2 * Copyright (c) 1995-2000, Index Data.
3 * See the file LICENSE for details.
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.34 2000-12-05 10:06:23 adam
8 * Added support for null-data rules like <tag/>.
10 * Revision 1.33 2000/11/29 14:22:47 adam
11 * Implemented XML/SGML attributes for data1 so that d1_read reads them
12 * and d1_write generates proper attributes for XML/SGML records. Added
13 * register locking for threaded version.
15 * Revision 1.32 2000/01/06 11:25:59 adam
16 * Added case to prevent warning.
18 * Revision 1.31 1999/12/21 14:16:20 ian
19 * Changed retrieval module to allow data1 trees with no associated absyn.
20 * Also added a simple interface for extracting values from data1 trees using
21 * a string based tagpath.
23 * Revision 1.30 1999/11/30 13:47:12 adam
24 * Improved installation. Moved header files to include/yaz.
26 * Revision 1.29 1999/10/21 12:06:29 adam
27 * Retrieval module no longer uses ctype.h - functions.
29 * Revision 1.28 1999/10/21 09:50:33 adam
30 * SGML reader uses own isspace - it doesn't do 8-bit on WIN32!
32 * Revision 1.27 1999/08/27 09:40:32 adam
33 * Renamed logf function to yaz_log. Removed VC++ project files.
35 * Revision 1.26 1999/07/13 13:23:48 adam
36 * Non-recursive version of data1_read_node. data1_read_nodex reads
37 * stream of bytes (instead of buffer in memory).
39 * Revision 1.25 1999/04/20 09:56:48 adam
40 * Added 'name' paramter to encoder/decoder routines (typedef Odr_fun).
41 * Modified all encoders/decoders to reflect this change.
43 * Revision 1.24 1998/10/28 15:10:09 adam
44 * Added --with-yc option to configure. For the data1_node in data1.h:
45 * decreased size of localdata and removed member "line" which wasn't useful.
47 * Revision 1.23 1998/03/12 11:28:45 adam
48 * Fix: didn't set root member of tagged node in function.
49 * data1_add_insert_taggeddata.
51 * Revision 1.22 1998/03/05 08:15:32 adam
52 * Implemented data1_add_insert_taggeddata utility which is more flexible
53 * than data1_insert_taggeddata.
55 * Revision 1.21 1998/02/27 14:08:05 adam
56 * Added const to some char pointer arguments.
57 * Reworked data1_read_node so that it doesn't create a tree with
58 * pointers to original "SGML"-buffer.
60 * Revision 1.20 1998/02/11 11:53:35 adam
61 * Changed code so that it compiles as C++.
63 * Revision 1.19 1997/12/09 16:17:09 adam
64 * Fix bug regarding variants. Tags with prefix "var" was incorrectly
65 * interpreted as "start of variants". Now, only "var" indicates such
67 * Cleaned up data1_read_node so tag names and variant names are
68 * copied and not pointed to by the generated data1 tree. Data nodes
69 * still point to old buffer.
71 * Revision 1.18 1997/11/18 09:51:09 adam
72 * Removed element num_children from data1_node. Minor changes in
75 * Revision 1.17 1997/11/05 09:20:51 adam
78 * Revision 1.16 1997/09/17 12:10:37 adam
81 * Revision 1.15 1997/09/05 09:50:57 adam
82 * Removed global data1_tabpath - uses data1_get_tabpath() instead.
84 * Revision 1.14 1997/05/14 06:54:04 adam
87 * Revision 1.13 1996/10/29 13:35:38 adam
88 * Implemented data1_set_tabpath and data1_get_tabpath.
90 * Revision 1.12 1996/10/11 10:35:38 adam
91 * Fixed a bug that caused data1_read_node to core dump when no abstract
92 * syntax was defined in a "sgml"-record.
94 * Revision 1.11 1996/07/06 19:58:35 quinn
95 * System headerfiles gathered in yconfig
97 * Revision 1.10 1996/01/19 15:41:47 quinn
98 * Fixed uninitialized boolean.
100 * Revision 1.9 1996/01/17 14:52:47 adam
101 * Changed prototype for reader function parsed to data1_read_record.
