+ zh->records_inserted++;
+ }
+ else
+ {
+ /* record already exists */
+#if NEW_REC_KEYS
+ zebra_rec_keys_t delkeys = zebra_rec_keys_open();
+#else
+ struct recKeys delkeys;
+#endif
+#if NATTR
+ struct recKeys sortKeys;
+#else
+ struct sortKeys sortKeys;
+#endif
+
+ if (!allow_update)
+ {
+ if (show_progress)
+ yaz_log (YLOG_LOG, "skipped %s %s %ld",
+ recordType, pr_fname, (long) recordOffset);
+ logRecord(zh);
+ return ZEBRA_FAIL;
+ }
+
+ rec = rec_get (zh->reg->records, *sysno);
+ assert (rec);
+
+ recordAttr = rec_init_attr (zh->reg->zei, rec);
+
+#if NEW_REC_KEYS
+ zebra_rec_keys_set_buf(delkeys,
+ rec->info[recInfo_delKeys],
+ rec->size[recInfo_delKeys],
+ 0);
+#else
+ delkeys.buf_used = rec->size[recInfo_delKeys];
+ delkeys.buf = rec->info[recInfo_delKeys];
+#endif
+ sortKeys.buf_used = rec->size[recInfo_sortKeys];
+ sortKeys.buf = rec->info[recInfo_sortKeys];
+
+#if NEW_REC_KEYS && NATTR
+ extract_flushSortKeys (zh, *sysno, 0, sortKeys);
+#else
+ extract_flushSortKeys (zh, *sysno, 0, &sortKeys);
+#endif
+#if NEW_REC_KEYS
+ extract_flushRecordKeys (zh, *sysno, 0, delkeys,
+ recordAttr->staticrank);
+#else
+ extract_flushRecordKeys (zh, *sysno, 0, &delkeys,
+ recordAttr->staticrank);
+#endif
+ if (delete_flag)
+ {
+ /* record going to be deleted */
+ if (
+#if NEW_REC_KEYS
+ zebra_rec_keys_empty(delkeys)
+#else
+ !delkeys.buf_used
+#endif
+ )
+ {
+ if (show_progress)
+ {
+ yaz_log (YLOG_LOG, "delete %s %s %ld", recordType,
+ pr_fname, (long) recordOffset);
+ yaz_log (YLOG_WARN, "cannot delete file above, "
+ "storeKeys false");
+ }
+ }
+ else
+ {
+ if (show_progress)
+ yaz_log (YLOG_LOG, "delete %s %s %ld", recordType,
+ pr_fname, (long) recordOffset);
+ zh->records_deleted++;
+ if (matchStr)
+ dict_delete (zh->reg->matchDict, matchStr);
+ rec_del (zh->reg->records, &rec);
+ }
+ rec_rm (&rec);
+ logRecord(zh);
+ return ZEBRA_OK;
+ }
+ else
+ {
+ /* record going to be updated */
+ if (
+#if NEW_REC_KEYS
+ zebra_rec_keys_empty(delkeys)
+#else
+ !delkeys.buf_used
+#endif
+ )
+ {
+ if (show_progress)
+ {
+ yaz_log (YLOG_LOG, "update %s %s %ld", recordType,
+ pr_fname, (long) recordOffset);
+ yaz_log (YLOG_WARN, "cannot update file above, storeKeys false");
+ }
+ }
+ else
+ {
+ if (show_progress)
+ yaz_log (YLOG_LOG, "update %s %s %ld", recordType,
+ pr_fname, (long) recordOffset);
+ recordAttr->staticrank = extractCtrl.staticrank;
+ extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys);
+#if NEW_REC_KEYS
+ extract_flushRecordKeys (zh, *sysno, 1, zh->reg->keys,
+ recordAttr->staticrank);
+#else
+ extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys,
+ recordAttr->staticrank);
+#endif
+ zh->records_updated++;
+ }
+ }
+#if NEW_REC_KEYS
+ zebra_rec_keys_close(delkeys);
+#endif
+#if NEW_REC_KEYS && NATTR
+ zebra_rec_keys_close(sortKeys);
+#endif
+ }
+ /* update file type */
+ xfree (rec->info[recInfo_fileType]);
+ rec->info[recInfo_fileType] =
+ rec_strdup (recordType, &rec->size[recInfo_fileType]);
+
+ /* update filename */
+ xfree (rec->info[recInfo_filename]);
+ rec->info[recInfo_filename] =
+ rec_strdup (fname, &rec->size[recInfo_filename]);
+
+ /* update delete keys */
+ xfree (rec->info[recInfo_delKeys]);
+ if (
+#if NEW_REC_KEYS
+ !zebra_rec_keys_empty(zh->reg->keys)
+#else
+ zh->reg->keys.buf_used > 0
+#endif
+ && zh->m_store_keys == 1)
+ {
+#if NEW_REC_KEYS
+ zebra_rec_keys_get_buf(zh->reg->keys,
+ &rec->info[recInfo_delKeys],
+ &rec->size[recInfo_delKeys]);
+#else
+
+ rec->size[recInfo_delKeys] = zh->reg->keys.buf_used;
+ rec->info[recInfo_delKeys] = zh->reg->keys.buf;
+ zh->reg->keys.buf = NULL;
+ zh->reg->keys.buf_max = 0;
+#endif
+ }
+ else
+ {
+ rec->info[recInfo_delKeys] = NULL;
+ rec->size[recInfo_delKeys] = 0;
+ }
+ /* update sort keys */
+ xfree (rec->info[recInfo_sortKeys]);
+
+ rec->size[recInfo_sortKeys] = zh->reg->sortKeys.buf_used;
+ rec->info[recInfo_sortKeys] = zh->reg->sortKeys.buf;
+ zh->reg->sortKeys.buf = NULL;
