rset/rsprox.c

   1 /* This file is part of the Zebra server.
   2    Copyright (C) 1994-2011 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 int r_write(RSFD rfd, const void *buf);
  41 static void r_pos(RSFD rfd, double *current, double *total);
  42 static void r_get_terms(RSET ct, TERMID *terms, int maxterms, int *curterm);
  43
  44 static const struct rset_control control =
  45 {
  46     "prox",
  47     r_delete,
  48     r_get_terms,
  49     r_open,
  50     r_close,
  51     r_forward,
  52     r_pos,
  53     r_read,
  54     r_write,
  55 };
  56
  57 struct rset_prox_info {
  58     int ordered;
  59     int exclusion;
  60     int relation;
  61     int distance;
  62 };
  63
  64 struct rset_prox_rfd {
  65     RSFD *rfd;
  66     char **buf;  /* lookahead key buffers */
  67     char *more;  /* more in each lookahead? */
  68     TERMID *terms; /* lookahead terms */
  69     zint hits;
  70 };
  71
  72
  73 RSET rset_create_prox(NMEM nmem, struct rset_key_control *kcontrol,
  74                       int scope,
  75                       int rset_no, RSET *rset,
  76                       int ordered, int exclusion,
  77                       int relation, int distance)
  78 {
  79     RSET rnew = rset_create_base(&control, nmem, kcontrol, scope, 0,
  80                                  rset_no, rset);
  81     struct rset_prox_info *info;
  82     info = (struct rset_prox_info *) nmem_malloc(rnew->nmem,sizeof(*info));
  83     info->ordered = ordered;
  84     info->exclusion = exclusion;
  85     info->relation = relation;
  86     info->distance = distance;
  87     rnew->priv = info;
  88     return rnew;
  89 }
  90
  91 static void r_delete(RSET ct)
  92 {
  93 }
  94
  95 static RSFD r_open(RSET ct, int flag)
  96 {
  97     RSFD rfd;
  98     struct rset_prox_rfd *p;
  99     int i;
 100
 101     if (flag & RSETF_WRITE)
 102     {
 103         yaz_log(YLOG_FATAL, "prox set type is read-only");
 104         return NULL;
 105     }
 106     rfd = rfd_create_base(ct);
 107     if (rfd->priv)
 108         p = (struct rset_prox_rfd *)(rfd->priv);
 109     else
 110     {
 111         p = (struct rset_prox_rfd *) nmem_malloc(ct->nmem,sizeof(*p));
 112         rfd->priv = p;
 113         p->more = nmem_malloc(ct->nmem,sizeof(*p->more) * ct->no_children);
 114         p->buf = nmem_malloc(ct->nmem,sizeof(*p->buf) * ct->no_children);
 115         p->terms = nmem_malloc(ct->nmem,sizeof(*p->terms) * ct->no_children);
 116         for (i = 0; i < ct->no_children; i++)
 117         {
 118             p->buf[i] = nmem_malloc(ct->nmem,ct->keycontrol->key_size);
 119             p->terms[i] = 0;
 120         }
 121         p->rfd = nmem_malloc(ct->nmem,sizeof(*p->rfd) * ct->no_children);
 122     }
 123     yaz_log(YLOG_DEBUG,"rsprox (%s) open [%p] n=%d",
 124             ct->control->desc, rfd, ct->no_children);
 125
 126     for (i = 0; i < ct->no_children; i++)
 127     {
 128         p->rfd[i] = rset_open(ct->children[i], RSETF_READ);
 129         p->more[i] = rset_read(p->rfd[i], p->buf[i], &p->terms[i]);
 130     }
 131     p->hits = 0;
 132     return rfd;
 133 }
 134
 135 static void r_close(RSFD rfd)
 136 {
 137     RSET ct = rfd->rset;
 138     struct rset_prox_rfd *p = (struct rset_prox_rfd *)(rfd->priv);
 139
 140     int i;
 141     for (i = 0; i < ct->no_children; i++)
 142         rset_close(p->rfd[i]);
 143 }
 144
 145 static int r_forward(RSFD rfd, void *buf, TERMID *term, const void *untilbuf)
 146 {
 147     RSET ct = rfd->rset;
 148     struct rset_prox_info *info = (struct rset_prox_info *)(ct->priv);
 149     struct rset_prox_rfd *p = (struct rset_prox_rfd *)(rfd->priv);
 150     const struct rset_key_control *kctrl = ct->keycontrol;
 151     int cmp = 0;
 152     int i;
 153
 154     if (untilbuf)
 155     {
 156         /* it is enough to forward first one. Other will follow. */
 157         if (p->more[0] &&   /* was: cmp >=2 */
 158             ((kctrl->cmp)(untilbuf, p->buf[0]) >= rfd->rset->scope) )
 159             p->more[0] = rset_forward(p->rfd[0], p->buf[0],
