2 * Copyright (c) 1995-2000, Index Data.
3 * See the file LICENSE for details.
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.35 2000-12-05 14:44:25 adam
8 * Readers skips <! ...> sections.
10 * Revision 1.34 2000/12/05 10:06:23 adam
11 * Added support for null-data rules like <tag/>.
13 * Revision 1.33 2000/11/29 14:22:47 adam
14 * Implemented XML/SGML attributes for data1 so that d1_read reads them
15 * and d1_write generates proper attributes for XML/SGML records. Added
16 * register locking for threaded version.
18 * Revision 1.32 2000/01/06 11:25:59 adam
19 * Added case to prevent warning.
21 * Revision 1.31 1999/12/21 14:16:20 ian
22 * Changed retrieval module to allow data1 trees with no associated absyn.
23 * Also added a simple interface for extracting values from data1 trees using
24 * a string based tagpath.
26 * Revision 1.30 1999/11/30 13:47:12 adam
27 * Improved installation. Moved header files to include/yaz.
29 * Revision 1.29 1999/10/21 12:06:29 adam
30 * Retrieval module no longer uses ctype.h - functions.
32 * Revision 1.28 1999/10/21 09:50:33 adam
33 * SGML reader uses own isspace - it doesn't do 8-bit on WIN32!
35 * Revision 1.27 1999/08/27 09:40:32 adam
36 * Renamed logf function to yaz_log. Removed VC++ project files.
38 * Revision 1.26 1999/07/13 13:23:48 adam
39 * Non-recursive version of data1_read_node. data1_read_nodex reads
40 * stream of bytes (instead of buffer in memory).
42 * Revision 1.25 1999/04/20 09:56:48 adam
43 * Added 'name' paramter to encoder/decoder routines (typedef Odr_fun).
44 * Modified all encoders/decoders to reflect this change.
46 * Revision 1.24 1998/10/28 15:10:09 adam
47 * Added --with-yc option to configure. For the data1_node in data1.h:
48 * decreased size of localdata and removed member "line" which wasn't useful.
50 * Revision 1.23 1998/03/12 11:28:45 adam
51 * Fix: didn't set root member of tagged node in function.
52 * data1_add_insert_taggeddata.
54 * Revision 1.22 1998/03/05 08:15:32 adam
55 * Implemented data1_add_insert_taggeddata utility which is more flexible
56 * than data1_insert_taggeddata.
58 * Revision 1.21 1998/02/27 14:08:05 adam
59 * Added const to some char pointer arguments.
60 * Reworked data1_read_node so that it doesn't create a tree with
61 * pointers to original "SGML"-buffer.
63 * Revision 1.20 1998/02/11 11:53:35 adam
64 * Changed code so that it compiles as C++.
66 * Revision 1.19 1997/12/09 16:17:09 adam
67 * Fix bug regarding variants. Tags with prefix "var" was incorrectly
68 * interpreted as "start of variants". Now, only "var" indicates such
70 * Cleaned up data1_read_node so tag names and variant names are
71 * copied and not pointed to by the generated data1 tree. Data nodes
72 * still point to old buffer.
74 * Revision 1.18 1997/11/18 09:51:09 adam
75 * Removed element num_children from data1_node. Minor changes in
78 * Revision 1.17 1997/11/05 09:20:51 adam
81 * Revision 1.16 1997/09/17 12:10:37 adam
84 * Revision 1.15 1997/09/05 09:50:57 adam
85 * Removed global data1_tabpath - uses data1_get_tabpath() instead.
87 * Revision 1.14 1997/05/14 06:54:04 adam
90 * Revision 1.13 1996/10/29 13:35:38 adam
91 * Implemented data1_set_tabpath and data1_get_tabpath.
93 * Revision 1.12 1996/10/11 10:35:38 adam
94 * Fixed a bug that caused data1_read_node to core dump when no abstract
95 * syntax was defined in a "sgml"-record.
97 * Revision 1.11 1996/07/06 19:58:35 quinn
98 * System headerfiles gathered in yconfig
100 * Revision 1.10 1996/01/19 15:41:47 quinn
101 * Fixed uninitialized boolean.
103 * Revision 1.9 1996/01/17 14:52:47 adam
104 * Changed prototype for reader function parsed to data1_read_record.
