+
+ if (sset->cache_rfd &&
+ num_i < num && positions[num_i] > sset->cache_position)
+ {
+ position = sset->cache_position;
+ rfd = sset->cache_rfd;
+ psysno = sset->cache_psysno;
+ }
+ else
+ {
+ if (sset->cache_rfd)
+ rset_close(sset->cache_rfd);
+ rfd = rset_open(rset, RSETF_READ);
+ }
+ while (num_i < num && rset_read(rfd, &key, 0))
+ {
+ zint this_sys = key.mem[sysno_mem_index];
+ if (this_sys != psysno)
+ {
+ psysno = this_sys;
+ if (sort_info)
+ {
+ /* determine we alreay have this in our set */
+ for (i = sort_info->num_entries; --i >= 0; )
+ if (psysno == sort_info->entries[i]->sysno)
+ break;
+ if (i >= 0)
+ continue;
+ }
+ position++;
+ assert(num_i < num);
+ if (position == positions[num_i])
+ {
+ sr[num_i].sysno = psysno;
+ yaz_log(log_level_sort, "got pos=" ZINT_FORMAT " (unsorted)", position);
+ sr[num_i].score = -1;
+ num_i++;
+ }
+ }
+ }
+ sset->cache_position = position;
+ sset->cache_psysno = psysno;
+ sset->cache_rfd = rfd;
+ }
+ }
+ return sr;
+}
+
+void zebra_meta_records_destroy(ZebraHandle zh, ZebraMetaRecord *records,
+ int num)
+{
+ assert(zh); /* compiler shut up about unused arg */
+ xfree(records);
+}
+
+struct sortKeyInfo {
+ int relation;
+ int *ord; /* array of ord for each database searched */
+ int *numerical; /* array of ord for each database searched */
+ const char *index_type;
+};
+
+void resultSetInsertSort(ZebraHandle zh, ZebraSet sset,
+ int database_no,
+ struct sortKeyInfo *criteria, int num_criteria,
+ zint sysno,
+ char *cmp_buf[], char *tmp_cmp_buf[])
+{
+ struct zset_sort_entry *new_entry = NULL;
+ struct zset_sort_info *sort_info = sset->sort_info;
+ int i, j;
+ WRBUF w = wrbuf_alloc();
+
+ zebra_sort_sysno(zh->reg->sort_index, sysno);
+ for (i = 0; i<num_criteria; i++)
+ {
+ char *this_entry_buf = tmp_cmp_buf[i];
+ memset(this_entry_buf, '\0', SORT_IDX_ENTRYSIZE);
+
+ if (criteria[i].ord[database_no] != -1)
+ {
+ yaz_log(log_level_sort, "pre zebra_sort_type ord is %d",
+ criteria[i].ord[database_no]);
+ zebra_sort_type(zh->reg->sort_index, criteria[i].ord[database_no]);
+ wrbuf_rewind(w);
+ if (zebra_sort_read(zh->reg->sort_index, 0, w))
+ {
+ /* consider each sort entry and take lowest/highest one
+ of the one as sorting key depending on whether sort is
+ ascending/descending */
+ int off = 0;
+ while (off != wrbuf_len(w))
+ {
+ size_t l = strlen(wrbuf_buf(w)+off);
+ assert(off < wrbuf_len(w));
+
+ if (l >= SORT_IDX_ENTRYSIZE)
+ l = SORT_IDX_ENTRYSIZE-1;
+ if ( (off == 0)
+ || (criteria[i].relation == 'A'
+ && strcmp(wrbuf_buf(w)+off, this_entry_buf) < 0)
+ || (criteria[i].relation == 'D'
+ && strcmp(wrbuf_buf(w)+off, this_entry_buf) > 0)
+ )
+ {
+ memcpy(this_entry_buf, wrbuf_buf(w)+off, l);
+ this_entry_buf[l] = '\0';
+ }