 160                                       &p->terms[0], untilbuf);
 161     }
 162     if (info->ordered && info->relation <= 3 && info->exclusion == 0)
 163     {
 164         while (p->more[0])
 165         {
 166             for (i = 1; i < ct->no_children; i++)
 167             {
 168                 if (!p->more[i])
 169                 {
 170                     p->more[0] = 0; /* saves us a goto out of while loop. */
 171                     break;
 172                 }
 173                 cmp = (*kctrl->cmp)(p->buf[i], p->buf[i-1]);
 174                 if (cmp >= rfd->rset->scope)  /* not same record */
 175                 {
 176                     p->more[i-1] = rset_forward(p->rfd[i-1],
 177                                                 p->buf[i-1],
 178                                                 &p->terms[i-1],
 179                                                 p->buf[i]);
 180                     break;
 181                 }
 182                 else if (cmp > 0) /* within record and ordered */
 183                 {
 184                     zint diff = (*kctrl->getseq)(p->buf[i]) -
 185                         (*kctrl->getseq)(p->buf[i-1]);
 186                     if (info->relation == 3 && diff == info->distance)
 187                         continue;
 188                     else if (info->relation == 2 && diff <= info->distance)
 189                         continue;
 190                     else if (info->relation == 1 && diff < info->distance)
 191                         continue;
 192
 193                     p->more[i-1] = rset_read(p->rfd[i-1], p->buf[i-1],
 194                                              &p->terms[i-1]);
 195                     break;
 196                 }
 197                 else  /* within record - wrong order */
 198                 {
 199                     p->more[i] = rset_forward(p->rfd[i], p->buf[i],
 200                                               &p->terms[i], p->buf[i-1]);
 201                     break;
 202                 }
 203             }
 204             if (i == ct->no_children)
 205             {
 206                 i = ct->no_children-1;
 207                 memcpy(buf, p->buf[i], kctrl->key_size);
 208                 if (term)
 209                     *term = p->terms[i];
 210                 p->more[i] = rset_read(p->rfd[i], p->buf[i], &p->terms[i]);
 211                 p->hits++;
 212                 return 1;
 213             }
 214         }
 215     }
 216     else if (ct->no_children == 2)
 217     {
 218         while (p->more[0] && p->more[1])
 219         {
 220             int cmp = (*kctrl->cmp)(p->buf[0], p->buf[1]);
 221             if ( cmp <= - rfd->rset->scope) /* cmp<-1*/
 222                 p->more[0] = rset_forward(p->rfd[0], p->buf[0],
 223                                           &p->terms[0],p->buf[1]);
 224             else if ( cmp >= rfd->rset->scope ) /* cmp>1 */
 225                 p->more[1] = rset_forward(p->rfd[1], p->buf[1],
 226                                           &p->terms[1],p->buf[0]);
 227             else
 228             {
 229                 zint seqno[500]; /* FIXME - why 500 ?? */
 230                 int n = 0;
 231
 232                 seqno[n++] = (*kctrl->getseq)(p->buf[0]);
 233                 while ((p->more[0] = rset_read(p->rfd[0],
 234                                                p->buf[0], &p->terms[0])))
 235                 {
 236                     cmp = (*kctrl->cmp)(p->buf[0], p->buf[1]);
 237                     if (cmp <= - rfd->rset->scope || cmp >= rfd->rset->scope)
 238                         break;
 239                     if (n < 500)
 240                         seqno[n++] = (*kctrl->getseq)(p->buf[0]);
 241                 }
 242                 /* set up return buffer.. (save buf[1]) */
 243                 memcpy(buf, p->buf[1], kctrl->key_size);
 244                 if (term)
 245                     *term = p->terms[1];
 246                 while (1)
 247                 {
