X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fextract.c;h=a93373ce8f259656e1c6d37299c33ae9bb483729;hp=0eec47e898066ea36e1652fcf0cf49dc487785d2;hb=90d4902e0921f68842031a4aa1e743d5df06047f;hpb=4b903c542156253ebb2a4f004b528fe9e3af5212 diff --git a/index/extract.c b/index/extract.c index 0eec47e..a93373c 100644 --- a/index/extract.c +++ b/index/extract.c @@ -1,8 +1,5 @@ -/* $Id: extract.c,v 1.264 2007-10-29 13:43:57 adam Exp $ - Copyright (C) 1995-2007 - Index Data ApS - -This file is part of the Zebra server. +/* This file is part of the Zebra server. + Copyright (C) 1994-2011 Index Data Zebra is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free @@ -24,6 +21,9 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA \brief indexes records and extract tokens for indexing and sorting */ +#if HAVE_CONFIG_H +#include +#endif #include #include #include @@ -50,11 +50,18 @@ static int log_level_initialized = 0; /* eventually this the 0-case code will be removed */ #define FLUSH2 1 -void extract_flush_record_keys2(ZebraHandle zh, zint sysno, - zebra_rec_keys_t ins_keys, - zint ins_rank, - zebra_rec_keys_t del_keys, - zint del_rank); +#if FLUSH2 +static void extract_flush_record_keys2(ZebraHandle zh, zint sysno, + zebra_rec_keys_t ins_keys, + zint ins_rank, + zebra_rec_keys_t del_keys, + zint del_rank); +#else +static void extract_flush_record_keys(ZebraHandle zh, zint sysno, + int cmd, + zebra_rec_keys_t reckeys, + zint staticrank); +#endif static void zebra_init_log_level(void) { @@ -67,14 +74,25 @@ static void zebra_init_log_level(void) } } -static void extract_flush_record_keys(ZebraHandle zh, zint sysno, - int cmd, zebra_rec_keys_t reckeys, - zint staticrank); +static WRBUF wrbuf_hex_str(const char *cstr) +{ + size_t i; + WRBUF w = wrbuf_alloc(); + for (i = 0; cstr[i]; i++) + { + if (cstr[i] < ' ' || cstr[i] > 126) + wrbuf_printf(w, "\\%02X", cstr[i] & 0xff); + else + wrbuf_putc(w, cstr[i]); + } + return w; +} + + static void extract_flush_sort_keys(ZebraHandle zh, zint sysno, int cmd, zebra_rec_keys_t skp); static void extract_schema_add(struct recExtractCtrl *p, Odr_oid *oid); static void extract_token_add(RecWord *p); -static void extract_token_add2(RecWord *p); static void check_log_limit(ZebraHandle zh) { @@ -100,14 +118,6 @@ static void logRecord(ZebraHandle zh) static void init_extractCtrl(ZebraHandle zh, struct recExtractCtrl *ctrl) { - int i; - for (i = 0; i<256; i++) - { - if (zebra_maps_is_positioned(zh->reg->zebra_maps, i)) - ctrl->seqno[i] = 1; - else - ctrl->seqno[i] = 0; - } ctrl->flagShowRecords = !zh->m_flag_rw; } @@ -122,7 +132,7 @@ static void extract_init(struct recExtractCtrl *p, RecWord *w) { w->seqno = 1; w->index_name = "any"; - w->index_type = 'w'; + w->index_type = "w"; w->extractCtrl = p; w->record_id = 0; w->section_id = 0; @@ -135,81 +145,19 @@ struct snip_rec_info { }; -static void snippet_add_complete_field(RecWord *p, int ord) +static void snippet_add_complete_field(RecWord *p, int ord, + zebra_map_t zm) { struct snip_rec_info *h = p->extractCtrl->handle; - ZebraHandle zh = h->zh; - - const char *b = p->term_buf; - char buf[IT_MAX_WORD+1]; - const char **map = 0; - int i = 0, remain = p->term_len; - const char *start = b; - const char *last = 0; - - if (remain > 0) - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, remain, 1); - - while (remain > 0 && i < IT_MAX_WORD) - { - while (map && *map && **map == *CHR_SPACE) - { - remain = p->term_len - (b - p->term_buf); - - if (i == 0) - start = b; /* set to first non-ws area */ - if (remain > 0) - { - int first = i ? 