106 * Revision 1.8 1995/12/15 16:20:41 quinn
107 * Added formatted text.
109 * Revision 1.7 1995/12/13 13:44:32 quinn
110 * Modified Data1-system to use nmem
112 * Revision 1.6 1995/12/12 16:37:08 quinn
113 * Added destroy element to data1_node.
115 * Revision 1.5 1995/12/11 15:22:37 quinn
116 * Added last_child field to the node.
117 * Rewrote schema-mapping.
119 * Revision 1.4 1995/11/13 09:27:36 quinn
120 * Fiddling with the variant stuff.
122 * Revision 1.3 1995/11/01 16:34:57 quinn
123 * Making data1 look for tables in data1_tabpath
125 * Revision 1.2 1995/11/01 13:54:48 quinn
128 * Revision 1.1 1995/11/01 11:56:09 quinn
129 * Added Retrieval (data management) functions en masse.
131 * Revision 1.14 1995/10/30 12:40:55 quinn
132 * Fixed a couple of bugs.
134 * Revision 1.13 1995/10/25 16:00:47 quinn
135 * USMARC support is now almost operational
137 * Revision 1.12 1995/10/16 14:02:55 quinn
138 * Changes to support element set names and espec1
140 * Revision 1.11 1995/10/13 16:05:08 quinn
141 * Adding Espec1-processing
143 * Revision 1.10 1995/10/11 14:53:44 quinn
146 * Revision 1.9 1995/10/06 16:56:50 quinn
147 * Fixed ranked result.
149 * Revision 1.8 1995/10/06 16:44:13 quinn
150 * Work on attribute set mapping, etc.
152 * Revision 1.7 1995/10/06 12:58:35 quinn
155 * Revision 1.6 1995/10/04 09:29:49 quinn
156 * Adjustments to support USGS test data
158 * Revision 1.5 1995/10/03 17:56:43 quinn
161 * Revision 1.4 1995/10/02 15:53:19 quinn
164 * Revision 1.3 1995/10/02 14:55:21 quinn
165 * *** empty log message ***
167 * Revision 1.2 1995/09/14 15:18:13 quinn
170 * Revision 1.1 1995/09/12 11:24:30 quinn
171 * Beginning to add code for structured records.
180 #include <yaz/xmalloc.h>
182 #include <yaz/data1.h>
185 * get the tag which is the immediate parent of this node (this may mean
186 * traversing intermediate things like variants and stuff.
188 data1_node *get_parent_tag (data1_handle dh, data1_node *n)
190 for (; n && n->which != DATA1N_root; n = n->parent)
191 if (n->which == DATA1N_tag)
196 data1_node *data1_mk_node (data1_handle dh, NMEM m)
200 r = (data1_node *)nmem_malloc(m, sizeof(*r));
201 r->next = r->child = r->last_child = r->parent = 0;
206 data1_node *data1_mk_node_type (data1_handle dh, NMEM m, int type)
210 r = data1_mk_node(dh, m);
216 r->u.tag.element = 0;
217 r->u.tag.no_data_requested = 0;
218 r->u.tag.node_selected = 0;
219 r->u.tag.make_variantlist = 0;
220 r->u.tag.get_bytes = -1;
221 #if DATA1_USING_XATTR
222 r->u.tag.attributes = 0;
233 r->u.data.formatted_text = 0;
236 logf (LOG_WARN, "data_mk_node_type. bad type = %d\n", type);
241 void data1_free_tree (data1_handle dh, data1_node *t)
243 data1_node *p = t->child, *pn;
248 data1_free_tree (dh, p);
255 char *data1_insert_string (data1_handle dh, data1_node *res,
256 NMEM m, const char *str)
258 int len = strlen(str);
260 if (len >= DATA1_LOCALDATA)
261 return nmem_strdup (m, str);
264 strcpy (res->lbuf, str);
269 data1_node *data1_add_insert_taggeddata(data1_handle dh, data1_node *root,
271 const char *tagname, NMEM m,
272 int first_flag, int local_allowed)
274 data1_node *partag = get_parent_tag (dh, at);
275 data1_node *tagn = data1_mk_node_type (dh, m, DATA1N_tag);
276 data1_element *e = NULL;
279 tagn->u.tag.tag = data1_insert_string (dh, tagn, m, tagname);
282 e = partag->u.tag.element;
283 tagn->u.tag.element =
284 data1_getelementbytagname (dh, root->u.root.absyn, e, tagname);
285 if (!local_allowed && !tagn->u.tag.element)
287 tagn->last_child = tagn->child = datn = data1_mk_node_type (dh, m, DATA1N_data);
291 datn->u.data.formatted_text = 0;
296 tagn->next = at->child;
298 at->last_child = tagn;
307 assert (at->last_child);
308 at->last_child->next = tagn;
310 at->last_child = tagn;
315 data1_node *data1_add_taggeddata(data1_handle dh, data1_node *root,
317 const char *tagname, NMEM m)
319 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 0, 1);
324 * Insert a tagged node into the record root as first child of the node at
325 * which should be root or tag itself). Returns pointer to the data node,
326 * which can then be modified.
