1 | /**********************************************************************
|
---|
2 | *
|
---|
3 | * phrasesearch.cpp -- tools to search for a phrase in a larger text
|
---|
4 | * Copyright (C) 1999 DigiLib Systems Limited
|
---|
5 | *
|
---|
6 | * PUT COPYRIGHT NOTICE HERE
|
---|
7 | *
|
---|
8 | * $Id: phrasesearch.cpp 351 1999-07-07 06:19:47Z rjmcnab $
|
---|
9 | *
|
---|
10 | *********************************************************************/
|
---|
11 |
|
---|
12 | /*
|
---|
13 | $Log$
|
---|
14 | Revision 1.3 1999/07/07 06:19:45 rjmcnab
|
---|
15 | Added ability to combine two or more independant queries.
|
---|
16 |
|
---|
17 | Revision 1.2 1999/07/01 09:25:54 rjmcnab
|
---|
18 | fixed bug :-^
|
---|
19 |
|
---|
20 | Revision 1.1 1999/07/01 04:01:46 rjmcnab
|
---|
21 | Initial revision.
|
---|
22 |
|
---|
23 | */
|
---|
24 |
|
---|
25 |
|
---|
26 | #include "phrasesearch.h"
|
---|
27 | #include "gsdlunicode.h"
|
---|
28 |
|
---|
29 | inline unsigned char *parse_nonword_word (unsigned char *here, unsigned char *end,
|
---|
30 | text_t &word) {
|
---|
31 | int c_len = 0;
|
---|
32 | unsigned short c = 0;
|
---|
33 |
|
---|
34 | word.clear();
|
---|
35 |
|
---|
36 | // parse non word
|
---|
37 | while (here <= end) {
|
---|
38 | c_len = parse_utf8_char (here, end, &c);
|
---|
39 | if (is_unicode_letdig(c)) {
|
---|
40 | while (c_len > 0) {
|
---|
41 | word.push_back(*here);
|
---|
42 | here++; c_len--;
|
---|
43 | }
|
---|
44 | break;
|
---|
45 | }
|
---|
46 | here += c_len;
|
---|
47 | }
|
---|
48 |
|
---|
49 | // parse word
|
---|
50 | while (here <= end) {
|
---|
51 | c_len = parse_utf8_char (here, end, &c);
|
---|
52 | if (!is_unicode_letdig(c)) {
|
---|
53 | here += c_len; // it is ok to skip a nonword character
|
---|
54 | break;
|
---|
55 | }
|
---|
56 | while (c_len > 0) {
|
---|
57 | word.push_back(*here);
|
---|
58 | here++; c_len--;
|
---|
59 | }
|
---|
60 | }
|
---|
61 |
|
---|
62 | return here;
|
---|
63 | }
|
---|
64 |
|
---|
65 |
|
---|
66 | bool doc_phrase_search (unsigned char *doc, int doclen,
|
---|
67 | const vector<termfreqclass> &phrase) {
|
---|
68 | // note: this uses the most braindead search routine :-)
|
---|
69 | // however its not so bad as there shouldn't be many partial
|
---|
70 | // matches
|
---|
71 |
|
---|
72 | // a null phrase matches anything
|
---|
73 | if (phrase.empty()) return true;
|
---|
74 |
|
---|
75 | // if there is nothing then there can't be a match
|
---|
76 | if (doc == NULL || doclen == 0) return false;
|
---|
77 |
|
---|
78 | text_t doc_word;
|
---|
79 | doc_word.reserve (16);
|
---|
80 |
|
---|
81 | bool first = true;
|
---|
82 |
|
---|
83 | unsigned char *doc_here = doc;
|
---|
84 | unsigned char *doc_herefirstword = doc;
|
---|
85 | unsigned char *doc_end = doc+doclen-1; // unitool conventions :-/
|
---|
86 |
|
---|
87 | while (doc_here <= doc_end) {
|
---|
88 | first = true;
|
---|
89 |
|
---|
90 | // there will be at least one member of phrase (see above)
|
---|
91 | vector<termfreqclass>::const_iterator phrase_here = phrase.begin();
|
---|
