relevance-h now calculates some kind of round-robin score

diff --git a/heikki/README-HEIKKI b/heikki/README-HEIKKI
index 77977e8..a6ca02d 100644 (file)
--- a/heikki/README-HEIKKI
+++ b/heikki/README-HEIKKI
@@ -18,3 +18,23 @@ Wed 20-Nov-2013 Make a script that tests ranking against yaz-zserver
 (as that is the default config). Mostly to have a script to build on later.
 
 Thu 21-Nov-2013. Start my own complete config
+
+Fri 22-Nov-2013. Adam defined a new sort type, relevance_h, and put it place
+in the code. Now I have a place to implement my stuff. Relevant places:
+ pazpar2_config.c:1020  - minor
+ session.c:1318 - call relevance_prepare_read also for my type
+ reclists.c:104 - parse params
+ reclists.c:166 - compare function (for quicksort)
+ relevance.c:417 - calculate score
+         (same function as for relevance, but with extra arg for type)
+
+The compare function compares positions, when sorting by Metadata_sortkey_position
+This loops through the records (in the cluster) and finds the smallest rec->pos
+and then compares those.
+
+Next: See if I can implement a round robin.
+ - clients.h declares int clients_count(void)
+ - rec->client is a pointer to the client, but we don't have an ordinal from that
+   - keep an array of structs with the pointer, and locate the client number that way
+ - robin-score = pos * n_clients + client_num
+

diff --git a/heikki/service.xml b/heikki/service.xml
index 5651f30..088ba98 100644 (file)
--- a/heikki/service.xml
+++ b/heikki/service.xml
@@ -3,6 +3,7 @@
   <timeout session="60" z3950_operation="30" z3950_session="180"/>
   
   <settings src="ztest.xml"/>
+  <settings src="opencontent.xml"/>
   
   <icu_chain id="relevance" locale="en">
     <transform rule="[:Control:] Any-Remove"/>

diff --git a/heikki/test1.sh b/heikki/test1.sh
index 0814bf0..a742ce0 100755 (executable)
--- a/heikki/test1.sh
+++ b/heikki/test1.sh
@@ -12,6 +12,7 @@ then
   exit
 fi
 
+URL="http://localhost:9017/"
 CFG="test1.cfg"
 
 PZ="../src/pazpar2"
@@ -26,7 +27,7 @@ rm -f *.out *.log
 $PZ -f $CFG  -l pz2.log -p $PIDFILE &
 sleep 0.2 # make sure it has time to start
 echo "Init"
-curl -s "http://localhost:9017/?command=init" > init.out
+curl -s "$URL?command=init" > init.out
 SESSION=`xml_grep --text_only "//session" init.out `
 # cat init.out; echo
 echo "Got session $SESSION"
@@ -38,14 +39,35 @@ QRY="query=computer"
 #SEARCH="command=search$SES&$QRY"
 SEARCH="command=search$SES&$QRY&sort=relevance"
 echo $SEARCH
-curl -s "http://localhost:9017/?$SEARCH" > search.out
+curl -s "$URL?$SEARCH" > search.out
 cat search.out | grep search
 echo
 
-SHOW="command=show$SES&sort=relevance"
+STAT="command=stat&$SES"
+echo "" > stat.out
+LOOPING=1
+while [ $LOOPING = 1 ]
+do
+  sleep 0.5
+  curl -s "$URL?$STAT" > stat.out
+  ACT=`xml_grep --text_only "//activeclients" stat.out`
+  HIT=`xml_grep --text_only "//hits" stat.out`
+  REC=`xml_grep --text_only "//records" stat.out`
+  echo "$ACT $HIT $REC"
+  if grep -q "<activeclients>0</activeclients>" stat.out
+  then
+    LOOPING=0
+  fi
+  echo >> stats.out
+  cat stat.out >> stats.out
+done
+
+
+SHOW="command=show$SES&sort=relevance_h&start=0&num=1000"
 echo $SHOW
 curl -s "http://localhost:9017/?$SHOW" > show.out
-grep "relevance" show.out | grep += | grep -v "(0)"
+#grep "relevance" show.out | grep += | grep -v "(0)"
+grep "round-robin" show.out
 echo
 
 echo "All done"

diff --git a/src/relevance.c b/src/relevance.c
index 2fd7414..a17c4a1 100644 (file)
--- a/src/relevance.c
+++ b/src/relevance.c
@@ -28,6 +28,7 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 #include "pazpar2_config.h"
 #include "relevance.h"
 #include "session.h"
+#include "client.h"
 
 #ifdef WIN32
 #define log2(x) (log(x)/log(2))
@@ -358,6 +359,12 @@ void relevance_prepare_read(struct relevance *rel, struct reclist *reclist,
 {
     int i;
     float *idfvec = xmalloc(rel->vec_len * sizeof(float));
+    int n_clients = clients_count();
+    struct client * clients[n_clients];
+    yaz_log(YLOG_LOG,"round-robin: have %d clients", n_clients);
+    for (i = 0; i < n_clients; i++)
+        clients[i] = 0;
+
 
     reclist_enter(reclist);
     // Calculate document frequency vector for each term.
@@ -414,8 +421,34 @@ void relevance_prepare_read(struct relevance *rel, struct reclist *reclist,
         {
             wrbuf_printf(w, "score = relevance(%d);\n", relevance);
         }
-        if (0 && type == Metadata_sortkey_relevance_h)
-            relevance *= 2;
+        // Experimental round-robin
+        // Overwrites the score calculated above, but I keep it there to
+        // get the log entries
+        if (type == Metadata_sortkey_relevance_h) {
+            struct record *record;
+            int thisclient = 0;
+            struct record *bestrecord = 0;
+            int nclust = 0;
+            for (record = rec->records; record; record = record->next) {
+                if ( bestrecord == 0 || bestrecord->position < record->position )
+                    bestrecord = record;
+                nclust++;
+            }
+            while ( clients[thisclient] != 0
+                    && clients[thisclient] != bestrecord->client )
+                thisclient++;
+            if ( clients[thisclient] == 0 )
+            {
+                yaz_log(YLOG_LOG,"round-robin: found new client at %d: p=%p\n", thisclient, bestrecord->client);
+                clients[thisclient] = bestrecord->client;
+            }
+
+            relevance = -(bestrecord->position * n_clients + thisclient) ;
+            wrbuf_printf(w,"round-robin score: pos=%d client=%d ncl=%d score=%d\n",
+                         bestrecord->position, thisclient, nclust, relevance );
+            yaz_log(YLOG_LOG,"round-robin score: pos=%d client=%d ncl=%d score=%d",
+                         bestrecord->position, thisclient, nclust, relevance );
+        }
         rec->relevance_score = relevance;
     }
     reclist_leave(reclist);