2 * Copyright (c) 1995-1998, Index Data.
3 * See the file LICENSE for details.
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.23 1998-03-12 11:28:45 adam
8 * Fix: didn't set root member of tagged node in function.
9 * data1_add_insert_taggeddata.
11 * Revision 1.22 1998/03/05 08:15:32 adam
12 * Implemented data1_add_insert_taggeddata utility which is more flexible
13 * than data1_insert_taggeddata.
15 * Revision 1.21 1998/02/27 14:08:05 adam
16 * Added const to some char pointer arguments.
17 * Reworked data1_read_node so that it doesn't create a tree with
18 * pointers to original "SGML"-buffer.
20 * Revision 1.20 1998/02/11 11:53:35 adam
21 * Changed code so that it compiles as C++.
23 * Revision 1.19 1997/12/09 16:17:09 adam
24 * Fix bug regarding variants. Tags with prefix "var" was incorrectly
25 * interpreted as "start of variants". Now, only "var" indicates such
27 * Cleaned up data1_read_node so tag names and variant names are
28 * copied and not pointed to by the generated data1 tree. Data nodes
29 * still point to old buffer.
31 * Revision 1.18 1997/11/18 09:51:09 adam
32 * Removed element num_children from data1_node. Minor changes in
35 * Revision 1.17 1997/11/05 09:20:51 adam
38 * Revision 1.16 1997/09/17 12:10:37 adam
41 * Revision 1.15 1997/09/05 09:50:57 adam
42 * Removed global data1_tabpath - uses data1_get_tabpath() instead.
44 * Revision 1.14 1997/05/14 06:54:04 adam
47 * Revision 1.13 1996/10/29 13:35:38 adam
48 * Implemented data1_set_tabpath and data1_get_tabpath.
50 * Revision 1.12 1996/10/11 10:35:38 adam
51 * Fixed a bug that caused data1_read_node to core dump when no abstract
52 * syntax was defined in a "sgml"-record.
54 * Revision 1.11 1996/07/06 19:58:35 quinn
55 * System headerfiles gathered in yconfig
57 * Revision 1.10 1996/01/19 15:41:47 quinn
58 * Fixed uninitialized boolean.
60 * Revision 1.9 1996/01/17 14:52:47 adam
61 * Changed prototype for reader function parsed to data1_read_record.
63 * Revision 1.8 1995/12/15 16:20:41 quinn
64 * Added formatted text.
66 * Revision 1.7 1995/12/13 13:44:32 quinn
67 * Modified Data1-system to use nmem
69 * Revision 1.6 1995/12/12 16:37:08 quinn
70 * Added destroy element to data1_node.
72 * Revision 1.5 1995/12/11 15:22:37 quinn
73 * Added last_child field to the node.
74 * Rewrote schema-mapping.
76 * Revision 1.4 1995/11/13 09:27:36 quinn
77 * Fiddling with the variant stuff.
79 * Revision 1.3 1995/11/01 16:34:57 quinn
80 * Making data1 look for tables in data1_tabpath
82 * Revision 1.2 1995/11/01 13:54:48 quinn
85 * Revision 1.1 1995/11/01 11:56:09 quinn
86 * Added Retrieval (data management) functions en masse.
88 * Revision 1.14 1995/10/30 12:40:55 quinn
89 * Fixed a couple of bugs.
91 * Revision 1.13 1995/10/25 16:00:47 quinn
92 * USMARC support is now almost operational
94 * Revision 1.12 1995/10/16 14:02:55 quinn
95 * Changes to support element set names and espec1
97 * Revision 1.11 1995/10/13 16:05:08 quinn
98 * Adding Espec1-processing
100 * Revision 1.10 1995/10/11 14:53:44 quinn
103 * Revision 1.9 1995/10/06 16:56:50 quinn
104 * Fixed ranked result.
106 * Revision 1.8 1995/10/06 16:44:13 quinn
107 * Work on attribute set mapping, etc.