+ zh->reg->sortKeys.buf_max = 0;
+
+ /* save file size of original record */
+ zebraExplain_recordBytesIncrement (zh->reg->zei,
+ - recordAttr->recordSize);
+#if 0
+ recordAttr->recordSize = fi->file_moffset - recordOffset;
+ if (!recordAttr->recordSize)
+ recordAttr->recordSize = fi->file_max - recordOffset;
+#else
+ recordAttr->recordSize = buf_size;
+#endif
+ zebraExplain_recordBytesIncrement (zh->reg->zei,
+ recordAttr->recordSize);
+
+ /* set run-number for this record */
+ recordAttr->runNumber =
+ zebraExplain_runNumberIncrement (zh->reg->zei, 0);
+
+ /* update store data */
+ xfree (rec->info[recInfo_storeData]);
+
+ /* update store data */
+ if (zh->store_data_buf)
+ {
+ rec->size[recInfo_storeData] = zh->store_data_size;
+ rec->info[recInfo_storeData] = zh->store_data_buf;
+ zh->store_data_buf = 0;
+ }
+ else if (zh->m_store_data)
+ {
+ rec->size[recInfo_storeData] = recordAttr->recordSize;
+ rec->info[recInfo_storeData] = (char *)
+ xmalloc (recordAttr->recordSize);
+ memcpy (rec->info[recInfo_storeData], buf, recordAttr->recordSize);
+ }
+ else
+ {
+ rec->info[recInfo_storeData] = NULL;
+ rec->size[recInfo_storeData] = 0;
+ }
+ /* update database name */
+ xfree (rec->info[recInfo_databaseName]);
+ rec->info[recInfo_databaseName] =
+ rec_strdup (zh->basenames[0], &rec->size[recInfo_databaseName]);
+
+ /* update offset */
+ recordAttr->recordOffset = recordOffset;
+
+ /* commit this record */
+ rec_put (zh->reg->records, &rec);
+ logRecord(zh);
+ return ZEBRA_OK;
+}
+
+int explain_extract (void *handle, Record rec, data1_node *n)
+{
+ ZebraHandle zh = (ZebraHandle) handle;
+ struct recExtractCtrl extractCtrl;
+
+ if (zebraExplain_curDatabase (zh->reg->zei,
+ rec->info[recInfo_databaseName]))
+ {
+ abort();
+ if (zebraExplain_newDatabase (zh->reg->zei,
+ rec->info[recInfo_databaseName], 0))
+ abort ();
+ }
+
+#if NEW_REC_KEYS
+ zebra_rec_keys_reset(zh->reg->keys);
+#else
+ create_rec_keys_codec(&zh->reg->keys);
+#endif
+
+#if NATTR
+#if NEW_REC_KEYS
+ zebra_rec_keys_reset(zh->reg->sortKeys);
+#else
+ create_rec_keys_codec(&zh->reg->sortKeys);
+#endif
+#else
+ zh->reg->sortKeys.buf_used = 0;
+#endif
+ extractCtrl.init = extract_init;
+ extractCtrl.tokenAdd = extract_token_add;
+ extractCtrl.schemaAdd = extract_schema_add;
+ extractCtrl.dh = zh->reg->dh;
+
+ init_extractCtrl(zh, &extractCtrl);
+
+ extractCtrl.flagShowRecords = 0;
+ extractCtrl.match_criteria[0] = '\0';
+ extractCtrl.staticrank = 0;
+ extractCtrl.handle = handle;
+ extractCtrl.first_record = 1;
+
+ extract_set_store_data_prepare(&extractCtrl);
+
+ if (n)
+ grs_extract_tree(&extractCtrl, n);
+
+ if (rec->size[recInfo_delKeys])
+ {
+#if NEW_REC_KEYS
+ zebra_rec_keys_t delkeys = zebra_rec_keys_open();
+#else
+ struct recKeys delkeys;
+#endif
+
+#if NATTR
+#if NEW_REC_KEYS
+ zebra_rec_keys_t sortkeys = zzebra_rec_keys_open();
+#else
+ struct recKeys sortkeys;
+#endif
+#else
+ struct sortKeys sortkeys;
+#endif
+
+#if NEW_REC_KEYS
+ zebra_rec_keys_set_buf(delkeys, rec->info[recInfo_delKeys],
+ rec->size[recInfo_delKeys],
+ 0);
+ extract_flushRecordKeys (zh, rec->sysno, 0, delkeys, 0);
+ zebra_rec_keys_close(delkeys);
+#else
+ delkeys.buf_used = rec->size[recInfo_delKeys];
+ delkeys.buf = rec->info[recInfo_delKeys];
+ extract_flushRecordKeys (zh, rec->sysno, 0, &delkeys, 0);
+#endif
+#if NATTR && NEW_REC_KEYS
+ zebra_rec_keys_set_buf(sortkeys, rec->info[recInfo_sortKeys],
+ rec->size[recInfo_sortKeys],
+ 0);
+
+ extract_flushSortKeys (zh, rec->sysno, 0, sortkeys);
+ zebra_rec_keys_close(sortkeys);
+#else
+ sortkeys.buf_used = rec->size[recInfo_sortKeys];
+ sortkeys.buf = rec->info[recInfo_sortKeys];
+ extract_flushSortKeys (zh, rec->sysno, 0, &sortkeys);
+#endif
+ }
+#if NEW_REC_KEYS
+ extract_flushRecordKeys (zh, rec->sysno, 1, zh->reg->keys, 0);
+#else
+ extract_flushRecordKeys (zh, rec->sysno, 1, &zh->reg->keys, 0);
+#endif
+#if NATTR && NEW_REC_KEYS
