2 * Copyright (c) 1995-2001, Index Data.
3 * See the file LICENSE for details.
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.37 2001-02-28 09:00:06 adam
8 * Fixed problem with stack overflow for very nested records.
10 * Revision 1.36 2001/02/21 13:46:53 adam
13 * Revision 1.35 2000/12/05 14:44:25 adam
14 * Readers skips <! ...> sections.
16 * Revision 1.34 2000/12/05 10:06:23 adam
17 * Added support for null-data rules like <tag/>.
19 * Revision 1.33 2000/11/29 14:22:47 adam
20 * Implemented XML/SGML attributes for data1 so that d1_read reads them
21 * and d1_write generates proper attributes for XML/SGML records. Added
22 * register locking for threaded version.
24 * Revision 1.32 2000/01/06 11:25:59 adam
25 * Added case to prevent warning.
27 * Revision 1.31 1999/12/21 14:16:20 ian
28 * Changed retrieval module to allow data1 trees with no associated absyn.
29 * Also added a simple interface for extracting values from data1 trees using
30 * a string based tagpath.
32 * Revision 1.30 1999/11/30 13:47:12 adam
33 * Improved installation. Moved header files to include/yaz.
35 * Revision 1.29 1999/10/21 12:06:29 adam
36 * Retrieval module no longer uses ctype.h - functions.
38 * Revision 1.28 1999/10/21 09:50:33 adam
39 * SGML reader uses own isspace - it doesn't do 8-bit on WIN32!
41 * Revision 1.27 1999/08/27 09:40:32 adam
42 * Renamed logf function to yaz_log. Removed VC++ project files.
44 * Revision 1.26 1999/07/13 13:23:48 adam
45 * Non-recursive version of data1_read_node. data1_read_nodex reads
46 * stream of bytes (instead of buffer in memory).
48 * Revision 1.25 1999/04/20 09:56:48 adam
49 * Added 'name' paramter to encoder/decoder routines (typedef Odr_fun).
50 * Modified all encoders/decoders to reflect this change.
52 * Revision 1.24 1998/10/28 15:10:09 adam
53 * Added --with-yc option to configure. For the data1_node in data1.h:
54 * decreased size of localdata and removed member "line" which wasn't useful.
56 * Revision 1.23 1998/03/12 11:28:45 adam
57 * Fix: didn't set root member of tagged node in function.
58 * data1_add_insert_taggeddata.
60 * Revision 1.22 1998/03/05 08:15:32 adam
61 * Implemented data1_add_insert_taggeddata utility which is more flexible
62 * than data1_insert_taggeddata.
64 * Revision 1.21 1998/02/27 14:08:05 adam
65 * Added const to some char pointer arguments.
66 * Reworked data1_read_node so that it doesn't create a tree with
67 * pointers to original "SGML"-buffer.
69 * Revision 1.20 1998/02/11 11:53:35 adam
70 * Changed code so that it compiles as C++.
72 * Revision 1.19 1997/12/09 16:17:09 adam
73 * Fix bug regarding variants. Tags with prefix "var" was incorrectly
74 * interpreted as "start of variants". Now, only "var" indicates such
76 * Cleaned up data1_read_node so tag names and variant names are
77 * copied and not pointed to by the generated data1 tree. Data nodes
78 * still point to old buffer.
80 * Revision 1.18 1997/11/18 09:51:09 adam
81 * Removed element num_children from data1_node. Minor changes in
84 * Revision 1.17 1997/11/05 09:20:51 adam
87 * Revision 1.16 1997/09/17 12:10:37 adam
90 * Revision 1.15 1997/09/05 09:50:57 adam
91 * Removed global data1_tabpath - uses data1_get_tabpath() instead.
93 * Revision 1.14 1997/05/14 06:54:04 adam
96 * Revision 1.13 1996/10/29 13:35:38 adam
97 * Implemented data1_set_tabpath and data1_get_tabpath.
99 * Revision 1.12 1996/10/11 10:35:38 adam
100 * Fixed a bug that caused data1_read_node to core dump when no abstract
101 * syntax was defined in a "sgml"-record.
103 * Revision 1.11 1996/07/06 19:58:35 quinn
104 * System headerfiles gathered in yconfig
106 * Revision 1.10 1996/01/19 15:41:47 quinn
107 * Fixed uninitialized boolean.
