rset/rsprox.c

   1 /* This file is part of the Zebra server.
   2    Copyright (C) Index Data
   3
   4 Zebra is free software; you can redistribute it and/or modify it under
   5 the terms of the GNU General Public License as published by the Free
   6 Software Foundation; either version 2, or (at your option) any later
   7 version.
   8
   9 Zebra is distributed in the hope that it will be useful, but WITHOUT ANY
  10 WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11 FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12 for more details.
  13
  14 You should have received a copy of the GNU General Public License
  15 along with this program; if not, write to the Free Software
  16 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  17
  18 */
  19
  20 #if HAVE_CONFIG_H
  21 #include <config.h>
  22 #endif
  23 #include <stdio.h>
  24 #include <stdlib.h>
  25 #include <string.h>
  26 #include <assert.h>
  27
  28 #include <idzebra/util.h>
  29 #include <rset.h>
  30
  31 #ifndef RSET_DEBUG
  32 #define RSET_DEBUG 0
  33 #endif
  34
  35 static RSFD r_open(RSET ct, int flag);
  36 static void r_close(RSFD rfd);
  37 static void r_delete(RSET ct);
  38 static int r_forward(RSFD rfd, void *buf, TERMID *term, const void *untilbuf);
  39 static int r_read(RSFD rfd, void *buf, TERMID *term);
  40 static void r_pos(RSFD rfd, double *current, double *total);
  41 static void r_get_terms(RSET ct, TERMID *terms, int maxterms, int *curterm);
  42
  43 static const struct rset_control control =
  44 {
  45     "prox",
  46     r_delete,
  47     r_get_terms,
  48     r_open,
  49     r_close,
  50     r_forward,
  51     r_pos,
  52     r_read,
  53     rset_no_write,
  54 };
  55
  56 struct rset_prox_info {
  57     int ordered;
  58     int exclusion;
  59     int relation;
  60     int distance;
  61 };
  62
  63 struct rset_prox_rfd {
  64     RSFD *rfd;
  65     char **buf;  /* lookahead key buffers */
  66     char *more;  /* more in each lookahead? */
  67     TERMID *terms; /* lookahead terms */
  68     zint hits;
  69 };
  70
  71
  72 RSET rset_create_prox(NMEM nmem, struct rset_key_control *kcontrol,
  73                       int scope,
  74                       int rset_no, RSET *rset,
  75                       int ordered, int exclusion,
  76                       int relation, int distance)
  77 {
  78     RSET rnew = rset_create_base(&control, nmem, kcontrol, scope, 0,
  79                                  rset_no, rset);
  80     struct rset_prox_info *info;
  81     info = (struct rset_prox_info *) nmem_malloc(rnew->nmem,sizeof(*info));
  82     info->ordered = ordered;
  83     info->exclusion = exclusion;
  84     info->relation = relation;
  85     info->distance = distance;
  86     rnew->priv = info;
  87     return rnew;
  88 }
  89
  90 static void r_delete(RSET ct)
  91 {
  92 }
  93
  94 static RSFD r_open(RSET ct, int flag)
  95 {
  96     RSFD rfd;
  97     struct rset_prox_rfd *p;
  98     int i;
  99
 100     if (flag & RSETF_WRITE)
 101     {
 102         yaz_log(YLOG_FATAL, "prox set type is read-only");
 103         return NULL;
 104     }
 105     rfd = rfd_create_base(ct);
 106     if (rfd->priv)
 107         p = (struct rset_prox_rfd *)(rfd->priv);
 108     else
 109     {
 110         p = (struct rset_prox_rfd *) nmem_malloc(ct->nmem,sizeof(*p));
 111         rfd->priv = p;
 112         p->more = nmem_malloc(ct->nmem,sizeof(*p->more) * ct->no_children);
 113         p->buf = nmem_malloc(ct->nmem,sizeof(*p->buf) * ct->no_children);
 114         p->terms = nmem_malloc(ct->nmem,sizeof(*p->terms) * ct->no_children);
 115         for (i = 0; i < ct->no_children; i++)
 116         {
 117             p->buf[i] = nmem_malloc(ct->nmem,ct->keycontrol->key_size);
 118             p->terms[i] = 0;
 119         }
 120         p->rfd = nmem_malloc(ct->nmem,sizeof(*p->rfd) * ct->no_children);
 121     }
 122     yaz_log(YLOG_DEBUG,"rsprox (%s) open [%p] n=%d",
 123             ct->control->desc, rfd, ct->no_children);
 124
 125     for (i = 0; i < ct->no_children; i++)
 126     {
 127         p->rfd[i] = rset_open(ct->children[i], RSETF_READ);
 128         p->more[i] = rset_read(p->rfd[i], p->buf[i], &p->terms[i]);
 129     }
 130     p->hits = 0;
 131     return rfd;
 132 }
 133
 134 static void r_close(RSFD rfd)
 135 {
 136     RSET ct = rfd->rset;
