2 * Copyright (c) 1995-2000, Index Data.
3 * See the file LICENSE for details.
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.36 2001-02-21 13:46:53 adam
10 * Revision 1.35 2000/12/05 14:44:25 adam
11 * Readers skips <! ...> sections.
13 * Revision 1.34 2000/12/05 10:06:23 adam
14 * Added support for null-data rules like <tag/>.
16 * Revision 1.33 2000/11/29 14:22:47 adam
17 * Implemented XML/SGML attributes for data1 so that d1_read reads them
18 * and d1_write generates proper attributes for XML/SGML records. Added
19 * register locking for threaded version.
21 * Revision 1.32 2000/01/06 11:25:59 adam
22 * Added case to prevent warning.
24 * Revision 1.31 1999/12/21 14:16:20 ian
25 * Changed retrieval module to allow data1 trees with no associated absyn.
26 * Also added a simple interface for extracting values from data1 trees using
27 * a string based tagpath.
29 * Revision 1.30 1999/11/30 13:47:12 adam
30 * Improved installation. Moved header files to include/yaz.
32 * Revision 1.29 1999/10/21 12:06:29 adam
33 * Retrieval module no longer uses ctype.h - functions.
35 * Revision 1.28 1999/10/21 09:50:33 adam
36 * SGML reader uses own isspace - it doesn't do 8-bit on WIN32!
38 * Revision 1.27 1999/08/27 09:40:32 adam
39 * Renamed logf function to yaz_log. Removed VC++ project files.
41 * Revision 1.26 1999/07/13 13:23:48 adam
42 * Non-recursive version of data1_read_node. data1_read_nodex reads
43 * stream of bytes (instead of buffer in memory).
45 * Revision 1.25 1999/04/20 09:56:48 adam
46 * Added 'name' paramter to encoder/decoder routines (typedef Odr_fun).
47 * Modified all encoders/decoders to reflect this change.
49 * Revision 1.24 1998/10/28 15:10:09 adam
50 * Added --with-yc option to configure. For the data1_node in data1.h:
51 * decreased size of localdata and removed member "line" which wasn't useful.
53 * Revision 1.23 1998/03/12 11:28:45 adam
54 * Fix: didn't set root member of tagged node in function.
55 * data1_add_insert_taggeddata.
57 * Revision 1.22 1998/03/05 08:15:32 adam
58 * Implemented data1_add_insert_taggeddata utility which is more flexible
59 * than data1_insert_taggeddata.
61 * Revision 1.21 1998/02/27 14:08:05 adam
62 * Added const to some char pointer arguments.
63 * Reworked data1_read_node so that it doesn't create a tree with
64 * pointers to original "SGML"-buffer.
66 * Revision 1.20 1998/02/11 11:53:35 adam
67 * Changed code so that it compiles as C++.
69 * Revision 1.19 1997/12/09 16:17:09 adam
70 * Fix bug regarding variants. Tags with prefix "var" was incorrectly
71 * interpreted as "start of variants". Now, only "var" indicates such
73 * Cleaned up data1_read_node so tag names and variant names are
74 * copied and not pointed to by the generated data1 tree. Data nodes
75 * still point to old buffer.
77 * Revision 1.18 1997/11/18 09:51:09 adam
78 * Removed element num_children from data1_node. Minor changes in
81 * Revision 1.17 1997/11/05 09:20:51 adam
84 * Revision 1.16 1997/09/17 12:10:37 adam
87 * Revision 1.15 1997/09/05 09:50:57 adam
88 * Removed global data1_tabpath - uses data1_get_tabpath() instead.
90 * Revision 1.14 1997/05/14 06:54:04 adam
93 * Revision 1.13 1996/10/29 13:35:38 adam
94 * Implemented data1_set_tabpath and data1_get_tabpath.
96 * Revision 1.12 1996/10/11 10:35:38 adam
97 * Fixed a bug that caused data1_read_node to core dump when no abstract
98 * syntax was defined in a "sgml"-record.
100 * Revision 1.11 1996/07/06 19:58:35 quinn
101 * System headerfiles gathered in yconfig
103 * Revision 1.10 1996/01/19 15:41:47 quinn
104 * Fixed uninitialized boolean.