92 | vector<termfreqclass>::const_iterator phrase_end = phrase.end();
|
---|
93 | do {
|
---|
94 | // get the next non-word ... and ignore it, then get the next word
|
---|
95 | doc_here = parse_nonword_word (doc_here, doc_end, doc_word);
|
---|
96 | if (first) {doc_herefirstword = doc_here; first = false;}
|
---|
97 |
|
---|
98 | // break if this word is not the next in the phrase
|
---|
99 | if ((*phrase_here).utf8equivterms.find (doc_word) ==
|
---|
100 | (*phrase_here).utf8equivterms.end()) break;
|
---|
101 |
|
---|
102 | phrase_here++;
|
---|
103 | } while (doc_here <= doc_end && phrase_here != phrase_end);
|
---|
104 |
|
---|
105 | // see if we found a phrase
|
---|
106 | if (phrase_here == phrase_end) return true;
|
---|
107 |
|
---|
108 | doc_here = doc_herefirstword; // set the counter back
|
---|
109 | }
|
---|
110 |
|
---|
111 | return false;
|
---|
112 | }
|
---|
113 |
|
---|
114 |
|
---|
115 | // looks for the stemmed phrase in the metadata or text associated with
|
---|
116 | // an OID. This function has not been coded with all situations in mind
|
---|
117 | bool OID_phrase_search (mgsearchclass &mgsearch,
|
---|
118 | gdbmclass &gdbm,
|
---|
119 | const text_t &index,
|
---|
120 | const text_t &subcollection,
|
---|
121 | const text_t &language,
|
---|
122 | const text_t &longindex,
|
---|
123 | const text_t &collection,
|
---|
124 | const vector<termfreqclass> &phrase,
|
---|
125 | int docnum) {
|
---|
126 | // disect the long index to find out where the text should come from
|
---|
127 | text_t level, gran;
|
---|
128 | text_t::const_iterator longindex_here = longindex.begin();
|
---|
129 | text_t::const_iterator longindex_end = longindex.end();
|
---|
130 | longindex_here = getdelimitstr (longindex_here, longindex_end, ':', level);
|
---|
131 | longindex_here = getdelimitstr (longindex_here, longindex_end, ':', gran);
|
---|
132 |
|
---|
133 | if (gran.empty()) return false;
|
---|
134 |
|
---|
135 | if (gran == "text") {
|
---|
136 | char *doc = NULL;
|
---|
137 | int doclen = 0;
|
---|
138 |
|
---|
139 | // get text from mg.
|
---|
140 | if (!mgsearch.mgdocument (index, subcollection, language, collection,
|
---|
141 | docnum, doc, doclen)) return false;
|
---|
142 | return doc_phrase_search ((unsigned char *)doc, doclen, phrase);
|
---|
143 | }
|
---|
144 |
|
---|
145 | // get OID
|
---|
146 | char *metadata = NULL;
|
---|
147 | text_t::size_type metadata_len = 0;
|
---|
148 | infodbclass docnum_info;
|
---|
149 | infodbclass OID_info;
|
---|
150 |
|
---|
151 | if (!gdbm.getinfo (docnum, docnum_info)) return false;
|
---|
152 | text_t &OID = docnum_info["section"];
|
---|
153 | if (OID.empty()) return false;
|
---|
154 |
|
---|
155 | // get field
|
---|
156 | if (!gdbm.getinfo (OID, OID_info)) return false;
|
---|
157 |
|
---|
158 | metadata = (to_utf8(OID_info[gran])).getcarr(metadata_len);
|
---|
159 | bool result = doc_phrase_search ((unsigned char *)metadata, metadata_len, phrase);
|
---|
160 | delete [] metadata;
|
---|
161 |
|
---|
162 | return result;
|
---|
163 | }
|
---|