109 * Revision 1.7 1995/10/06 12:58:35 quinn
112 * Revision 1.6 1995/10/04 09:29:49 quinn
113 * Adjustments to support USGS test data
115 * Revision 1.5 1995/10/03 17:56:43 quinn
118 * Revision 1.4 1995/10/02 15:53:19 quinn
121 * Revision 1.3 1995/10/02 14:55:21 quinn
122 * *** empty log message ***
124 * Revision 1.2 1995/09/14 15:18:13 quinn
127 * Revision 1.1 1995/09/12 11:24:30 quinn
128 * Beginning to add code for structured records.
143 * get the tag which is the immediate parent of this node (this may mean
144 * traversing intermediate things like variants and stuff.
146 data1_node *get_parent_tag (data1_handle dh, data1_node *n)
148 for (; n && n->which != DATA1N_root; n = n->parent)
149 if (n->which == DATA1N_tag)
154 data1_node *data1_mk_node (data1_handle dh, NMEM m)
158 r = (data1_node *)nmem_malloc(m, sizeof(*r));
159 r->next = r->child = r->last_child = r->parent = 0;
164 void data1_free_tree (data1_handle dh, data1_node *t)
166 data1_node *p = t->child, *pn;
171 data1_free_tree (dh, p);
178 char *data1_insert_string (data1_handle dh, data1_node *res,
179 NMEM m, const char *str)
181 int len = strlen(str);
183 if (len >= DATA1_LOCALDATA)
184 return nmem_strdup (m, str);
187 strcpy (res->lbuf, str);
193 data1_node *data1_add_insert_taggeddata(data1_handle dh, data1_node *root,
195 const char *tagname, NMEM m,
196 int first_flag, int local_allowed)
198 data1_node *partag = get_parent_tag (dh, at);
199 data1_node *tagn = data1_mk_node (dh, m);
200 data1_element *e = NULL;
203 tagn->which = DATA1N_tag;
205 tagn->u.tag.tag = data1_insert_string (dh, tagn, m, tagname);
206 tagn->u.tag.node_selected = 0;
207 tagn->u.tag.make_variantlist = 0;
208 tagn->u.tag.no_data_requested = 0;
209 tagn->u.tag.get_bytes = -1;
212 e = partag->u.tag.element;
213 tagn->u.tag.element =
214 data1_getelementbytagname (dh, root->u.root.absyn, e, tagname);
215 if (!local_allowed && !tagn->u.tag.element)
217 tagn->last_child = tagn->child = datn = data1_mk_node (dh, m);
221 datn->which = DATA1N_data;
222 datn->u.data.formatted_text = 0;
227 tagn->next = at->child;
229 at->last_child = tagn;
238 assert (at->last_child);
239 at->last_child->next = tagn;
241 at->last_child = tagn;
246 data1_node *data1_add_taggeddata(data1_handle dh, data1_node *root,
248 const char *tagname, NMEM m)
250 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 0, 1);
255 * Insert a tagged node into the record root as first child of the node at
256 * which should be root or tag itself). Returns pointer to the data node,
257 * which can then be modified.
259 data1_node *data1_insert_taggeddata(data1_handle dh, data1_node *root,
261 const char *tagname, NMEM m)
263 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 1, 0);
267 * Ugh. Sometimes functions just grow and grow on you. This one reads a
268 * 'node' and its children.
270 data1_node *data1_read_node (data1_handle dh, const char **buf,
271 data1_node *parent, int *line,
272 data1_absyn *absyn, NMEM m)
276 while (**buf && isspace(**buf))
285 if (**buf == '<') /* beginning of tag */
290 const char *t = (*buf) + 1;
292 data1_element *elem = 0;
294 for (i = 0; *t && *t != '>' && !isspace(*t); t++)
295 if (i < (sizeof(tag)-1))
300 for (i = 0; *t && *t != '>'; t++)
301 if (i < (sizeof(args)-1))
304 if (*t != '>' && !isspace(*t))
306 logf(LOG_WARN, "d1: %d: Malformed tag", *line);
310 * if end-tag, see if we terminate parent. If so, consume and return.
