X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=index%2Fzrpn.c;h=c647c83a7a70f68037016a1d439c5df5a4bf0b76;hb=b8844c65622c12fb105bf7fc6bab2fc237af7c14;hp=0073e1037c5ae0f8b83f02588a1441c6bea14dd8;hpb=29357d2be4bc20d412d38bcb29876daedda28637;p=idzebra-moved-to-github.git diff --git a/index/zrpn.c b/index/zrpn.c index 0073e10..c647c83 100644 --- a/index/zrpn.c +++ b/index/zrpn.c @@ -1,10 +1,97 @@ /* - * Copyright (C) 1994-1995, Index Data I/S + * Copyright (C) 1994-1996, Index Data I/S * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * * $Log: zrpn.c,v $ - * Revision 1.24 1995-10-09 16:18:37 adam + * Revision 1.49 1996-06-04 10:18:11 adam + * Search/scan uses character mapping module. + * + * Revision 1.48 1996/05/28 15:15:01 adam + * Bug fix: Didn't handle unknown database correctly. + * + * Revision 1.47 1996/05/15 18:36:28 adam + * Function trans_term transforms unsearchable characters to blanks. + * + * Revision 1.46 1996/05/15 11:57:56 adam + * Fixed bug introduced by set/field mapping in search operations. + * + * Revision 1.45 1996/05/14 11:34:00 adam + * Scan support in multiple registers/databases. + * + * Revision 1.44 1996/05/14 06:16:44 adam + * Compact use/set bytes used in search service. + * + * Revision 1.43 1996/05/09 09:54:43 adam + * Server supports maps from one logical attributes to a list of physical + * attributes. + * The extraction process doesn't make space consuming 'any' keys. + * + * Revision 1.42 1996/05/09 07:28:56 quinn + * Work towards phrases and multiple registers + * + * Revision 1.41 1996/03/20 09:36:43 adam + * Function dict_lookup_grep got extra parameter, init_pos, which marks + * from which position in pattern approximate pattern matching should occur. + * Approximate pattern matching is used in relevance=re-2. + * + * Revision 1.40 1996/02/02 13:44:44 adam + * The public dictionary functions simply use char instead of Dict_char + * to represent search strings. Dict_char is used internally only. + * + * Revision 1.39 1996/01/03 16:22:13 quinn + * operator->roperator + * + * Revision 1.38 1995/12/11 09:12:55 adam + * The rec_get function returns NULL if record doesn't exist - will + * happen in the server if the result set records have been deleted since + * the creation of the set (i.e. the search). + * The server saves a result temporarily if it is 'volatile', i.e. the + * set is register dependent. + * + * Revision 1.37 1995/12/06 15:05:28 adam + * More verbose in count_set. + * + * Revision 1.36 1995/12/06 12:41:27 adam + * New command 'stat' for the index program. + * Filenames can be read from stdin by specifying '-'. + * Bug fix/enhancement of the transformation from terms to regular + * expressons in the search engine. + * + * Revision 1.35 1995/11/27 09:29:00 adam + * Bug fixes regarding conversion to regular expressions. + * + * Revision 1.34 1995/11/16 17:00:56 adam + * Better logging of rpn query. + * + * Revision 1.33 1995/11/01 13:58:28 quinn + * Moving data1 to yaz/retrieval + * + * Revision 1.32 1995/10/27 14:00:11 adam + * Implemented detection of database availability. + * + * Revision 1.31 1995/10/17 18:02:10 adam + * New feature: databases. Implemented as prefix to words in dictionary. + * + * Revision 1.30 1995/10/16 09:32:38 adam + * More work on relational op. + * + * Revision 1.29 1995/10/13 16:01:49 adam + * Work on relations. + * + * Revision 1.28 1995/10/13 12:26:43 adam + * Optimization of truncation. + * + * Revision 1.27 1995/10/12 17:07:22 adam + * Truncation works. + * + * Revision 1.26 1995/10/12 12:40:54 adam + * Bug fixes in rpn_prox. + * + * Revision 1.25 1995/10/10 13:59:24 adam + * Function rset_open changed its wflag parameter to general flags. + * + * Revision 1.24 1995/10/09 16:18:37 adam * Function dict_lookup_grep got extra client data parameter. * * Revision 1.23 1995/10/06 16:33:37 adam @@ -86,68 +173,18 @@ #include #include #include +#include #include "zserver.h" -#include +#include "attribute.h" +#include #include #include #include #include #include -int index_word_prefix_map (char *string, oid_value attrSet, int attrUse) -{ - attent *attp; - - logf (LOG_DEBUG, "oid_value attrSet = %d, attrUse = %d", attrSet, attrUse); - attp = att_getentbyatt (attrSet, attrUse); - if (!attp) - return -1; - logf (LOG_DEBUG, "ord=%d", attp->attset_ordinal); - return index_word_prefix (string, attp->attset_ordinal, - attp->local_attribute); -} - -/* - * attr_print: log attributes - */ -static void attr_print (Z_AttributesPlusTerm *t) -{ - int of, i; - for (of = 0; of < t->num_attributes; of++) - { - Z_AttributeElement *element; - element = t->attributeList[of]; - - switch (element->which) - { - case Z_AttributeValue_numeric: - logf (LOG_DEBUG, "attributeType=%d value=%d", - *element->attributeType, - *element->value.numeric); - break; - case Z_AttributeValue_complex: - logf (LOG_DEBUG, "attributeType=%d complex", - *element->attributeType); - for (i = 0; ivalue.complex->num_list; i++) - { - if (element->value.complex->list[i]->which == - Z_StringOrNumeric_string) - logf (LOG_DEBUG, " string: '%s'", - element->value.complex->list[i]->u.string); - else if (element->value.complex->list[i]->which == - Z_StringOrNumeric_numeric) - logf (LOG_DEBUG, " numeric: '%d'", - *element->value.complex->list[i]->u.numeric); - } - break; - default: - assert (0); - } - } -} - typedef struct { int type; int major; @@ -210,6 +247,7 @@ static void attr_init (AttrType *src, Z_AttributesPlusTerm *zapt, } struct trunc_info { + int *ptr; int *indx; char **heap; int heapnum; @@ -224,27 +262,23 @@ static void heap_swap (struct trunc_info *ti, int i1, int i2) { int swap; - memcpy (ti->swapbuf, ti->heap[i1], ti->keysize); - memcpy (ti->heap[i1], ti->heap[i2], ti->keysize); - memcpy (ti->heap[i2], ti->swapbuf, ti->keysize); - - swap = ti->indx[i1]; - ti->indx[i1] = ti->indx[i2]; - ti->indx[i2] = swap; + swap = ti->ptr[i1]; + ti->ptr[i1] = ti->ptr[i2]; + ti->ptr[i2] = swap; } static void heap_delete (struct trunc_info *ti) { int cur = 1, child = 2; - assert (ti->heapnum > 0); - memcpy (ti->heap[1], ti->heap[ti->heapnum], ti->keysize); - ti->indx[1] = ti->indx[ti->heapnum--]; + heap_swap (ti, 1, ti->heapnum--); while (child <= ti->heapnum) { if (child < ti->heapnum && - (*ti->cmp)(ti->heap[child], ti->heap[1+child]) > 0) + (*ti->cmp)(ti->heap[ti->ptr[child]], + ti->heap[ti->ptr[1+child]]) > 0) child++; - if ((*ti->cmp)(ti->heap[cur], ti->heap[child]) > 0) + if ((*ti->cmp)(ti->heap[ti->ptr[cur]], + ti->heap[ti->ptr[child]]) > 0) { heap_swap (ti, cur, child); cur = child; @@ -260,10 +294,11 @@ static void heap_insert (struct trunc_info *ti, const char *buf, int indx) int cur, parent; cur = ++(ti->heapnum); - memcpy (ti->heap[cur], buf, ti->keysize); - ti->indx[cur] = indx; + memcpy (ti->heap[ti->ptr[cur]], buf, ti->keysize); + ti->indx[ti->ptr[cur]] = indx; parent = cur/2; - while (parent && (*ti->cmp)(ti->heap[parent], ti->heap[cur]) > 0) + while (parent && (*ti->cmp)(ti->heap[ti->ptr[parent]], + ti->heap[ti->ptr[cur]]) > 0) { heap_swap (ti, cur, parent); cur = parent; @@ -284,16 +319,21 @@ struct trunc_info *heap_init (int size, int key_size, ti->cmp = cmp; ti->indx = xmalloc (size * sizeof(*ti->indx)); ti->heap = xmalloc (size * sizeof(*ti->heap)); + ti->ptr = xmalloc (size * sizeof(*ti->ptr)); ti->swapbuf = xmalloc (ti->keysize); ti->tmpbuf = xmalloc (ti->keysize); ti->buf = xmalloc (size * ti->keysize); for (i = size; --i >= 0; ) + { + ti->ptr[i] = i; ti->heap[i] = ti->buf + ti->keysize * i; + } return ti; } static void heap_close (struct trunc_info *ti) { + xfree (ti->ptr); xfree (ti->indx); xfree (ti->heap); xfree (ti->swapbuf); @@ -301,27 +341,85 @@ static void heap_close (struct trunc_info *ti) xfree (ti); } -static RSET rset_trunc (ISAM isam, ISAM_P *isam_p, int from, int to, - int merge_chunk) +static RSET rset_trunc_r (ISAM isam, ISAM_P *isam_p, int from, int to, + int merge_chunk) { - logf (LOG_DEBUG, "rset_trunc, range=%d-%d", from, to-1); + RSET result; + RSFD result_rsfd; + rset_temp_parms parms; + + parms.key_size = sizeof(struct it_key); + result = rset_create (rset_kind_temp, &parms); + result_rsfd = rset_open (result, RSETF_WRITE|RSETF_SORT_SYSNO); + if (to - from > merge_chunk) { - return NULL; + RSFD *rsfd; + RSET *rset; + int i, i_add = (to-from)/merge_chunk + 1; + struct trunc_info *ti; + int rscur = 0; + int rsmax = (to-from)/i_add + 1; + + rset = xmalloc (sizeof(*rset) * rsmax); + rsfd = xmalloc (sizeof(*rsfd) * rsmax); + + for (i = from; i < to; i += i_add) + { + if (i_add <= to - i) + rset[rscur] = rset_trunc_r (isam, isam_p, i, i+i_add, + merge_chunk); + else + rset[rscur] = rset_trunc_r (isam, isam_p, i, to, + merge_chunk); + rscur++; + } + ti = heap_init (rscur, sizeof(struct it_key), key_compare); + for (i = rscur; --i >= 0; ) + { + rsfd[i] = rset_open (rset[i], RSETF_READ|RSETF_SORT_SYSNO); + if (rset_read (rset[i], rsfd[i], ti->tmpbuf)) + heap_insert (ti, ti->tmpbuf, i); + else + { + rset_close (rset[i], rsfd[i]); + rset_delete (rset[i]); + } + } + while (ti->heapnum) + { + int n = ti->indx[ti->ptr[1]]; + + rset_write (result, result_rsfd, ti->heap[ti->ptr[1]]); + + while (1) + { + if (!rset_read (rset[n], rsfd[n], ti->tmpbuf)) + { + heap_delete (ti); + rset_close (rset[n], rsfd[n]); + rset_delete (rset[n]); + break; + } + if ((*ti->cmp)(ti->tmpbuf, ti->heap[ti->ptr[1]]) > 1) + { + heap_delete (ti); + heap_insert (ti, ti->tmpbuf, n); + break; + } + } + } + xfree (rset); + xfree (rsfd); + heap_close (ti); } else { ISPT *ispt; int i; struct trunc_info *ti; - RSET result; - RSFD rsfd; - rset_temp_parms parms; ispt = xmalloc (sizeof(*ispt) * (to-from)); - parms.key_size = sizeof (struct