106 * Revision 1.9 1996/01/17 14:52:47 adam
107 * Changed prototype for reader function parsed to data1_read_record.
109 * Revision 1.8 1995/12/15 16:20:41 quinn
110 * Added formatted text.
112 * Revision 1.7 1995/12/13 13:44:32 quinn
113 * Modified Data1-system to use nmem
115 * Revision 1.6 1995/12/12 16:37:08 quinn
116 * Added destroy element to data1_node.
118 * Revision 1.5 1995/12/11 15:22:37 quinn
119 * Added last_child field to the node.
120 * Rewrote schema-mapping.
122 * Revision 1.4 1995/11/13 09:27:36 quinn
123 * Fiddling with the variant stuff.
125 * Revision 1.3 1995/11/01 16:34:57 quinn
126 * Making data1 look for tables in data1_tabpath
128 * Revision 1.2 1995/11/01 13:54:48 quinn
131 * Revision 1.1 1995/11/01 11:56:09 quinn
132 * Added Retrieval (data management) functions en masse.
134 * Revision 1.14 1995/10/30 12:40:55 quinn
135 * Fixed a couple of bugs.
137 * Revision 1.13 1995/10/25 16:00:47 quinn
138 * USMARC support is now almost operational
140 * Revision 1.12 1995/10/16 14:02:55 quinn
141 * Changes to support element set names and espec1
143 * Revision 1.11 1995/10/13 16:05:08 quinn
144 * Adding Espec1-processing
146 * Revision 1.10 1995/10/11 14:53:44 quinn
149 * Revision 1.9 1995/10/06 16:56:50 quinn
150 * Fixed ranked result.
152 * Revision 1.8 1995/10/06 16:44:13 quinn
153 * Work on attribute set mapping, etc.
155 * Revision 1.7 1995/10/06 12:58:35 quinn
158 * Revision 1.6 1995/10/04 09:29:49 quinn
159 * Adjustments to support USGS test data
161 * Revision 1.5 1995/10/03 17:56:43 quinn
164 * Revision 1.4 1995/10/02 15:53:19 quinn
167 * Revision 1.3 1995/10/02 14:55:21 quinn
168 * *** empty log message ***
170 * Revision 1.2 1995/09/14 15:18:13 quinn
173 * Revision 1.1 1995/09/12 11:24:30 quinn
174 * Beginning to add code for structured records.
183 #include <yaz/xmalloc.h>
185 #include <yaz/data1.h>
188 * get the tag which is the immediate parent of this node (this may mean
189 * traversing intermediate things like variants and stuff.
191 data1_node *get_parent_tag (data1_handle dh, data1_node *n)
193 for (; n && n->which != DATA1N_root; n = n->parent)
194 if (n->which == DATA1N_tag)
199 data1_node *data1_mk_node (data1_handle dh, NMEM m)
203 r = (data1_node *)nmem_malloc(m, sizeof(*r));
204 r->next = r->child = r->last_child = r->parent = 0;
209 data1_node *data1_mk_node_type (data1_handle dh, NMEM m, int type)
213 r = data1_mk_node(dh, m);
219 r->u.tag.element = 0;
220 r->u.tag.no_data_requested = 0;
221 r->u.tag.node_selected = 0;
222 r->u.tag.make_variantlist = 0;
223 r->u.tag.get_bytes = -1;
224 #if DATA1_USING_XATTR
225 r->u.tag.attributes = 0;
236 r->u.data.formatted_text = 0;
239 logf (LOG_WARN, "data_mk_node_type. bad type = %d\n", type);
244 void data1_free_tree (data1_handle dh, data1_node *t)
246 data1_node *p = t->child, *pn;
251 data1_free_tree (dh, p);
258 char *data1_insert_string (data1_handle dh, data1_node *res,
259 NMEM m, const char *str)
261 int len = strlen(str);
263 if (len >= DATA1_LOCALDATA)
264 return nmem_strdup (m, str);
267 strcpy (res->lbuf, str);
272 data1_node *data1_add_insert_taggeddata(data1_handle dh, data1_node *root,
274 const char *tagname, NMEM m,
275 int first_flag, int local_allowed)
277 data1_node *partag = get_parent_tag (dh, at);
278 data1_node *tagn = data1_mk_node_type (dh, m, DATA1N_tag);
279 data1_element *e = NULL;
282 tagn->u.tag.tag = data1_insert_string (dh, tagn, m, tagname);
285 e = partag->u.tag.element;
286 tagn->u.tag.element =
287 data1_getelementbytagname (dh, root->u.root.absyn, e, tagname);
288 if (!local_allowed && !tagn->u.tag.element)
290 tagn->last_child = tagn->child = datn = data1_mk_node_type (dh, m, DATA1N_data);
294 datn->u.data.formatted_text = 0;
299 tagn->next = at->child;
301 at->last_child = tagn;
310 assert (at->last_child);
311 at->last_child->next = tagn;
313 at->last_child = tagn;
318 data1_node *data1_add_taggeddata(data1_handle dh, data1_node *root,
320 const char *tagname, NMEM m)
322 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 0, 1);
327 * Insert a tagged node into the record root as first child of the node at
328 * which should be root or tag itself). Returns pointer to the data node,
329 * which can then be modified.
