X-Git-Url: http://jsfdemo.indexdata.com/?a=blobdiff_plain;f=rset%2Frsbetween.c;h=0d6301d3bbc8d424e6ad9d67e85988631ae21108;hb=bb714813849a0ddc109f8b8028d4c80604a7f860;hp=4222c3081a9908300f1640c0d3751ef8a8d687aa;hpb=dbf7ead21f60be189f0d532175c92924a8eb73fd;p=idzebra-moved-to-github.git
diff --git a/rset/rsbetween.c b/rset/rsbetween.c
index 4222c30..0d6301d 100644
--- a/rset/rsbetween.c
+++ b/rset/rsbetween.c
@@ -1,10 +1,35 @@
-/*
- * Copyright (C) 1994-2002, Index Data
- * All rights reserved.
- * Heikki Levanto
- *
- * $Id: rsbetween.c,v 1.6 2002-08-01 08:53:35 adam Exp $
- */
+/* $Id: rsbetween.c,v 1.15.2.4 2004-12-17 13:43:09 heikki Exp $
+ Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002
+ Index Data Aps
+
+This file is part of the Zebra server.
+
+Zebra is free software; you can redistribute it and/or modify it under
+the terms of the GNU General Public License as published by the Free
+Software Foundation; either version 2, or (at your option) any later
+version.
+
+Zebra is distributed in the hope that it will be useful, but WITHOUT ANY
+WARRANTY; without even the implied warranty of MERCHANTABILITY or
+FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+for more details.
+
+You should have received a copy of the GNU General Public License
+along with Zebra; see the file LICENSE.zebra. If not, write to the
+Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
+02111-1307, USA.
+*/
+
+
+/* rsbetween is (mostly) used for xml searches. It returns the hits of the
+ * "middle" rset, that are in between the "left" and "right" rsets. For
+ * example "Shakespeare" in between "
" and . The thing is
+ * complicated by the inclusion of attributes (from their own rset). If attrs
+ * specified, they must match the "left" rset (start tag). "Hamlet" between
+ * "" and "". (This assumes that the attributes are
+ * indexed to the same seqno as the tags).
+
+*/
#include
#include
@@ -14,13 +39,20 @@
#include
#include
+#define RSBETWEEN_DEBUG 0
+#if RSBETWEEN_DEBUG
+#define log_level LOG_DEBUG
+#endif
+
static void *r_create_between(RSET ct, const struct rset_control *sel, void *parms);
static RSFD r_open_between (RSET ct, int flag);
static void r_close_between (RSFD rfd);
static void r_delete_between (RSET ct);
static void r_rewind_between (RSFD rfd);
-static int r_count_between (RSET ct);
-static int r_read_between (RSFD rfd, void *buf, int *term_index);
+static int r_forward_between(RSET ct, RSFD rfd, void *buf, int *term_index,
+ int (*cmpfunc)(const void *p1, const void *p2),
+ const void *untilbuf);
+static int r_read (RSFD rfd, void *buf, int *term_index);
static int r_write_between (RSFD rfd, const void *buf);
static const struct rset_control control_between =
@@ -31,14 +63,20 @@ static const struct rset_control control_between =
r_close_between,
r_delete_between,
r_rewind_between,
- r_count_between,
- r_read_between,
+ r_forward_between, /* rset_default_forward, */
+ rset_default_pos,
+ r_read,
r_write_between,
};
const struct rset_control *rset_kind_between = &control_between;
+#define WHICH_L 0
+#define WHICH_M 1
+#define WHICH_R 2
+#define WHICH_A 3
+
struct rset_between_info {
int key_size;
RSET rset_l;
@@ -70,8 +108,37 @@ struct rset_between_rfd {
int level;