103 * Revision 1.8 1995/12/15 16:20:41 quinn
104 * Added formatted text.
106 * Revision 1.7 1995/12/13 13:44:32 quinn
107 * Modified Data1-system to use nmem
109 * Revision 1.6 1995/12/12 16:37:08 quinn
110 * Added destroy element to data1_node.
112 * Revision 1.5 1995/12/11 15:22:37 quinn
113 * Added last_child field to the node.
114 * Rewrote schema-mapping.
116 * Revision 1.4 1995/11/13 09:27:36 quinn
117 * Fiddling with the variant stuff.
119 * Revision 1.3 1995/11/01 16:34:57 quinn
120 * Making data1 look for tables in data1_tabpath
122 * Revision 1.2 1995/11/01 13:54:48 quinn
125 * Revision 1.1 1995/11/01 11:56:09 quinn
126 * Added Retrieval (data management) functions en masse.
128 * Revision 1.14 1995/10/30 12:40:55 quinn
129 * Fixed a couple of bugs.
131 * Revision 1.13 1995/10/25 16:00:47 quinn
132 * USMARC support is now almost operational
134 * Revision 1.12 1995/10/16 14:02:55 quinn
135 * Changes to support element set names and espec1
137 * Revision 1.11 1995/10/13 16:05:08 quinn
138 * Adding Espec1-processing
140 * Revision 1.10 1995/10/11 14:53:44 quinn
143 * Revision 1.9 1995/10/06 16:56:50 quinn
144 * Fixed ranked result.
146 * Revision 1.8 1995/10/06 16:44:13 quinn
147 * Work on attribute set mapping, etc.
149 * Revision 1.7 1995/10/06 12:58:35 quinn
152 * Revision 1.6 1995/10/04 09:29:49 quinn
153 * Adjustments to support USGS test data
155 * Revision 1.5 1995/10/03 17:56:43 quinn
158 * Revision 1.4 1995/10/02 15:53:19 quinn
161 * Revision 1.3 1995/10/02 14:55:21 quinn
162 * *** empty log message ***
164 * Revision 1.2 1995/09/14 15:18:13 quinn
167 * Revision 1.1 1995/09/12 11:24:30 quinn
168 * Beginning to add code for structured records.
177 #include <yaz/xmalloc.h>
179 #include <yaz/data1.h>
182 * get the tag which is the immediate parent of this node (this may mean
183 * traversing intermediate things like variants and stuff.
185 data1_node *get_parent_tag (data1_handle dh, data1_node *n)
187 for (; n && n->which != DATA1N_root; n = n->parent)
188 if (n->which == DATA1N_tag)
193 data1_node *data1_mk_node (data1_handle dh, NMEM m)
197 r = (data1_node *)nmem_malloc(m, sizeof(*r));
198 r->next = r->child = r->last_child = r->parent = 0;
203 data1_node *data1_mk_node_type (data1_handle dh, NMEM m, int type)
207 r = data1_mk_node(dh, m);
213 r->u.tag.element = 0;
214 r->u.tag.no_data_requested = 0;
215 r->u.tag.node_selected = 0;
216 r->u.tag.make_variantlist = 0;
217 r->u.tag.get_bytes = -1;
218 #if DATA1_USING_XATTR
219 r->u.tag.attributes = 0;
230 r->u.data.formatted_text = 0;
233 logf (LOG_WARN, "data_mk_node_type. bad type = %d\n", type);
238 void data1_free_tree (data1_handle dh, data1_node *t)
240 data1_node *p = t->child, *pn;
245 data1_free_tree (dh, p);
252 char *data1_insert_string (data1_handle dh, data1_node *res,
253 NMEM m, const char *str)
255 int len = strlen(str);
257 if (len >= DATA1_LOCALDATA)
258 return nmem_strdup (m, str);
261 strcpy (res->lbuf, str);
266 data1_node *data1_add_insert_taggeddata(data1_handle dh, data1_node *root,
268 const char *tagname, NMEM m,
269 int first_flag, int local_allowed)
271 data1_node *partag = get_parent_tag (dh, at);
272 data1_node *tagn = data1_mk_node_type (dh, m, DATA1N_tag);
273 data1_element *e = NULL;
276 tagn->u.tag.tag = data1_insert_string (dh, tagn, m, tagname);
279 e = partag->u.tag.element;
280 tagn->u.tag.element =
281 data1_getelementbytagname (dh, root->u.root.absyn, e, tagname);
282 if (!local_allowed && !tagn->u.tag.element)
284 tagn->last_child = tagn->child = datn = data1_mk_node_type (dh, m, DATA1N_data);
288 datn->u.data.formatted_text = 0;
293 tagn->next = at->child;
295 at->last_child = tagn;
304 assert (at->last_child);
305 at->last_child->next = tagn;
307 at->last_child = tagn;
312 data1_node *data1_add_taggeddata(data1_handle dh, data1_node *root,
314 const char *tagname, NMEM m)
316 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 0, 1);
321 * Insert a tagged node into the record root as first child of the node at
322 * which should be root or tag itself). Returns pointer to the data node,
323 * which can then be modified.