331 data1_node *data1_insert_taggeddata(data1_handle dh, data1_node *root,
333 const char *tagname, NMEM m)
335 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 1, 0);
338 #if DATA1_USING_XATTR
339 data1_xattr *data1_read_xattr (data1_handle dh, NMEM m,
340 int (*get_byte)(void *fh), void *fh,
341 WRBUF wrbuf, int *ch)
343 data1_xattr *p_first = 0;
344 data1_xattr **pp = &p_first;
350 while (c && d1_isspace(c))
352 if (!c || c == '>' || c == '/')
354 *pp = p = (data1_xattr *) nmem_malloc (m, sizeof(*p));
360 while (c && c != '=' && c != '>' && c != '/' && !d1_isspace(c))
362 wrbuf_putc (wrbuf, c);
365 wrbuf_putc (wrbuf, '\0');
366 len = wrbuf_len(wrbuf);
367 p->name = (char*) nmem_malloc (m, len);
368 strcpy (p->name, wrbuf_buf(wrbuf));
375 while (c && c != '"' && c != '>' && c != '/')
377 wrbuf_putc (wrbuf, c);
380 wrbuf_putc (wrbuf, '\0');
381 len = wrbuf_len(wrbuf);
382 p->value = (char*) nmem_malloc (m, len);
383 strcpy (p->value, wrbuf_buf(wrbuf));
394 * Ugh. Sometimes functions just grow and grow on you. This one reads a
395 * 'node' and its children.
397 data1_node *data1_read_nodex (data1_handle dh, NMEM m,
398 int (*get_byte)(void *fh), void *fh, WRBUF wrbuf)
400 data1_absyn *absyn = 0;
401 data1_node *d1_stack[256];
411 data1_node *parent = level ? d1_stack[level-1] : 0;
412 while (c != '\0' && d1_isspace(c))
421 if (c == '<') /* beginning of tag */
423 #if DATA1_USING_XATTR
438 else if (c == '!') /* tags/comments that we don't deal with yet */
440 while (c && c != '>')
446 while (c && c != '>' && c != '/' && !d1_isspace(c))
448 if (i < (sizeof(tag)-1))
453 #if DATA1_USING_XATTR
454 xattr = data1_read_xattr (dh, m, get_byte, fh, wrbuf, &c);
457 while (d1_isspace(c))
459 for (i = 0; c && c != '>' && c != '/'; c = (*get_byte)(fh))
460 if (i < (sizeof(args)-1))
465 { /* <tag attrs/> or <tag/> */
471 yaz_log(LOG_WARN, "d1: %d: Malformed tag", line);
487 parent = d1_stack[--i];
488 if ((parent->which == DATA1N_root &&
489 !strcmp(tag, parent->u.root.type)) ||
490 (parent->which == DATA1N_tag &&
491 !strcmp(tag, parent->u.tag.tag)))
499 yaz_log (LOG_WARN, "%d: no begin tag for %s",
508 if (level == 0) /* root ? */
510 if (!(absyn = data1_get_absyn (dh, tag)))
512 yaz_log(LOG_WARN, "Unable to acquire abstract syntax " "for '%s'", tag);
513 /* It's now OK for a record not to have an absyn */
515 res = data1_mk_node_type (dh, m, DATA1N_root);
516 res->u.root.type = data1_insert_string (dh, res, m, tag);
517 res->u.root.absyn = absyn;
520 else if (!strcmp(tag, "var"))
522 char tclass[DATA1_MAX_SYMBOL], type[DATA1_MAX_SYMBOL];
526 if (sscanf(args, "%s %s %n", tclass, type, &val_offset) != 2)
528 yaz_log(LOG_WARN, "Malformed variant triple at '%s'", tag);
532 data1_getvartypebyct(dh,
533 parent->root->u.root.absyn->varset,
537 * If we're the first variant in this group, create a parent
538 * variant, and insert it before the current variant.