0 : 1; /* first position */ - - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, - &b, remain, first); - } - else - map = 0; - } - if (!map) - break; - - if (i && i < IT_MAX_WORD) - buf[i++] = *CHR_SPACE; - while (map && *map && **map != *CHR_SPACE) - { - const char *cp = *map; - - if (**map == *CHR_CUT) - { - i = 0; - } - else - { - if (i >= IT_MAX_WORD) - break; - while (i < IT_MAX_WORD && *cp) - buf[i++] = *(cp++); - } - last = b; - remain = p->term_len - (b - p->term_buf); - if (remain > 0) - { - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, - remain, 0); - } - else - map = 0; - } - } - if (!i) - return; - if (last && start != last) + if (p->term_len && p->term_buf && zebra_maps_is_index(zm)) zebra_snippets_appendn(h->snippets, p->seqno, 0, ord, - start, last - start); + p->term_buf, p->term_len); + p->seqno++; } -static void snippet_add_incomplete_field(RecWord *p, int ord) +static void snippet_add_incomplete_field(RecWord *p, int ord, zebra_map_t zm) { struct snip_rec_info *h = p->extractCtrl->handle; - ZebraHandle zh = h->zh; const char *b = p->term_buf; int remain = p->term_len; int first = 1; @@ -218,12 +166,11 @@ static void snippet_add_incomplete_field(RecWord *p, int ord) const char *last = b; if (remain > 0) - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, remain, 0); + map = zebra_maps_input(zm, &b, remain, 0); while (map) { - char buf[IT_MAX_WORD+1]; - int i, remain; + int remain; /* Skip spaces */ while (map && *map && **map == *CHR_SPACE) @@ -231,48 +178,40 @@ static void snippet_add_incomplete_field(RecWord *p, int ord) remain = p->term_len - (b - p->term_buf); last = b; if (remain > 0) - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, - remain, 0); + map = zebra_maps_input(zm, &b, remain, 0); else map = 0; } if (!map) break; - if (start != last) + if (start != last && zebra_maps_is_index(zm)) { zebra_snippets_appendn(h->snippets, p->seqno, 1, ord, start, last - start); - } start = last; - - i = 0; while (map && *map && **map != *CHR_SPACE) { - const char *cp = *map; - - while (i < IT_MAX_WORD && *cp) - buf[i++] = *(cp++); remain = p->term_len - (b - p->term_buf); last = b; if (remain > 0) - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, remain, 0); + map = zebra_maps_input(zm, &b, remain, 0); else map = 0; } - if (!i) - return; + if (start == last) + return ; if (first) { first = 0; - if (zebra_maps_is_first_in_field(zh->reg->zebra_maps, p->index_type)) + if (zebra_maps_is_first_in_field(zm)) { /* first in field marker */ p->seqno++; } } - if (start != last) + if (start != last && zebra_maps_is_index(zm)) zebra_snippets_appendn(h->snippets, p->seqno, 0, ord, start, last - start); start = last; @@ -281,21 +220,48 @@ static void snippet_add_incomplete_field(RecWord *p, int ord) } +static void snippet_add_icu(RecWord *p, int ord, zebra_map_t zm) +{ + struct snip_rec_info *h = p->extractCtrl->handle; + + const char *res_buf = 0; + size_t res_len = 0; + + const char *display_buf = 0; + size_t display_len = 0; + + zebra_map_tokenize_start(zm, p->term_buf, p->term_len); + while (zebra_map_tokenize_next(zm, &res_buf, &res_len, + &display_buf, &display_len)) + { + if (zebra_maps_is_index(zm)) + zebra_snippets_appendn(h->snippets, p->seqno, 0, ord, + display_buf, display_len); + p->seqno++; + } +} + static void snippet_token_add(RecWord *p) { struct snip_rec_info *h = p->extractCtrl->handle; ZebraHandle zh = h->zh; + zebra_map_t zm = zebra_map_get_or_add(zh->reg->zebra_maps, p->index_type); - if (zebra_maps_is_index(zh->reg->zebra_maps, p->index_type)) + if (zm) { ZebraExplainInfo zei = zh->reg->zei; int ch = zebraExplain_lookup_attr_str( zei, zinfo_index_category_index, p->index_type, p->index_name); - if(zebra_maps_is_complete (h->zh->reg->zebra_maps, p->index_type)) - snippet_add_complete_field(p, ch); + if (zebra_maps_is_icu(zm)) + snippet_add_icu(p, ch, zm); else - snippet_add_incomplete_field(p, ch); + { + if (zebra_maps_is_complete(zm)) + snippet_add_complete_field(p, ch, zm); + else + snippet_add_incomplete_field(p, ch, zm); + } } } @@ -311,7 +277,6 @@ void extract_snippet(ZebraHandle zh, zebra_snippets *sn, { struct recExtractCtrl extractCtrl; struct snip_rec_info info; - int r; extractCtrl.stream = stream; extractCtrl.first_record = 1; @@ -334,8 +299,7 @@ void extract_snippet(ZebraHandle zh, zebra_snippets *sn, extractCtrl.setStoreData = 0; - r = (*rt->extract)(recTypeClientData, &extractCtrl); - + (*rt->extract)(recTypeClientData, &extractCtrl); } static void searchRecordKey(ZebraHandle zh, @@ -351,11 +315,11 @@ static void searchRecordKey(ZebraHandle zh, ws[i] = NULL; if (ch < 0) - ch = zebraExplain_lookup_attr_str(zh->reg->zei, cat, '0', index_name); + ch = zebraExplain_lookup_attr_str(zh->reg->zei, cat, "0", index_name); if (ch < 0) - ch = zebraExplain_lookup_attr_str(zh->reg->zei, cat, 'p', index_name); + ch = zebraExplain_lookup_attr_str(zh->reg->zei, cat, "p", index_name); if (ch < 0) - ch = zebraExplain_lookup_attr_str(zh->reg->zei, cat, 'w', index_name); + ch = zebraExplain_lookup_attr_str(zh->reg->zei, cat, "w", index_name); if (ch < 0) return ; @@ -432,9 +396,6 @@ static char *get_match_from_spec(ZebraHandle zh, attname_str[i++] = *s; attname_str[i] = '\0'; } - - searchRecordKey(zh, reckeys, attname_str, ws, 32); - if (*s != ')') { yaz_log(YLOG_WARN, "Missing ) in match criteria %s in group %s", @@ -443,6 +404,20 @@ static char *get_match_from_spec(ZebraHandle zh, } s++; + searchRecordKey(zh, reckeys, attname_str, ws, 32); + if (0) /* for debugging */ + { + for (i = 0; i<32; i++) + { + if (ws[i]) + { + WRBUF w = wrbuf_hex_str(ws[i]); + yaz_log(YLOG_LOG, "ws[%d] = %s", i, wrbuf_cstr(w)); + wrbuf_destroy(w); + } + } + } + for (i = 0; i<32; i++) if (ws[i]) { @@ -526,6 +501,14 @@ static char *get_match_from_spec(ZebraHandle zh, return NULL; } *dst = '\0'; + + if (0) /* for debugging */ + { + WRBUF w = wrbuf_hex_str(dstBuf); + yaz_log(YLOG_LOG, "get_match_from_spec %s", wrbuf_cstr(w)); + wrbuf_destroy(w); + } + return dstBuf; } @@ -535,21 +518,37 @@ struct recordLogInfo { struct recordGroup *rGroup; }; -static void all_matches_add(struct recExtractCtrl *ctrl) +/** \brief add the always-matches index entry and map to real record ID + \param ctrl record control + \param record_id custom record ID + \param sysno system record ID + + This function serves two purposes.. It adds the always matches + entry and makes a pointer from the custom record ID (if defined) + back to the system record ID (sysno) + See zebra_recid_to_sysno . + */ +static void all_matches_add(struct recExtractCtrl *ctrl, zint record_id, + zint sysno) { RecWord word; extract_init(ctrl, &word); + word.record_id = record_id; + /* we use the seqno as placeholder for a way to get back to + record database from _ALLRECORDS.. This is used if a custom + RECORD was defined */ + word.seqno = sysno; word.index_name = "_ALLRECORDS"; - word.index_type = 'w'; - word.seqno = 1; + word.index_type = "w"; + extract_add_index_string(&word, zinfo_index_category_alwaysmatches, "", 0); } +/* forward declaration */ ZEBRA_RES zebra_extract_records_stream(ZebraHandle zh, struct ZebraRecStream *stream, enum zebra_recctrl_action_t action, - int test_mode, const char *recordType, zint *sysno, const char *match_criteria, @@ -559,14 +558,13 @@ ZEBRA_RES zebra_extract_records_stream(ZebraHandle zh, ZEBRA_RES zebra_extract_file(ZebraHandle zh, zint *sysno, const char *fname, - int deleteFlag) + enum zebra_recctrl_action_t action) { ZEBRA_RES r = ZEBRA_OK; int i, fd; char gprefix[128]; char ext[128]; char ext_res[128]; - struct file_read_info *fi = 0; const char *original_record_type = 0; RecType recType; void *recTypeClientData; @@ -629,10 +627,9 @@ ZEBRA_RES zebra_extract_file(ZebraHandle zh, zint *sysno, const char *fname, default: yaz_log(YLOG_WARN, "Bad filter version: %s", zh->m_record_type); } - if (sysno && deleteFlag) + if (sysno && (action == action_delete || action == action_a_delete)) { streamp = 0; - fi = 0; } else { @@ -657,9 +654,7 @@ ZEBRA_RES zebra_extract_file(ZebraHandle zh, zint *sysno, const char *fname, zebra_create_stream_fd(streamp, fd, 0); } r = zebra_extract_records_stream(zh, streamp, - deleteFlag ? - action_delete : action_update, - 0, /* tst_mode */ + action, zh->m_record_type, sysno, 0, /*match_criteria */ @@ -681,7 +676,6 @@ ZEBRA_RES zebra_extract_file(ZebraHandle zh, zint *sysno, const char *fname, ZEBRA_RES zebra_buffer_extract_record(ZebraHandle zh, const char *buf, size_t buf_size, enum zebra_recctrl_action_t action, - int test_mode, const char *recordType, zint *sysno, const char *match_criteria, @@ -723,7 +717,6 @@ ZEBRA_RES zebra_buffer_extract_record(ZebraHandle zh, res = zebra_extract_records_stream(zh, &stream, action, - test_mode, recordType, sysno, match_criteria, @@ -733,69 +726,17 @@ ZEBRA_RES zebra_buffer_extract_record(ZebraHandle zh, return res; } -ZEBRA_RES zebra_extract_records_stream(ZebraHandle zh, - struct ZebraRecStream *stream, - enum zebra_recctrl_action_t action, - int test_mode, - const char *recordType, - zint *sysno, - const char *match_criteria, - const char *fname, - RecType recType, - void *recTypeClientData) -{ - ZEBRA_RES res = ZEBRA_OK; - while (1) - { - int more = 0; - res = zebra_extract_record_stream(zh, stream, - action, - test_mode, - recordType, - sysno, - match_criteria, - fname, - recType, recTypeClientData, &more); - if (!more) - { - res = ZEBRA_OK; - break; - } - if (res != ZEBRA_OK) - break; - if (sysno) - break; - } - return res; -} - - -static WRBUF wrbuf_hex_str(const char *cstr) -{ - size_t i; - WRBUF w = wrbuf_alloc(); - for (i = 0; cstr[i]; i++) - { - if (cstr[i] < ' ' || cstr[i] > 126) - wrbuf_printf(w, "\\%02X", cstr[i] & 0xff); - else - wrbuf_putc(w, cstr[i]); - } - return w; -} - -ZEBRA_RES zebra_extract_record_stream(ZebraHandle zh, - struct ZebraRecStream *stream, - enum zebra_recctrl_action_t action, - int test_mode, - const char *recordType, - zint *sysno, - const char *match_criteria, - const char *fname, - RecType recType, - void *recTypeClientData, - int *more) - +static ZEBRA_RES zebra_extract_record_stream(ZebraHandle zh, + struct ZebraRecStream *stream, + enum zebra_recctrl_action_t action, + const char *recordType, + zint *sysno, + const char *match_criteria, + const char *fname, + RecType recType, + void *recTypeClientData, + int *more) + { zint sysno0 = 0; RecordAttr *recordAttr; @@ -835,14 +776,7 @@ ZEBRA_RES zebra_extract_record_stream(ZebraHandle zh, stream->endf(stream, &null_offset);; extractCtrl.init = extract_init; - if (zh->reg->index_types) - { - extractCtrl.tokenAdd = extract_token_add2; - } - else - { - extractCtrl.tokenAdd = extract_token_add; - } + extractCtrl.tokenAdd = extract_token_add; extractCtrl.schemaAdd = extract_schema_add; extractCtrl.dh = zh->reg->dh; extractCtrl.handle = zh; @@ -896,21 +830,30 @@ ZEBRA_RES zebra_extract_record_stream(ZebraHandle zh, else end_offset = stream->tellf(stream); - all_matches_add(&extractCtrl); - if (extractCtrl.match_criteria[0]) match_criteria = extractCtrl.match_criteria; } *more = 1; + + if (zh->m_flag_rw == 0) + { + yaz_log(YLOG_LOG, "test %s %s " ZINT_FORMAT, recordType, + pr_fname, (zint) start_offset); + /* test mode .. Do not perform match */ + return ZEBRA_OK; + } + if (!sysno) { sysno = &sysno0; - - if (match_criteria && *match_criteria) { + + if (match_criteria && *match_criteria) matchStr = match_criteria; - } else { - if (zh->m_record_id && *zh->m_record_id) { + else + { + if (zh->m_record_id && *zh->m_record_id) + { matchStr = get_match_from_spec(zh, zh->reg->keys, pr_fname, zh->m_record_id); if (!matchStr) @@ -919,6 +862,17 @@ ZEBRA_RES zebra_extract_record_stream(ZebraHandle zh, pr_fname, (zint) start_offset); return ZEBRA_FAIL; } + if (0 && matchStr) + { + WRBUF w = wrbuf_alloc(); + size_t i; + for (i = 0; i < strlen(matchStr); i++) + { + wrbuf_printf(w, "%02X", matchStr[i] & 0xff); + } + yaz_log(YLOG_LOG, "Got match %s", wrbuf_cstr(w)); + wrbuf_destroy(w); + } } } if (matchStr) @@ -941,24 +895,24 @@ ZEBRA_RES zebra_extract_record_stream(ZebraHandle