+ off += 1 + strlen(wrbuf_buf(w)+off);
+ }
+ }
+ }
+ else
+ {
+ yaz_log(log_level_sort, "criteria[i].ord is -1 so not reading from sort index");
+ }
+ }
+ wrbuf_destroy(w);
+ i = sort_info->num_entries;
+ while (--i >= 0)
+ {
+ int rel = 0;
+ for (j = 0; j<num_criteria; j++)
+ {
+ char *this_entry_buf = tmp_cmp_buf[j];
+ char *other_entry_buf =
+ cmp_buf[j] + i * SORT_IDX_ENTRYSIZE;
+ if (criteria[j].numerical[database_no])
+ {
+ char this_entry_org[1024];
+ char other_entry_org[1024];
+ double diff;
+ const char *index_type = criteria[j].index_type;
+ zebra_term_untrans(zh, index_type, this_entry_org,
+ this_entry_buf);
+ zebra_term_untrans(zh, index_type, other_entry_org,
+ other_entry_buf);
+ diff = atof(this_entry_org) - atof(other_entry_org);
+
+ if (diff > 0.0)
+ rel = 1;
+ else if (diff < 0.0)
+ rel = -1;
+ else
+ rel = 0;
+ }
+ else
+ {
+ rel = memcmp(this_entry_buf, other_entry_buf,
+ SORT_IDX_ENTRYSIZE);
+ }
+ /* when the compare is equal, continue to next criteria,
+ else break out */
+ if (rel)
+ break;
+ }
+ if (!rel)
+ break;
+ if (criteria[j].relation == 'A')
+ {
+ if (rel > 0)
+ break;
+ }
+ else if (criteria[j].relation == 'D')
+ {
+ if (rel < 0)
+ break;
+ }
+ }
+ ++i;
+ yaz_log(log_level_sort, "ok, we want to insert record at position %d",i);
+ j = sort_info->max_entries;
+ if (i == j){
+ yaz_log(log_level_sort, "sort_info->max_entries reached (%d) abort sort",j);
+ return;
+ }
+
+ if (sort_info->num_entries == j)
+ --j;
+ else
+ j = (sort_info->num_entries)++;
+ new_entry = sort_info->entries[j];
+ /* move up all higher entries (to make room) */
+ while (j != i)
+ {
+ int k;
+ for (k = 0; k<num_criteria; k++)
+ {
+ char *j_buf = cmp_buf[k] + j * SORT_IDX_ENTRYSIZE;
+ char *j_1_buf = cmp_buf[k] + (j-1) * SORT_IDX_ENTRYSIZE;
+ memcpy(j_buf, j_1_buf, SORT_IDX_ENTRYSIZE);
+ }
+ sort_info->entries[j] = sort_info->entries[j-1];
+ --j;
+ }
+ /* and insert the new entry at the correct place */
+ sort_info->entries[i] = new_entry;
+ assert(new_entry);
+ /* and add this to the compare buffer */
+ for (i = 0; i<num_criteria; i++)
+ {
+ char *new_entry_buf = cmp_buf[i] + j * SORT_IDX_ENTRYSIZE;
+ char *this_entry_buf = tmp_cmp_buf[i];
+ memcpy(new_entry_buf, this_entry_buf, SORT_IDX_ENTRYSIZE);
+ }
+ new_entry->sysno = sysno;
+ new_entry->score = -1;
+}
+
+void resultSetInsertRank(ZebraHandle zh, struct zset_sort_info *sort_info,
+ zint sysno, int score, int relation)
+{
+ struct zset_sort_entry *new_entry = NULL;
+ int i, j;
+ assert(zh); /* compiler shut up about unused arg */
+
+ i = sort_info->num_entries;
+ while (--i >= 0)
+ {
+ int rel = 0;
+
+ rel = score - sort_info->entries[i]->score;
+
+ if (relation == 'D')
+ {
+ if (rel >= 0)
+ break;
+ }
+ else if (relation == 'A')
