|
+
/*
|
1062
|
+
* terms don't get tokenized as it doesn't really make sense to do so for
|
1063
|
+
* range queries.
|
1064
|
+
|
1065
|
+
if (from) {
|
1066
|
+
FrtTokenStream *stream = get_cached_ts(qp, field, from);
|
1067
|
+
FrtToken *token = frt_ts_next(stream);
|
1068
|
+
from = token ? frt_estrdup(token->text) : NULL;
|
1069
|
+
}
|
1070
|
+
if (to) {
|
1071
|
+
FrtTokenStream *stream = get_cached_ts(qp, field, to);
|
1072
|
+
FrtToken *token = frt_ts_next(stream);
|
1073
|
+
to = token ? frt_estrdup(token->text) : NULL;
|
1074
|
+
}
|
1075
|
+
*/
|
1076
|
+
|
1077
|
+
rq = qp->use_typed_range_query ?
|
1078
|
+
frt_trq_new(field, from, to, inc_lower, inc_upper) :
|
1079
|
+
frt_rq_new(field, from, to, inc_lower, inc_upper);
|
1080
|
+
return rq;
|
1081
|
+
}
|
1082
|
+
|
1083
|
+
/**
|
1084
|
+
* Every time the query parser sees a new field modifier ("field1|field2:")
|
1085
|
+
* it pushes a new FieldStack object onto the stack and sets its fields to the
|
1086
|
+
* fields specified in the fields modifier. If the field modifier is '*',
|
1087
|
+
* fs->fields is set to all_fields. fs->fields is set to +qp->def_field+ at
|
1088
|
+
* the bottom of the stack (ie the very first set of fields pushed onto the
|
1089
|
+
* stack).
|
1090
|
+
*/
|
1091
|
+
static void qp_push_fields(QParser *self, FrtHashSet *fields, bool destroy)
|
1092
|
+
{
|
1093
|
+
FieldStack *fs = FRT_ALLOC(FieldStack);
|
1094
|
+
|
1095
|
+
fs->next = self->fields_top;
|
1096
|
+
fs->fields = fields;
|
1097
|
+
fs->destroy = destroy;
|
1098
|
+
|
1099
|
+
self->fields_top = fs;
|
1100
|
+
self->fields = fields;
|
1101
|
+
}
|
1102
|
+
|
1103
|
+
/**
|
1104
|
+
* Pops the top of the fields stack and frees any memory used by it. This will
|
1105
|
+
* get called when query modified by a field modifier ("field1|field2:") has
|
1106
|
+
* been fully parsed and the field specifier no longer applies.
|
1107
|
+
*/
|
1108
|
+
static void qp_pop_fields(QParser *self)
|
1109
|
+
{
|
1110
|
+
FieldStack *fs = self->fields_top;
|
1111
|
+
|
1112
|
+
if (fs->destroy) {
|
1113
|
+
frt_hs_destroy(fs->fields);
|
1114
|
+
}
|
1115
|
+
self->fields_top = fs->next;
|
1116
|
+
if (self->fields_top) {
|
1117
|
+
self->fields = self->fields_top->fields;
|
1118
|
+
}
|
1119
|
+
free(fs);
|
1120
|
+
}
|
1121
|
+
|
1122
|
+
/**
|
1123
|
+
* Free all memory allocated by the QueryParser.
|
1124
|
+
*/
|
1125
|
+
void frt_qp_destroy(QParser *self)
|
1126
|
+
{
|
1127
|
+
if (self->tokenized_fields != self->all_fields) {
|
1128
|
+
frt_hs_destroy(self->tokenized_fields);
|
1129
|
+
}
|
1130
|
+
if (self->def_fields != self->all_fields) {
|
1131
|
+
frt_hs_destroy(self->def_fields);
|
1132
|
+
}
|
1133
|
+
frt_hs_destroy(self->all_fields);
|
1134
|
+
|
1135
|
+
qp_pop_fields(self);
|
1136
|
+
assert(NULL == self->fields_top);
|
1137
|
+
|
1138
|
+
frt_h_destroy(self->ts_cache);
|
1139
|
+
frt_tk_destroy(self->non_tokenizer);
|
1140
|
+
frt_a_deref(self->analyzer);
|
1141
|
+
free(self);
|
1142
|
+
}
|
1143
|
+
|
1144
|
+
/**
|
1145
|
+
* Creates a new QueryParser setting all boolean parameters to their defaults.