it_key); - result = rset_create (rset_kind_temp, &parms); - rsfd = rset_open (result, 1); ti = heap_init (to-from, sizeof(struct it_key), key_compare); @@ -330,23 +428,64 @@ static RSET rset_trunc (ISAM isam, ISAM_P *isam_p, int from, int to, ispt[i] = is_position (isam, isam_p[from+i]); if (is_readkey (ispt[i], ti->tmpbuf)) heap_insert (ti, ti->tmpbuf, i); + else + is_pt_free (ispt[i]); } while (ti->heapnum) { - int n = ti->indx[1]; + int n = ti->indx[ti->ptr[1]]; - rset_write (result, rsfd, ti->heap[1]); + rset_write (result, result_rsfd, ti->heap[ti->ptr[1]]); +#if 0 +/* section that preserve all keys */ heap_delete (ti); if (is_readkey (ispt[n], ti->tmpbuf)) heap_insert (ti, ti->tmpbuf, n); + else + is_pt_free (ispt[n]); +#else +/* section that preserve all keys with unique sysnos */ + while (1) + { + if (!is_readkey (ispt[n], ti->tmpbuf)) + { + heap_delete (ti); + is_pt_free (ispt[n]); + break; + } + if ((*ti->cmp)(ti->tmpbuf, ti->heap[ti->ptr[1]]) > 1) + { + heap_delete (ti); + heap_insert (ti, ti->tmpbuf, n); + break; + } + } +#endif } - for (i = to-from; --i >= 0; ) - is_pt_free (ispt[i]); - rset_close (result, rsfd); heap_close (ti); xfree (ispt); - return result; } + rset_close (result, result_rsfd); + return result; +} + +static int isam_trunc_cmp (const void *p1, const void *p2) +{ + ISAM_P i1 = *(ISAM_P*) p1; + ISAM_P i2 = *(ISAM_P*) p2; + int d; + + d = is_type (i1) - is_type (i2); + if (d) + return d; + return is_block (i1) - is_block (i2); +} + +static RSET rset_trunc (ISAM isam, ISAM_P *isam_p, int no) +{ + + qsort (isam_p, no, sizeof(*isam_p), isam_trunc_cmp); + return rset_trunc_r (isam, isam_p, 0, no, 100); } struct grep_info { @@ -377,20 +516,163 @@ static void add_isam_p (const char *info, struct grep_info *p) (p->isam_p_indx)++; } -static int grep_handle (Dict_char *name, const char *info, void *p) +static int grep_handle (char *name, const char *info, void *p) { logf (LOG_DEBUG, "dict name: %s", name); add_isam_p (info, p); return 0; } -static int trunc_term (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, - const char *term_sub, - oid_value attributeSet, struct grep_info *grep_info) +static void gen_regular_rel (char *dst, int val, int islt) +{ + int dst_p = 1; + int w, d, i; + int pos = 0; + char numstr[20]; + + *dst = '('; + sprintf (numstr, "%d", val); + for (w = strlen(numstr); --w >= 0; pos++) + { + d = numstr[w]; + if (pos > 0) + { + if (islt) + { + if (d == '0') + continue; + d--; + } + else + { + if (d == '9') + continue; + d++; + } + } + + strcpy (dst + dst_p, numstr); + dst_p = strlen(dst) - pos - 1; + + if (islt) + { + if (d != '0') + { + dst[dst_p++] = '['; + dst[dst_p++] = '0'; + dst[dst_p++] = '-'; + dst[dst_p++] = d; + dst[dst_p++] = ']'; + } + else + dst[dst_p++] = d; + } + else + { + if (d != '9') + { + dst[dst_p++] = '['; + dst[dst_p++] = d; + dst[dst_p++] = '-'; + dst[dst_p++] = '9'; + dst[dst_p++] = ']'; + } + else + dst[dst_p++] = d; + } + for (i = 0; i="); + gen_regular_rel (term_dict + strlen(term_dict), term_value, 0); + break; + case 5: + if (term_value < 0) + term_value = 0; + logf (LOG_DEBUG, "Relation >"); + gen_regular_rel (term_dict + strlen(term_dict), term_value+1, 0); + break; + default: + return 0; + } + logf (LOG_DEBUG, "dict_lookup_grep: %s", term_dict); + r = dict_lookup_grep (zi->wordDict, term_dict, 0, grep_info, max_pos, + 0, grep_handle); + if (r) + logf (LOG_WARN, "dict_lookup_grep fail, rel=gt: %d", r); + logf (LOG_DEBUG, "%d positions", grep_info->isam_p_indx); + return 1; +} + +static void verbatim_char (int ch, int *indx, char *dst) +{ + if (!isalnum (ch)) + dst[(*indx)++] = '\\'; + dst[(*indx)++] = ch; +} + +static int field_term (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, + const char *term_sub, int regType, + oid_value attributeSet, struct grep_info *grep_info, + int num_bases, char **basenames) { char term_dict[2*IT_MAX_WORD+2]; - int i, j; - const char *info; + int i, j, r, base_no; AttrType truncation; int truncation_value; AttrType use; @@ -406,49 +688,125 @@ static int trunc_term (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, if (use_value == -1) use_value = 1016; - i = index_word_prefix_map (term_dict, curAttributeSet, use_value); - if (i < 0) + + for (base_no = 0; base_no < num_bases; base_no++) { - zi->errCode = 114; - return -1; - } - - switch (truncation_value) - { - case -1: /* not specified */ - case 100: /* do not truncate */ - strcat (term_dict, term_sub); - logf (LOG_DEBUG, "dict_lookup: %s", term_dict); - if ((info = dict_lookup (zi->wordDict, term_dict))) - add_isam_p (info, grep_info); - break; - case 1: /* right truncation */ - strcat (term_dict, term_sub); - strcat (term_dict, ".