325 data1_node *data1_insert_taggeddata(data1_handle dh, data1_node *root,
327 const char *tagname, NMEM m)
329 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 1, 0);
332 #if DATA1_USING_XATTR
333 data1_xattr *data1_read_xattr (data1_handle dh, NMEM m,
334 int (*get_byte)(void *fh), void *fh,
335 WRBUF wrbuf, int *ch)
337 data1_xattr *p_first = 0;
338 data1_xattr **pp = &p_first;
344 while (c && d1_isspace(c))
346 if (!c || c == '>' || c == '/')
348 *pp = p = nmem_malloc (m, sizeof(*p));
354 while (c && c != '=' && c != '>' && c != '/' && !d1_isspace(c))
356 wrbuf_putc (wrbuf, c);
359 wrbuf_putc (wrbuf, '\0');
360 len = wrbuf_len(wrbuf);
361 p->name = nmem_malloc (m, len);
362 strcpy (p->name, wrbuf_buf(wrbuf));
369 while (c && c != '"' && c != '>' && c != '/')
371 wrbuf_putc (wrbuf, c);
374 wrbuf_putc (wrbuf, '\0');
375 len = wrbuf_len(wrbuf);
376 p->value = nmem_malloc (m, len);
377 strcpy (p->value, wrbuf_buf(wrbuf));
388 * Ugh. Sometimes functions just grow and grow on you. This one reads a
389 * 'node' and its children.
391 data1_node *data1_read_nodex (data1_handle dh, NMEM m,
392 int (*get_byte)(void *fh), void *fh, WRBUF wrbuf)
394 data1_absyn *absyn = 0;
395 data1_node *d1_stack[256];
405 data1_node *parent = level ? d1_stack[level-1] : 0;
406 while (c != '\0' && d1_isspace(c))
415 if (c == '<') /* beginning of tag */
417 #if DATA1_USING_XATTR
432 while (c && c != '>' && c != '/' && !d1_isspace(c))
434 if (i < (sizeof(tag)-1))
439 #if DATA1_USING_XATTR
440 xattr = data1_read_xattr (dh, m, get_byte, fh, wrbuf, &c);
443 while (d1_isspace(c))
445 for (i = 0; c && c != '>' && c != '/'; c = (*get_byte)(fh))
446 if (i < (sizeof(args)-1))
451 { /* <tag attrs/> or <tag/> */
457 yaz_log(LOG_WARN, "d1: %d: Malformed tag", line);
473 parent = d1_stack[--i];
474 if ((parent->which == DATA1N_root &&
475 !strcmp(tag, parent->u.root.type)) ||
476 (parent->which == DATA1N_tag &&
477 !strcmp(tag, parent->u.tag.tag)))
485 yaz_log (LOG_WARN, "%d: no begin tag for %s",
494 if (level == 0) /* root ? */
496 if (!(absyn = data1_get_absyn (dh, tag)))
498 yaz_log(LOG_WARN, "Unable to acquire abstract syntax " "for '%s'", tag);
499 /* It's now OK for a record not to have an absyn */
501 res = data1_mk_node_type (dh, m, DATA1N_root);
502 res->u.root.type = data1_insert_string (dh, res, m, tag);
503 res->u.root.absyn = absyn;
506 else if (!strcmp(tag, "var"))
508 char tclass[DATA1_MAX_SYMBOL], type[DATA1_MAX_SYMBOL];
512 if (sscanf(args, "%s %s %n", tclass, type, &val_offset) != 2)
514 yaz_log(LOG_WARN, "Malformed variant triple at '%s'", tag);
518 data1_getvartypebyct(dh,
519 parent->root->u.root.absyn->varset,
523 * If we're the first variant in this group, create a parent
524 * variant, and insert it before the current variant.