zh, } } } - if (zebra_rec_keys_empty(zh->reg->keys)) - { - /* the extraction process returned no information - the record - is probably empty - unless flagShowRecords is in use */ - if (test_mode) - return ZEBRA_OK; - } if (! *sysno) { - /* new record */ + /* new record AKA does not exist already */ if (action == action_delete) { - yaz_log(YLOG_LOG, "delete %s %s " ZINT_FORMAT, recordType, - pr_fname, (zint) start_offset); + yaz_log(YLOG_LOG, "delete %s %s " ZINT_FORMAT, recordType, + pr_fname, (zint) start_offset); yaz_log(YLOG_WARN, "cannot delete record above (seems new)"); return ZEBRA_FAIL; } + else if (action == action_a_delete) + { + if (show_progress) + yaz_log(YLOG_LOG, "adelete %s %s " ZINT_FORMAT, recordType, + pr_fname, (zint) start_offset); + return ZEBRA_OK; + } else if (action == action_replace) { yaz_log(YLOG_LOG, "update %s %s " ZINT_FORMAT, recordType, @@ -973,6 +927,15 @@ ZEBRA_RES zebra_extract_record_stream(ZebraHandle zh, *sysno = rec->sysno; + + if (stream) + { + all_matches_add(&extractCtrl, + zebra_rec_keys_get_custom_record_id(zh->reg->keys), + *sysno); + } + + recordAttr = rec_init_attr(zh->reg->zei, rec); if (extractCtrl.staticrank < 0) { @@ -1014,6 +977,13 @@ ZEBRA_RES zebra_extract_record_stream(ZebraHandle zh, rec = rec_get(zh->reg->records, *sysno); assert(rec); + + if (stream) + { + all_matches_add(&extractCtrl, + zebra_rec_keys_get_custom_record_id(zh->reg->keys), + *sysno); + } recordAttr = rec_init_attr(zh->reg->zei, rec); @@ -1035,7 +1005,7 @@ ZEBRA_RES zebra_extract_record_stream(ZebraHandle zh, extract_flush_record_keys(zh, *sysno, 0, delkeys, recordAttr->staticrank); #endif - if (action == action_delete) + if (action == action_delete || action == action_a_delete) { /* record going to be deleted */ #if FLUSH2 @@ -1172,6 +1142,52 @@ ZEBRA_RES zebra_extract_record_stream(ZebraHandle zh, return ZEBRA_OK; } +/** \brief extracts records from stream + \param zh Zebra Handle + \param stream stream that we read from + \param action (action_insert, action_replace, action_delete, ..) + \param recordType Record filter type "grs.xml", etc. + \param sysno pointer to sysno if already known; NULL otherwise + \param match_criteria (NULL if not already given) + \param fname filename that we read from (for logging purposes only) + \param recType record type + \param recTypeClientData client data for record type + \returns ZEBRA_OK for success; ZEBRA_FAIL for failure +*/ +ZEBRA_RES zebra_extract_records_stream(ZebraHandle zh, + struct ZebraRecStream *stream, + enum zebra_recctrl_action_t action, + const char *recordType, + zint *sysno, + const char *match_criteria, + const char *fname, + RecType recType, + void *recTypeClientData) +{ + ZEBRA_RES res = ZEBRA_OK; + while (1) + { + int more = 0; + res = zebra_extract_record_stream(zh, stream, + action, + recordType, + sysno, + match_criteria, + fname, + recType, recTypeClientData, &more); + if (!more) + { + res = ZEBRA_OK; + break; + } + if (res != ZEBRA_OK) + break; + if (sysno) + break; + } + return res; +} + ZEBRA_RES zebra_extract_explain(void *handle, Record rec, data1_node *n) { ZebraHandle zh = (ZebraHandle) handle; @@ -1256,6 +1272,56 @@ ZEBRA_RES zebra_extract_explain(void *handle, Record rec, data1_node *n) return ZEBRA_OK; } +void zebra_it_key_str_dump(ZebraHandle zh, struct it_key *key, + const char *str, size_t slen, NMEM nmem, int level) +{ + char keystr[200]; /* room for zints to print */ + char *dst_term = 0; + int ord = CAST_ZINT_TO_INT(key->mem[0]); + const char *index_type; + int i; + const char *string_index; + + zebraExplain_lookup_ord(zh->reg->zei, ord, &index_type, + 0/* db */, &string_index); + assert(index_type); + zebra_term_untrans_iconv(zh, nmem, index_type, + &dst_term, str); + *keystr = '\0'; + for (i = 0; i < key->len; i++) + { + sprintf(keystr + strlen(keystr), ZINT_FORMAT " ", key->mem[i]); + } + + if (*str < CHR_BASE_CHAR) + { + int