*"); - dict_lookup_grep (zi->wordDict, term_dict, 0, grep_info, grep_handle); - break; - case 2: /* left truncation */ - case 3: /* left&right truncation */ - zi->errCode = 120; - return -1; - case 101: /* process # in term */ - for (j = strlen(term_dict), i = 0; term_sub[i] && i < 2; i++) - term_dict[j++] = term_sub[i]; - for (; term_sub[i]; i++) - if (term_sub[i] == '#') + attent *attp; + data1_local_attribute *local_attr; + int max_pos, prefix_len = 0; + + attp = att_getentbyatt (curAttributeSet, use_value); + if (!attp) + { + logf (LOG_DEBUG, "att_getentbyatt fail. set=%d use=%d", + curAttributeSet, use_value); + zi->errCode = 114; + return -1; + } + if (zebTargetInfo_curDatabase (zi->zti, basenames[base_no])) + { + zi->errCode = 109; /* Database unavailable */ + zi->errString = basenames[base_no]; + return -1; + } + for (local_attr = attp->local_attributes; local_attr; + local_attr = local_attr->next) + { + int ord; + + ord = zebTargetInfo_lookupSU (zi->zti, attp->attset_ordinal, + local_attr->local); + if (ord < 0) + continue; + if (prefix_len) + term_dict[prefix_len++] = '|'; + else + term_dict[prefix_len++] = '('; + term_dict[prefix_len++] = 1; + term_dict[prefix_len++] = ord; + } + if (!prefix_len) + { + zi->errCode = 114; + return -1; + } + term_dict[prefix_len++] = ')'; + term_dict[prefix_len++] = 1; + term_dict[prefix_len++] = regType; + term_dict[prefix_len] = '\0'; + if (!relational_term (zi, zapt, term_sub, term_dict, + attributeSet, grep_info, &max_pos)) + { + const char *cp; + + j = prefix_len; + switch (truncation_value) { - term_dict[j++] = '.'; - term_dict[j++] = '*'; + case -1: /* not specified */ + case 100: /* do not truncate */ + term_dict[j++] = '('; + for (i = 0; term_sub[i]; i++) + verbatim_char (term_sub[i], &j, term_dict); + strcpy (term_dict+j, ")"); + r = dict_lookup_grep (zi->wordDict, term_dict, 0, grep_info, + &max_pos, 0, grep_handle); + if (r) + logf (LOG_WARN, "dict_lookup_grep err, trunc=none:%d", r); + break; + case 1: /* right truncation */ + term_dict[j++] = '('; + for (i = 0; term_sub[i]; i++) + verbatim_char (term_sub[i], &j, term_dict); + strcpy (term_dict+j, ".*)"); + dict_lookup_grep (zi->wordDict, term_dict, 0, grep_info, + &max_pos, 0, grep_handle); + break; + case 2: /* left truncation */ + case 3: /* left&right truncation */ + zi->errCode = 120; + return -1; + case 101: /* process # in term */ + term_dict[j++] = '('; + for (i=0; term_sub[i]; i++) + if (term_sub[i] == '#' && i > 2) + { + term_dict[j++] = '.'; + term_dict[j++] = '*'; + } + else + verbatim_char (term_sub[i], &j, term_dict); + strcpy (term_dict+j, ")"); + r = dict_lookup_grep (zi->wordDict, term_dict, 0, grep_info, + &max_pos, 0, grep_handle); + if (r) + logf (LOG_WARN, "dict_lookup_grep err, trunc=#: %d", + r); + break; + case 102: /* regular expression */ + sprintf (term_dict + j, "(%s)", term_sub); + r = dict_lookup_grep (zi->wordDict, term_dict, 0, grep_info, + &max_pos, 0, grep_handle); + if (r) + logf (LOG_WARN, "dict_lookup_grep err, trunc=regular: %d", + r); + break; + case 103: /* regular expression with error correction */ + cp = term_sub; + r = 0; + if (*cp == '*' && cp[1] && cp[2]) + { + r = atoi (cp+1); + cp += 2; + } + sprintf (term_dict + j, "(%s)", cp); + r = dict_lookup_grep (zi->wordDict, term_dict, r, grep_info, + &max_pos, j, grep_handle); + if (r) + logf (LOG_WARN, "dict_lookup_grep err, trunc=eregular: %d", + r); + break; } - else - term_dict[j++] = term_sub[i]; - term_dict[j] = '\0'; - dict_lookup_grep (zi->wordDict, term_dict, 0, grep_info, grep_handle); - break; - case 102: /* regular expression */ - strcat (term_dict, term_sub); - dict_lookup_grep (zi->wordDict, term_dict, 0, grep_info, grep_handle); - break; + } } logf (LOG_DEBUG, "%d positions", grep_info->isam_p_indx); return 0; @@ -457,20 +815,49 @@ static int trunc_term (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, static void trans_term (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, char *termz) { - size_t i, sizez; + size_t sizez; Z_Term *term = zapt->term; sizez = term->u.general->len; - if (sizez > IT_MAX_WORD) - sizez = IT_MAX_WORD; - for (i = 0; i < sizez; i++) - termz[i] = index_char_cvt (term->u.general->buf[i]); + if (sizez > IT_MAX_WORD-1) + sizez = IT_MAX_WORD-1; + memcpy (termz, term->u.general->buf, sizez); + termz[sizez] = '\0'; +} + +static void trans_scan_term (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, + char *termz) +{ + Z_Term *term = zapt->term; + char **map; + const char *cp = (const char *) term->u.general->buf; + const char *cp_end = cp + term->u.general->len; + const char *src; + int i = 0; + int prev_space = 0; + int len; + + while ((len = (cp_end - cp)) > 0) + { + map = map_chrs_input (&cp, len); + if (**map == *CHR_SPACE) + { + if (prev_space) + continue; + prev_space = 1; + } + else + prev_space = 0; + for (src = *map; *src; src++) + termz[i++] = *src; + } termz[i] = '\0'; } static RSET rpn_search_APT_relevance (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, - oid_value attributeSet) + oid_value attributeSet, + int num_bases, char **basenames) { rset_relevance_parms parms; char termz[IT_MAX_WORD+1]; @@ -490,25 +877,38 @@ static RSET rpn_search_APT_relevance (ZServerInfo *zi, return NULL; } trans_term (zi, zapt, termz); + grep_info.isam_p_indx = 0; grep_info.isam_p_size = 0; grep_info.isam_p_buf = NULL; while (1) { - if ((p1 = strchr (p0, ' '))) + char **map; + char *p2 = NULL; + p1 = p0; + while (*p1) { - memcpy (term_sub, p0, p1-p0); - term_sub[p1-p0] = '\0'; + map = map_chrs_input (&p1, strlen(p1)); + if (**map == *CHR_SPACE) + break; + p2 = p1; } - else - strcpy (term_sub, p0); - if (trunc_term (zi, zapt, term_sub, attributeSet, &grep_info)) + if (p1 == p0) + break; + memcpy (term_sub, p0, p2-p0); + term_sub[p2-p0] = '\0'; + while (*p1) + { + map = map_chrs_input (&p1, strlen(p1)); + if (**map == *CHR_SPACE) + break; + } + if (field_term (zi, zapt, term_sub, 'w', attributeSet, &grep_info, + num_bases, basenames)) return NULL; if (!p1) break; p0 = p1; - while (*++p0 == ' ') - ; } parms.isam_positions = grep_info.isam_p_buf; parms.no_isam_positions = grep_info.isam_p_indx; @@ -520,9 +920,10 @@ static RSET rpn_search_APT_relevance (ZServerInfo *zi, return result; } -static RSET rpn_search_APT_word (ZServerInfo *zi, - Z_AttributesPlusTerm *zapt, - oid_value attributeSet) +static RSET rpn_search_APT_cphrase (ZServerInfo *zi, + Z_AttributesPlusTerm *zapt, + oid_value attributeSet, + int num_bases, char **basenames) { rset_isam_parms parms; char termz[IT_MAX_WORD+1]; @@ -540,7 +941,8 @@ static RSET rpn_search_APT_word (ZServerInfo *zi, grep_info.isam_p_size = 0; grep_info.isam_p_buf = NULL; - if (trunc_term (zi, zapt, termz, attributeSet, &grep_info)) + if (field_term (zi, zapt, termz, 'p', attributeSet, &grep_info, + num_bases, basenames)) return NULL; if (grep_info.isam_p_indx < 1) result = rset_create (rset_kind_null, NULL); @@ -551,8 +953,8 @@ static RSET rpn_search_APT_word (ZServerInfo *zi, result = rset_create (rset_kind_isam, &parms); } else - result = rset_trunc (zi->wordIsam, grep_info.isam_p_buf, 0, - grep_info.isam_p_indx, 400); + result = rset_trunc (zi->wordIsam, grep_info.isam_p_buf, + grep_info.isam_p_indx); xfree (grep_info.isam_p_buf); return result; } @@ -570,16 +972,28 @@ static RSET rpn_prox (RSET *rset, int rset_no) rsfd = xmalloc (sizeof(*rsfd)*rset_no); more = xmalloc (sizeof(*more)*rset_no); buf = xmalloc (sizeof(*buf)*rset_no); - + for (i = 0; i= 0) + { + rset_close (rset[i], rsfd[i]); + xfree (buf[i]); + --i; + } + xfree (rsfd); + xfree (more); + xfree (buf); + return rset_create (rset_kind_null, NULL); + } } parms.key_size = sizeof (struct it_key); result = rset_create (rset_kind_temp, &parms); - rsfd_result = rset_open (result, 1); + rsfd_result = rset_open (result, RSETF_WRITE|RSETF_SORT_SYSNO); while (*more) { @@ -633,7 +1047,8 @@ static RSET rpn_prox (RSET *rset, int rset_no) static RSET rpn_search_APT_phrase (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, - oid_value attributeSet) + oid_value attributeSet, + int num_bases, char **basenames) { char termz[IT_MAX_WORD+1]; char term_sub[IT_MAX_WORD+1]; @@ -654,40 +1069,51 @@ static RSET rpn_search_APT_phrase (ZServerInfo *zi, while (1) { - if ((p1 = strchr (p0, ' '))) + char **map; + char *p2 = NULL; + p1 = p0; + while (*p1) { - memcpy (term_sub, p0, p1-p0); - term_sub[p1-p0] = '\0'; + map = map_chrs_input (&p1, strlen(p1)); + if (**map == *CHR_SPACE) + break; + p2 = p1; + } + if (p1 == p0) + break; + memcpy (term_sub, p0, p2-p0); + term_sub[p2-p0] = '\0'; + p0 = p1; + while (*p1) + { + map = map_chrs_input (&p1, strlen(p1)); + if (**map != *CHR_SPACE) + break; + p0 = p1; } - else - strcpy (term_sub, p0); - grep_info.isam_p_indx = 0; - if (trunc_term (zi, zapt, term_sub, attributeSet, &grep_info)) + if (field_term (zi, zapt, term_sub, 'w', attributeSet, &grep_info, + num_bases, basenames)) return NULL; - if (grep_info.isam_p_indx > 0) + if (grep_info.isam_p_indx == 0) + rset[rset_no] = rset_create (rset_kind_null, NULL); + else if (grep_info.isam_p_indx > 1) + rset[rset_no] = rset_trunc (zi->wordIsam, + grep_info.isam_p_buf, + grep_info.isam_p_indx); + else { - if (grep_info.isam_p_indx > 1) - rset[rset_no] = rset_trunc (zi->wordIsam, - grep_info.isam_p_buf, 0, - grep_info.isam_p_indx, 400); - else - { - rset_isam_parms parms; - - parms.is = zi->wordIsam; - parms.pos = *grep_info.isam_p_buf; - rset[rset_no] = rset_create (rset_kind_isam, &parms); - } - rset_no++; - if (rset_no >= sizeof(rset)/sizeof(*rset)) - break; + rset_isam_parms parms; + + parms.is = zi->wordIsam; + parms.pos = *grep_info.isam_p_buf; + rset[rset_no] = rset_create (rset_kind_isam, &parms); } + assert (rset[rset_no]); + if (++rset_no >= sizeof(rset)/sizeof(*rset)) + break; if (!p1) break; - p0 = p1; - while (*++p0 == ' ') - ; } xfree (grep_info.isam_p_buf); if (rset_no == 0) @@ -717,9 +1143,10 @@ static RSET rpn_search_APT_local (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, } parms.key_size = sizeof (struct it_key); result = rset_create (rset_kind_temp, &parms); - rsfd = rset_open (result, 1); + rsfd = rset_open (result, RSETF_WRITE|RSETF_SORT_SYSNO); trans_term (zi, zapt, termz); + key.sysno = atoi (termz); if (key.sysno <= 0) key.sysno = 1; @@ -728,34 +1155,52 @@ static RSET rpn_search_APT_local (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, return result; } - static RSET rpn_search_APT (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, - oid_value attributeSet) + oid_value attributeSet, + int num_bases, char **basenames) { AttrType relation; AttrType structure; - int relation_value, structure_value; + AttrType completeness; + int relation_value, structure_value, completeness_value; attr_init (&relation, zapt, 2); attr_init (&structure, zapt, 4); + attr_init (&completeness, zapt, 6); relation_value = attr_find (&relation, NULL); structure_value = attr_find (&structure, NULL); + completeness_value = attr_find (&completeness, NULL); switch (structure_value) { case -1: if (relation_value == 102) /* relevance relation */ - return rpn_search_APT_relevance (zi, zapt, attributeSet); - return rpn_search_APT_phrase (zi, zapt, attributeSet); + return rpn_search_APT_relevance (zi, zapt, attributeSet, + num_bases, basenames); + if (completeness_value == 2 || completeness_value == 3) + return rpn_search_APT_cphrase (zi, zapt, attributeSet, + num_bases, basenames); + return rpn_search_APT_phrase (zi, zapt, attributeSet, + num_bases, basenames); case 1: /* phrase */ if (relation_value == 102) /* relevance relation */ - return rpn_search_APT_relevance (zi, zapt, attributeSet); - return rpn_search_APT_phrase (zi, zapt, attributeSet); + return rpn_search_APT_relevance (zi, zapt, attributeSet, + num_bases, basenames); + if (completeness_value == 2 || completeness_value == 3) + return rpn_search_APT_cphrase (zi, zapt, attributeSet, + num_bases, basenames); + return rpn_search_APT_phrase (zi, zapt, attributeSet, + num_bases, basenames); break; case 2: /* word */ if (relation_value == 102) /* relevance relation */ - return rpn_search_APT_relevance (zi, zapt, attributeSet); - return rpn_search_APT_word (zi, zapt, attributeSet); + return rpn_search_APT_relevance (zi, zapt, attributeSet, + num_bases, basenames); + if (completeness_value == 2 || completeness_value == 3) + return rpn_search_APT_cphrase (zi, zapt, attributeSet, + num_bases, basenames); + return rpn_search_APT_phrase (zi, zapt, attributeSet, + num_bases, basenames); case 3: /* key */ break; case 4: /* year */ @@ -763,7 +1208,8 @@ static RSET rpn_search_APT (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, case 5: /* date - normalized */ break; case 6: /* word list */ - return rpn_search_APT_relevance (zi, zapt, attributeSet); + return rpn_search_APT_relevance (zi, zapt, attributeSet, + num_bases, basenames); case 100: /* date - un-normalized */ break; case 101: /* name - normalized */ @@ -775,13 +1221,16 @@ static RSET rpn_search_APT (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, case 104: /* urx */ break; case 105: /* free-form-text */ - return rpn_search_APT_relevance (zi, zapt, attributeSet); + return rpn_search_APT_relevance (zi, zapt, attributeSet, + num_bases, basenames); case 106: /* document-text */ - return rpn_search_APT_relevance (zi, zapt, attributeSet); + return rpn_search_APT_relevance (zi, zapt, attributeSet, + num_bases, basenames); case 107: /* local-number */ return rpn_search_APT_local (zi, zapt, attributeSet); case 108: /* string */ - return rpn_search_APT_word (zi, zapt, attributeSet); + return rpn_search_APT_phrase (zi, zapt, attributeSet, + num_bases, basenames); case 109: /* numeric string */ break; } @@ -799,37 +1248,45 @@ static RSET rpn_search_ref (ZServerInfo *zi, Z_ResultSetId *resultSetId) } static RSET rpn_search_structure (ZServerInfo *zi, Z_RPNStructure *zs, - oid_value attributeSet) + oid_value attributeSet, + int num_bases, char **basenames) { RSET r = NULL; if (zs->which == Z_RPNStructure_complex) { rset_bool_parms bool_parms; + int soft = 0; bool_parms.rset_l = rpn_search_structure (zi, zs->u.complex->s1, - attributeSet); + attributeSet, + num_bases, basenames); if (bool_parms.rset_l == NULL) return NULL; + if (rset_is_ranked(bool_parms.rset_l)) + soft = 1; bool_parms.rset_r = rpn_search_structure (zi, zs->u.complex->s2, - attributeSet); + attributeSet, + num_bases, basenames); if (bool_parms.rset_r == NULL) { rset_delete (bool_parms.rset_l); return NULL; } + if (rset_is_ranked(bool_parms.rset_r)) + soft = 1; bool_parms.key_size = sizeof(struct it_key); bool_parms.cmp = key_compare; - switch (zs->u.complex->operator->which) + switch (zs->u.complex->roperator->which) { case Z_Operator_and: - r = rset_create (rset_kind_and, &bool_parms); + r = rset_create (soft ? rset_kind_sand:rset_kind_and, &bool_parms); break; case Z_Operator_or: - r = rset_create (rset_kind_or, &bool_parms); + r = rset_create (soft ? rset_kind_sor:rset_kind_or, &bool_parms); break; case Z_Operator_and_not: - r = rset_create (rset_kind_not, &bool_parms); + r = rset_create (soft ? rset_kind_snot:rset_kind_not, &bool_parms); break; default: assert (0); @@ -841,7 +1298,7 @@ static RSET rpn_search_structure (ZServerInfo *zi, Z_RPNStructure *zs, { logf (LOG_DEBUG, "rpn_search_APT"); r = rpn_search_APT (zi, zs->u.simple->u.attributesPlusTerm, - attributeSet); + attributeSet, num_bases, basenames); } else if (zs->u.simple->which == Z_Operand_resultSetId) { @@ -860,15 +1317,48 @@ static RSET rpn_search_structure (ZServerInfo *zi, Z_RPNStructure *zs, return r; } +void count_set_save (RSET *r, int *count) +{ + int psysno = 0; + int kno = 0; + struct it_key key; + RSFD rfd, wfd; + RSET w; + rset_temp_parms parms; + + logf (LOG_DEBUG, "count_set_save"); + *count = 0; + parms.key_size = sizeof(struct it_key); + w = rset_create (rset_kind_temp, &parms); + wfd = rset_open (w, RSETF_WRITE|RSETF_SORT_SYSNO); + rfd = rset_open (*r, RSETF_READ|RSETF_SORT_SYSNO); + while (rset_read (*r, rfd, &key)) + { + if (key.sysno != psysno) + { + rset_write (w, wfd, &key); + psysno = key.sysno; + (*count)++; + } + kno++; + } + rset_close (*r, rfd); + rset_delete (*r); + rset_close (w, wfd); + *r = w; + logf (LOG_DEBUG, "%d keys, %d distinct sysnos", kno, *count); +} + static void count_set (RSET r, int *count) { int psysno = 0; + int kno = 0; struct it_key key; RSFD rfd; - logf (LOG_DEBUG, "rpn_save_set"); + logf (LOG_DEBUG, "count_set"); *count = 0; - rfd = rset_open (r, 0); + rfd = rset_open (r, RSETF_READ|RSETF_SORT_SYSNO); while (rset_read (r, rfd, &key)) { if (key.sysno != psysno) @@ -876,9 +1366,10 @@ static void count_set (RSET r, int *count) psysno = key.sysno; (*count)++; } + kno++; } rset_close (r, rfd); - logf (LOG_DEBUG, "%d distinct sysnos", *count); + logf (LOG_DEBUG, "%d keys, %d distinct sysnos", kno, *count); } int rpn_search (ZServerInfo *zi, @@ -889,40 +1380,45 @@ int rpn_search (ZServerInfo *zi, oident *attrset; oid_value attributeSet; + dict_grep_cmap (zi->wordDict, map_chrs_input); + zlog_rpn (rpn); + zi->errCode = 0; zi->errString = NULL; - + attrset = oid_getentbyoid (rpn->attributeSetId); attributeSet = attrset->value; - - rset = rpn_search_structure (zi, rpn->RPNStructure, attributeSet); + rset = rpn_search_structure (zi, rpn->RPNStructure, attributeSet, + num_bases, basenames); if (!rset) return zi->errCode; - count_set (rset, hits); + if (rset_is_volatile(rset)) + count_set_save(&rset,hits); + else + count_set (rset, hits); resultSetAdd (zi, setname, 1, rset); if (zi->errCode) logf (LOG_DEBUG, "search error: %d", zi->errCode); return zi->errCode; } +struct scan_info_entry { + char *term; + ISAM_P isam_p; +}; + struct scan_info { - struct scan_entry *list; + struct scan_info_entry *list; ODR odr; int before, after; - ISAM isam; char prefix[20]; }; -static int scan_handle (Dict_char *name, const char *info, int pos, - void *client) +static int scan_handle (char *name, const char *info, int pos, void *client) { int len_prefix, idx; - ISAM_P isam_p; - RSET rset; struct scan_info *scan_info = client; - rset_isam_parms parms; - len_prefix = strlen(scan_info->prefix); if (memcmp (name, scan_info->prefix, len_prefix)) return 1; @@ -930,94 +1426,261 @@ static int scan_handle (Dict_char *name, const char *info, int pos, idx = scan_info->after - pos + scan_info->before; else idx = - pos - 1; + logf (LOG_DEBUG, "%-3d %s", idx, name+len_prefix); scan_info->list[idx].term = odr_malloc (scan_info->odr, strlen(name + len_prefix)+1); strcpy (scan_info->list[idx].term, name + len_prefix); - assert (*info == sizeof(isam_p)); - memcpy (&isam_p, info+1, sizeof(isam_p)); - parms.is = scan_info->isam; - parms.pos = isam_p; -#if 1 - rset = rset_create (rset_kind_isam, &parms); - count_set (rset, &scan_info->list[idx].occurrences); - rset_delete (rset); -#else - scan_info->list[idx].occurrences = 1; -#endif - logf (LOG_DEBUG, "pos=%3d idx=%3d name=%s", pos, idx, name); + assert (*info == sizeof(ISAM_P)); + memcpy (&scan_info->list[idx].isam_p, info+1, sizeof(ISAM_P)); return 0; } -int rpn_scan (ZServerInfo *zi, ODR odr, Z_AttributesPlusTerm *zapt, + +static void scan_term_untrans (ODR odr, char **dstp, const char *src) +{ + char *dst = odr_malloc (odr, strlen(src)*2+1); + *dstp = dst; + + while (*src) + { + const char *cp = map_chrs_output (&src); + while (*cp) + *dst++ = *cp++; + } + *dst = '\0'; +} + +int rpn_scan (ZServerInfo *zi, Z_AttributesPlusTerm *zapt, + oid_value attributeset, + int