struct rset_between_rfd *next;
struct rset_between_info *info;
+ /* 1.4 backport */
+ void *recbuf; /* a key that tells which record we are in */
+ void *startbuf; /* the start tag */
+ int startbufok; /* we have seen the first start tag */
+ void *attrbuf; /* the attr tag. If these two match, we have attr match */
+ int attrbufok; /* we have seen the first attr tag, can compare */
+ int depth; /* number of start-tags without end-tags */
+ int attrdepth; /* on what depth the attr matched */
+ int hits; /* number of hits returned so far */
+ /* Stuff for the multi-and read, also backported */
+ /* uses more_l/m/r/a etc from above */
+ int tailcount;
+ int tailbits[4];
+ int and_eof;
};
+#if RSBETWEEN_DEBUG
+static void log2 (struct rset_between_rfd *p, char *msg, int cmp_l, int cmp_r)
+{
+ char buf_l[32];
+ char buf_m[32];
+ char buf_r[32];
+ logf(LOG_DEBUG,"btw: %s l=%s(%d/%d) m=%s(%d) r=%s(%d/%d), lev=%d",
+ msg,
+ (*p->info->printer)(p->buf_l, buf_l), p->more_l, cmp_l,
+ (*p->info->printer)(p->buf_m, buf_m), p->more_m,
+ (*p->info->printer)(p->buf_r, buf_r), p->more_r, cmp_r,
+ p->level);
+}
+#endif
+
static void *r_create_between (RSET ct, const struct rset_control *sel,
void *parms)
{
@@ -135,8 +202,8 @@ static RSFD r_open_between (RSET ct, int flag)
if (flag & RSETF_WRITE)
{
- logf (LOG_FATAL, "between set type is read-only");
- return NULL;
+ logf (LOG_FATAL, "between set type is read-only");
+ return NULL;
}
rfd = (struct rset_between_rfd *) xmalloc (sizeof(*rfd));
rfd->next = info->rfd_list;
@@ -153,11 +220,11 @@ static RSFD r_open_between (RSET ct, int flag)
rfd->rfd_r = rset_open (info->rset_r, RSETF_READ);
rfd->more_l = rset_read (info->rset_l, rfd->rfd_l, rfd->buf_l,
- &rfd->term_index_l);
+ &rfd->term_index_l);
rfd->more_m = rset_read (info->rset_m, rfd->rfd_m, rfd->buf_m,
- &rfd->term_index_m);
+ &rfd->term_index_m);
rfd->more_r = rset_read (info->rset_r, rfd->rfd_r, rfd->buf_r,
- &rfd->term_index_r);
+ &rfd->term_index_r);
if (info->rset_attr)
{
int dummy;
@@ -166,6 +233,20 @@ static RSFD r_open_between (RSET ct, int flag)
rfd->buf_attr, &dummy);
}
rfd->level=0;
+ rfd->recbuf=xmalloc (info->key_size);
+ rfd->startbuf=xmalloc (info->key_size);
+ rfd->attrbuf=xmalloc (info->key_size);
+ rfd->startbufok=0;
+ rfd->attrbufok=0;
+ rfd->depth=0;
+ rfd->attrdepth=0;
+ rfd->hits=-1; /* signals uninitialized */
+ rfd->tailcount=0;
+ rfd->tailbits[WHICH_L]=0;
+ rfd->tailbits[WHICH_M]=0;
+ rfd->tailbits[WHICH_R]=0;
+ rfd->tailbits[WHICH_A]=0;
+ rfd->and_eof=0;
return rfd;
}
@@ -181,6 +262,9 @@ static void r_close_between (RSFD rfd)
xfree ((*rfdp)->buf_m);
xfree ((*rfdp)->buf_r);
xfree ((*rfdp)->buf_attr);
+ xfree ((*rfdp)->recbuf);
+ xfree ((*rfdp)->startbuf);
+ xfree ((*rfdp)->attrbuf);
rset_close (info->rset_l, (*rfdp)->rfd_l);
rset_close (info->rset_m, (*rfdp)->rfd_m);
rset_close (info->rset_r, (*rfdp)->rfd_r);
@@ -214,7 +298,9 @@ static void r_rewind_between (RSFD rfd)
struct rset_between_info *info = ((struct rset_between_rfd*)rfd)->info;
struct rset_between_rfd *p = (struct rset_between_rfd *) rfd;
+#if RSBETWEEN_DEBUG
logf (LOG_DEBUG, "rsbetween_rewind");
+#endif
rset_rewind (info->rset_l, p->rfd_l);
rset_rewind (info->rset_m, p->rfd_m);
rset_rewind (info->rset_r, p->rfd_r);
