X-Git-Url: http://jsfdemo.indexdata.com/?a=blobdiff_plain;f=index%2Fzvrank.c;h=a99327605866525ce86745dc58e8ee263f3616ab;hb=67700507285e0823f80e097fa0b1aae5968451bb;hp=622278ef49eb0c78243ad25cc487b0f491689ef7;hpb=00c427dace60183f4ba7bacdcdf004ac419527ce;p=idzebra-moved-to-github.git

diff --git a/index/zvrank.c b/index/zvrank.c
index 622278e..a993276 100644
--- a/index/zvrank.c
+++ b/index/zvrank.c
@@ -1,4 +1,4 @@
-/* $Id: zvrank.c,v 1.6 2004-06-08 13:27:48 marc Exp $
+/* $Id: zvrank.c,v 1.12 2004-10-28 10:37:15 heikki Exp $
    Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002,2003
    Index Data Aps
 
@@ -40,6 +40,7 @@ fernuni-hagen.de>
 ** "ntc-atn", "atc-atn", etc.
 */
 
+
 #include <math.h>  /* for log */
 
 #include <stdio.h>
@@ -61,8 +62,7 @@ static double blog(double x) {
 
 /* structures */
 
-struct rank_class_info { /* now we need this */
-    int dummy;
+struct rank_class_info {
     char rscheme[8];    /* name of weighting scheme */
 };
 
@@ -76,6 +76,7 @@ struct rs_info {      /* for result set */
     char rscheme[8];    /* name of weighting scheme */
     /**/
     int veclen;
+    NMEM nmem;
     void (*d_tf_fct)(void *, void *);   /* doc term frequency function */
     void (*d_idf_fct)(void *, void *);  /* doc idf function */
     void (*d_norm_fct)(void *, void *); /* doc normalization function */
@@ -131,7 +132,7 @@ struct ts_info {       /* term info */
     char *name;
     int *id;
     /**/
-    int gocc;
+    zint gocc;
     int locc;
     double tf;
     double idf;
@@ -267,7 +268,7 @@ static void idf_none(void *rsi, void *dsi) {
 static void idf_tfidf(void *rsi, void *dsi) {
     RS rs=(RS)rsi;
     DS ds=(DS)dsi;
-    int num_docs, gocc;
+    zint num_docs, gocc;
     int i, veclen;
     double idf;
     /* normal tfidf weight */
@@ -278,7 +279,7 @@ static void idf_tfidf(void *rsi, void *dsi) {
         if (gocc==0) 
             idf=0.0; 
         else
-            idf=blog(num_docs/gocc);
+            idf=blog((double) (num_docs/gocc));
         ds->terms[i].idf=idf;
     }
     return;
@@ -287,7 +288,7 @@ static void idf_tfidf(void *rsi, void *dsi) {
 static void idf_prob(void *rsi, void *dsi) {
     RS rs=(RS)rsi;
     DS ds=(DS)dsi;
-    int num_docs, gocc;
+    zint num_docs, gocc;
     int i, veclen;
     double idf;
     /* probabilistic formulation */
@@ -298,7 +299,7 @@ static void idf_prob(void *rsi, void *dsi) {
         if (gocc==0)
             idf=0.0; 
         else
-            idf=blog((num_docs-gocc)/gocc);
+            idf=blog((double) ((num_docs-gocc)/gocc));
         ds->terms[i].idf=idf;
     }
     return;
@@ -326,7 +327,7 @@ static void idf_freq(void *rsi, void *dsi) {
 static void idf_squared(void *rsi, void *dsi) {
     RS rs=(RS)rsi;
     DS ds=(DS)dsi;
-    int num_docs, gocc;
+    zint num_docs, gocc;
     int i, veclen;
     double idf;
     /* idf ^ 2 */
@@ -643,6 +644,7 @@ static void zv_init(RS rs, const char *rscheme) {
     rs->db_terms=500000;  /* assign correct value here (for debugging) */
     rs->db_f_max=50;      /* assign correct value here */
     rs->db_f_max_str="a"; /* assign correct value here (for debugging) */
+    /* FIXME - get those values from somewhere */
     zv_init_scheme(rs, rscheme);
     return;
 }
@@ -660,9 +662,10 @@ static void *zv_create (ZebraHandle zh) {
     struct rank_class_info *ci = (struct rank_class_info *)
         xmalloc (sizeof(*ci));
     yaz_log(LOG_DEBUG, "zv_create");
-    wscheme=res_get(res, "zvrank.weighting-scheme");
-    for (i=0; (i < strlen(wscheme)) && (i < 8); i++) 
+    wscheme=res_get_def(res, "zvrank.weighting-scheme", "");
+    for (i=0; wscheme[i] && i < 8; i++) 
         ci->rscheme[i]=wscheme[i];
+    ci->rscheme[i] = '\0';
     return ci;
 }
 