+ extract_flushSortKeys (zh, rec->sysno, 1, zh->reg->sortKeys);
+#else
+ extract_flushSortKeys (zh, rec->sysno, 1, &zh->reg->sortKeys);
+#endif
+
+ xfree (rec->info[recInfo_delKeys]);
+#if NEW_REC_KEYS
+ zebra_rec_keys_get_buf(zh->reg->keys,
+ &rec->info[recInfo_delKeys],
+ &rec->size[recInfo_delKeys]);
+#else
+ rec->size[recInfo_delKeys] = zh->reg->keys.buf_used;
+ rec->info[recInfo_delKeys] = zh->reg->keys.buf;
+ zh->reg->keys.buf = NULL;
+ zh->reg->keys.buf_max = 0;
+#endif
+
+ xfree (rec->info[recInfo_sortKeys]);
+ rec->size[recInfo_sortKeys] = zh->reg->sortKeys.buf_used;
+ rec->info[recInfo_sortKeys] = zh->reg->sortKeys.buf;
+ zh->reg->sortKeys.buf = NULL;
+ zh->reg->sortKeys.buf_max = 0;
+
+ return 0;
+}
+
+void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno,
+ int cmd,
+#if NEW_REC_KEYS
+ zebra_rec_keys_t reckeys,
+#else
+ struct recKeys *reckeys,
+#endif
+ zint staticrank)
+{
+#if NEW_REC_KEYS
+#else
+ void *decode_handle = iscz1_start();
+ int off = 0;
+ int ch = 0;
+#endif
+ ZebraExplainInfo zei = zh->reg->zei;
+
+ if (!zh->reg->key_buf)
+ {
+ int mem= 1024*1024* atoi( res_get_def( zh->res, "memmax", "8"));
+ if (mem <= 0)
+ {
+ yaz_log(YLOG_WARN, "Invalid memory setting, using default 8 MB");
+ mem= 1024*1024*8;
+ }
+ /* FIXME: That "8" should be in a default settings include */
+ /* not hard-coded here! -H */
+ zh->reg->key_buf = (char**) xmalloc (mem);
+ zh->reg->ptr_top = mem/sizeof(char*);
+ zh->reg->ptr_i = 0;
+ zh->reg->key_buf_used = 0;
+ zh->reg->key_file_no = 0;
+ }
+ zebraExplain_recordCountIncrement (zei, cmd ? 1 : -1);
+
+#if NEW_REC_KEYS
+ 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))
+ {
+ int ch = 0;
+ struct it_key key_out;
+ zint *keyp = key_out.mem;
+
+ assert(key_in.len == 4);
+
+ /* check for buffer overflow */
+ if (zh->reg->key_buf_used + 1024 >
+ (zh->reg->ptr_top -zh->reg->ptr_i)*sizeof(char*))
+ extract_flushWriteKeys (zh, 0);
+
+ ++(zh->reg->ptr_i);
+ assert(zh->reg->ptr_i > 0);
+ (zh->reg->key_buf)[zh->reg->ptr_top - zh->reg->ptr_i] =
+ (char*)zh->reg->key_buf + zh->reg->key_buf_used;
+
+ /* encode the ordinal value (field/use/attribute) .. */
+ ch = (int) key_in.mem[0];
+ zh->reg->key_buf_used +=
+ key_SU_encode(ch, (char*)zh->reg->key_buf +
+ zh->reg->key_buf_used);
+
+ /* copy the 0-terminated stuff from str to output */
+ memcpy((char*)zh->reg->key_buf + zh->reg->key_buf_used, str, slen);
+ zh->reg->key_buf_used += slen;
+ ((char*)zh->reg->key_buf)[(zh->reg->key_buf_used)++] = '\0';
+
+ /* the delete/insert indicator */
+ ((char*)zh->reg->key_buf)[(zh->reg->key_buf_used)++] = cmd;
+
+ if (zh->m_staticrank) /* rank config enabled ? */
+ {
+ *keyp++ = staticrank;
+ key_out.len = 4;
+ }
+ else
+ key_out.len = 3;
+
+ if (key_in.mem[1]) /* filter specified record ID */
+ *keyp++ = key_in.mem[1];
+ else
+ *keyp++ = sysno;
+ *keyp++ = key_in.mem[2]; /* section_id */
+ *keyp++ = key_in.mem[3]; /* sequence .. */
+
+ memcpy((char*)zh->reg->key_buf + zh->reg->key_buf_used,
+ &key_out, sizeof(key_out));
+ (zh->reg->key_buf_used) += sizeof(key_out);
+ }
+ }
+#else
+ while (off < reckeys->buf_used)
+ {
+ const char *src = reckeys->buf + off;
+ struct it_key key_in;
+ struct it_key key_out;
+ char *dst = (char*) &key_in;
+ zint *keyp = key_out.mem;
+
+ iscz1_decode(decode_handle, &dst, &src);
+ assert(key_in.len == 4);
+
+ if (zh->reg->key_buf_used + 1024 >
+ (zh->reg->ptr_top -zh->reg->ptr_i)*sizeof(char*))
+ extract_flushWriteKeys(zh, 0);
+ ++(zh->reg->ptr_i);
+ assert(zh->reg->ptr_i > 0);
+ (zh->reg->key_buf)[zh->reg->ptr_top - zh->reg->ptr_i] =
+ (char*)zh->reg->key_buf + zh->reg->key_buf_used;
+
+ ch = (int) key_in.mem[0]; /* ordinal for field/use/attribute */
+
+ zh->reg->key_buf_used +=
+ key_SU_encode(ch, ((char*)zh->reg->key_buf) +
+ zh->reg->key_buf_used);
+ while (*src)
+ ((char*)zh->reg->key_buf) [(zh->reg->key_buf_used)++] = *src++;
+ src++;
+ ((char*)(zh->reg->key_buf))[(zh->reg->key_buf_used)++] = '\0';
