| 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 328 1999-07-01 04:01:47Z rjmcnab $
|
|---|
| 9 | *
|
|---|
| 10 | *********************************************************************/
|
|---|
| 11 |
|
|---|
| 12 | /*
|
|---|
| 13 | $Log$
|
|---|
| 14 | Revision 1.1 1999/07/01 04:01:46 rjmcnab
|
|---|
| 15 | Initial revision.
|
|---|
| 16 |
|
|---|
| 17 | */
|
|---|
| 18 |
|
|---|
| 19 |
|
|---|
| 20 | #include "phrasesearch.h"
|
|---|
| 21 | #include "gsdlunicode.h"
|
|---|
| 22 |
|
|---|
| 23 | inline unsigned char *parse_nonword_word (unsigned char *here, unsigned char *end,
|
|---|
| 24 | text_t &word) {
|
|---|
| 25 | int c_len = 0;
|
|---|
| 26 | unsigned short c = 0;
|
|---|
| 27 |
|
|---|
| 28 | word.clear();
|
|---|
| 29 |
|
|---|
| 30 | // parse non word
|
|---|
| 31 | while (here <= end) {
|
|---|
| 32 | c_len = parse_utf8_char (here, end, &c);
|
|---|
| 33 | here += c_len;
|
|---|
| 34 | if (is_unicode_letdig(c)) {
|
|---|
| 35 | word.push_back(c);
|
|---|
| 36 | break;
|
|---|
| 37 | }
|
|---|
| 38 | }
|
|---|
| 39 |
|
|---|
| 40 | // parse word
|
|---|
| 41 | while (here <= end) {
|
|---|
| 42 | c_len = parse_utf8_char (here, end, &c);
|
|---|
| 43 | here += c_len; // it is ok to skip a nonword character
|
|---|
| 44 | if (!is_unicode_letdig(c)) break;
|
|---|
| 45 | word.push_back(c);
|
|---|
| 46 | }
|
|---|
| 47 |
|
|---|
| 48 | return here;
|
|---|
| 49 | }
|
|---|
| 50 |
|
|---|
| 51 |
|
|---|
| 52 | bool doc_phrase_search (unsigned char *doc, int doclen,
|
|---|
| 53 | const vector<termfreqclass> &phrase) {
|
|---|
| 54 | // note: this uses the most braindead search routine :-)
|
|---|
| 55 | // however its not so bad as there shouldn't be many partial
|
|---|
| 56 | // matches
|
|---|
| 57 |
|
|---|
| 58 | // a null phrase matches anything
|
|---|
| 59 | if (phrase.empty()) return true;
|
|---|
| 60 |
|
|---|
| 61 | // if there is nothing then there can't be a match
|
|---|
| 62 | if (doc == NULL || doclen == 0) return false;
|
|---|
| 63 |
|
|---|
| 64 | text_t doc_word;
|
|---|
| 65 | doc_word.reserve (16);
|
|---|
| 66 |
|
|---|
| 67 | bool first = true;
|
|---|
| 68 |
|
|---|
| 69 | unsigned char *doc_here = doc;
|
|---|
| 70 | unsigned char *doc_herefirstword = doc;
|
|---|
| 71 | unsigned char *doc_end = doc+doclen-1; // unitool conventions :-/
|
|---|
| 72 |
|
|---|
| 73 | while (doc_here <= doc_end) {
|
|---|
| 74 | first = true;
|
|---|
| 75 |
|
|---|
| 76 | // there will be at least one member of phrase (see above)
|
|---|
| 77 | vector<termfreqclass>::const_iterator phrase_here = phrase.begin();
|
|---|
| 78 | vector<termfreqclass>::const_iterator phrase_end = phrase.end();
|
|---|
| 79 | do {
|
|---|
| 80 | // get the next non-word ... and ignore it, then get the next word
|
|---|
| 81 | doc_here = parse_nonword_word (doc_here, doc_end, doc_word);
|
|---|