526 if (parent->which != DATA1N_variant)
528 res = data1_mk_node (dh, m);
529 res->which = DATA1N_variant;
530 res->u.variant.type = 0;
531 res->u.variant.value = 0;
536 * now determine if one of our ancestor triples is of
537 * same type. If so, we break here.
540 for (i = level-1; d1_stack[i]->which==DATA1N_variant; --i)
541 if (d1_stack[i]->u.variant.type == tp)
546 res = data1_mk_node (dh, m);
547 res->which = DATA1N_variant;
548 res->u.variant.type = tp;
549 res->u.variant.value =
550 data1_insert_string (dh, res, m, args + val_offset);
553 else /* tag.. acquire our element in the abstract syntax */
555 data1_node *partag = get_parent_tag (dh, parent);
556 data1_element *elem, *e = 0;
559 if (parent->which == DATA1N_variant)
562 if (!(e = partag->u.tag.element))
563 localtag = 1; /* our parent is a local tag */
565 elem = data1_getelementbytagname(dh, absyn, e, tag);
566 res = data1_mk_node_type (dh, m, DATA1N_tag);
567 res->u.tag.tag = data1_insert_string (dh, res, m, tag);
568 res->u.tag.element = elem;
569 #if DATA1_USING_XATTR
570 res->u.tag.attributes = xattr;
575 parent->last_child = res;
576 res->root = parent->root;
578 res->parent = parent;
580 d1_stack[level]->next = res;
583 d1_stack[level] = res;
584 d1_stack[level+1] = 0;
588 else /* != '<'... this is a body of text */
592 int len, prev_char = 0;
599 res = data1_mk_node_type (dh, m, DATA1N_data);
600 res->parent = parent;
601 res->u.data.what = DATA1I_text;
602 res->u.data.formatted_text = 0;
603 res->root = parent->root;
604 parent->last_child = res;
606 d1_stack[level]->next = res;
609 d1_stack[level] = res;
613 while (c && c != '<')
615 wrbuf_putc (wrbuf, c);
618 len = wrbuf_len(wrbuf);
620 /* use local buffer of nmem if too large */
621 if (len >= DATA1_LOCALDATA)
622 res->u.data.data = (char*) nmem_malloc (m, len);
624 res->u.data.data = res->lbuf;
626 /* read "data" and transfer while removing white space */
627 dst = res->u.data.data;
628 for (src = wrbuf_buf(wrbuf); --len >= 0; src++)
632 if (d1_isspace (*src))
644 res->u.data.len = dst - res->u.data.data;
650 int getc_mem (void *fh)
652 const char **p = (const char **) fh;
658 data1_node *data1_read_node (data1_handle dh, const char **buf, NMEM m)
660 WRBUF wrbuf = wrbuf_alloc();
663 node = data1_read_nodex(dh, m, getc_mem, (void *) (buf), wrbuf);
664 wrbuf_free (wrbuf, 1);
669 * Read a record in the native syntax.
671 data1_node *data1_read_record(data1_handle dh,
672 int (*rf)(void *, char *, size_t), void *fh,
676 char **buf = data1_get_read_buf (dh, &size);
681 *buf = (char *)xmalloc(*size = 4096);
685 if (rd + 2048 >= *size && !(*buf =(char *)xrealloc(*buf, *size *= 2)))
687 if ((res = (*rf)(fh, *buf + rd, 2048)) <= 0)
693 return data1_read_node(dh, &bp, m);
702 data1_node *data1_read_sgml (data1_handle dh, NMEM m, const char *buf)
704 const char *bp = buf;
705 return data1_read_node (dh, &bp, m);