@@ -231,138 +317,279 @@ static void r_rewind_between (RSFD rfd)
p->level=0;
}
-static int r_count_between (RSET ct)
+
+
+static int r_forward_between(RSET ct, RSFD rfd, void *buf, int *term_index,
+ int (*cmpfunc)(const void *p1, const void *p2),
+ const void *untilbuf)
{
- return 0;
+ struct rset_between_info *info = ((struct rset_between_rfd*)rfd)->info;
+ struct rset_between_rfd *p = (struct rset_between_rfd *) rfd;
+ int rc;
+#if RSBETWEEN_DEBUG
+ log2( p, "fwd: before forward", 0,0);
+#endif
+ /* It is enough to forward the m pointer here, the read will */
+ /* naturally forward the l, m, and attr pointers */
+ if (p->more_m)
+ p->more_m=rset_forward(info->rset_m,p->rfd_m, p->buf_m,
+ &p->term_index_m, info->cmp,untilbuf);
+#if RSBETWEEN_DEBUG
+ log2( p, "fwd: after forward M", 0,0);
+#endif
+ /* rc = r_read_between(rfd, buf, term_index);*/
+ rc = r_read(rfd, buf, term_index);
+#if RSBETWEEN_DEBUG
+ log2( p, "fwd: after forward", 0,0);
+#endif
+ return rc;
}
-static void logit( struct rset_between_info *info, char *prefix, void *l, void *m, void *r)
+
+
+static int r_write_between (RSFD rfd, const void *buf)
{
- char buf_l[32];
- char buf_m[32];
- char buf_r[32];
- logf(LOG_DEBUG,"btw: %s l=%s m=%s r=%s",
- prefix,
- (*info->printer)(l, buf_l),
- (*info->printer)(m, buf_m),
- (*info->printer)(r, buf_r) );
+ logf (LOG_FATAL, "between set type is read-only");
+ return -1;
}
-static int r_read_between (RSFD rfd, void *buf, int *term_index)
-{
- struct rset_between_rfd *p = (struct rset_between_rfd *) rfd;
- struct rset_between_info *info = p->info;
- int cmp_l;
- int cmp_r;
- int attr_match;
- while (p->more_m)
+
+/* Heikki's back-port of the cleaner 1.4 algorithm */
+
+static void checkattr(struct rset_between_rfd *p)
+{
+ struct rset_between_info *info =p->info;
+ int cmp;
+ if (p->attrdepth)
+ return; /* already found one */
+ if (!info->rset_attr)
{
- logit( info, "start of loop", p->buf_l, p->buf_m, p->buf_r);
-
- /* forward L until past m, count levels, note rec boundaries */
- if (p->more_l)
- cmp_l= (*info->cmp)(p->buf_l, p->buf_m);
- else
- cmp_l=2; /* past this record */
- logf(LOG_DEBUG, "cmp_l=%d", cmp_l);
-
- while (cmp_l < 0) /* l before m */
- {
- if (cmp_l == -2)
- p->level=0; /* earlier record */
- if (cmp_l == -1)
- {
- p->level++; /* relevant start tag */
+ p->attrdepth=-1; /* matches always */
+ return;
+ }
+ if ( p->startbufok && p->attrbufok )
+ { /* have buffers to compare */
+ cmp=(*info->cmp)(p->startbuf,p->attrbuf);
+ if (0==cmp) /* and the keys match */
+ {
+ p->attrdepth=p->depth;
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level, "found attribute match at depth %d",p->attrdepth);
+#endif
+ }
+ }
+}
+
- if (!info->rset_attr)
- attr_match = 1;
- else
+/* Implements a multi-and between the l,r, and m rfds. Returns all */
+/* hits from those, provided that all three point to the same record */
+/* See rsmultiandor.c in zebra 1.4 for details */
+static int read_anded(struct rset_between_rfd *p,void *buf,
+ int *term, int *which) {
+ RSFD rfds[4];
+ RSET rsets[4];
+ void *bufs[4];
+ int *terms[4];
+ int dummyterm;
+ struct rset_between_info *info =p->info;
+ int i, mintail;
+ int n;
+ int cmp;
+