+ ((char*)(zh->reg->key_buf))[(zh->reg->key_buf_used)++] = cmd;
+
+ if (zh->m_staticrank) /* rank config enabled ? */
+ {
+ *keyp++ = staticrank;
+ key_out.len = 4;
+ }
+ else
+ key_out.len = 3;
+
+ if (key_in.mem[1]) /* filter specified record ID */
+ *keyp++ = key_in.mem[1];
+ else
+ *keyp++ = sysno;
+ *keyp++ = key_in.mem[2]; /* section_id */
+ *keyp++ = key_in.mem[3]; /* sequence .. */
+
+ memcpy ((char*)zh->reg->key_buf + zh->reg->key_buf_used,
+ &key_out, sizeof(key_out));
+ (zh->reg->key_buf_used) += sizeof(key_out);
+ off = src - reckeys->buf;
+ }
+ assert (off == reckeys->buf_used);
+ iscz1_stop(decode_handle);
+#endif
+}
+
+void extract_flushWriteKeys (ZebraHandle zh, int final)
+ /* optimizing: if final=1, and no files written yet */
+ /* push the keys directly to merge, sidestepping the */
+ /* temp file altogether. Speeds small updates */
+{
+ FILE *outf;
+ char out_fname[200];
+ char *prevcp, *cp;
+ struct encode_info encode_info;
+ int ptr_i = zh->reg->ptr_i;
+ int temp_policy;
+#if SORT_EXTRA
+ int i;
+#endif
+ if (!zh->reg->key_buf || ptr_i <= 0)
+ {
+ yaz_log (YLOG_DEBUG, " nothing to flush section=%d buf=%p i=%d",
+ zh->reg->key_file_no, zh->reg->key_buf, ptr_i);
+ yaz_log (YLOG_DEBUG, " buf=%p ",
+ zh->reg->key_buf);
+ yaz_log (YLOG_DEBUG, " ptr=%d ",zh->reg->ptr_i);
+ yaz_log (YLOG_DEBUG, " reg=%p ",zh->reg);
+
+ return;
+ }
+
+ (zh->reg->key_file_no)++;
+ yaz_log (YLOG_LOG, "sorting section %d", (zh->reg->key_file_no));
+ yaz_log (YLOG_DEBUG, " sort_buff at %p n=%d",
+ zh->reg->key_buf + zh->reg->ptr_top - ptr_i,ptr_i);
+#if !SORT_EXTRA
+ qsort (zh->reg->key_buf + zh->reg->ptr_top - ptr_i, ptr_i,
+ sizeof(char*), key_qsort_compare);
+
+ /* zebra.cfg: tempfiles:
+ Y: always use temp files (old way)
+ A: use temp files, if more than one (auto)
+ = if this is both the last and the first
+ N: never bother with temp files (new) */
+
+ temp_policy=toupper(res_get_def(zh->res,"tempfiles","auto")[0]);
+ if (temp_policy != 'Y' && temp_policy != 'N' && temp_policy != 'A') {
+ yaz_log (YLOG_WARN, "Illegal tempfiles setting '%c'. using 'Auto' ",
+ temp_policy);
+ temp_policy='A';
+ }
+
+ if ( ( temp_policy =='N' ) || /* always from memory */
+ ( ( temp_policy =='A' ) && /* automatic */
+ (zh->reg->key_file_no == 1) && /* this is first time */
+ (final) ) ) /* and last (=only) time */
+ { /* go directly from memory */
+ zh->reg->key_file_no =0; /* signal not to read files */
+ zebra_index_merge(zh);
+ zh->reg->ptr_i = 0;
+ zh->reg->key_buf_used = 0;
+ return;
+ }
+
+ /* Not doing directly from memory, write into a temp file */
+ extract_get_fname_tmp (zh, out_fname, zh->reg->key_file_no);
+
+ if (!(outf = fopen (out_fname, "wb")))
+ {
+ yaz_log (YLOG_FATAL|YLOG_ERRNO, "fopen %s", out_fname);
+ exit (1);
+ }
+ yaz_log (YLOG_LOG, "writing section %d", zh->reg->key_file_no);
+ prevcp = cp = (zh->reg->key_buf)[zh->reg->ptr_top - ptr_i];
+
+ encode_key_init (&encode_info);
+ encode_key_write (cp, &encode_info, outf);
+
+ while (--ptr_i > 0)
+ {
+ cp = (zh->reg->key_buf)[zh->reg->ptr_top - ptr_i];
+ if (strcmp (cp, prevcp))
+ {
+ encode_key_flush ( &encode_info, outf);
+ encode_key_init (&encode_info);
+ encode_key_write (cp, &encode_info, outf);
+ prevcp = cp;
+ }
+ else
+ encode_key_write (cp + strlen(cp), &encode_info, outf);
+ }
+ encode_key_flush ( &encode_info, outf);
+#else
+ qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_x_compare);
+ extract_get_fname_tmp (out_fname, key_file_no);
+
+ if (!(outf = fopen (out_fname, "wb")))
+ {
+ yaz_log (YLOG_FATAL|YLOG_ERRNO, "fopen %s", out_fname);
+ exit (1);
+ }
+ yaz_log (YLOG_LOG, "writing section %d", key_file_no);
+ i = ptr_i;
+ prevcp = key_buf[ptr_top-i];
+ while (1)
+ if (!--i || strcmp (prevcp, key_buf[ptr_top-i]))
+ {
+ key_y_len = strlen(prevcp)+1;
+#if 0
+ yaz_log (YLOG_LOG, "key_y_len: %2d %02x %02x %s",
+ key_y_len, prevcp[0], prevcp[1], 2+prevcp);
+#endif