328 data1_node *data1_insert_taggeddata(data1_handle dh, data1_node *root,
330 const char *tagname, NMEM m)
332 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 1, 0);
335 #if DATA1_USING_XATTR
336 data1_xattr *data1_read_xattr (data1_handle dh, NMEM m,
337 int (*get_byte)(void *fh), void *fh,
338 WRBUF wrbuf, int *ch)
340 data1_xattr *p_first = 0;
341 data1_xattr **pp = &p_first;
347 while (c && d1_isspace(c))
349 if (!c || c == '>' || c == '/')
351 *pp = p = nmem_malloc (m, sizeof(*p));
357 while (c && c != '=' && c != '>' && c != '/' && !d1_isspace(c))
359 wrbuf_putc (wrbuf, c);
362 wrbuf_putc (wrbuf, '\0');
363 len = wrbuf_len(wrbuf);
364 p->name = nmem_malloc (m, len);
365 strcpy (p->name, wrbuf_buf(wrbuf));
372 while (c && c != '"' && c != '>' && c != '/')
374 wrbuf_putc (wrbuf, c);
377 wrbuf_putc (wrbuf, '\0');
378 len = wrbuf_len(wrbuf);
379 p->value = nmem_malloc (m, len);
380 strcpy (p->value, wrbuf_buf(wrbuf));
391 * Ugh. Sometimes functions just grow and grow on you. This one reads a
392 * 'node' and its children.
394 data1_node *data1_read_nodex (data1_handle dh, NMEM m,
395 int (*get_byte)(void *fh), void *fh, WRBUF wrbuf)
397 data1_absyn *absyn = 0;
398 data1_node *d1_stack[256];
408 data1_node *parent = level ? d1_stack[level-1] : 0;
409 while (c != '\0' && d1_isspace(c))
418 if (c == '<') /* beginning of tag */
420 #if DATA1_USING_XATTR
435 else if (c == '!') /* tags/comments that we don't deal with yet */
437 while (c && c != '>')
443 while (c && c != '>' && c != '/' && !d1_isspace(c))
445 if (i < (sizeof(tag)-1))
450 #if DATA1_USING_XATTR
451 xattr = data1_read_xattr (dh, m, get_byte, fh, wrbuf, &c);
454 while (d1_isspace(c))
456 for (i = 0; c && c != '>' && c != '/'; c = (*get_byte)(fh))
457 if (i < (sizeof(args)-1))
462 { /* <tag attrs/> or <tag/> */
468 yaz_log(LOG_WARN, "d1: %d: Malformed tag", line);
484 parent = d1_stack[--i];
485 if ((parent->which == DATA1N_root &&
486 !strcmp(tag, parent->u.root.type)) ||
487 (parent->which == DATA1N_tag &&
488 !strcmp(tag, parent->u.tag.tag)))
496 yaz_log (LOG_WARN, "%d: no begin tag for %s",
505 if (level == 0) /* root ? */
507 if (!(absyn = data1_get_absyn (dh, tag)))
509 yaz_log(LOG_WARN, "Unable to acquire abstract syntax " "for '%s'", tag);
510 /* It's now OK for a record not to have an absyn */
512 res = data1_mk_node_type (dh, m, DATA1N_root);
513 res->u.root.type = data1_insert_string (dh, res, m, tag);
514 res->u.root.absyn = absyn;
517 else if (!strcmp(tag, "var"))
519 char tclass[DATA1_MAX_SYMBOL], type[DATA1_MAX_SYMBOL];
523 if (sscanf(args, "%s %s %n", tclass, type, &val_offset) != 2)
525 yaz_log(LOG_WARN, "Malformed variant triple at '%s'", tag);
529 data1_getvartypebyct(dh,
530 parent->root->u.root.absyn->varset,
534 * If we're the first variant in this group, create a parent
535 * variant, and insert it before the current variant.