|
1146
|
+
* If +def_fields+ is NULL then +all_fields+ is used in place of +def_fields+.
|
1147
|
+
* Not also that this method ensures that all fields that exist in
|
1148
|
+
* +def_fields+ must also exist in +all_fields+. This should make sense.
|
1149
|
+
*/
|
1150
|
+
QParser *qp_new(FrtAnalyzer *analyzer)
|
1151
|
+
{
|
1152
|
+
FrtQParser *self = FRT_ALLOC(QParser);
|
1153
|
+
self->or_default = true;
|
1154
|
+
self->wild_lower = true;
|
1155
|
+
self->clean_str = false;
|
1156
|
+
self->max_clauses = QP_MAX_CLAUSES;
|
1157
|
+
self->handle_parse_errors = false;
|
1158
|
+
self->allow_any_fields = false;
|
1159
|
+
self->use_keywords = true;
|
1160
|
+
self->use_typed_range_query = false;
|
1161
|
+
self->def_slop = 0;
|
1162
|
+
|
1163
|
+
self->tokenized_fields = frt_hs_new_ptr(NULL);
|
1164
|
+
self->all_fields = frt_hs_new_ptr(NULL);
|
1165
|
+
self->def_fields = frt_hs_new_ptr(NULL);
|
1166
|
+
|
1167
|
+
self->fields_top = NULL;
|
1168
|
+
qp_push_fields(self, self->def_fields, false);
|
1169
|
+
|
1170
|
+
/* make sure all_fields contains the default fields */
|
1171
|
+
self->analyzer = analyzer;
|
1172
|
+
self->ts_cache = frt_h_new_ptr((frt_free_ft)&ts_deref);
|
1173
|
+
self->buf_index = 0;
|
1174
|
+
self->dynbuf = NULL;
|
1175
|
+
self->non_tokenizer = non_tokenizer_new();
|
1176
|
+
frt_mutex_init(&self->mutex, NULL);
|
1177
|
+
return self;
|
1178
|
+
}
|
1179
|
+
|
1180
|
+
void frt_qp_add_field(QParser *self,
|
1181
|
+
FrtSymbol field,
|
1182
|
+
bool is_default,
|
1183
|
+
bool is_tokenized)
|
1184
|
+
{
|
1185
|
+
frt_hs_add(self->all_fields, field);
|
1186
|
+
if (is_default) {
|
1187
|
+
frt_hs_add(self->def_fields, field);
|
1188
|
+
}
|
1189
|
+
if (is_tokenized) {
|
1190
|
+
frt_hs_add(self->tokenized_fields, field);
|
1191
|
+
}
|
1192
|
+
}
|
1193
|
+
|
1194
|
+
/* these chars have meaning within phrases */
|
1195
|
+
static const char *PHRASE_CHARS = "<>|\"";
|
1196
|
+
|
1197
|
+
/**
|
1198
|
+
* +str_insert_char+ inserts a character at the beginning of a string by
|
1199
|
+
* shifting the rest of the string right.