109 * Revision 1.9 1996/01/17 14:52:47 adam
110 * Changed prototype for reader function parsed to data1_read_record.
112 * Revision 1.8 1995/12/15 16:20:41 quinn
113 * Added formatted text.
115 * Revision 1.7 1995/12/13 13:44:32 quinn
116 * Modified Data1-system to use nmem
118 * Revision 1.6 1995/12/12 16:37:08 quinn
119 * Added destroy element to data1_node.
121 * Revision 1.5 1995/12/11 15:22:37 quinn
122 * Added last_child field to the node.
123 * Rewrote schema-mapping.
125 * Revision 1.4 1995/11/13 09:27:36 quinn
126 * Fiddling with the variant stuff.
128 * Revision 1.3 1995/11/01 16:34:57 quinn
129 * Making data1 look for tables in data1_tabpath
131 * Revision 1.2 1995/11/01 13:54:48 quinn
134 * Revision 1.1 1995/11/01 11:56:09 quinn
135 * Added Retrieval (data management) functions en masse.
137 * Revision 1.14 1995/10/30 12:40:55 quinn
138 * Fixed a couple of bugs.
140 * Revision 1.13 1995/10/25 16:00:47 quinn
141 * USMARC support is now almost operational
143 * Revision 1.12 1995/10/16 14:02:55 quinn
144 * Changes to support element set names and espec1
146 * Revision 1.11 1995/10/13 16:05:08 quinn
147 * Adding Espec1-processing
149 * Revision 1.10 1995/10/11 14:53:44 quinn
152 * Revision 1.9 1995/10/06 16:56:50 quinn
153 * Fixed ranked result.
155 * Revision 1.8 1995/10/06 16:44:13 quinn
156 * Work on attribute set mapping, etc.
158 * Revision 1.7 1995/10/06 12:58:35 quinn
161 * Revision 1.6 1995/10/04 09:29:49 quinn
162 * Adjustments to support USGS test data
164 * Revision 1.5 1995/10/03 17:56:43 quinn
167 * Revision 1.4 1995/10/02 15:53:19 quinn
170 * Revision 1.3 1995/10/02 14:55:21 quinn
171 * *** empty log message ***
173 * Revision 1.2 1995/09/14 15:18:13 quinn
176 * Revision 1.1 1995/09/12 11:24:30 quinn
177 * Beginning to add code for structured records.
186 #include <yaz/xmalloc.h>
188 #include <yaz/data1.h>
191 * get the tag which is the immediate parent of this node (this may mean
192 * traversing intermediate things like variants and stuff.
194 data1_node *get_parent_tag (data1_handle dh, data1_node *n)
196 for (; n && n->which != DATA1N_root; n = n->parent)
197 if (n->which == DATA1N_tag)
202 data1_node *data1_mk_node (data1_handle dh, NMEM m)
206 r = (data1_node *)nmem_malloc(m, sizeof(*r));
207 r->next = r->child = r->last_child = r->parent = 0;
212 data1_node *data1_mk_node_type (data1_handle dh, NMEM m, int type)
216 r = data1_mk_node(dh, m);
222 r->u.tag.element = 0;
223 r->u.tag.no_data_requested = 0;
224 r->u.tag.node_selected = 0;
225 r->u.tag.make_variantlist = 0;
226 r->u.tag.get_bytes = -1;
227 #if DATA1_USING_XATTR
228 r->u.tag.attributes = 0;
239 r->u.data.formatted_text = 0;
242 logf (LOG_WARN, "data_mk_node_type. bad type = %d\n", type);
247 void data1_free_tree (data1_handle dh, data1_node *t)
249 data1_node *p = t->child, *pn;
254 data1_free_tree (dh, p);
261 char *data1_insert_string (data1_handle dh, data1_node *res,
262 NMEM m, const char *str)
264 int len = strlen(str);
266 if (len >= DATA1_LOCALDATA)
267 return nmem_strdup (m, str);
270 strcpy (res->lbuf, str);
275 data1_node *data1_add_insert_taggeddata(data1_handle dh, data1_node *root,
277 const char *tagname, NMEM m,
278 int first_flag, int local_allowed)
280 data1_node *partag = get_parent_tag (dh, at);
281 data1_node *tagn = data1_mk_node_type (dh, m, DATA1N_tag);
282 data1_element *e = NULL;
285 tagn->u.tag.tag = data1_insert_string (dh, tagn, m, tagname);
288 e = partag->u.tag.element;
289 tagn->u.tag.element =
290 data1_getelementbytagname (dh, root->u.root.absyn, e, tagname);
291 if (!local_allowed && !tagn->u.tag.element)
293 tagn->last_child = tagn->child = datn = data1_mk_node_type (dh, m, DATA1N_data);
297 datn->u.data.formatted_text = 0;
302 tagn->next = at->child;
304 at->last_child = tagn;
313 assert (at->last_child);
314 at->last_child->next = tagn;
316 at->last_child = tagn;
321 data1_node *data1_add_taggeddata(data1_handle dh, data1_node *root,
323 const char *tagname, NMEM m)
325 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 0, 1);
330 * Insert a tagged node into the record root as first child of the node at
331 * which should be root or tag itself). Returns pointer to the data node,
332 * which can then be modified.