| 82 | if (first) {doc_herefirstword = doc_here; first = false;}
|
|---|
| 83 |
|
|---|
| 84 | // break if this word is not the next in the phrase
|
|---|
| 85 | if ((*phrase_here).utf8equivterms.find (doc_word) ==
|
|---|
| 86 | (*phrase_here).utf8equivterms.end()) break;
|
|---|
| 87 |
|
|---|
| 88 | phrase_here++;
|
|---|
| 89 | } while (doc_here <= doc_end && phrase_here != phrase_end);
|
|---|
| 90 |
|
|---|
| 91 | // see if we found a phrase
|
|---|
| 92 | if (phrase_here == phrase_end) return true;
|
|---|
| 93 |
|
|---|
| 94 | doc_here = doc_herefirstword; // set the counter back
|
|---|
| 95 | }
|
|---|
| 96 |
|
|---|
| 97 | return false;
|
|---|
| 98 | }
|
|---|
| 99 |
|
|---|
| 100 |
|
|---|
| 101 | // looks for the stemmed phrase in the metadata or text associated with
|
|---|
| 102 | // an OID. This function has not been coded with all situations in mind
|
|---|
| 103 | bool OID_phrase_search (mgsearchclass &mgsearch,
|
|---|
| 104 | gdbmclass &gdbm,
|
|---|
| 105 | const text_t &index,
|
|---|
| 106 | const text_t &longindex,
|
|---|
| 107 | const text_t &collection,
|
|---|
| 108 | const vector<termfreqclass> &phrase,
|
|---|
| 109 | int docnum) {
|
|---|
| 110 | // disect the long index to find out where the text should come from
|
|---|
| 111 | text_t level, gran;
|
|---|
| 112 | text_t::const_iterator longindex_here = longindex.begin();
|
|---|
| 113 | text_t::const_iterator longindex_end = longindex.end();
|
|---|
| 114 | longindex_here = getdelimitstr (longindex_here, longindex_end, ':', level);
|
|---|
| 115 | longindex_here = getdelimitstr (longindex_here, longindex_end, ':', gran);
|
|---|
| 116 |
|
|---|
| 117 | if (gran.empty()) return false;
|
|---|
| 118 |
|
|---|
| 119 | if (gran == "text") {
|
|---|
| 120 | char *doc = NULL;
|
|---|
| 121 | int doclen = 0;
|
|---|
| 122 |
|
|---|
| 123 | // get text from mg.
|
|---|
| 124 | if (!mgsearch.mgdocument (index, collection, docnum, doc, doclen)) return false;
|
|---|
| 125 | return doc_phrase_search ((unsigned char *)doc, doclen, phrase);
|
|---|
| 126 | }
|
|---|
| 127 |
|
|---|
| 128 | // get OID
|
|---|
| 129 | char *metadata = NULL;
|
|---|
| 130 | text_t::size_type metadata_len = 0;
|
|---|
| 131 | infodbclass docnum_info;
|
|---|
| 132 | infodbclass OID_info;
|
|---|
| 133 |
|
|---|
| 134 | if (!gdbm.getinfo (docnum, docnum_info)) return false;
|
|---|
| 135 | text_t &OID = docnum_info["section"];
|
|---|
| 136 | if (OID.empty()) return false;
|
|---|
| 137 |
|
|---|
| 138 | // get field
|
|---|
| 139 | if (!gdbm.getinfo (OID, OID_info)) return false;
|
|---|
| 140 |
|
|---|
| 141 | metadata = (to_utf8(OID_info[gran])).getcarr(metadata_len);
|
|---|
| 142 | bool result = doc_phrase_search ((unsigned char *)metadata, metadata_len, phrase);
|
|---|
| 143 | delete [] metadata;
|
|---|
| 144 |
|
|---|
| 145 | return result;
|
|---|
| 146 | }
|
|---|
