-/* $Id: ranksimilarity.c,v 1.1 2006-05-03 09:31:26 marc Exp $
+/* $Id: ranksimilarity.c,v 1.10 2006-06-07 10:50:08 adam Exp $
Copyright (C) 1995-2005
Index Data ApS
static int log_initialized = 0;
struct ranksimilarity_class_info {
- int dummy;
+ int dummy;
};
+/** term specific info and statistics to be used under ranking */
struct ranksimilarity_term_info {
- int local_occur;
- zint global_occur;
- int global_inv;
- int rank_flag;
- int rank_weight;
- TERMID term;
- int term_index;
+
+ /** frequency of term within document field */
+ int freq_term_docfield;
+
+ /** frequency of term within result set of given term */
+ zint freq_term_resset;
+
+ /** number of docs within result set */
+ zint no_docs_resset;
+
+ /** number of docs with this fieldindex in database */
+ zint no_docs_fieldindex;
+
+ /** number of terms in this fieldindex */
+ zint no_terms_fieldindex;
+
+ /** rank flag is one if term is to be included in ranking */
+ int rank_flag;
+
+ /** relative ranking weight of term fieldindex */
+ int fieldindex_weight;
+
+ /** term id used to access term name and other info */
+ TERMID term;
+
+ /** index number in terms[i] array */
+ int term_index;
};
struct ranksimilarity_set_info {
- int last_pos;
- int no_entries;
- int no_rank_entries;
- struct ranksimilarity_term_info *entries;
- NMEM nmem;
+ int last_pos;
+
+ /** number of terms in query */
+ int no_terms_query;
+
+ /** number of terms in query which are included in ranking */
+ int no_ranked_terms_query;
+
+ /** number of documents in entire database */
+ zint no_docs_database;
+
+ /** number of terms in entire database */
+ zint no_terms_database;
+
+ /** array of size no_terms_query with statistics gathered per term */
+ struct ranksimilarity_term_info *entries;
+
+ NMEM nmem;
};
+/* local clean-up function */
+static void ranksimilar_rec_reset(struct ranksimilarity_set_info *si)
+{
+ int i;
+
+ for (i = 0; i < si->no_terms_query; i++)
+ {
+ si->entries[i].freq_term_docfield = 0;
+ }
+}
+
+
/*
* create: Creates/Initialises this rank handler. This routine is
* called exactly once. The routine returns the class_handle.
*/
static void *create (ZebraHandle zh)
{
- struct ranksimilarity_class_info *ci =
- (struct ranksimilarity_class_info *) xmalloc (sizeof(*ci));
+ struct ranksimilarity_class_info *ci =
+ (struct ranksimilarity_class_info *) xmalloc (sizeof(*ci));
- if (!log_initialized)
+ if (!log_initialized)
{
- log_level = yaz_log_module_level("ranksimilarity");
- log_initialized = 1;
+ log_level = yaz_log_module_level("rank-similarity");
+ log_initialized = 1;
}
- yaz_log(log_level, "create()");
- return 0;
+ yaz_log(log_level, "create()");
+ return 0;
}
/*
*/
static void destroy (struct zebra_register *reg, void *class_handle)
{
- struct ranksimilarity_class_info *ci
- = (struct ranksimilarity_class_info *) class_handle;
- yaz_log(log_level, "destroy()");
- xfree (ci);
+ struct ranksimilarity_class_info *ci
+ = (struct ranksimilarity_class_info *) class_handle;
+ yaz_log(log_level, "destroy()");
+ xfree (ci);
}
void *class_handle, RSET rset, NMEM nmem,
TERMID *terms, int numterms)
{
- struct ranksimilarity_set_info *si =
- (struct ranksimilarity_set_info *) nmem_malloc (nmem, sizeof(*si));
- int i;
+ struct ranksimilarity_set_info *si =
+ (struct ranksimilarity_set_info *) nmem_malloc (nmem, sizeof(*si));
+ int i;
+
+ yaz_log(log_level, "begin() numterms=%d", numterms);
+
+ /* setting database global statistics */
+ si->no_docs_database = -1; /* TODO */
+ si->no_terms_database = -1; /* TODO */
- yaz_log(log_level, "begin()");
+ /* setting query statistics */
+ si->no_terms_query = numterms;
+ si->no_ranked_terms_query = 0;
- /* count how many terms are ranked (2=102 or similar) */
- si->no_entries = numterms;
- si->no_rank_entries = 0;
- si->nmem=nmem;
- si->entries = (struct ranksimilarity_term_info *)
- nmem_malloc (si->nmem, sizeof(*si->entries)*numterms);
+ /* setting internal data structures */
+ si->nmem=nmem;