334 data1_node *data1_insert_taggeddata(data1_handle dh, data1_node *root,
336 const char *tagname, NMEM m)
338 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 1, 0);
341 #if DATA1_USING_XATTR
342 data1_xattr *data1_read_xattr (data1_handle dh, NMEM m,
343 int (*get_byte)(void *fh), void *fh,
344 WRBUF wrbuf, int *ch)
346 data1_xattr *p_first = 0;
347 data1_xattr **pp = &p_first;
353 while (c && d1_isspace(c))
355 if (!c || c == '>' || c == '/')
357 *pp = p = (data1_xattr *) nmem_malloc (m, sizeof(*p));
363 while (c && c != '=' && c != '>' && c != '/' && !d1_isspace(c))
365 wrbuf_putc (wrbuf, c);
368 wrbuf_putc (wrbuf, '\0');
369 len = wrbuf_len(wrbuf);
370 p->name = (char*) nmem_malloc (m, len);
371 strcpy (p->name, wrbuf_buf(wrbuf));
378 while (c && c != '"' && c != '>' && c != '/')
380 wrbuf_putc (wrbuf, c);
383 wrbuf_putc (wrbuf, '\0');
384 len = wrbuf_len(wrbuf);
385 p->value = (char*) nmem_malloc (m, len);
386 strcpy (p->value, wrbuf_buf(wrbuf));
397 * Ugh. Sometimes functions just grow and grow on you. This one reads a
398 * 'node' and its children.
400 data1_node *data1_read_nodex (data1_handle dh, NMEM m,
401 int (*get_byte)(void *fh), void *fh, WRBUF wrbuf)
403 data1_absyn *absyn = 0;
404 data1_node *d1_stack[256];
414 data1_node *parent = level ? d1_stack[level-1] : 0;
415 while (c != '\0' && d1_isspace(c))
424 if (c == '<') /* beginning of tag */
426 #if DATA1_USING_XATTR
441 else if (c == '!') /* tags/comments that we don't deal with yet */
443 while (c && c != '>')
449 while (c && c != '>' && c != '/' && !d1_isspace(c))
451 if (i < (sizeof(tag)-1))
456 #if DATA1_USING_XATTR
457 xattr = data1_read_xattr (dh, m, get_byte, fh, wrbuf, &c);
460 while (d1_isspace(c))
462 for (i = 0; c && c != '>' && c != '/'; c = (*get_byte)(fh))
463 if (i < (sizeof(args)-1))
468 { /* <tag attrs/> or <tag/> */
474 yaz_log(LOG_WARN, "d1: %d: Malformed tag", line);
490 parent = d1_stack[--i];
491 if ((parent->which == DATA1N_root &&
492 !strcmp(tag, parent->u.root.type)) ||
493 (parent->which == DATA1N_tag &&
494 !strcmp(tag, parent->u.tag.tag)))
502 yaz_log (LOG_WARN, "%d: no begin tag for %s",
511 if (level == 0) /* root ? */
513 if (!(absyn = data1_get_absyn (dh, tag)))
515 yaz_log(LOG_WARN, "Unable to acquire abstract syntax " "for '%s'", tag);
516 /* It's now OK for a record not to have an absyn */
518 res = data1_mk_node_type (dh, m, DATA1N_root);
519 res->u.root.type = data1_insert_string (dh, res, m, tag);
520 res->u.root.absyn = absyn;
523 else if (!strcmp(tag, "var"))
525 char tclass[DATA1_MAX_SYMBOL], type[DATA1_MAX_SYMBOL];
529 if (sscanf(args, "%s %s %n", tclass, type, &val_offset) != 2)
531 yaz_log(LOG_WARN, "Malformed variant triple at '%s'", tag);
535 data1_getvartypebyct(dh,
536 parent->root->u.root.absyn->varset,
540 * If we're the first variant in this group, create a parent
541 * variant, and insert it before the current variant.