|
1200
|
+
*/
|
1201
|
+
static void str_insert_char(char *str, int len, char chr)
|
1202
|
+
{
|
1203
|
+
memmove(str+1, str, len*sizeof(char));
|
1204
|
+
*str = chr;
|
1205
|
+
}
|
1206
|
+
|
1207
|
+
/**
|
1208
|
+
* +qp_clean_str+ basically scans the query string and ensures that all open
|
1209
|
+
* and close parentheses '()' and quotes '"' are balanced. It does this by
|
1210
|
+
* inserting or appending extra parentheses or quotes to the string. This
|
1211
|
+
* obviously won't necessarily be exactly what the user wanted but we are
|
1212
|
+
* never going to know that anyway. The main job of this method is to help the
|
1213
|
+
* query at least parse correctly.
|
1214
|
+
*
|
1215
|
+
* It also checks that all special characters within phrases (ie between
|
1216
|
+
* quotes) are escaped correctly unless they have meaning within a phrase
|
1217
|
+
* ( <>,|," ). Note that '<' and '>' will also be escaped unless the appear
|
1218
|
+
* together like so; '<>'.
|
1219
|
+
*/
|
1220
|
+
char *qp_clean_str(char *str)
|
1221
|
+
{
|
1222
|
+
int b, pb = -1;
|
1223
|
+
int br_cnt = 0;
|
1224
|
+
bool quote_open = false;
|
1225
|
+
char *sp, *nsp;
|
1226
|
+
|
1227
|
+
/* leave a little extra */
|
1228
|
+
char *new_str = FRT_ALLOC_N(char, strlen(str)*2 + 1);
|
1229
|
+
|
1230
|
+
for (sp = str, nsp = new_str; *sp; sp++) {
|
1231
|
+
b = *sp;
|
1232
|
+
/* ignore escaped characters */
|
1233
|
+
if (pb == '\\') {
|
1234
|
+
if (quote_open && strrchr(PHRASE_CHARS, b)) {
|
1235
|
+
*nsp++ = '\\'; /* this was left off the first time through */
|
1236
|
+
}
|
1237
|
+
*nsp++ = b;
|
1238
|
+
/* \ has escaped itself so has no power. Assign pb random char 'r' */
|
1239
|
+
pb = ((b == '\\') ? 'r' : b);
|
1240
|
+
continue;
|
1241
|
+
}
|
1242
|
+
switch (b) {
|
1243
|
+
case '\\':
|
1244
|
+
if (!quote_open) { /* We do our own escaping below */
|
1245
|
+
*nsp++ = b;
|
1246
|
+
}
|
1247
|
+
break;
|
1248
|
+
case '"':
|
1249
|
+
quote_open = !quote_open;
|
1250
|
+
*nsp++ = b;
|
1251
|
+
break;
|
1252
|
+
case '(':
|
1253
|
+
if (!quote_open) {
|
1254
|
+
br_cnt++;
|
1255
|
+
}
|
1256
|
+
else {
|
1257
|
+
*nsp++ = '\\';
|
1258
|
+
}
|
1259
|
+
*nsp++ = b;
|
1260
|
+
break;
|
1261
|
+
case ')':
|
1262
|
+
if (!quote_open) {
|
1263
|
+
if (br_cnt == 0) {
|
1264
|
+
str_insert_char(new_str, (int)(nsp - new_str), '(');
|
1265
|
+
nsp++;
|
1266
|
+
}
|
1267
|
+
else {
|
1268
|
+
br_cnt--;
|
1269
|
+
}
|
1270
|
+
}
|
1271
|
+
else {
|
1272
|
+
*nsp++ = '\\';
|
1273
|
+
}
|
1274
|
+
*nsp++ = b;
|
1275
|
+
break;
|
1276
|
+
case '>':
|
1277
|
+
if (quote_open) {
|
1278
|
+
if (pb == '<') {
|
1279
|
+
/* remove the escape character */
|
1280
|
+
nsp--;
|
1281
|
+
nsp[-1] = '<';
|
1282
|
+
}
|
1283
|
+
else {
|
1284
|
+
*nsp++ = '\\';
|
1285
|
+
}
|
1286
|
+
}
|
1287
|
+
*nsp++ = b;
|
1288
|
+
break;
|
1289
|
+
default:
|
1290
|
+
if (quote_open) {
|
1291
|
+
if (strrchr(special_char, b) && b != '|') {
|
1292
|
+
*nsp++ = '\\';
|
1293
|
+
}
|
1294
|
+
}
|
1295
|
+
*nsp++ = b;
|
1296
|
+
}
|
1297
|
+
pb = b;
|
1298
|
+
}
|
1299
|
+
if (quote_open) {
|
1300
|
+
*nsp++ = '"';
|
1301
|
+
}
|
1302
|
+
for (;br_cnt > 0; br_cnt--) {
|
1303
|
+
*nsp++ = ')';
|
1304
|
+
}
|
1305
|
+
*nsp = '\0';
|
1306
|
+
return new_str;
|
1307
|
+
}
|
1308
|
+
|
1309
|
+
/**
|
1310
|
+
* Takes a string and finds whatever tokens it can using the QueryParser's
|
1311
|
+
* analyzer. It then turns these tokens (if any) into a boolean query. If it
|
1312
|
+
* fails to find any tokens, this method will return NULL.