315 if (parent && (!*(tag +1) ||
316 (parent->which == DATA1N_root &&
317 !strcmp(tag + 1,parent->u.root.type)) ||
318 (parent->which == DATA1N_tag &&
319 !strcmp(tag + 1, parent->u.tag.tag))))
323 if (!absyn) /* parent node - what are we? */
325 if (!(absyn = data1_get_absyn (dh, tag)))
327 logf(LOG_WARN, "Unable to acquire abstract syntax for '%s'",
331 res = data1_mk_node (dh, m);
332 res->which = DATA1N_root;
333 res->u.root.type = data1_insert_string (dh, res, m, tag);
334 res->u.root.absyn = absyn;
338 else if (!strcmp(tag, "var"))
340 char tclass[DATA1_MAX_SYMBOL], type[DATA1_MAX_SYMBOL];
345 if (sscanf(args, "%s %s %n", tclass, type, &val_offset) != 2)
347 logf(LOG_WARN, "Malformed variant triple at '%s'", tag);
351 data1_getvartypebyct(dh, parent->root->u.root.absyn->varset,
356 * If we're the first variant in this group, create a parent var,
357 * and insert it before the current variant.
359 if (parent->which != DATA1N_variant)
361 res = data1_mk_node (dh, m);
362 res->which = DATA1N_variant;
363 res->u.variant.type = 0;
364 res->u.variant.value = 0;
365 res->root = parent->root;
370 * now determine if one of our ancestor triples is of same type.
371 * If so, we break here. This will make the parser unwind until
372 * we become a sibling (alternate variant) to the aforementioned
373 * triple. It stinks that we re-parse these tags on every
374 * iteration of this. This is a function in need of a rewrite.
376 for (p = parent; p->which == DATA1N_variant; p = p->parent)
377 if (p->u.variant.type == tp)
379 res = data1_mk_node (dh, m);
380 res->which = DATA1N_variant;
381 res->root = parent->root;
382 res->u.variant.type = tp;
383 res->u.variant.value =
384 data1_insert_string (dh, res, m, args + val_offset);
388 else /* tag.. acquire our element in the abstract syntax */
390 data1_node *partag = get_parent_tag (dh, parent);
391 data1_element *e = 0;
394 if (parent->which == DATA1N_variant)
397 if (!(e = partag->u.tag.element))
398 localtag = 1; /* our parent is a local tag */
400 elem = data1_getelementbytagname(dh, absyn, e, tag);
401 res = data1_mk_node (dh, m);
402 res->which = DATA1N_tag;
403 res->u.tag.tag = data1_insert_string (dh, res, m, tag);
404 res->u.tag.element = elem;
405 res->u.tag.node_selected = 0;
406 res->u.tag.make_variantlist = 0;
407 res->u.tag.no_data_requested = 0;
408 res->u.tag.get_bytes = -1;
409 res->root = parent->root;
413 res->parent = parent;
418 while ((*pp = data1_read_node(dh, buf, res, line, absyn, m)))
420 res->last_child = *pp;
424 else /* != '<'... this is a body of text */
428 int len, prev_char = 0;
433 res = data1_mk_node(dh, m);
434 res->parent = parent;
435 res->which = DATA1N_data;
436 res->u.data.what = DATA1I_text;
437 res->u.data.formatted_text = 0;
438 res->root = parent->root;
440 /* determine length of "data" */
441 src = strchr (*buf, '<');
447 /* use local buffer of nmem if too large */
448 if (len >= DATA1_LOCALDATA)
449 res->u.data.data = (char*) nmem_malloc (m, len);
451 res->u.data.data = res->lbuf;
453 /* read "data" and transfer while removing white space */
454 dst = res->u.data.data;
455 for (src = *buf; --len >= 0; src++)
472 res->u.data.len = dst - res->u.data.data;
478 * Read a record in the native syntax.
480 data1_node *data1_read_record(data1_handle dh,
481 int (*rf)(void *, char *, size_t), void *fh,
485 char **buf = data1_get_read_buf (dh, &size);
491 *buf = (char *)xmalloc(*size = 4096);
495 if (rd + 4096 > *size && !(*buf =(char *)xrealloc(*buf, *size *= 2)))
497 if ((res = (*rf)(fh, *buf + rd, 4096)) <= 0)
502 return data1_read_node(dh, &bp, 0, &line, 0, m);
511 data1_node *data1_read_sgml (data1_handle dh, NMEM m, const char *buf)
513 const char *bp = buf;
515 return data1_read_node (dh, &bp, 0, &line, 0, m);