+ {
+ if (rel <= 0)
+ break;
+ }
+ }
+ ++i;
+ j = sort_info->max_entries;
+ if (i == j)
+ return;
+
+ if (sort_info->num_entries == j)
+ --j;
+ else
+ j = (sort_info->num_entries)++;
+
+ new_entry = sort_info->entries[j];
+ while (j != i)
+ {
+ sort_info->entries[j] = sort_info->entries[j-1];
+ --j;
+ }
+ sort_info->entries[i] = new_entry;
+ assert(new_entry);
+ new_entry->sysno = sysno;
+ new_entry->score = score;
+}
+
+static Z_RPNQuery *copy_RPNQuery(Z_RPNQuery *src, NMEM nmem)
+{
+ Z_RPNQuery *dst = 0;
+ ODR encode = odr_createmem(ODR_ENCODE);
+ ODR decode = odr_createmem(ODR_DECODE);
+
+ if (z_RPNQuery(encode, &src, 0, 0))
+ {
+ int len;
+ char *buf = odr_getbuf(encode, &len, 0);
+
+ if (buf)
+ {
+ odr_setbuf(decode, buf, len, 0);
+ z_RPNQuery(decode, &dst, 0, 0);
+ }
+ }
+ nmem_transfer(nmem, decode->mem);
+ odr_destroy(encode);
+ odr_destroy(decode);
+ return dst;
+}
+
+static Z_SortKeySpecList *copy_SortKeySpecList(Z_SortKeySpecList *src, NMEM nmem)
+{
+ Z_SortKeySpecList *dst = 0;
+ ODR encode = odr_createmem(ODR_ENCODE);
+ ODR decode = odr_createmem(ODR_DECODE);
+
+ if (z_SortKeySpecList(encode, &src, 0, 0))
+ {
+ int len;
+ char *buf = odr_getbuf(encode, &len, 0);
+
+ if (buf)
+ {
+ odr_setbuf(decode, buf, len, 0);
+ z_SortKeySpecList(decode, &dst, 0, 0);
+ }
+ }
+ nmem_transfer(nmem, decode->mem);
+ odr_destroy(encode);
+ odr_destroy(decode);
+ return dst;
+}
+
+ZebraSet resultSetClone(ZebraHandle zh, const char *setname,
+ ZebraSet rset)
+{
+ ZebraSet nset;
+ int i;
+
+ nset = resultSetAdd(zh, setname, 1);
+ if (!nset)
+ return 0;
+
+ nset->nmem = nmem_create();
+
+ nset->num_bases = rset->num_bases;
+ nset->basenames =
+ nmem_malloc(nset->nmem, nset->num_bases * sizeof(*rset->basenames));
+ for (i = 0; i<rset->num_bases; i++)
+ nset->basenames[i] = nmem_strdup(nset->nmem, rset->basenames[i]);
+
+ if (rset->rset)
+ nset->rset = rset_dup(rset->rset);
+ if (rset->rpn)
+ nset->rpn = copy_RPNQuery(rset->rpn, nset->nmem);
+ return nset;
+}
+
+ZEBRA_RES resultSetSort(ZebraHandle zh, NMEM nmem,
+ int num_input_setnames, const char **input_setnames,
+ const char *output_setname,
+ Z_SortKeySpecList *sort_sequence, int *sort_status)
+{
+ ZebraSet sset;
+ RSET rset;
+
+ if (num_input_setnames == 0)
+ {
+ zebra_setError(zh, YAZ_BIB1_NO_RESULT_SET_NAME_SUPPLIED_ON_SORT, 0);
+ return ZEBRA_FAIL;
+ }
+ if (num_input_setnames > 1)
+ {
+ zebra_setError(zh, YAZ_BIB1_SORT_TOO_MANY_INPUT_RESULTS, 0);
+ return ZEBRA_FAIL;
+ }
+ if (!log_level_set)
+ loglevels();
+ yaz_log(log_level_sort, "result set sort input=%s output=%s",
+ *input_setnames, output_setname);
+ sset = resultSetGet(zh, input_setnames[0]);
+ if (!sset)
+ {
+ zebra_setError(zh, YAZ_BIB1_SPECIFIED_RESULT_SET_DOES_NOT_EXIST,
