2 * Copyright (c) 1995-1998, Index Data.
3 * See the file LICENSE for details.
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.21 1998-02-27 14:08:05 adam
8 * Added const to some char pointer arguments.
9 * Reworked data1_read_node so that it doesn't create a tree with
10 * pointers to original "SGML"-buffer.
12 * Revision 1.20 1998/02/11 11:53:35 adam
13 * Changed code so that it compiles as C++.
15 * Revision 1.19 1997/12/09 16:17:09 adam
16 * Fix bug regarding variants. Tags with prefix "var" was incorrectly
17 * interpreted as "start of variants". Now, only "var" indicates such
19 * Cleaned up data1_read_node so tag names and variant names are
20 * copied and not pointed to by the generated data1 tree. Data nodes
21 * still point to old buffer.
23 * Revision 1.18 1997/11/18 09:51:09 adam
24 * Removed element num_children from data1_node. Minor changes in
27 * Revision 1.17 1997/11/05 09:20:51 adam
30 * Revision 1.16 1997/09/17 12:10:37 adam
33 * Revision 1.15 1997/09/05 09:50:57 adam
34 * Removed global data1_tabpath - uses data1_get_tabpath() instead.
36 * Revision 1.14 1997/05/14 06:54:04 adam
39 * Revision 1.13 1996/10/29 13:35:38 adam
40 * Implemented data1_set_tabpath and data1_get_tabpath.
42 * Revision 1.12 1996/10/11 10:35:38 adam
43 * Fixed a bug that caused data1_read_node to core dump when no abstract
44 * syntax was defined in a "sgml"-record.
46 * Revision 1.11 1996/07/06 19:58:35 quinn
47 * System headerfiles gathered in yconfig
49 * Revision 1.10 1996/01/19 15:41:47 quinn
50 * Fixed uninitialized boolean.
52 * Revision 1.9 1996/01/17 14:52:47 adam
53 * Changed prototype for reader function parsed to data1_read_record.
55 * Revision 1.8 1995/12/15 16:20:41 quinn
56 * Added formatted text.
58 * Revision 1.7 1995/12/13 13:44:32 quinn
59 * Modified Data1-system to use nmem
61 * Revision 1.6 1995/12/12 16:37:08 quinn
62 * Added destroy element to data1_node.
64 * Revision 1.5 1995/12/11 15:22:37 quinn
65 * Added last_child field to the node.
66 * Rewrote schema-mapping.
68 * Revision 1.4 1995/11/13 09:27:36 quinn
69 * Fiddling with the variant stuff.
71 * Revision 1.3 1995/11/01 16:34:57 quinn
72 * Making data1 look for tables in data1_tabpath
74 * Revision 1.2 1995/11/01 13:54:48 quinn
77 * Revision 1.1 1995/11/01 11:56:09 quinn
78 * Added Retrieval (data management) functions en masse.
80 * Revision 1.14 1995/10/30 12:40:55 quinn
81 * Fixed a couple of bugs.
83 * Revision 1.13 1995/10/25 16:00:47 quinn
84 * USMARC support is now almost operational
86 * Revision 1.12 1995/10/16 14:02:55 quinn
87 * Changes to support element set names and espec1
89 * Revision 1.11 1995/10/13 16:05:08 quinn
90 * Adding Espec1-processing
92 * Revision 1.10 1995/10/11 14:53:44 quinn
95 * Revision 1.9 1995/10/06 16:56:50 quinn
96 * Fixed ranked result.
98 * Revision 1.8 1995/10/06 16:44:13 quinn
99 * Work on attribute set mapping, etc.
101 * Revision 1.7 1995/10/06 12:58:35 quinn
104 * Revision 1.6 1995/10/04 09:29:49 quinn
105 * Adjustments to support USGS test data
107 * Revision 1.5 1995/10/03 17:56:43 quinn
110 * Revision 1.4 1995/10/02 15:53:19 quinn
113 * Revision 1.3 1995/10/02 14:55:21 quinn
114 * *** empty log message ***
116 * Revision 1.2 1995/09/14 15:18:13 quinn
119 * Revision 1.1 1995/09/12 11:24:30 quinn
120 * Beginning to add code for structured records.
135 * get the tag which is the immediate parent of this node (this may mean
136 * traversing intermediate things like variants and stuff.
138 data1_node *get_parent_tag (data1_handle dh, data1_node *n)
140 for (; n && n->which != DATA1N_root; n = n->parent)
141 if (n->which == DATA1N_tag)
146 data1_node *data1_mk_node (data1_handle dh, NMEM m)
150 r = (data1_node *)nmem_malloc(m, sizeof(*r));
151 r->next = r->child = r->last_child = r->parent = 0;
156 void data1_free_tree (data1_handle dh, data1_node *t)
158 data1_node *p = t->child, *pn;
163 data1_free_tree (dh, p);
170 char *data1_insert_string (data1_handle dh, data1_node *res,
171 NMEM m, const char *str)
173 int len = strlen(str);
175 if (len >= DATA1_LOCALDATA)
176 return nmem_strdup (m, str);
179 strcpy (res->lbuf, str);
185 data1_node *data1_add_insert_taggeddata(data1_handle dh, data1_node *root,
187 const char *tagname, NMEM m,
190 data1_node *partag = get_parent_tag (dh, at);
191 data1_node *tagn = data1_mk_node (dh, m);
192 data1_element *e = NULL;
195 tagn->which = DATA1N_tag;
197 tagn->u.tag.tag = data1_insert_string (dh, tagn, m, tagname);
198 tagn->u.tag.node_selected = 0;
199 tagn->u.tag.make_variantlist = 0;
200 tagn->u.tag.no_data_requested = 0;
201 tagn->u.tag.get_bytes = -1;
204 e = partag->u.tag.element;
205 tagn->u.tag.element =
206 data1_getelementbytagname (dh, root->u.root.absyn, e, tagname);
207 tagn->last_child = tagn->child = datn = data1_mk_node (dh, m);
210 datn->which = DATA1N_data;
211 datn->u.data.formatted_text = 0;
216 tagn->next = at->child;
218 at->last_child = tagn;
227 assert (at->last_child);
228 at->last_child->next = tagn;
230 at->last_child = tagn;
235 data1_node *data1_add_taggeddata(data1_handle dh, data1_node *root,
237 const char *tagname, NMEM m)
239 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 0);
244 * Insert a tagged node into the record root as first child of the node at
245 * which should be root or tag itself). Returns pointer to the data node,
246 * which can then be modified.