+ /* make the individual args into arrays, to match rsmultiandor */
+ rfds[WHICH_L]=p->rfd_l;
+ rfds[WHICH_M]=p->rfd_m;
+ rfds[WHICH_R]=p->rfd_r;
+ rfds[WHICH_A]=p->rfd_attr;
+ rsets[WHICH_L]=info->rset_l;
+ rsets[WHICH_M]=info->rset_m;
+ rsets[WHICH_R]=info->rset_r;
+ rsets[WHICH_A]=info->rset_attr;
+ bufs[WHICH_L]=p->buf_l;
+ bufs[WHICH_M]=p->buf_m;
+ bufs[WHICH_R]=p->buf_r;
+ bufs[WHICH_A]=p->buf_attr;
+ terms[WHICH_L]=&(p->term_index_l);
+ terms[WHICH_M]=&(p->term_index_m);
+ terms[WHICH_R]=&(p->term_index_r);
+ terms[WHICH_A]=&dummyterm;
+ dummyterm=0;
+ if (info->rset_attr)
+ n=4;
+ else
+ n=3;
+ while (1) {
+ if (p->tailcount)
+ { /* we are tailing */
+ mintail=0;
+ while ((mintailtailbits[mintail])
+ mintail++; /* first tail */
+ for (i=mintail+1;itailbits[i])
{
- int cmp_attr;
- int dummy_term;
- attr_match = 0;
- while (p->more_attr)
- {
- cmp_attr = (*info->cmp)(p->buf_attr, p->buf_l);
- if (cmp_attr == 0)
- {
- attr_match = 1;
- break;
- }
- else if (cmp_attr > 0)
- break;
- p->more_attr = rset_read (info->rset_attr, p->rfd_attr,
- p->buf_attr, &dummy_term);
- }
+ cmp=(*info->cmp)(bufs[i],bufs[mintail]);
+ if (cmp<0)
+ mintail=i;
}
}
- if (p->more_l)
+ /* return the lowest tail */
+ memcpy(buf, bufs[mintail], info->key_size);
+ *which=mintail;
+ *term=*terms[mintail];
+ /* and read that one further */
+ if (!rset_read(rsets[mintail],rfds[mintail],
+ bufs[mintail],terms[mintail]))
{
- p->more_l = rset_read (info->rset_l, p->rfd_l, p->buf_l,
- &p->term_index_l);
- cmp_l= (*info->cmp)(p->buf_l, p->buf_m);
- logit( info, "forwarded L", p->buf_l, p->buf_m, p->buf_r);
- logf(LOG_DEBUG, " cmp_l=%d", cmp_l);
+ p->and_eof=1;
+ p->tailbits[mintail]=0;
+ (p->tailcount)--;
+ return 1;
}
- else
- cmp_l=2;
- } /* forward L */
-
-
- /* forward R until past m, count levels */
- if (p->more_r)
- cmp_r= (*info->cmp)(p->buf_r, p->buf_m);
- else
- cmp_r=2;
- logf(LOG_DEBUG, "cmp_r=%d", cmp_r);
- while (cmp_r < 0) /* r before m */
- {
- /* -2, earlier record, doesn't matter */
- if (cmp_r == -1)
- p->level--; /* relevant end tag */
- if (p->more_r)
+ /* still a tail? */
+ cmp=(info->cmp)(bufs[mintail],buf);
+ if (cmp>=2)
{
- p->more_r = rset_read (info->rset_r, p->rfd_r, p->buf_r,
- &p->term_index_r);
- cmp_r= (*info->cmp)(p->buf_r, p->buf_m);
- logit( info, "forwarded R", p->buf_l, p->buf_m, p->buf_r);
- logf(LOG_DEBUG, " cmp_r=%d", cmp_r);
+ p->tailbits[mintail]=0;
+ (p->tailcount)--;
}
- else
- cmp_r=2;
- } /* forward R */
-
- if ( ( p->level <= 0 ) && ! p->more_l)
- return 0; /* no more start tags, nothing more to find */
-
- if ( attr_match && p->level > 0) /* within a tag pair (or deeper) */
- {
- memcpy (buf, p->buf_m, info->key_size);
- *term_index = p->term_index_m;
- logit( info, "Returning a hit (m)", p->buf_l, p->buf_m, p->buf_r);
- p->more_m = rset_read (info->rset_m, p->rfd_m, p->buf_m,
- &p->term_index_m);
- return 1;
- }
- else
- if ( ! p->more_l ) /* not in data, no more starts */
- return 0; /* ergo, nothing can be found. stop scanning */
-
- p->more_m = rset_read (info->rset_m, p->rfd_m, p->buf_m,
- &p->term_index_m);
- } /* while more_m */
-
- logf(LOG_DEBUG,"Exiting, no more stuff in m");
- return 0; /* no more data possible */
+ return 1;
+ }
+ else
+ { /* not tailing */
+ if (p->and_eof)