 248                     for (i = 0; i < n; i++)
 249                     {
 250                         zint diff = (*kctrl->getseq)(p->buf[1]) - seqno[i];
 251                         int excl = info->exclusion;
 252                         if (!info->ordered && diff < 0)
 253                             diff = -diff;
 254                         switch (info->relation)
 255                         {
 256                         case 1:      /* < */
 257                             if (diff < info->distance && diff >= 0)
 258                                 excl = !excl;
 259                             break;
 260                         case 2:      /* <= */
 261                             if (diff <= info->distance && diff >= 0)
 262                                 excl = !excl;
 263                             break;
 264                         case 3:      /* == */
 265                             if (diff == info->distance && diff >= 0)
 266                                 excl = !excl;
 267                             break;
 268                         case 4:      /* >= */
 269                             if (diff >= info->distance && diff >= 0)
 270                                 excl = !excl;
 271                             break;
 272                         case 5:      /* > */
 273                             if (diff > info->distance && diff >= 0)
 274                                 excl = !excl;
 275                             break;
 276                         case 6:      /* != */
 277                             if (diff != info->distance && diff >= 0)
 278                                 excl = !excl;
 279                             break;
 280                         }
 281                         if (excl)
 282                         {
 283                             p->more[1] = rset_read( p->rfd[1], p->buf[1],
 284                                                     &p->terms[1]);
 285                             p->hits++;
 286                             return 1;
 287                         }
 288                     }
 289                     p->more[1] = rset_read(p->rfd[1], p->buf[1], &p->terms[1]);
 290                     if (!p->more[1])
 291                         break;
 292                     cmp = (*kctrl->cmp)(buf, p->buf[1]);
 293                     if (cmp <= - rfd->rset->scope || cmp >= rfd->rset->scope)
 294                         break;
 295                 }
 296             }
 297         }
 298     }
 299     return 0;
 300 }
 301
 302
 303 static int r_read(RSFD rfd, void *buf, TERMID *term)
 304 {
 305     return r_forward(rfd, buf, term, 0);
 306 }
 307
 308 static int r_write(RSFD rfd, const void *buf)
 309 {
 310     yaz_log(YLOG_FATAL, "prox set type is read-only");
 311     return -1;
 312 }
 313
 314 static void r_pos(RSFD rfd, double *current, double *total)
 315 {
 316     RSET ct = rfd->rset;
 317     struct rset_prox_rfd *p = (struct rset_prox_rfd *)(rfd->priv);
 318     int i;
 319     double ratio = 0.0;
 320
 321     for (i = 0; i < ct->no_children; i++)
 322     {
 323         double cur, tot;
 324         rset_pos(p->rfd[i], &cur, &tot);
 325         if (tot > 0.0)
 326         {
 327             double nratio = cur / tot;
 328             if (ratio < nratio)
 329                 ratio = nratio;
 330         }
 331     }
 332     *current = (double) p->hits;
 333     if (ratio > 0.0)
 334         *total = *current/ratio;
 335     else
 336         *total = 0.0;
 337
 338     yaz_log(YLOG_DEBUG, "prox_pos: [%d] %0.1f/%0.1f= %0.4f ",
 339             i, *current, *total, ratio);
 340 }
 341
 342 static void r_get_terms(RSET ct, TERMID *terms, int maxterms, int *curterm)
 343 {
 344     int i;
 345     for (i = 0; i < ct->no_children; i++)
 346         rset_getterms(ct->children[i], terms, maxterms, curterm);
 347 }
 348
 349 /*
 350  * Local variables:
 351  * c-basic-offset: 4
 352  * c-file-style: "Stroustrup"
 353  * indent-tabs-mode: nil
 354  * End:
 355  * vim: shiftwidth=4 tabstop=8 expandtab
 356  */
 357