+ qsort (key_buf + ptr_top-ptr_i, ptr_i - i,
+ sizeof(char*), key_y_compare);
+ cp = key_buf[ptr_top-ptr_i];
+ --key_y_len;
+ encode_key_init (&encode_info);
+ encode_key_write (cp, &encode_info, outf);
+ while (--ptr_i > i)
+ {
+ cp = key_buf[ptr_top-ptr_i];
+ encode_key_write (cp+key_y_len, &encode_info, outf);
+ }
+ encode_key_flush ( &encode_info, outf);
+ if (!i)
+ break;
+ prevcp = key_buf[ptr_top-ptr_i];
+ }
+#endif
+ if (fclose (outf))
+ {
+ yaz_log (YLOG_FATAL|YLOG_ERRNO, "fclose %s", out_fname);
+ exit (1);
+ }
+ yaz_log (YLOG_LOG, "finished section %d", zh->reg->key_file_no);
+ zh->reg->ptr_i = 0;
+ zh->reg->key_buf_used = 0;
+}
+
+#if NEW_REC_KEYS
+void extract_add_it_key (ZebraHandle zh,
+ zebra_rec_keys_t *keys,
+ int reg_type,
+ const char *str, int slen, struct it_key *key)
+{
+ zebra_rec_keys_write(*keys, reg_type, str, slen, key);
+}
+#else
+void extract_add_it_key (ZebraHandle zh,
+ struct recKeys *keys,
+ int reg_type,
+ const char *str, int slen, struct it_key *key)
+{
+ char *dst;
+ const char *src = (char*) key;
+
+ if (keys->buf_used+1024 > keys->buf_max)
+ {
+ char *b = (char *) xmalloc (keys->buf_max += 128000);
+ if (keys->buf_used > 0)
+ memcpy (b, keys->buf, keys->buf_used);
+ xfree (keys->buf);
+ keys->buf = b;
+ }
+ dst = keys->buf + keys->buf_used;
+
+ iscz1_encode(keys->codec_handle, &dst, &src);
+
+#if REG_TYPE_PREFIX
+ *dst++ = reg_type;
+#endif
+ memcpy (dst, str, slen);
+ dst += slen;
+ *dst++ = '\0';
+ keys->buf_used = dst - keys->buf;
+}
+#endif
+
+ZEBRA_RES zebra_snippets_rec_keys(ZebraHandle zh,
+#if NEW_REC_KEYS
+ zebra_rec_keys_t reckeys,
+#else
+ struct recKeys *reckeys,
+#endif
+ zebra_snippets *snippets)
+{
+ NMEM nmem = nmem_create();
+
+#if NEW_REC_KEYS
+ if (zebra_rec_keys_rewind(reckeys))
+ {
+ const char *str;
+ size_t slen;
+ struct it_key key;
+ while (zebra_rec_keys_read(reckeys, &str, &slen, &key))
+ {
+ char dst_buf[IT_MAX_WORD];
+ char *dst_term = dst_buf;
+ int ord, seqno;
+ int index_type;
+ assert(key.len <= 4 && key.len > 2);
+ seqno = (int) key.mem[key.len-1];
+ ord = key.mem[0];
+
+ zebraExplain_lookup_ord(zh->reg->zei, ord, &index_type,
+ 0/* db */, 0/* set */, 0/* use */);
+ assert(index_type);
+ zebra_term_untrans_iconv(zh, nmem, index_type,
+ &dst_term, str);
+ zebra_snippets_append(snippets, seqno, ord, dst_term);
+ nmem_reset(nmem);
+ }
+ }
+#else
+ int off = 0;
+ int seqno = 0;
+ void *decode_handle = iscz1_start();
+ assert(reckeys->buf);
+ while (off < reckeys->buf_used)
+ {
+ const char *src = reckeys->buf + off;
+ struct it_key key;
+ char *dst = (char*) &key;
+ char dst_buf[IT_MAX_WORD];
+ char *dst_term = dst_buf;
+ int index_type = 0, ord;
+
+ iscz1_decode(decode_handle, &dst, &src);
+ assert(key.len <= 4 && key.len > 2);
+
+ seqno = (int) key.mem[key.len-1];
+ ord = key.mem[0];
+
+ zebraExplain_lookup_ord(zh->reg->zei, ord, &index_type,
+ 0/* db */, 0/* set */, 0/* use */);
+ assert(index_type);
+ zebra_term_untrans_iconv(zh, nmem, index_type,
+ &dst_term, src);
+ zebra_snippets_append(snippets, seqno, ord, dst_term);
+ while (*src++)
+ ;
+ off = src - reckeys->buf;
+ nmem_reset(nmem);
+ }
+ iscz1_stop(decode_handle);
+#endif
+ nmem_destroy(nmem);
+ return ZEBRA_OK;
+}
+
+void print_rec_keys(ZebraHandle zh,
+#if NEW_REC_KEYS
+ zebra_rec_keys_t reckeys
+#else
+ struct recKeys *reckeys
+#endif
+)
+{
+#if NEW_REC_KEYS
+ yaz_log(YLOG_LOG, "print_rec_keys");
+ if (zebra_rec_keys_rewind(reckeys))
+ {
+ const char *str;
+ size_t slen;
+ struct it_key key;
+ while (zebra_rec_keys_read(reckeys, &str, &slen, &key))
+ {
+ char dst_buf[IT_MAX_WORD];
+ int seqno;
+ int index_type;
+ const char *db = 0;
+ assert(key.len <= 4 && key.len > 2);
+
+ zebraExplain_lookup_ord(zh->reg->zei,
+ key.mem[0], &index_type, &db, 0, 0);
+
+ seqno = (int) key.mem[key.len-1];
+
+ zebra_term_untrans(zh, index_type, dst_buf, str);
+
+ yaz_log(YLOG_LOG, "ord=" ZINT_FORMAT " seqno=%d term=%s",
+ key.mem[0], seqno, dst_buf);