+ return 0; /* game over */
+ i=1; /* assume items[0] is highest up */
+ while (icmp)(bufs[0],bufs[i]);
+ if (cmp<=-2)
+ { /* [0] was behid, forward it and start anew */
+ if (!rset_forward(rsets[0],rfds[0], bufs[0],
+ terms[0], info->cmp, bufs[i]))
+ {
+ p->and_eof=1;
+ return 0;
+ }
+ i=0; /* all over again */
+ }
+ else if (cmp>=2)
+ {/* [0] was ahead, forward i */
+ if (!rset_forward(rsets[i],rfds[i], bufs[i],
+ terms[i], info->cmp, bufs[0]))
+ {
+ p->and_eof=1;
+ return 0;
+ }
+ } else
+ i++;
+ } /* while i */
+ /* now all are within the same records, set tails and let upper
+ * if return hits */
+ for (i=0; itailbits[i]=1;
+ p->tailcount=n;
+ }
+ }
+} /* read_anded */
-} /* r_read */
+static int r_read (RSFD rfd, void *buf, int *term)
+{
+ struct rset_between_rfd *p=(struct rset_between_rfd *)rfd;
+ struct rset_between_info *info =p->info;
+ int cmp;
+ int thisterm=0;
+ int which;
+ *term=0; /* just in case, should not be necessary */
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level,"btw: == read: term=%p",term);
+#endif
+ while ( read_anded(p,buf,&thisterm,&which) )
+ {
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level,"btw: read loop term=%p d=%d ad=%d",
+ term, p->depth, p->attrdepth);
+#endif
+ if (p->hits<0)
+ {/* first time? */
+ memcpy(p->recbuf,buf,info->key_size);
+ p->hits=0;
+ cmp=2;
+ }
+ else {
+ cmp=(*info->cmp)(buf,p->recbuf);
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level, "btw: cmp=%d",cmp);
+#endif
+ }
+ if (cmp>=2)
+ {
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level,"btw: new record");
+#endif
+ p->depth=0;
+ p->attrdepth=0;
+ memcpy(p->recbuf,buf,info->key_size);
+ }
-static int r_write_between (RSFD rfd, const void *buf)
-{
- logf (LOG_FATAL, "between set type is read-only");
- return -1;
-}
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level,"btw: which: %d", which);
+#endif
+ if (which==WHICH_L)
+ {
+ p->depth++;
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level,"btw: read start tag. d=%d",p->depth);
+#endif
+ memcpy(p->startbuf,buf,info->key_size);
+ p->startbufok=1;
+ checkattr(rfd); /* in case we already saw the attr here */
+ }
+ else if (which==WHICH_R)
+ {
+ if (p->depth == p->attrdepth)
+ p->attrdepth=0; /* ending the tag with attr match */
+ p->depth--;
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level,"btw: read end tag. d=%d ad=%d",
+ p->depth, p->attrdepth);
+#endif
+ }
+ else if (which==WHICH_A)
+ {
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level,"btw: read attr");
+#endif
+ memcpy(p->attrbuf,buf,info->key_size);
+ p->attrbufok=1;
+ checkattr(rfd); /* in case the start tag came first */
+ }
+ else
+ { /* mut be a real hit */
+ if (p->depth && p->attrdepth)
+ {
+ p->hits++;
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level,"btw: got a hit h=%d d=%d ad=%d t=%d+%d",
+ p->hits,p->depth,p->attrdepth,
+ info->rset_m->no_rset_terms,thisterm);
+#endif
+ *term= info->rset_m->no_rset_terms + thisterm;
+ return 1; /* everything else is in place already */
+ } else
+ {
+#if RSBETWEEN_DEBUG
+ yaz_log(log_level, "btw: Ignoring hit. h=%d d=%d ad=%d",
+ p->hits,p->depth,p->attrdepth);
+#endif
+ }
+ }
+ } /* while read */
+
+ return 0;
+
+} /* r_read */