1 /* $Id: rank1.c,v 1.21 2004-11-29 21:55:25 adam Exp $
2 Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002,2003,2004
5 This file is part of the Zebra server.
7 Zebra is free software; you can redistribute it and/or modify it under
8 the terms of the GNU General Public License as published by the Free
9 Software Foundation; either version 2, or (at your option) any later
12 Zebra is distributed in the hope that it will be useful, but WITHOUT ANY
13 WARRANTY; without even the implied warranty of MERCHANTABILITY or
14 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
17 You should have received a copy of the GNU General Public License
18 along with Zebra; see the file LICENSE.zebra. If not, write to the
19 Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
34 static int log_level=0;
35 static int log_initialized=0;
37 struct rank_class_info {
41 struct rank_term_info {
51 struct rank_set_info {
55 struct rank_term_info *entries;
59 static int log2_int (unsigned g)
68 * create: Creates/Initialises this rank handler. This routine is
69 * called exactly once. The routine returns the class_handle.
71 static void *create (ZebraHandle zh)
73 struct rank_class_info *ci =
74 (struct rank_class_info *) xmalloc (sizeof(*ci));
78 log_level=yaz_log_module_level("rank1");
81 yaz_log (log_level, "rank-1 create");
86 * destroy: Destroys this rank handler. This routine is called
87 * when the handler is no longer needed - i.e. when the server
88 * dies. The class_handle was previously returned by create.
90 static void destroy (struct zebra_register *reg, void *class_handle)
92 struct rank_class_info *ci = (struct rank_class_info *) class_handle;
94 yaz_log (log_level, "rank-1 destroy");
100 * begin: Prepares beginning of "real" ranking. Called once for
101 * each result set. The returned handle is a "set handle" and
102 * will be used in each of the handlers below.
104 static void *begin (struct zebra_register *reg,
105 void *class_handle, RSET rset, NMEM nmem,
106 TERMID *terms, int numterms)
108 struct rank_set_info *si =
109 (struct rank_set_info *) nmem_malloc (nmem,sizeof(*si));
112 yaz_log (log_level, "rank-1 begin");
113 si->no_entries = numterms;
114 si->no_rank_entries = 0;
116 si->entries = (struct rank_term_info *)
117 nmem_malloc (si->nmem, sizeof(*si->entries)*numterms);
118 for (i = 0; i < numterms; i++)
120 zint g = rset_count(terms[i]->rset);
121 yaz_log(log_level, "i=%d flags=%s '%s'", i,
122 terms[i]->flags, terms[i]->name );
123 if (!strncmp (terms[i]->flags, "rank,", 5))
125 const char *cp = strstr(terms[i]->flags+4, ",w=");
126 si->entries[i].rank_flag = 1;
128 si->entries[i].rank_weight = atoi (cp+3);
130 si->entries[i].rank_weight = 34;
131 yaz_log (log_level, " i=%d weight=%d g="ZINT_FORMAT, i,
132 si->entries[i].rank_weight, g);
133 (si->no_rank_entries)++;
136 si->entries[i].rank_flag = 0;
137 si->entries[i].local_occur = 0; /* FIXME */
138 si->entries[i].global_occur = g;
139 si->entries[i].global_inv = 32 - log2_int (g);
140 yaz_log (log_level, " global_inv = %d g = " ZINT_FORMAT,
141 (int) (32-log2_int (g)), g);
142 si->entries[i].term=terms[i];
143 si->entries[i].term_index=i;
144 terms[i]->rankpriv=&(si->entries[i]);
150 * end: Terminates ranking process. Called after a result set
153 static void end (struct zebra_register *reg, void *set_handle)
155 yaz_log (log_level, "rank-1 end");
156 /* no need to free anything, they are in nmems */
161 * add: Called for each word occurence in a result set. This routine
162 * should be as fast as possible. This routine should "incrementally"
165 static void add (void *set_handle, int seqno, TERMID term)
167 struct rank_set_info *si = (struct rank_set_info *) set_handle;
168 struct rank_term_info *ti;
172 yaz_log (log_level, "rank-1 add NULL term");
175 ti= (struct rank_term_info *) term->rankpriv;
177 si->last_pos = seqno;
179 yaz_log (log_level, "rank-1 add seqno=%d term=%s count=%d",
180 seqno, term->name,ti->local_occur);
184 * calc: Called for each document in a result. This handler should
185 * produce a score based on previous call(s) to the add handler. The
186 * score should be between 0 and 1000. If score cannot be obtained
187 * -1 should be returned.
189 static int calc (void *set_handle, zint sysno)
191 int i, lo, divisor, score = 0;
192 struct rank_set_info *si = (struct rank_set_info *) set_handle;
194 if (!si->no_rank_entries)
197 for (i = 0; i < si->no_entries; i++)
199 yaz_log(log_level, "calc: i=%d rank_flag=%d lo=%d",
200 i, si->entries[i].rank_flag, si->entries[i].local_occur);
201 if (si->entries[i].rank_flag && (lo = si->entries[i].local_occur))
202 score += (8+log2_int (lo)) * si->entries[i].global_inv *
203 si->entries[i].rank_weight;
205 divisor = si->no_rank_entries * (8+log2_int (si->last_pos/si->no_entries));
206 score = score / divisor;
207 yaz_log (log_level, "calc sysno=" ZINT_FORMAT " score=%d", sysno, score);
210 /* reset the counts for the next term */
211 for (i = 0; i < si->no_entries; i++)
212 si->entries[i].local_occur = 0;
217 * Pseudo-meta code with sequence of calls as they occur in a
218 * server. Handlers are prefixed by --:
234 static struct rank_control rank_control = {
244 struct rank_control *rank1_class = &rank_control;