+ }
+ }
+#else
+ int off = 0;
+ int seqno = 0;
+ void *decode_handle = iscz1_start();
+ yaz_log(YLOG_LOG, "print_rec_keys buf=%p sz=%d", reckeys->buf,
+ reckeys->buf_used);
+ assert(reckeys->buf);
+ while (off < reckeys->buf_used)
+ {
+ const char *src = reckeys->buf + off;
+ struct it_key key;
+ char *dst = (char*) &key;
+ char dst_buf[IT_MAX_WORD];
+ int index_type;
+ const char *db = 0;
+
+ iscz1_decode(decode_handle, &dst, &src);
+ assert(key.len <= 4 && key.len > 2);
+
+ seqno = (int) key.mem[key.len-1];
+
+ zebraExplain_lookup_ord(zh->reg->zei,
+ key.mem[0], &index_type, &db, 0, 0);
+
+ zebra_term_untrans(zh, index_type, dst_buf, src);
+
+ yaz_log(YLOG_LOG, "ord=" ZINT_FORMAT " seqno=%d term=%s",
+ key.mem[0], seqno, dst_buf);
+ while (*src++)
+ ;
+ off = src - reckeys->buf;
+ }
+ iscz1_stop(decode_handle);
+#endif
+}
+
+void extract_add_index_string (RecWord *p, const char *str, int length)
+{
+ struct it_key key;
+
+ ZebraHandle zh = p->extractCtrl->handle;
+ ZebraExplainInfo zei = zh->reg->zei;
+ int ch;
+
+ if (p->index_name)
+ {
+ ch = zebraExplain_lookup_attr_str(zei, p->index_type, p->index_name);
+ if (ch < 0)
+ ch = zebraExplain_add_attr_str(zei, p->index_type, p->index_name);
+ }
+ else
+ {
+#if NATTR
+ return;
+#else
+ ch = zebraExplain_lookup_attr_su(zei, p->index_type,
+ p->attrSet, p->attrUse);
+ if (ch < 0)
+ ch = zebraExplain_add_attr_su(zei, p->index_type,
+ p->attrSet, p->attrUse);
+#endif
+ }
+ key.len = 4;
+ key.mem[0] = ch;
+ key.mem[1] = p->record_id;
+ key.mem[2] = p->section_id;
+ key.mem[3] = p->seqno;
+
+#if 0
+ /* just for debugging .. */
+ yaz_log(YLOG_LOG, "add: set=%d use=%d "
+ "record_id=%lld section_id=%lld seqno=%lld",
+ p->attrSet, p->attrUse, p->record_id, p->section_id, p->seqno);
+#endif
+
+ extract_add_it_key(p->extractCtrl->handle,
+ &zh->reg->keys,
+ p->index_type, str,
+ length, &key);
+}
+
+#if NATTR
+static void extract_add_sort_string (RecWord *p, const char *str, int length)
+{
+ struct it_key key;
+
+ ZebraHandle zh = p->extractCtrl->handle;
+ ZebraExplainInfo zei = zh->reg->zei;
+ int ch;
+
+ if (p->index_name)
+ {
+ ch = zebraExplain_lookup_attr_str(zei, p->index_type, p->index_name);
+ if (ch < 0)
+ ch = zebraExplain_add_attr_str(zei, p->index_type, p->index_name);
+ }
+ else
+ {
+ return;
+ }
+ key.len = 4;
+ key.mem[0] = ch;
+ key.mem[1] = p->record_id;
+ key.mem[2] = p->section_id;
+ key.mem[3] = p->seqno;
+
+ extract_add_it_key(p->extractCtrl->handle,
+ &zh->reg->sortKeys,
+ p->index_type, str,
+ length, &key);
+}
+#else
+static void extract_add_sort_string (RecWord *p, const char *str, int length)
+{
+ ZebraHandle zh = p->extractCtrl->handle;
+ struct sortKeys *sk = &zh->reg->sortKeys;
+ int off = 0;
+
+ while (off < sk->buf_used)
+ {
+ int set, use, slen;
+
+ off += key_SU_decode(&set, (unsigned char *) sk->buf + off);
+ off += key_SU_decode(&use, (unsigned char *) sk->buf + off);
+ off += key_SU_decode(&slen, (unsigned char *) sk->buf + off);
+ off += slen;
+ if (p->attrSet == set && p->attrUse == use)
+ return;
+ }
+ assert (off == sk->buf_used);
+
+ if (sk->buf_used + IT_MAX_WORD > sk->buf_max)
+ {
+ char *b;
+
+ b = (char *) xmalloc (sk->buf_max += 128000);
+ if (sk->buf_used > 0)
+ memcpy (b, sk->buf, sk->buf_used);
+ xfree (sk->buf);
+ sk->buf = b;
+ }
+ off += key_SU_encode(p->attrSet, sk->buf + off);
+ off += key_SU_encode(p->attrUse, sk->buf + off);
+ off += key_SU_encode(length, sk->buf + off);
+ memcpy (sk->buf + off, str, length);
+ sk->buf_used = off + length;
+}
+#endif
+
+void extract_add_string (RecWord *p, const char *string, int length)
+{
+ assert (length > 0);
+ if (zebra_maps_is_sort (p->zebra_maps, p->index_type))
+ extract_add_sort_string (p, string, length);
+ else
+ extract_add_index_string (p, string, length);
+}
+
+static void extract_add_incomplete_field (RecWord *p)
+{
+ const char *b = p->term_buf;
+ int remain = p->term_len;
+ const char **map = 0;