540 if (parent->which != DATA1N_variant)
542 res = data1_mk_node (dh, m);
543 res->which = DATA1N_variant;
544 res->u.variant.type = 0;
545 res->u.variant.value = 0;
550 * now determine if one of our ancestor triples is of
551 * same type. If so, we break here.
554 for (i = level-1; d1_stack[i]->which==DATA1N_variant; --i)
555 if (d1_stack[i]->u.variant.type == tp)
560 res = data1_mk_node (dh, m);
561 res->which = DATA1N_variant;
562 res->u.variant.type = tp;
563 res->u.variant.value =
564 data1_insert_string (dh, res, m, args + val_offset);
567 else /* tag.. acquire our element in the abstract syntax */
569 data1_node *partag = get_parent_tag (dh, parent);
570 data1_element *elem, *e = 0;
573 if (parent->which == DATA1N_variant)
576 if (!(e = partag->u.tag.element))
577 localtag = 1; /* our parent is a local tag */
579 elem = data1_getelementbytagname(dh, absyn, e, tag);
580 res = data1_mk_node_type (dh, m, DATA1N_tag);
581 res->u.tag.tag = data1_insert_string (dh, res, m, tag);
582 res->u.tag.element = elem;
583 #if DATA1_USING_XATTR
584 res->u.tag.attributes = xattr;
589 parent->last_child = res;
590 res->root = parent->root;
592 res->parent = parent;
594 d1_stack[level]->next = res;
597 d1_stack[level] = res;
598 d1_stack[level+1] = 0;
602 else /* != '<'... this is a body of text */
606 int len, prev_char = 0;
613 res = data1_mk_node_type (dh, m, DATA1N_data);
614 res->parent = parent;
615 res->u.data.what = DATA1I_text;
616 res->u.data.formatted_text = 0;
617 res->root = parent->root;
618 parent->last_child = res;
620 d1_stack[level]->next = res;
623 d1_stack[level] = res;
627 while (c && c != '<')
629 wrbuf_putc (wrbuf, c);
632 len = wrbuf_len(wrbuf);
634 /* use local buffer of nmem if too large */
635 if (len >= DATA1_LOCALDATA)
636 res->u.data.data = (char*) nmem_malloc (m, len);
638 res->u.data.data = res->lbuf;
640 /* read "data" and transfer while removing white space */
641 dst = res->u.data.data;
642 for (src = wrbuf_buf(wrbuf); --len >= 0; src++)
646 if (d1_isspace (*src))
658 res->u.data.len = dst - res->u.data.data;
664 int getc_mem (void *fh)
666 const char **p = (const char **) fh;
672 data1_node *data1_read_node (data1_handle dh, const char **buf, NMEM m)
674 WRBUF wrbuf = wrbuf_alloc();
677 node = data1_read_nodex(dh, m, getc_mem, (void *) (buf), wrbuf);
678 wrbuf_free (wrbuf, 1);
683 * Read a record in the native syntax.
685 data1_node *data1_read_record(data1_handle dh,
686 int (*rf)(void *, char *, size_t), void *fh,
690 char **buf = data1_get_read_buf (dh, &size);
695 *buf = (char *)xmalloc(*size = 4096);
699 if (rd + 2048 >= *size && !(*buf =(char *)xrealloc(*buf, *size *= 2)))
701 if ((res = (*rf)(fh, *buf + rd, 2048)) <= 0)
707 return data1_read_node(dh, &bp, m);
716 data1_node *data1_read_sgml (data1_handle dh, NMEM m, const char *buf)
718 const char *bp = buf;
719 return data1_read_node (dh, &bp, m);