+ si->entries = (struct ranksimilarity_term_info *)
+ nmem_malloc (si->nmem, sizeof(*si->entries)*numterms);
- /* looping all terms in a specific field of query */
- for (i = 0; i < numterms; i++)
+ /* reset the counts for the next term */
+ ranksimilar_rec_reset(si);
+
+
+ /* looping all terms in a specific fieldindex of query */
+ for (i = 0; i < numterms; i++)
{
- struct ord_list *ol = terms[i]->ol;
-
- yaz_log(log_level, "begin() term i=%d flags=%s '%s'", i,
- terms[i]->flags, terms[i]->name );
-
- for (; ol; ol = ol->next)
- {
- int index_type = 0;
- const char *db = 0;
- const char *string_index = 0;
- int set = -1;
- int use = -1;
-
- zebraExplain_lookup_ord(reg->zei,
- ol->ord, &index_type, &db, &set, &use,
- &string_index);
-
- if (string_index)
- yaz_log(log_level, "begin() ord=%d index_type=%c db=%s str-index=%s",
- ol->ord, index_type, db, string_index);
- else
- yaz_log(log_level, "begin() ord=%d index_type=%c db=%s set=%d use=%d",
- ol->ord, index_type, db, set, use);
- }
- if (!strncmp (terms[i]->flags, "rank,", 5))
- (si->no_rank_entries)++;
-
- /* setting next entry in term */
- terms[i]->rankpriv = &(si->entries[i]);
+ struct ord_list *ol = NULL;
+
+
+ /* adding to number of rank entries */
+ if (strncmp (terms[i]->flags, "rank,", 5))
+ {
+ si->entries[i].rank_flag = 0;
+ yaz_log(log_level, "begin() terms[%d]: '%s' flags=%s not ranked",
+ i, terms[i]->name, terms[i]->flags);
+ }
+ else
+ {
+ const char *cp = strstr(terms[i]->flags+4, ",w=");
+
+ zint no_docs_fieldindex = 0;
+ zint no_terms_fieldindex = 0;
+
+ yaz_log(log_level, "begin() terms[%d]: '%s' flags=%s",
+ i, terms[i]->name, terms[i]->flags);
+
+ (si->no_ranked_terms_query)++;
+ ol = terms[i]->ol;
+
+ si->entries[i].rank_flag = 1;
+ /* notice that the call to rset_count(rset) has he side-effect
+ of setting rset->hits_limit = rset_count(rset) ??? */
+ si->entries[i].freq_term_resset = rset_count(terms[i]->rset);
+ si->entries[i].no_docs_resset = terms[i]->rset->hits_count;
+
+
+ if (cp)
+ si->entries[i].fieldindex_weight = atoi (cp+3);
+ else
+ si->entries[i].fieldindex_weight = 34; /* sqrroot of 1000 */
+
+
+ /*
+ yaz_log(log_level, "begin() rset_count(terms[%d]->rset) = "
+ ZINT_FORMAT, i, rset_count(terms[i]->rset));
+ yaz_log(log_level, "begin() terms[%d]->rset->hits_limit = "
+ ZINT_FORMAT, i, terms[i]->rset->hits_limit);
+ yaz_log(log_level, "begin() terms[%d]->rset->hits_count = "
+ ZINT_FORMAT, i, terms[i]->rset->hits_count);
+ yaz_log(log_level, "begin() terms[%d]->rset->hits_round = "
+ ZINT_FORMAT, i, terms[i]->rset->hits_round);
+ yaz_log(log_level, "begin() terms[%d]->rset->hits_approx = %d",
+ i, terms[i]->rset->hits_approx);
+ */
+
+ /* looping indexes where term terms[i] is found */
+
+ for (; ol; ol = ol->next)
+ {
+ int index_type = 0;
+ const char *db = 0;
+ const char *string_index = 0;
+
+ zebraExplain_lookup_ord(reg->zei,
+ ol->ord, &index_type, &db,
+ &string_index);
+
+ no_docs_fieldindex
+ += zebraExplain_ord_get_doc_occurrences(reg->zei, ol->ord);
+ no_terms_fieldindex
+ += zebraExplain_ord_get_term_occurrences(reg->zei, ol->ord);
+
+ if (string_index)
+ yaz_log(log_level,
+ "begin() index: ord=%d type=%c db=%s str-index=%s",
+ ol->ord, index_type, db, string_index);
+ else
+ yaz_log(log_level,
+ "begin() index: ord=%d type=%c db=%s",
+ ol->ord, index_type, db);
+ }
+
+ si->entries[i].no_docs_fieldindex = no_docs_fieldindex;
+ si->entries[i].no_terms_fieldindex = no_terms_fieldindex;
+ }
+
+ si->entries[i].term = terms[i];
+ si->entries[i].term_index=i;
+
+ /* setting next entry in term */
+ terms[i]->rankpriv = &(si->entries[i]);
}
- return si;
+
+ return si;
}
/*
*/
static void end (struct zebra_register *reg, void *set_handle)
{
- yaz_log(log_level, "end()");
+ yaz_log(log_level, "end()");
}
+
/**
* add: Called for each word occurence in a result set. This routine
* should be as fast as possible. This routine should "incrementally"
*/
static void add (void *set_handle, int seqno, TERMID term)