i; + char dst_buf[200]; /* room for special chars */ + + strcpy(dst_buf , "?"); + + if (!strcmp(str, "")) + strcpy(dst_buf, "alwaysmatches"); + if (!strcmp(str, FIRST_IN_FIELD_STR)) + strcpy(dst_buf, "firstinfield"); + else if (!strcmp(str, CHR_UNKNOWN)) + strcpy(dst_buf, "unknown"); + else if (!strcmp(str, CHR_SPACE)) + strcpy(dst_buf, "space"); + + for (i = 0; ireg->zei, ord, &index_type, - 0/* db */, &string_index); - assert(index_type); - zebra_term_untrans_iconv(zh, nmem, index_type, - &dst_term, str); - *keystr = '\0'; - for (i = 0; ireg->zei; int normal = 0; @@ -1444,10 +1468,11 @@ void extract_flush_record_keys2(ZebraHandle zh, zint sysno, } yaz_log(log_level_extract, "normal=%d optimized=%d", normal, optimized); } - -void extract_flush_record_keys(ZebraHandle zh, zint sysno, int cmd, - zebra_rec_keys_t reckeys, - zint staticrank) +#else +static void extract_flush_record_keys( + ZebraHandle zh, zint sysno, int cmd, + zebra_rec_keys_t reckeys, + zint staticrank) { ZebraExplainInfo zei = zh->reg->zei; @@ -1462,7 +1487,7 @@ void extract_flush_record_keys(ZebraHandle zh, zint sysno, int cmd, if (!zh->reg->key_block) { - int mem = 1024*1024 * atoi( res_get_def( zh->res, "memmax", "8")); + int mem = 1024*1024 * atoi( res_get_def( zh->res, "memmax", "8")); const char *key_tmp_dir = res_get_def(zh->res, "keyTmpDir", "."); int use_threads = atoi(res_get_def(zh->res, "threads", "1")); zh->reg->key_block = key_block_create(mem, key_tmp_dir, use_threads); @@ -1475,19 +1500,20 @@ void extract_flush_record_keys(ZebraHandle zh, zint sysno, int cmd, #endif if (zebra_rec_keys_rewind(reckeys)) { - size_t slen; - const char *str; - struct it_key key_in; - while(zebra_rec_keys_read(reckeys, &str, &slen, &key_in)) - { + size_t slen; + const char *str; + struct it_key key_in; + while(zebra_rec_keys_read(reckeys, &str, &slen, &key_in)) + { key_block_write(zh->reg->key_block, sysno, &key_in, cmd, str, slen, staticrank, zh->m_staticrank); - } + } } } +#endif -ZEBRA_RES zebra_rec_keys_to_snippets(ZebraHandle zh, +ZEBRA_RES zebra_rec_keys_to_snippets1(ZebraHandle zh, zebra_rec_keys_t reckeys, zebra_snippets *snippets) { @@ -1502,7 +1528,7 @@ ZEBRA_RES zebra_rec_keys_to_snippets(ZebraHandle zh, char *dst_term = 0; int ord; zint seqno; - int index_type; + const char *index_type; assert(key.len <= IT_KEY_LEVEL_MAX && key.len > 2); seqno = key.mem[key.len-1]; @@ -1533,7 +1559,7 @@ void print_rec_keys(ZebraHandle zh, zebra_rec_keys_t reckeys) { char dst_buf[IT_MAX_WORD]; zint seqno; - int index_type; + const char *index_type; int ord = CAST_ZINT_TO_INT(key.mem[0]); const char *db = 0; assert(key.len <= IT_KEY_LEVEL_MAX && key.len > 2); @@ -1586,9 +1612,10 @@ static void extract_add_sort_string(RecWord *p, const char *str, int length) ch = zebraExplain_lookup_attr_str(zei, cat, p->index_type, p->index_name); if (ch < 0) ch = zebraExplain_add_attr_str(zei, cat, p->index_type, p->index_name); - key.len = 2; + key.len = 3; key.mem[0] = ch; key.mem[1] = p->record_id; + key.mem[2] = p->section_id; zebra_rec_keys_write(zh->reg->sortKeys, str, length, &key); } @@ -1607,19 +1634,27 @@ static void extract_add_staticrank_string(RecWord *p, ctrl->staticrank = atozint(valz); } -static void extract_add_string(RecWord *p, const char *string, int length) +static void extract_add_string(RecWord *p, zebra_map_t zm, + const char *string, int length) { - ZebraHandle zh = p->extractCtrl->handle; assert(length > 0); if (!p->index_name) return; + if (log_level_details) + { - if (zebra_maps_is_index(zh->reg->zebra_maps, p->index_type)) + WRBUF w = wrbuf_alloc(); + + wrbuf_write_escaped(w, string, length); + yaz_log(log_level_details, "extract_add_string: %s", wrbuf_cstr(w)); + wrbuf_destroy(w); + } + if (zebra_maps_is_index(zm)) { extract_add_index_string(p, zinfo_index_category_index, string, length); - if (zebra_maps_is_alwaysmatches(zh->reg->zebra_maps, p->index_type)) + if (zebra_maps_is_alwaysmatches(zm)) { RecWord word; memcpy(&word, p, sizeof(word)); @@ -1629,26 +1664,25 @@ static void extract_add_string(RecWord *p, const char *string, int length) &word, zinfo_index_category_alwaysmatches, "", 0); } } - else if (zebra_maps_is_sort(zh->reg->zebra_maps, p->index_type)) + else if (zebra_maps_is_sort(zm)) { extract_add_sort_string(p, string, length); } - else if (zebra_maps_is_staticrank(zh->reg->zebra_maps, p->index_type)) + else if (zebra_maps_is_staticrank(zm)) { extract_add_staticrank_string(p, string, length); } } -static void extract_add_incomplete_field(RecWord *p) +static void extract_add_incomplete_field(RecWord *p, zebra_map_t zm) { - ZebraHandle zh = p->extractCtrl->handle; const char *b = p->term_buf; int remain = p->term_len; int first = 1; const char **map = 0; if (remain > 0) - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, remain, 0); + map = zebra_maps_input(zm, &b, remain, 0); while (map) { @@ -1660,8 +1694,7 @@ static void extract_add_incomplete_field(RecWord *p) { remain = p->term_len - (b - p->term_buf); if (remain > 0) - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, - remain, 0); + map = zebra_maps_input(zm, &b, remain, 0); else map = 0; } @@ -1676,7 +1709,7 @@ static void extract_add_incomplete_field(RecWord *p) buf[i++] = *(cp++); remain = p->term_len - (b - p->term_buf); if (remain > 0) - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, remain, 0); + map = zebra_maps_input(zm, &b, remain, 0); else map = 0; } @@ -1686,28 +1719,27 @@ static void extract_add_incomplete_field(RecWord *p) if (first) { first = 0; - if (zebra_maps_is_first_in_field(zh->reg->zebra_maps, p->index_type)) + if (zebra_maps_is_first_in_field(zm)) { /* first in field marker */ - extract_add_string(p, FIRST_IN_FIELD_STR, FIRST_IN_FIELD_LEN); + extract_add_string(p, zm, FIRST_IN_FIELD_STR, FIRST_IN_FIELD_LEN); p->seqno++; } } - extract_add_string(p, buf, i); + extract_add_string(p, zm, buf, i); p->seqno++; } } -static void extract_add_complete_field(RecWord *p) +static void extract_add_complete_field(RecWord *p, zebra_map_t zm) { - ZebraHandle zh = p->extractCtrl->handle; const char *b = p->term_buf; char buf[IT_MAX_WORD+1]; const char **map = 0; int i = 0, remain = p->term_len; if (remain > 0) - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, remain, 1); + map = zebra_maps_input(zm, &b, remain, 1); while (remain > 0 && i < IT_MAX_WORD) { @@ -1718,7 +1750,7 @@ static void extract_add_complete_field(RecWord *p) if (remain > 0) { int first = i ? 0 : 1; /* first position */ - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, remain, first); + map = zebra_maps_input(zm, &b, remain, first); } else map = 0; @@ -1746,8 +1778,7 @@ static void extract_add_complete_field(RecWord *p) remain = p->term_len - (b - p->term_buf); if (remain > 0) { - map = zebra_maps_input(zh->reg->zebra_maps, p->index_type, &b, - remain, 0); + map = zebra_maps_input(zm, &b, remain, 0); } else map = 0; @@ -1755,69 +1786,35 @@ static void extract_add_complete_field(RecWord *p) } if (!i) return; - extract_add_string(p, buf, i); + extract_add_string(p, zm, buf, i); + p->seqno++; } -static void extract_token_add2_index(ZebraHandle zh, zebra_index_type_t type, - RecWord *p) +static void extract_add_icu(RecWord *p, zebra_map_t zm) { - struct it_key key; const char *res_buf = 0; size_t res_len = 0; - int r = zebra_index_type_tokenize(type, p->term_buf, p->term_len, - &res_buf, &res_len); - int cat = zinfo_index_category_index; - int ch = zebraExplain_lookup_attr_str(zh->reg->zei, cat, p->index_type, p->index_name); - if (ch < 0) - ch = zebraExplain_add_attr_str(zh->reg->zei, cat, p->index_type, p->index_name); - while (r) - { - int i = 0; - key.mem[i++] = ch; - key.mem[i++] = p->record_id; - key.mem[i++] = p->section_id; - - if (zh->m_segment_indexing) - key.mem[i++] = p->segment; - key.mem[i++] = p->seqno; - key.len = i; - - yaz_log(YLOG_LOG, "keys_write %.