 137     struct rset_prox_rfd *p = (struct rset_prox_rfd *)(rfd->priv);
 138
 139     int i;
 140     for (i = 0; i < ct->no_children; i++)
 141         rset_close(p->rfd[i]);
 142 }
 143
 144 static int r_forward(RSFD rfd, void *buf, TERMID *term, const void *untilbuf)
 145 {
 146     RSET ct = rfd->rset;
 147     struct rset_prox_info *info = (struct rset_prox_info *)(ct->priv);
 148     struct rset_prox_rfd *p = (struct rset_prox_rfd *)(rfd->priv);
 149     const struct rset_key_control *kctrl = ct->keycontrol;
 150     int cmp = 0;
 151     int i;
 152
 153     if (untilbuf)
 154     {
 155         /* it is enough to forward first one. Other will follow. */
 156         if (p->more[0] &&   /* was: cmp >=2 */
 157             ((kctrl->cmp)(untilbuf, p->buf[0]) >= rfd->rset->scope) )
 158             p->more[0] = rset_forward(p->rfd[0], p->buf[0],
 159                                       &p->terms[0], untilbuf);
 160     }
 161     if (info->ordered && info->relation <= 3 && info->exclusion == 0)
 162     {
 163         while (p->more[0])
 164         {
 165             for (i = 1; i < ct->no_children; i++)
 166             {
 167                 if (!p->more[i])
 168                 {
 169                     p->more[0] = 0; /* saves us a goto out of while loop. */
 170                     break;
 171                 }
 172                 cmp = (*kctrl->cmp)(p->buf[i], p->buf[i-1]);
 173                 if (cmp >= rfd->rset->scope)  /* not same record */
 174                 {
 175                     p->more[i-1] = rset_forward(p->rfd[i-1],
 176                                                 p->buf[i-1],
 177                                                 &p->terms[i-1],
 178                                                 p->buf[i]);
 179                     break;
 180                 }
 181                 else if (cmp > 0) /* within record and ordered */
 182                 {
 183                     zint diff = (*kctrl->getseq)(p->buf[i]) -
 184                         (*kctrl->getseq)(p->buf[i-1]);
 185                     if (info->relation == 3 && diff == info->distance)
 186                         continue;
 187                     else if (info->relation == 2 && diff <= info->distance)
 188                         continue;
 189                     else if (info->relation == 1 && diff < info->distance)
 190                         continue;
 191
 192                     p->more[i-1] = rset_read(p->rfd[i-1], p->buf[i-1],
 193                                              &p->terms[i-1]);
 194                     break;
 195                 }
 196                 else  /* within record - wrong order */
 197                 {
 198                     p->more[i] = rset_forward(p->rfd[i], p->buf[i],
 199                                               &p->terms[i], p->buf[i-1]);
 200                     break;
 201                 }
 202             }
 203             if (i == ct->no_children)
 204             {
 205                 i = ct->no_children-1;
 206                 memcpy(buf, p->buf[i], kctrl->key_size);
 207                 if (term)
 208                     *term = p->terms[i];
 209                 p->more[i] = rset_read(p->rfd[i], p->buf[i], &p->terms[i]);
 210                 p->hits++;
 211                 return 1;
 212             }
 213         }
 214     }
 215     else if (ct->no_children == 2)
 216     {
 217         while (p->more[0] && p->more[1])
 218         {
 219             int cmp = (*kctrl->cmp)(p->buf[0], p->buf[1]);
 220             if ( cmp <= - rfd->rset->scope) /* cmp<-1*/
 221                 p->more[0] = rset_forward(p->rfd[0], p->buf[0],
 222                                           &p->terms[0],p->buf[1]);
 223             else if ( cmp >= rfd->rset->scope ) /* cmp>1 */
 224                 p->more[1] = rset_forward(p->rfd[1], p->buf[1],
 225                                           &p->terms[1],p->buf[0]);
 226             else
 227             {
 228                 zint seqno[500]; /* FIXME - why 500 ?? */
 229                 int n = 0;
 230
 231                 seqno[n++] = (*kctrl->getseq)(p->buf[0]);
 232                 while ((p->more[0] = rset_read(p->rfd[0],
 233                                                p->buf[0], &p->terms[0])))
 234                 {
 235                     cmp = (*kctrl->cmp)(p->buf[0], p->buf[1]);