num_bases, char **basenames, int *position, int *num_entries, struct scan_entry **list, int *status) { - int i, j, sizez; + int i; int pos = *position; int num = *num_entries; int before; int after; + int base_no; char termz[IT_MAX_WORD+20]; AttrType use; int use_value; - Z_Term *term = zapt->term; - struct scan_info scan_info; + AttrType completeness; + int completeness_value; + struct scan_info *scan_info_array; + struct scan_entry *glist; + int ords[32], ord_no = 0; + int ptr[32]; logf (LOG_DEBUG, "scan, position = %d, num = %d", pos, num); - scan_info.before = before = pos-1; - scan_info.after = after = 1+num-pos; - scan_info.odr = odr; - logf (LOG_DEBUG, "scan, before = %d, after = %d", before, after); - - scan_info.isam = zi->wordIsam; - scan_info.list = odr_malloc (odr, (before+after)*sizeof(*scan_info.list)); - for (j = 0; ju.general->len; - if (sizez > IT_MAX_WORD) - sizez = IT_MAX_WORD; - for (j = 0; ju.general->buf[j]); - termz[j+i] = '\0'; - - dict_scan (zi->wordDict, termz, &before, &after, &scan_info, scan_handle); + for (base_no = 0; base_no < num_bases && ord_no < 32; base_no++) + { + attent *attp; + data1_local_attribute *local_attr; - *status = BEND_SCAN_SUCCESS; + attp = att_getentbyatt (attributeset, use_value); + if (!attp) + { + logf (LOG_DEBUG, "att_getentbyatt fail. set=%d use=%d", + attributeset, use_value); + return zi->errCode = 114; + } + if (zebTargetInfo_curDatabase (zi->zti, basenames[base_no])) + { + zi->errString = basenames[base_no]; + return zi->errCode = 109; /* Database unavailable */ + } + for (local_attr = attp->local_attributes; local_attr && ord_no < 32; + local_attr = local_attr->next) + { + int ord; - for (i = 0; izti, attp->attset_ordinal, + local_attr->local); + if (ord > 0) + ords[ord_no++] = ord; + } + } + if (ord_no == 0) + return zi->errCode = 113; + before = pos-1; + after = 1+num-pos; + scan_info_array = odr_malloc (zi->odr, ord_no * sizeof(*scan_info_array)); + for (i = 0; i < ord_no; i++) + { + int j, prefix_len = 0; + int before_tmp = before, after_tmp = after; + struct scan_info *scan_info = scan_info_array + i; + + scan_info->before = before; + scan_info->after = after; + scan_info->odr = zi->odr; + + scan_info->list = odr_malloc (zi->odr, (before+after)* + sizeof(*scan_info->list)); + for (j = 0; jlist[j].term = NULL; + termz[prefix_len++] = ords[i]; + termz[prefix_len++] = + (completeness_value==2 || completeness_value==3) ? 'p': 'w'; + termz[prefix_len] = 0; + strcpy (scan_info->prefix, termz); + + trans_scan_term (zi, zapt, termz+prefix_len); + + dict_scan (zi->wordDict, termz, &before_tmp, &after_tmp, scan_info, + scan_handle); + } + glist = odr_malloc (zi->odr, (before+after)*sizeof(*glist)); + for (i = 0; i < ord_no; i++) + ptr[i] = before; + + *status = BEND_SCAN_SUCCESS; + for (i = 0; iodr, &glist[i+before].term, mterm); + parms.is = zi->wordIsam; + parms.pos = scan_info_array[j0].list[ptr[j0]].isam_p; + rset = rset_create (rset_kind_isam, &parms); + + ptr[j0]++; + for (j = j0+1; jwordIsam; + parms.pos = scan_info_array[j].list[ptr[j]].isam_p; + rset2 = rset_create (rset_kind_isam, &parms); + + bool_parms.key_size = sizeof(struct it_key); + bool_parms.cmp = key_compare; + bool_parms.rset_l = rset; + bool_parms.rset_r = rset2; + + rset = rset_create (rset_kind_or, &bool_parms); + + ptr[j]++; + } + } + count_set (rset, &glist[i+before].occurrences); + rset_delete (rset); + } + if (i < after) + { + *num_entries -= (after-i); *status = BEND_SCAN_PARTIAL; + } - for (i = 0; i 0)) + { + j0 = j; + mterm = tst; + } + } + if (j0 == -1) break; - if (i) - *status = BEND_SCAN_PARTIAL; - *position -= i; - *num_entries -= i; - *list = scan_info.list+i; /* list is set to first 'real' entry */ + scan_term_untrans (zi->odr, &glist[before-1-i].term, mterm); + + parms.is = zi->wordIsam; + parms.pos = scan_info_array[j0].list[before-1-ptr[j0]].isam_p; + rset = rset_create (rset_kind_isam, &parms); + + ptr[j0]++; - if (*num_entries == 0) /* signal 'unsupported use-attribute' */ - zi->errCode = 114; /* if no entries was found */ + for (j = j0+1; jwordIsam; + parms.pos = scan_info_array[j].list[before-1-ptr[j]].isam_p; + rset2 = rset_create (rset_kind_isam, &parms); + + bool_parms.key_size = sizeof(struct it_key); + bool_parms.cmp = key_compare; + bool_parms.rset_l = rset; + bool_parms.rset_r = rset2; + + rset = rset_create (rset_kind_or, &bool_parms); + + ptr[j]++; + } + } + count_set (rset, &glist[before-1-i].occurrences); + rset_delete (rset); + } + i = before-i; + if (i) + { + *status = BEND_SCAN_PARTIAL; + *position -= i; + *num_entries -= i; + } + *list = glist + i; /* list is set to first 'real' entry */ + logf (LOG_DEBUG, "position = %d, num_entries = %d", *position, *num_entries); if (zi->errCode) logf (LOG_DEBUG, "scan error: %d", zi->errCode); - return 0; + return zi->errCode; } - -