+ input_setnames[0]);
+ return ZEBRA_FAIL;
+ }
+ if (!(rset = sset->rset))
+ {
+ zebra_setError(zh, YAZ_BIB1_SPECIFIED_RESULT_SET_DOES_NOT_EXIST,
+ input_setnames[0]);
+ return ZEBRA_FAIL;
+ }
+ if (strcmp(output_setname, input_setnames[0]))
+ sset = resultSetClone(zh, output_setname, sset);
+ sset->sortSpec = copy_SortKeySpecList(sort_sequence, sset->nmem);
+ return resultSetSortSingle(zh, nmem, sset, rset, sort_sequence,
+ sort_status);
+}
+
+ZEBRA_RES resultSetSortSingle(ZebraHandle zh, NMEM nmem,
+ ZebraSet sset, RSET rset,
+ Z_SortKeySpecList *sort_sequence,
+ int *sort_status)
+{
+ int i;
+ int ib;
+ int n = 0;
+ zint kno = 0;
+ zint psysno = 0;
+ struct it_key key;
+ struct sortKeyInfo sort_criteria[ZSET_SORT_MAX_LEVEL];
+ char *cmp_buf[ZSET_SORT_MAX_LEVEL];
+ char *tmp_cmp_buf[ZSET_SORT_MAX_LEVEL];
+ int num_criteria;
+ RSFD rfd;
+ TERMID termid;
+ TERMID *terms;
+ int numTerms = 0;
+ size_t sysno_mem_index = 0;
+
+ int numbases = zh->num_basenames;
+ yaz_log(log_level_sort, "searching %d databases",numbases);
+
+ if (zh->m_staticrank)
+ sysno_mem_index = 1;
+
+ assert(nmem); /* compiler shut up about unused param */
+ sset->sort_info->num_entries = 0;
+
+ rset_getterms(rset, 0, 0, &n);
+ terms = (TERMID *) nmem_malloc(nmem, sizeof(*terms)*n);
+ rset_getterms(rset, terms, n, &numTerms);
+
+ sset->hits = 0;
+ num_criteria = sort_sequence->num_specs;
+ if (num_criteria > ZSET_SORT_MAX_LEVEL)
+ num_criteria = ZSET_SORT_MAX_LEVEL;
+ /* set up the search criteria */
+ for (i = 0; i < num_criteria; i++)
+ {
+ Z_SortKeySpec *sks = sort_sequence->specs[i];
+ Z_SortKey *sk;
+ ZEBRA_RES res;
+
+ sort_criteria[i].ord = (int *)
+ nmem_malloc(nmem, sizeof(int)*numbases);
+ sort_criteria[i].numerical = (int *)
+ nmem_malloc(nmem, sizeof(int)*numbases);
+
+ /* initialize ord and numerical for each database */
+ for (ib = 0; ib < numbases; ib++)
+ {
+ sort_criteria[i].ord[ib] = -1;
+ sort_criteria[i].numerical[ib] = 0;
+ }
+
+ if (sks->which == Z_SortKeySpec_missingValueData)
+ {
+ zebra_setError(zh, YAZ_BIB1_UNSUPP_MISSING_DATA_ACTION, 0);
+ return ZEBRA_FAIL;
+ }
+ if (*sks->sortRelation == Z_SortKeySpec_ascending)
+ sort_criteria[i].relation = 'A';
+ else if (*sks->sortRelation == Z_SortKeySpec_descending)
+ sort_criteria[i].relation = 'D';
+ else
+ {
+ zebra_setError(zh, YAZ_BIB1_ILLEGAL_SORT_RELATION, 0);
+ return ZEBRA_FAIL;
+ }
+ if (sks->sortElement->which == Z_SortElement_databaseSpecific)
+ {
+ zebra_setError(zh, YAZ_BIB1_DATABASE_SPECIFIC_SORT_UNSUPP, 0);
+ return ZEBRA_FAIL;
+ }
+ else if (sks->sortElement->which != Z_SortElement_generic)
+ {
+ zebra_setError(zh, YAZ_BIB1_SORT_ILLEGAL_SORT, 0);
+ return ZEBRA_FAIL;
+ }
+ sk = sks->sortElement->u.generic;
+ switch (sk->which)