543 if (parent->which != DATA1N_variant)
545 res = data1_mk_node (dh, m);
546 res->which = DATA1N_variant;
547 res->u.variant.type = 0;
548 res->u.variant.value = 0;
553 * now determine if one of our ancestor triples is of
554 * same type. If so, we break here.
557 for (i = level-1; d1_stack[i]->which==DATA1N_variant; --i)
558 if (d1_stack[i]->u.variant.type == tp)
563 res = data1_mk_node (dh, m);
564 res->which = DATA1N_variant;
565 res->u.variant.type = tp;
566 res->u.variant.value =
567 data1_insert_string (dh, res, m, args + val_offset);
570 else /* tag.. acquire our element in the abstract syntax */
572 data1_node *partag = get_parent_tag (dh, parent);
573 data1_element *elem, *e = 0;
576 if (parent->which == DATA1N_variant)
579 if (!(e = partag->u.tag.element))
580 localtag = 1; /* our parent is a local tag */
582 elem = data1_getelementbytagname(dh, absyn, e, tag);
583 res = data1_mk_node_type (dh, m, DATA1N_tag);
584 res->u.tag.tag = data1_insert_string (dh, res, m, tag);
585 res->u.tag.element = elem;
586 #if DATA1_USING_XATTR
587 res->u.tag.attributes = xattr;
592 parent->last_child = res;
593 res->root = parent->root;
595 res->parent = parent;
597 d1_stack[level]->next = res;
600 d1_stack[level] = res;
601 d1_stack[level+1] = 0;
602 if (level < 250 && !null_tag)
605 else /* != '<'... this is a body of text */
609 int len, prev_char = 0;
616 res = data1_mk_node_type (dh, m, DATA1N_data);
617 res->parent = parent;
618 res->u.data.what = DATA1I_text;
619 res->u.data.formatted_text = 0;
620 res->root = parent->root;
621 parent->last_child = res;
623 d1_stack[level]->next = res;
626 d1_stack[level] = res;
630 while (c && c != '<')
632 wrbuf_putc (wrbuf, c);
635 len = wrbuf_len(wrbuf);
637 /* use local buffer of nmem if too large */
638 if (len >= DATA1_LOCALDATA)
639 res->u.data.data = (char*) nmem_malloc (m, len);
641 res->u.data.data = res->lbuf;
643 /* read "data" and transfer while removing white space */
644 dst = res->u.data.data;
645 for (src = wrbuf_buf(wrbuf); --len >= 0; src++)
649 if (d1_isspace (*src))
661 res->u.data.len = dst - res->u.data.data;
667 int getc_mem (void *fh)
669 const char **p = (const char **) fh;
675 data1_node *data1_read_node (data1_handle dh, const char **buf, NMEM m)
677 WRBUF wrbuf = wrbuf_alloc();
680 node = data1_read_nodex(dh, m, getc_mem, (void *) (buf), wrbuf);
681 wrbuf_free (wrbuf, 1);
686 * Read a record in the native syntax.
688 data1_node *data1_read_record(data1_handle dh,
689 int (*rf)(void *, char *, size_t), void *fh,
693 char **buf = data1_get_read_buf (dh, &size);
698 *buf = (char *)xmalloc(*size = 4096);
702 if (rd + 2048 >= *size && !(*buf =(char *)xrealloc(*buf, *size *= 2)))
704 if ((res = (*rf)(fh, *buf + rd, 2048)) <= 0)
710 return data1_read_node(dh, &bp, m);
719 data1_node *data1_read_sgml (data1_handle dh, NMEM m, const char *buf)
721 const char *bp = buf;
722 return data1_read_node (dh, &bp, m);