+
+ yaz_log(YLOG_DEBUG, "Incomplete field, w='%.*s'", p->term_len, p->term_buf);
+
+ if (remain > 0)
+ map = zebra_maps_input(p->zebra_maps, p->index_type, &b, remain, 0);
+
+ while (map)
+ {
+ char buf[IT_MAX_WORD+1];
+ int i, remain;
+
+ /* Skip spaces */
+ while (map && *map && **map == *CHR_SPACE)
+ {
+ remain = p->term_len - (b - p->term_buf);
+ if (remain > 0)
+ map = zebra_maps_input(p->zebra_maps, p->index_type, &b,
+ remain, 0);
+ else
+ map = 0;
+ }
+ if (!map)
+ break;
+ 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);
+ if (remain > 0)
+ map = zebra_maps_input(p->zebra_maps, p->index_type, &b, remain, 0);
+ else
+ map = 0;
+ }
+ if (!i)
+ return;
+ extract_add_string (p, buf, i);
+ p->seqno++;
+ }
+}
+
+static void extract_add_complete_field (RecWord *p)
+{
+ const char *b = p->term_buf;
+ char buf[IT_MAX_WORD+1];
+ const char **map = 0;
+ int i = 0, remain = p->term_len;
+
+ yaz_log(YLOG_DEBUG, "Complete field, w='%.*s'",
+ p->term_len, p->term_buf);
+
+ if (remain > 0)
+ map = zebra_maps_input (p->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 (remain > 0)
+ {
+ int first = i ? 0 : 1; /* first position */
+ map = zebra_maps_input(p->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;
+ yaz_log(YLOG_DEBUG, "Adding string to index '%d'", **map);
+ while (i < IT_MAX_WORD && *cp)
+ buf[i++] = *(cp++);
+ }
+ remain = p->term_len - (b - p->term_buf);
+ if (remain > 0)
+ {
+ map = zebra_maps_input (p->zebra_maps, p->index_type, &b,
+ remain, 0);
+ }
+ else
+ map = 0;
+ }
+ }
+ if (!i)
+ return;
+ extract_add_string (p, buf, i);
+}
+
+void extract_token_add (RecWord *p)
+{
+ WRBUF wrbuf;
+#if 0
+ yaz_log (YLOG_LOG, "token_add "
+ "reg_type=%c attrSet=%d attrUse=%d seqno=%d s=%.*s",
+ p->reg_type, p->attrSet, p->attrUse, p->seqno, p->length,
+ p->string);
+#endif
+ if ((wrbuf = zebra_replace(p->zebra_maps, p->index_type, 0,
+ p->term_buf, p->term_len)))
+ {
+ p->term_buf = wrbuf_buf(wrbuf);
+ p->term_len = wrbuf_len(wrbuf);
+ }
+ if (zebra_maps_is_complete (p->zebra_maps, p->index_type))
+ extract_add_complete_field (p);
+ else
+ extract_add_incomplete_field(p);
+}
+
+static void extract_set_store_data_cb(struct recExtractCtrl *p,
+ void *buf, size_t sz)
+{
+ ZebraHandle zh = (ZebraHandle) p->handle;
+
+ xfree(zh->store_data_buf);
+ zh->store_data_buf = 0;
+ zh->store_data_size = 0;
+ if (buf && sz)
+ {
+ zh->store_data_buf = xmalloc(sz);
+ zh->store_data_size = sz;
+ memcpy(zh->store_data_buf, buf, sz);
+ }
+}
+
+static void extract_set_store_data_prepare(struct recExtractCtrl *p)
+{
+ ZebraHandle zh = (ZebraHandle) p->handle;
+ xfree(zh->store_data_buf);
+ zh->store_data_buf = 0;
+ zh->store_data_size = 0;
+ p->setStoreData = extract_set_store_data_cb;
+}
+
+void extract_schema_add (struct recExtractCtrl *p, Odr_oid *oid)
+{
+ ZebraHandle zh = (ZebraHandle) p->handle;
+ zebraExplain_addSchema (zh->reg->zei, oid);
+}
+
+#if NATTR
+void extract_flushSortKeys (ZebraHandle zh, SYSNO sysno,
+ int cmd, struct recKeys *reckeys)
+{
+ SortIdx sortIdx = zh->reg->sortIdx;
+ void *decode_handle = iscz1_start();
+ int off = 0;
+ int ch = 0;
+
+ while (off < reckeys->buf_used)
+ {
+ const char *src = reckeys->buf + off;
+ struct it_key key;
+ char *dst = (char*) &key;
+
+ iscz1_decode(decode_handle, &dst, &src);
+ assert(key.len == 4);
+
+ ch = (int) key.mem[0]; /* ordinal for field/use/attribute */
+
+ sortIdx_type(sortIdx, ch);
+ if (cmd == 1)
+ sortIdx_add(sortIdx, src, strlen(src));
+ else
+ sortIdx_add(sortIdx, "", 1);
+
+ src += strlen(src);
+ src++;
+
+ off = src - reckeys->buf;
+ }
+ assert (off == reckeys->buf_used);
+ iscz1_stop(decode_handle);
+}
+#else
+void extract_flushSortKeys (ZebraHandle zh, SYSNO sysno,
+ int cmd, struct sortKeys *sk)
+{
+ SortIdx sortIdx = zh->reg->sortIdx;
+ int off = 0;
+
+ sortIdx_sysno (sortIdx, sysno);
+
+ while (off < sk->buf_used)
+ {
+ int set, use, slen;
+