{
- struct ranksimilarity_set_info *si = (struct ranksimilarity_set_info *) set_handle;
+ struct ranksimilarity_set_info *si
+ = (struct ranksimilarity_set_info *) set_handle;
struct ranksimilarity_term_info *ti;
- assert(si);
- if (!term)
+ assert(si);
+ if (!term)
{
- /* yaz_log(log_level, "add() NULL term"); */
- return;
+ /* yaz_log(log_level, "add() seqno=%d NULL term", seqno); */
+ return;
}
-
- ti= (struct ranksimilarity_term_info *) term->rankpriv;
- assert(ti);
- si->last_pos = seqno;
- ti->local_occur++;
- /* yaz_log(log_level, "add() seqno=%d term=%s count=%d",
- seqno, term->name,ti->local_occur); */
+ ti= (struct ranksimilarity_term_info *) term->rankpriv;
+ assert(ti);
+ si->last_pos = seqno;
+ ti->freq_term_docfield++;
+ /*yaz_log(log_level, "add() seqno=%d term=%s freq_term_docfield=%d",
+ seqno, term->name, ti->freq_term_docfield); */
}
/*
struct ranksimilarity_set_info *si
= (struct ranksimilarity_set_info *) set_handle;
- yaz_log(log_level, "calc()");
- if (!si->no_rank_entries)
- return -1; /* ranking not enabled for any terms */
+ yaz_log(log_level, "calc() sysno = " ZINT_FORMAT, sysno);
+ yaz_log(log_level, "calc() staticrank = " ZINT_FORMAT, staticrank);
+
+ yaz_log(log_level, "calc() si->no_terms_query = %d",
+ si->no_terms_query);
+ yaz_log(log_level, "calc() si->no_ranked_terms_query = %d",
+ si->no_ranked_terms_query);
+ yaz_log(log_level, "calc() si->no_docs_database = " ZINT_FORMAT,
+ si->no_docs_database);
+ yaz_log(log_level, "calc() si->no_terms_database = " ZINT_FORMAT,
+ si->no_terms_database);
- /* here you put in your own score function */
-
- /* reset the counts for the next term */
- for (i = 0; i < si->no_entries; i++)
- si->entries[i].local_occur = 0;
+ if (!si->no_ranked_terms_query)
+ return -1; /* ranking not enabled for any terms */
+
/* if we set *stop_flag = 1, we stop processing (of result set list) */
- /* staticrank = 0 is highest, MAXINT lowest */
/* here goes your formula to compute a scoring function */
/* you may use all the gathered statistics here */
+ for (i = 0; i < si->no_terms_query; i++)
+ {
+ yaz_log(log_level, "calc() entries[%d] termid %p",
+ i, si->entries[i].term);
+ if (si->entries[i].term){
+ yaz_log(log_level, "calc() entries[%d] term '%s' flags=%s",
+ i, si->entries[i].term->name, si->entries[i].term->flags);
+ yaz_log(log_level, "calc() entries[%d] rank_flag %d",
+ i, si->entries[i].rank_flag );
+ yaz_log(log_level, "calc() entries[%d] fieldindex_weight %d",
+ i, si->entries[i].fieldindex_weight );
+ yaz_log(log_level, "calc() entries[%d] freq_term_docfield %d",
+ i, si->entries[i].freq_term_docfield );
+ yaz_log(log_level, "calc() entries[%d] freq_term_resset " ZINT_FORMAT,
+ i, si->entries[i].freq_term_resset );
+ yaz_log(log_level, "calc() entries[%d] no_docs_resset " ZINT_FORMAT,
+ i, si->entries[i].no_docs_resset );
+ yaz_log(log_level, "calc() entries[%d] no_docs_fieldindex "
+ ZINT_FORMAT,
+ i, si->entries[i].no_docs_fieldindex );
+ yaz_log(log_level, "calc() entries[%d] no_terms_fieldindex "
+ ZINT_FORMAT,
+ i, si->entries[i].no_terms_fieldindex );
+ }
+ }
+
+
+ /* reset the counts for the next term */
+ ranksimilar_rec_reset(si);
+
- score = INT_MAX - staticrank; /* but score is reverse (logical) */
+ /* staticrank = 0 is highest, MAXINT lowest */
+ if (staticrank >= INT_MAX)
+ score = 0;
+ else
+ { /* but score is reverse (logical) */
+ score = INT_MAX - CAST_ZINT_TO_INT(staticrank);
+ }
+ /* debugging statistics output */
+ yaz_log(log_level, "calc() statistics: score = %d", score);
return score;
}
*/
static struct rank_control rank_control = {
- "rank-similarity",
- create,
- destroy,
- begin,
- end,
- calc,
- add,
+ "rank-similarity",
+ create,
+ destroy,
+ begin,
+ end,
+ calc,
+ add,
};
struct rank_control *rank_similarity_class = &rank_control;
+/*
+ * Local variables:
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ * vim: shiftwidth=4 tabstop=8 expandtab
+ */
+