|
1313
|
+
*/
|
1314
|
+
static FrtQuery *qp_get_bad_query(QParser *qp, char *str)
|
1315
|
+
{
|
1316
|
+
FrtQuery *volatile q = NULL;
|
1317
|
+
qp->recovering = true;
|
1318
|
+
assert(qp->fields_top->next == NULL);
|
1319
|
+
FLDS(q, get_term_q(qp, field, str));
|
1320
|
+
return q;
|
1321
|
+
}
|
1322
|
+
|
1323
|
+
/**
|
1324
|
+
* +qp_parse+ takes a string and turns it into a Query object using Ferret's
|
1325
|
+
* query language. It must either raise an error or return a query object. It
|
1326
|
+
* must not return NULL. If the yacc parser fails it will use a very basic
|
1327
|
+
* boolean query parser which takes whatever tokens it can find in the query
|
1328
|
+
* and terns them into a boolean query on the default fields.
|
1329
|
+
*/
|
1330
|
+
extern VALUE cQueryParseException;
|
1331
|
+
Query *qp_parse(QParser *self, char *qstr)
|
1332
|
+
{
|
1333
|
+
FrtQuery *result = NULL;
|
1334
|
+
frt_mutex_lock(&self->mutex);
|
1335
|
+
/* if qp->fields_top->next is not NULL we have a left over field-stack
|
1336
|
+
* object that was not popped during the last query parse */
|
1337
|
+
assert(NULL == self->fields_top->next);
|
1338
|
+
|
1339
|
+
self->recovering = self->destruct = false;
|
1340
|
+
if (self->clean_str) {
|
1341
|
+
self->qstrp = self->qstr = frt_qp_clean_str(qstr);
|
1342
|
+
}
|
1343
|
+
else {
|
1344
|
+
self->qstrp = self->qstr = qstr;
|
1345
|
+
}
|
1346
|
+
self->fields = self->def_fields;
|
1347
|
+
self->result = NULL;
|
1348
|
+
|
1349
|
+
if (0 == yyparse(self)) result = self->result;
|
1350
|
+
if (!result && self->handle_parse_errors) {
|
1351
|
+
self->destruct = false;
|
1352
|
+
result = qp_get_bad_query(self, self->qstr);
|
1353
|
+
}
|
1354
|
+
if (self->destruct && !self->handle_parse_errors) {
|
1355
|
+
rb_raise(cQueryParseException, xmsg_buffer);
|
1356
|
+
}
|
1357
|
+
if (!result) {
|
1358
|
+
result = frt_bq_new(false);
|
1359
|
+
}
|
1360
|
+
if (self->clean_str) {
|
1361
|
+
free(self->qstr);
|
1362
|
+
}
|
1363
|
+
|
1364
|
+
frt_mutex_unlock(&self->mutex);
|
1365
|
+
return result;
|
1366
|
+
}
|