537 if (parent->which != DATA1N_variant)
539 res = data1_mk_node (dh, m);
540 res->which = DATA1N_variant;
541 res->u.variant.type = 0;
542 res->u.variant.value = 0;
547 * now determine if one of our ancestor triples is of
548 * same type. If so, we break here.
551 for (i = level-1; d1_stack[i]->which==DATA1N_variant; --i)
552 if (d1_stack[i]->u.variant.type == tp)
557 res = data1_mk_node (dh, m);
558 res->which = DATA1N_variant;
559 res->u.variant.type = tp;
560 res->u.variant.value =
561 data1_insert_string (dh, res, m, args + val_offset);
564 else /* tag.. acquire our element in the abstract syntax */
566 data1_node *partag = get_parent_tag (dh, parent);
567 data1_element *elem, *e = 0;
570 if (parent->which == DATA1N_variant)
573 if (!(e = partag->u.tag.element))
574 localtag = 1; /* our parent is a local tag */
576 elem = data1_getelementbytagname(dh, absyn, e, tag);
577 res = data1_mk_node_type (dh, m, DATA1N_tag);
578 res->u.tag.tag = data1_insert_string (dh, res, m, tag);
579 res->u.tag.element = elem;
580 #if DATA1_USING_XATTR
581 res->u.tag.attributes = xattr;
586 parent->last_child = res;
587 res->root = parent->root;
589 res->parent = parent;
591 d1_stack[level]->next = res;
594 d1_stack[level] = res;
595 d1_stack[level+1] = 0;
599 else /* != '<'... this is a body of text */
603 int len, prev_char = 0;
610 res = data1_mk_node_type (dh, m, DATA1N_data);
611 res->parent = parent;
612 res->u.data.what = DATA1I_text;
613 res->u.data.formatted_text = 0;
614 res->root = parent->root;
615 parent->last_child = res;
617 d1_stack[level]->next = res;
620 d1_stack[level] = res;
624 while (c && c != '<')
626 wrbuf_putc (wrbuf, c);
629 len = wrbuf_len(wrbuf);
631 /* use local buffer of nmem if too large */
632 if (len >= DATA1_LOCALDATA)
633 res->u.data.data = (char*) nmem_malloc (m, len);
635 res->u.data.data = res->lbuf;
637 /* read "data" and transfer while removing white space */
638 dst = res->u.data.data;
639 for (src = wrbuf_buf(wrbuf); --len >= 0; src++)
643 if (d1_isspace (*src))
655 res->u.data.len = dst - res->u.data.data;
661 int getc_mem (void *fh)
663 const char **p = (const char **) fh;
669 data1_node *data1_read_node (data1_handle dh, const char **buf, NMEM m)
671 WRBUF wrbuf = wrbuf_alloc();
674 node = data1_read_nodex(dh, m, getc_mem, (void *) (buf), wrbuf);
675 wrbuf_free (wrbuf, 1);
680 * Read a record in the native syntax.
682 data1_node *data1_read_record(data1_handle dh,
683 int (*rf)(void *, char *, size_t), void *fh,
687 char **buf = data1_get_read_buf (dh, &size);
692 *buf = (char *)xmalloc(*size = 4096);
696 if (rd + 2048 >= *size && !(*buf =(char *)xrealloc(*buf, *size *= 2)))
698 if ((res = (*rf)(fh, *buf + rd, 2048)) <= 0)
704 return data1_read_node(dh, &bp, m);
713 data1_node *data1_read_sgml (data1_handle dh, NMEM m, const char *buf)
715 const char *bp = buf;
716 return data1_read_node (dh, &bp, m);