@@ -683,34 +686,42 @@ static void zv_destroy (struct zebra_register *reg, void *class_handle) {
  *  each result set. The returned handle is a "set handle" and
  *  will be used in each of the handlers below.
  */
-static void *zv_begin(struct zebra_register *reg, void *class_handle, RSET rset)
+static void *zv_begin(struct zebra_register *reg, void *class_handle, 
+                      RSET rset, NMEM nmem, TERMID *terms, int numterms)
 {
-    struct rs_info *rs=(struct rs_info *)xmalloc(sizeof(*rs));
+    struct rs_info *rs=(struct rs_info *)nmem_malloc(nmem,sizeof(*rs));
     struct rank_class_info *ci=(struct rank_class_info *)class_handle;
     int i;
-    int veclen, gocc;
+    int veclen;
+    int *ip;
+    zint gocc;
     /**/
     yaz_log(LOG_DEBUG, "zv_begin");
-    veclen=rset->no_rset_terms; /* smaller vector here */
+    veclen= numterms;
     zv_init(rs, ci->rscheme);
+    rs->nmem=nmem;
     rs->veclen=veclen;
     prn_rs(rs);
   
-    rs->qdoc=(struct ds_info *)xmalloc(sizeof(*rs->qdoc));
-    rs->qdoc->terms=(struct ts_info *)xmalloc(sizeof(*rs->qdoc->terms)*rs->veclen);
+    rs->qdoc=(struct ds_info *)nmem_malloc(nmem,sizeof(*rs->qdoc));
+    rs->qdoc->terms=(struct ts_info *)nmem_malloc(nmem,
+                                sizeof(*rs->qdoc->terms)*rs->veclen);
     rs->qdoc->veclen=veclen;
     rs->qdoc->d_f_max=1; /* no duplicates */ 
     rs->qdoc->d_f_max_str=""; 
 
-    rs->rdoc=(struct ds_info *)xmalloc(sizeof(*rs->rdoc));
-    rs->rdoc->terms=(struct ts_info *)xmalloc(sizeof(*rs->rdoc->terms)*rs->veclen);
+    rs->rdoc=(struct ds_info *)nmem_malloc(nmem,sizeof(*rs->rdoc));
+    rs->rdoc->terms=(struct ts_info *)nmem_malloc(nmem,
+                         sizeof(*rs->rdoc->terms)*rs->veclen);
     rs->rdoc->veclen=veclen;
     rs->rdoc->d_f_max=10; /* just a guess */
     rs->rdoc->d_f_max_str=""; 
     /* yaz_log(LOG_DEBUG, "zv_begin_init"); */
     for (i = 0; i < rs->veclen; i++)
     {
-        gocc=rset->rset_terms[i]->nn;
+        gocc= rset_count(terms[i]->rset);
+        terms[i]->rankpriv=ip=nmem_malloc(nmem, sizeof(int));
+        *ip=i; /* save the index for add() */
         /* yaz_log(LOG_DEBUG, "zv_begin_init i=%d gocc=%d", i, gocc); */
         rs->qdoc->terms[i].gocc=gocc;
         rs->qdoc->terms[i].locc=1;  /* assume query has no duplicate terms */
@@ -729,13 +740,8 @@ static void *zv_begin(struct zebra_register *reg, void *class_handle, RSET rset)
  */
 static void zv_end (struct zebra_register *reg, void *rsi)
 {
-    RS rs=(RS)rsi;
     yaz_log(LOG_DEBUG, "zv_end");
-    xfree(rs->qdoc->terms);
-    xfree(rs->rdoc->terms);
-    xfree(rs->qdoc);
-    xfree(rs->rdoc);
-    xfree(rs);
+    /* they all are nmem'd */
     return;
 }
 
@@ -744,10 +750,13 @@ static void zv_end (struct zebra_register *reg, void *rsi)
  *  should be as fast as possible. This routine should "incrementally"
  *  update the score.
  */
-static void zv_add (void *rsi, int seqno, int i) {
+static void zv_add (void *rsi, int seqno, TERMID term) {
     RS rs=(RS)rsi;
-    /* yaz_log(LOG_DEBUG, "zvrank zv_add seqno=%d term_index=%d", seqno, term_index);*/
+    int *ip = term->rankpriv;
+    int i=*ip;
     rs->rdoc->terms[i].locc++;
+    yaz_log(LOG_DEBUG, "zvrank zv_add seqno=%d '%s' term_index=%d cnt=%d", 
+             seqno, term->name, i, rs->rdoc->terms[i].locc );
 }
 
 /*
@@ -756,7 +765,7 @@ static void zv_add (void *rsi, int seqno, int i) {
  *  score should be between 0 and 1000. If score cannot be obtained
  *  -1 should be returned.
  */
-static int zv_calc (void *rsi, int sysno)
+static int zv_calc (void *rsi, zint sysno)
 {
     int i, veclen; 
     int score=0;
@@ -774,11 +783,12 @@ static int zv_calc (void *rsi, int sysno)
         (*rs->d_norm_fct)(rs, rs->rdoc);
         dscore=rs->sim_fct(rs->qdoc, rs->rdoc);
     }
-    score = dscore * 1000;
-    yaz_log (LOG_LOG, "sysno=%d score=%d", sysno, score);
+    score = (int) (dscore * 1000 +.5);
+    yaz_log (LOG_DEBUG, "zv_calc: sysno=" ZINT_FORMAT " score=%d", 
+            sysno, score);
     if (score > 1000) /* should not happen */
         score = 1000;
-    return score;
+    return (int) score;
 }
 
 /*