+ off += key_SU_decode(&set, (unsigned char *) sk->buf + off);
+ off += key_SU_decode(&use, (unsigned char *) sk->buf + off);
+ off += key_SU_decode(&slen, (unsigned char *) sk->buf + off);
+
+ sortIdx_type(sortIdx, use);
+ if (cmd == 1)
+ sortIdx_add(sortIdx, sk->buf + off, slen);
+ else
+ sortIdx_add(sortIdx, "", 1);
+ off += slen;
+ }
+}
+#endif
+
+void encode_key_init (struct encode_info *i)
+{
+ i->sysno = 0;
+ i->seqno = 0;
+ i->cmd = -1;
+ i->prevsys=0;
+ i->prevseq=0;
+ i->prevcmd=-1;
+ i->keylen=0;
+ i->encode_handle = iscz1_start();
+}
+
+#define OLDENCODE 1
+
+#ifdef OLDENCODE
+/* this is the old encode_key_write
+ * may be deleted once we are confident that the new works
+ * HL 15-oct-2002
+ */
+void encode_key_write (char *k, struct encode_info *i, FILE *outf)
+{
+ struct it_key key;
+ char *bp = i->buf, *bp0;
+ const char *src = (char *) &key;
+
+ /* copy term to output buf */
+ while ((*bp++ = *k++))
+ ;
+ /* and copy & align key so we can mangle */
+ memcpy (&key, k+1, sizeof(struct it_key)); /* *k is insert/delete */
+
+ bp0 = bp++;
+ iscz1_encode(i->encode_handle, &bp, &src);
+ *bp0 = (*k * 128) + bp - bp0 - 1; /* length and insert/delete combined */
+ if (fwrite (i->buf, bp - i->buf, 1, outf) != 1)
+ {
+ yaz_log (YLOG_FATAL|YLOG_ERRNO, "fwrite");
+ exit (1);
+ }
+}
+
+void encode_key_flush (struct encode_info *i, FILE *outf)
+{ /* dummy routine */
+ iscz1_stop(i->encode_handle);
+}
+
+#else
+
+/* new encode_key_write
+ * The idea is to buffer one more key, and compare them
+ * If we are going to delete and insert the same key,
+ * we may as well not bother. Should make a difference in
+ * updates with small modifications (appending to a mbox)
+ */
+void encode_key_write (char *k, struct encode_info *i, FILE *outf)
+{
+ struct it_key key;
+ char *bp;
+
+ if (*k) /* first time for new key */
+ {
+ bp = i->buf;
+ while ((*bp++ = *k++))
+ ;
+ i->keylen= bp - i->buf -1;
+ assert(i->keylen+1+sizeof(struct it_key) < ENCODE_BUFLEN);
+ }
+ else
+ {
+ bp=i->buf + i->keylen;
+ *bp++=0;
+ k++;
+ }
+
+ memcpy (&key, k+1, sizeof(struct it_key));
+ if (0==i->prevsys) /* no previous filter, fill up */
+ {
+ i->prevsys=key.sysno;
+ i->prevseq=key.seqno;
+ i->prevcmd=*k;
+ }
+ else if ( (i->prevsys==key.sysno) &&
+ (i->prevseq==key.seqno) &&
+ (i->prevcmd!=*k) )
+ { /* same numbers, diff cmd, they cancel out */
+ i->prevsys=0;
+ }
+ else
+ { /* different stuff, write previous, move buf */
+ bp = encode_key_int ( (i->prevsys - i->sysno) * 2 + i->prevcmd, bp);
+ if (i->sysno != i->prevsys)
+ {
+ i->sysno = i->prevsys;
+ i->seqno = 0;
+ }
+ else if (!i->seqno && !i->prevseq && i->cmd == i->prevcmd)
+ {
+ return; /* ??? Filters some sort of duplicates away */
+ /* ??? Can this ever happen -H 15oct02 */
+ }
+ bp = encode_key_int (i->prevseq - i->seqno, bp);
+ i->seqno = i->prevseq;
+ i->cmd = i->prevcmd;
+ if (fwrite (i->buf, bp - i->buf, 1, outf) != 1)
+ {
+ yaz_log (YLOG_FATAL|YLOG_ERRNO, "fwrite");
+ exit (1);
+ }
+ i->keylen=0; /* ok, it's written, forget it */
+ i->prevsys=key.sysno;
+ i->prevseq=key.seqno;
+ i->prevcmd=*k;
+ }
+}
+
+void encode_key_flush (struct encode_info *i, FILE *outf)
+{ /* flush the last key from i */
+ char *bp =i->buf + i->keylen;
+ if (0==i->prevsys)
+ {
+ return; /* nothing to flush */
+ }
+ *bp++=0;
+ bp = encode_key_int ( (i->prevsys - i->sysno) * 2 + i->prevcmd, bp);
+ if (i->sysno != i->prevsys)
+ {
+ i->sysno = i->prevsys;
+ i->seqno = 0;
+ }
+ else if (!i->seqno && !i->prevseq && i->cmd == i->prevcmd)
+ {
+ return; /* ??? Filters some sort of duplicates away */
+ /* ??? Can this ever happen -H 15oct02 */
+ }
+ bp = encode_key_int (i->prevseq - i->seqno, bp);
+ i->seqno = i->prevseq;
+ i->cmd = i->prevcmd;
+ if (fwrite (i->buf, bp - i->buf, 1, outf) != 1)
+ {
+ yaz_log (YLOG_FATAL|YLOG_ERRNO, "fwrite");
+ exit (1);
+ }
+ i->keylen=0; /* ok, it's written, forget it */
+ i->prevsys=0; /* forget the values too */
+ i->prevseq=0;
+}
+#endif