X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=index%2Franksimilarity.c;h=c45007863b6d188b2e641a6709aa443b56dc3f0c;hb=9b7dbbec2a6a910a2883ea3da5e6136e9240f820;hp=ace230bfce837d8861e1671c8764c24034f595a0;hpb=178199283530635a796aec2a90db8f57c1152bb0;p=idzebra-moved-to-github.git diff --git a/index/ranksimilarity.c b/index/ranksimilarity.c index ace230b..c450078 100644 --- a/index/ranksimilarity.c +++ b/index/ranksimilarity.c @@ -1,4 +1,4 @@ -/* $Id: ranksimilarity.c,v 1.2 2006-05-03 13:55:20 marc Exp $ +/* $Id: ranksimilarity.c,v 1.5 2006-05-08 14:03:16 marc Exp $ Copyright (C) 1995-2005 Index Data ApS @@ -40,24 +40,58 @@ struct ranksimilarity_class_info { int dummy; }; +/** term specific info and statistics to be used under ranking */ struct ranksimilarity_term_info { - /* frequency of term within document */ - int freq_term_doc; - /* frequency of term within result set */ + + /** frequency of term within document field */ + int freq_term_docfield; + + /** frequency of term within result set of given term */ zint freq_term_resset; - /* rank flag is one if term is to be included in ranking */ + + /** number of docs within result set */ + zint no_docs_resset; + + /** number of terms in this field */ + zint no_terms_field; + + /** number of docs with this field in database */ + zint no_docs_field; + + /** size of all docs with this field in database (in bytes ir terms ??) */ + zint size_docs_field; + + /** rank flag is one if term is to be included in ranking */ int rank_flag; - /* relative ranking weight of term */ - int term_weight; + + /** relative ranking weight of term field */ + int field_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; + + /** 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 collection */ + zint no_docs_database; + + /** size of all documents in entire collection (in bytes or terms ??)*/ + zint size_docs_database; + + /** array of size no_terms_query with statistics gathered per term */ struct ranksimilarity_term_info *entries; + NMEM nmem; }; @@ -67,9 +101,10 @@ static void ranksimilar_rec_reset(struct ranksimilarity_set_info *si) { int i; - for (i = 0; i < si->no_entries; i++){ - si->entries[i].freq_term_doc = 0; - } + for (i = 0; i < si->no_terms_query; i++) + { + si->entries[i].freq_term_docfield = 0; + } } @@ -120,9 +155,15 @@ static void *begin (struct zebra_register *reg, yaz_log(log_level, "begin() numterms=%d", numterms); - si->no_entries = numterms; - /* count how many terms are ranked (2=102 or similar) */ - si->no_rank_entries = 0; + /* setting database global statistics */ + si->no_docs_database = -1; /* TODO */ + si->size_docs_database = -1; /* TODO */ + + /* setting query statistics */ + si->no_terms_query = numterms; + si->no_ranked_terms_query = 0; + + /* setting internal data structures */ si->nmem=nmem; si->entries = (struct ranksimilarity_term_info *) nmem_malloc (si->nmem, sizeof(*si->entries)*numterms); @@ -148,17 +189,39 @@ static void *begin (struct zebra_register *reg, { const char *cp = strstr(terms[i]->flags+4, ",w="); - (si->no_rank_entries)++; + 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; + si->entries[i].no_docs_field = -1; /*TODO*/ + si->entries[i].size_docs_field = -1; /*TODO*/ + si->entries[i].no_terms_field = -1; /*TODO*/ + if (cp) - si->entries[i].term_weight = atoi (cp+3); + si->entries[i].field_weight = atoi (cp+3); else - si->entries[i].term_weight = 34; /* sqrroot of 1000 */ - - yaz_log(log_level, "begin() terms[%d]: '%s' flags=%s", - i, terms[i]->name, terms[i]->flags); + si->entries[i].field_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) @@ -175,11 +238,11 @@ static void *begin (struct zebra_register *reg, if (string_index) yaz_log(log_level, - "begin() index: ord=%d type=%c db=%s str-index=%s", + "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 set=%d use=%d", + "begin() index: ord=%d type=%c db=%s set=%d use=%d", ol->ord, index_type, db, set, use); } @@ -205,6 +268,7 @@ static void end (struct zebra_register *reg, void *set_handle) } + /** * add: Called for each word occurence in a result set. This routine * should be as fast as possible. This routine should "incrementally" @@ -218,16 +282,16 @@ static void add (void *set_handle, int seqno, TERMID term) assert(si); if (!term) { - yaz_log(log_level, "add() seqno=%d NULL term", seqno); + /* 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->freq_term_doc++; - yaz_log(log_level, "add() seqno=%d term=%s freq_term_doc=%d", - seqno, term->name, ti->freq_term_doc); + ti->freq_term_docfield++; + /* yaz_log(log_level, "add() seqno=%d term=%s freq_term_docfield=%d", + seqno, term->name, ti->freq_term_docfield); */ } /* @@ -244,16 +308,20 @@ static int calc (void *set_handle, zint sysno, zint staticrank, = (struct ranksimilarity_set_info *) set_handle; - yaz_log(log_level, "calc() sysno = %d", sysno); - yaz_log(log_level, "calc() staticrank = %d", staticrank); + 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_entries = %d", - si->no_entries); - yaz_log(log_level, "calc() si->no_rank_entries = %d", - si->no_rank_entries); + 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->size_docs_database = " ZINT_FORMAT, + si->size_docs_database); - if (!si->no_rank_entries) + if (!si->no_ranked_terms_query) return -1; /* ranking not enabled for any terms */ @@ -262,21 +330,29 @@ static int calc (void *set_handle, zint sysno, zint staticrank, /* here goes your formula to compute a scoring function */ /* you may use all the gathered statistics here */ - for (i = 0; i < si->no_entries; i++) + for (i = 0; i < si->no_terms_query; i++) { - yaz_log(log_level, "calc() entries[%d] termid %d", + 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] term_weight %d", - i, si->entries[i].term_weight ); - yaz_log(log_level, "calc() entries[%d] freq_term_doc %d", - i, si->entries[i].freq_term_doc ); - yaz_log(log_level, "calc() entries[%d] freq_term_resset %d", + yaz_log(log_level, "calc() entries[%d] field_weight %d", + i, si->entries[i].field_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_field " ZINT_FORMAT, + i, si->entries[i].no_docs_field ); + yaz_log(log_level, "calc() entries[%d] size_docs_field " ZINT_FORMAT, + i, si->entries[i].size_docs_field ); + yaz_log(log_level, "calc() entries[%d] no_terms_field " ZINT_FORMAT, + i, si->entries[i].no_terms_field ); } }