 236                     if (cmp <= - rfd->rset->scope || cmp >= rfd->rset->scope)
 237                         break;
 238                     if (n < 500)
 239                         seqno[n++] = (*kctrl->getseq)(p->buf[0]);
 240                 }
 241                 /* set up return buffer.. (save buf[1]) */
 242                 memcpy(buf, p->buf[1], kctrl->key_size);
 243                 if (term)
 244                     *term = p->terms[1];
 245                 while (1)
 246                 {
 247                     for (i = 0; i < n; i++)
 248                     {
 249                         zint diff = (*kctrl->getseq)(p->buf[1]) - seqno[i];
 250                         int excl = info->exclusion;
 251                         if (!info->ordered && diff < 0)
 252                             diff = -diff;
 253                         switch (info->relation)
 254                         {
 255                         case 1:      /* < */
 256                             if (diff < info->distance && diff >= 0)
 257                                 excl = !excl;
 258                             break;
 259                         case 2:      /* <= */
 260                             if (diff <= info->distance && diff >= 0)
 261                                 excl = !excl;
 262                             break;
 263                         case 3:      /* == */
 264                             if (diff == info->distance && diff >= 0)
 265                                 excl = !excl;
 266                             break;
 267                         case 4:      /* >= */
 268                             if (diff >= info->distance && diff >= 0)
 269                                 excl = !excl;
 270                             break;
 271                         case 5:      /* > */
 272                             if (diff > info->distance && diff >= 0)
 273                                 excl = !excl;
 274                             break;
 275                         case 6:      /* != */
 276                             if (diff != info->distance && diff >= 0)
 277                                 excl = !excl;
 278                             break;
 279                         }
 280                         if (excl)
 281                         {
 282                             p->more[1] = rset_read( p->rfd[1], p->buf[1],
 283                                                     &p->terms[1]);
 284                             p->hits++;
 285                             return 1;
 286                         }
 287                     }
 288                     p->more[1] = rset_read(p->rfd[1], p->buf[1], &p->terms[1]);
 289                     if (!p->more[1])
 290                         break;
 291                     cmp = (*kctrl->cmp)(buf, p->buf[1]);
 292                     if (cmp <= - rfd->rset->scope || cmp >= rfd->rset->scope)
 293                         break;
 294                 }
 295             }
 296         }
 297     }
 298     return 0;
 299 }
 300
 301
 302 static int r_read(RSFD rfd, void *buf, TERMID *term)
 303 {
 304     return r_forward(rfd, buf, term, 0);
 305 }
 306
 307 static void r_pos(RSFD rfd, double *current, double *total)
 308 {
 309     RSET ct = rfd->rset;
 310     struct rset_prox_rfd *p = (struct rset_prox_rfd *)(rfd->priv);
 311     int i;
 312     double ratio = 0.0;
 313
 314     for (i = 0; i < ct->no_children; i++)
 315     {
 316         double cur, tot;
 317         rset_pos(p->rfd[i], &cur, &tot);
 318         if (tot > 0.0)
 319         {
 320             double nratio = cur / tot;
 321             if (ratio < nratio)
 322                 ratio = nratio;
 323         }
 324     }
 325     *current = (double) p->hits;
 326     if (ratio > 0.0)
 327         *total = *current/ratio;
 328     else
 329         *total = 0.0;
 330
 331     yaz_log(YLOG_DEBUG, "prox_pos: [%d] %0.1f/%0.1f= %0.4f ",
 332             i, *current, *total, ratio);
 333 }
 334
 335 static void r_get_terms(RSET ct, TERMID *terms, int maxterms, int *curterm)
 336 {
 337     int i;
 338     for (i = 0; i < ct->no_children; i++)
 339         rset_getterms(ct->children[i], terms, maxterms, curterm);
 340 }
 341
 342 /*
 343  * Local variables:
 344  * c-basic-offset: 4
 345  * c-file-style: "Stroustrup"
 346  * indent-tabs-mode: nil
 347  * End:
 348  * vim: shiftwidth=4 tabstop=8 expandtab
 349  */
 350