+ {
+ case Z_SortKey_sortField:
+ yaz_log(log_level_sort, "key %d is of type sortField", i+1);
+ for (ib = 0; ib < numbases; ib++)
+ {
+ zebraExplain_curDatabase(zh->reg->zei, zh->basenames[ib]);
+ sort_criteria[i].numerical[ib] = 0;
+ sort_criteria[i].ord[ib] =
+ zebraExplain_lookup_attr_str(zh->reg->zei,
+ zinfo_index_category_sort,
+ 0, sk->u.sortField);
+ if (sks->which != Z_SortKeySpec_null
+ && sort_criteria[i].ord[ib] == -1)
+ {
+ zebra_setError(zh,
+ YAZ_BIB1_CANNOT_SORT_ACCORDING_TO_SEQUENCE, 0);
+ return ZEBRA_FAIL;
+ }
+ }
+ break;
+ case Z_SortKey_elementSpec:
+ yaz_log(log_level_sort, "key %d is of type elementSpec", i+1);
+ zebra_setError(zh, YAZ_BIB1_CANNOT_SORT_ACCORDING_TO_SEQUENCE, 0);
+ return ZEBRA_FAIL;
+ case Z_SortKey_sortAttributes:
+ yaz_log(log_level_sort, "key %d is of type sortAttributes", i+1);
+ /* for every database we searched, get the sort index file
+ id (ord) and its numerical indication and store them in
+ the sort_criteria */
+ for (ib = 0; ib < numbases; ib++)
+ {
+ zebraExplain_curDatabase(zh->reg->zei, zh->basenames[ib]);
+ res = zebra_sort_get_ord(zh, sk->u.sortAttributes,
+ &sort_criteria[i].ord[ib],
+ &sort_criteria[i].numerical[ib]);
+ }
+
+ if (sks->which != Z_SortKeySpec_null && res != ZEBRA_OK)
+ return ZEBRA_FAIL;
+ break;
+ }
+ /* right now we look up the index type based on the first database
+ if the index_type's can differ between the indexes of different
+ databases (which i guess they can?) then we have to store the
+ index types for each database, just like the ord and numerical */
+ if (zebraExplain_lookup_ord(zh->reg->zei, sort_criteria[i].ord[0],
+ &sort_criteria[i].index_type,
+ 0, 0))
+ {
+ zebra_setError(zh, YAZ_BIB1_CANNOT_SORT_ACCORDING_TO_SEQUENCE, 0);
+ return ZEBRA_FAIL;
+ }
+ }
+ /* allocate space for each cmpare buf + one extra for tmp comparison */
+ /* cmp_buf is an array of array, the first dimension is the criteria and the second dimension are
+ all other result entries to compare against. This is slowly filled when records are processed.
+ tmp_cmp_buf is an array with a value of the current record for each criteria
+ */
+ for (i = 0; i<num_criteria; i++)
+ {
+ cmp_buf[i] = xmalloc(sset->sort_info->max_entries
+ * SORT_IDX_ENTRYSIZE);
+ tmp_cmp_buf[i] = xmalloc(SORT_IDX_ENTRYSIZE);
+ }
+ rfd = rset_open(rset, RSETF_READ);
+ while (rset_read(rfd, &key, &termid))
+ {
+ zint this_sys = key.mem[sysno_mem_index];
+ if (log_level_searchhits)
+ key_logdump_txt(log_level_searchhits, &key, termid->name);
+ kno++;
+ if (this_sys != psysno)
+ {
+ int database_no = 0;
+ if ((sset->hits & 255) == 0 && zh->break_handler_func)
+ {
+ if (zh->break_handler_func(zh->break_handler_data))
+ {
+ rset_set_hits_limit(rset, 0);