*s", (int) res_len, res_buf); - zebra_rec_keys_write(zh->reg->keys, res_buf, res_len, &key); - - p->seqno++; - r = zebra_index_type_tokenize(type, 0, 0, &res_buf, &res_len); - } -} -static void extract_token_add2(RecWord *p) -{ - zebra_index_type_t type; - ZebraHandle zh = p->extractCtrl->handle; - char type_tmp[2]; - type_tmp[0] = p->index_type; - type_tmp[1] = '\0'; - type = zebra_index_type_get(zh->reg->index_types, type_tmp); - if (type) + zebra_map_tokenize_start(zm, p->term_buf, p->term_len); + while (zebra_map_tokenize_next(zm, &res_buf, &res_len, 0, 0)) { - if (zebra_index_type_is_index(type)) + if (res_len > IT_MAX_WORD) { - extract_token_add2_index(zh, type, p); - } - else if (zebra_index_type_is_sort(type)) - { - ; - + yaz_log(YLOG_LOG, "Truncating long term %ld", (long) res_len); + res_len = IT_MAX_WORD; } + extract_add_string(p, zm, res_buf, res_len); + p->seqno++; } } + /** \brief top-level indexing handler for recctrl system \param p token data to be indexed Call sequence: - extract_token - zebra_add_{in}_complete + extract_token_add + extract_add_{in}_complete / extract_add_icu extract_add_string extract_add_index_string @@ -1830,25 +1827,26 @@ static void extract_token_add2(RecWord *p) static void extract_token_add(RecWord *p) { ZebraHandle zh = p->extractCtrl->handle; - WRBUF wrbuf; + zebra_map_t zm = zebra_map_get_or_add(zh->reg->zebra_maps, p->index_type); if (log_level_details) { yaz_log(log_level_details, "extract_token_add " - "type=%c index=%s seqno=" ZINT_FORMAT " s=%.*s", + "type=%s index=%s seqno=" ZINT_FORMAT " s=%.*s", p->index_type, p->index_name, p->seqno, p->term_len, p->term_buf); } - if ((wrbuf = zebra_replace(zh->reg->zebra_maps, p->index_type, 0, - p->term_buf, p->term_len))) + if (zebra_maps_is_icu(zm)) { - p->term_buf = wrbuf_buf(wrbuf); - p->term_len = wrbuf_len(wrbuf); + extract_add_icu(p, zm); } - if (zebra_maps_is_complete(zh->reg->zebra_maps, p->index_type)) - extract_add_complete_field(p); else - extract_add_incomplete_field(p); + { + if (zebra_maps_is_complete(zm)) + extract_add_complete_field(p, zm); + else + extract_add_incomplete_field(p, zm); + } } static void extract_set_store_data_cb(struct recExtractCtrl *p, @@ -1898,24 +1896,68 @@ void extract_flush_sort_keys(ZebraHandle zh, zint sysno, const char *str; struct it_key key_in; - zebra_sort_sysno(si, sysno); + NMEM nmem = nmem_create(); + struct sort_add_ent { + int ord; + int cmd; + struct sort_add_ent *next; + WRBUF wrbuf; + zint sysno; + zint section_id; + }; + struct sort_add_ent *sort_ent_list = 0; while (zebra_rec_keys_read(reckeys, &str, &slen, &key_in)) { int ord = CAST_ZINT_TO_INT(key_in.mem[0]); + zint filter_sysno = key_in.mem[1]; + zint section_id = key_in.mem[2]; + + struct sort_add_ent **e = &sort_ent_list; + for (; *e; e = &(*e)->next) + if ((*e)->ord == ord && section_id == (*e)->section_id) + break; + if (!*e) + { + *e = nmem_malloc(nmem, sizeof(**e)); + (*e)->next = 0; + (*e)->wrbuf = wrbuf_alloc(); + (*e)->ord = ord; + (*e)->cmd = cmd; + (*e)->sysno = filter_sysno ? filter_sysno : sysno; + (*e)->section_id = section_id; + } - zebra_sort_type(si, ord); - if (cmd == 1) - zebra_sort_add(si, str, slen); - else - zebra_sort_delete(si); + wrbuf_write((*e)->wrbuf, str, slen); + wrbuf_putc((*e)->wrbuf, '\0'); + } + if (sort_ent_list) + { + zint last_sysno = 0; + struct sort_add_ent *e = sort_ent_list; + for (; e; e = e->next) + { + if (last_sysno != e->sysno) + { + zebra_sort_sysno(si, e->sysno); + last_sysno = e->sysno; + } + zebra_sort_type(si, e->ord); + if (e->cmd == 1) + zebra_sort_add(si, e->section_id, e->wrbuf); + else + zebra_sort_delete(si, e->section_id); + wrbuf_destroy(e->wrbuf); + } } + nmem_destroy(nmem); } } /* * Local variables: * c-basic-offset: 4 + * c-file-style: "Stroustrup" * indent-tabs-mode: nil * End: * vim: shiftwidth=4 tabstop=8 expandtab