248 data1_node *data1_insert_taggeddata(data1_handle dh, data1_node *root,
250 const char *tagname, NMEM m)
252 return data1_add_insert_taggeddata (dh, root, at, tagname, m, 1);
256 * Ugh. Sometimes functions just grow and grow on you. This one reads a
257 * 'node' and its children.
259 data1_node *data1_read_node (data1_handle dh, const char **buf,
260 data1_node *parent, int *line,
261 data1_absyn *absyn, NMEM m)
265 while (**buf && isspace(**buf))
274 if (**buf == '<') /* beginning of tag */
279 const char *t = (*buf) + 1;
281 data1_element *elem = 0;
283 for (i = 0; *t && *t != '>' && !isspace(*t); t++)
284 if (i < (sizeof(tag)-1))
289 for (i = 0; *t && *t != '>'; t++)
290 if (i < (sizeof(args)-1))
293 if (*t != '>' && !isspace(*t))
295 logf(LOG_WARN, "d1: %d: Malformed tag", *line);
299 * if end-tag, see if we terminate parent. If so, consume and return.
304 if (parent && (!*(tag +1) ||
305 (parent->which == DATA1N_root &&
306 !strcmp(tag + 1,parent->u.root.type)) ||
307 (parent->which == DATA1N_tag &&
308 !strcmp(tag + 1, parent->u.tag.tag))))
312 if (!absyn) /* parent node - what are we? */
314 if (!(absyn = data1_get_absyn (dh, tag)))
316 logf(LOG_WARN, "Unable to acquire abstract syntax for '%s'",
320 res = data1_mk_node (dh, m);
321 res->which = DATA1N_root;
322 res->u.root.type = data1_insert_string (dh, res, m, tag);
323 res->u.root.absyn = absyn;
327 else if (!strcmp(tag, "var"))
329 char tclass[DATA1_MAX_SYMBOL], type[DATA1_MAX_SYMBOL];
334 if (sscanf(args, "%s %s %n", tclass, type, &val_offset) != 2)
336 logf(LOG_WARN, "Malformed variant triple at '%s'", tag);
340 data1_getvartypebyct(dh, parent->root->u.root.absyn->varset,
345 * If we're the first variant in this group, create a parent var,
346 * and insert it before the current variant.
348 if (parent->which != DATA1N_variant)
350 res = data1_mk_node (dh, m);
351 res->which = DATA1N_variant;
352 res->u.variant.type = 0;
353 res->u.variant.value = 0;
354 res->root = parent->root;
359 * now determine if one of our ancestor triples is of same type.
360 * If so, we break here. This will make the parser unwind until
361 * we become a sibling (alternate variant) to the aforementioned
362 * triple. It stinks that we re-parse these tags on every
363 * iteration of this. This is a function in need of a rewrite.
365 for (p = parent; p->which == DATA1N_variant; p = p->parent)
366 if (p->u.variant.type == tp)
368 res = data1_mk_node (dh, m);
369 res->which = DATA1N_variant;
370 res->root = parent->root;
371 res->u.variant.type = tp;
372 res->u.variant.value =
373 data1_insert_string (dh, res, m, args + val_offset);
377 else /* tag.. acquire our element in the abstract syntax */
379 data1_node *partag = get_parent_tag (dh, parent);
380 data1_element *e = 0;
383 if (parent->which == DATA1N_variant)
386 if (!(e = partag->u.tag.element))
387 localtag = 1; /* our parent is a local tag */
389 elem = data1_getelementbytagname(dh, absyn, e, tag);
390 res = data1_mk_node (dh, m);
391 res->which = DATA1N_tag;
392 res->u.tag.tag = data1_insert_string (dh, res, m, tag);
393 res->u.tag.element = elem;
394 res->u.tag.node_selected = 0;
395 res->u.tag.make_variantlist = 0;
396 res->u.tag.no_data_requested = 0;
397 res->u.tag.get_bytes = -1;
398 res->root = parent->root;
402 res->parent = parent;
407 while ((*pp = data1_read_node(dh, buf, res, line, absyn, m)))
409 res->last_child = *pp;
413 else /* != '<'... this is a body of text */
417 int len, prev_char = 0;
422 res = data1_mk_node(dh, m);
423 res->parent = parent;
424 res->which = DATA1N_data;
425 res->u.data.what = DATA1I_text;
426 res->u.data.formatted_text = 0;
427 res->root = parent->root;
429 /* determine length of "data" */
430 src = strchr (*buf, '<');
436 /* use local buffer of nmem if too large */
437 if (len >= DATA1_LOCALDATA)
438 res->u.data.data = nmem_malloc (m, len);
440 res->u.data.data = res->lbuf;
442 /* read "data" and transfer while removing white space */
443 dst = res->u.data.data;
444 for (src = *buf; --len >= 0; src++)
461 res->u.data.len = dst - res->u.data.data;
467 * Read a record in the native syntax.
469 data1_node *data1_read_record(data1_handle dh,
470 int (*rf)(void *, char *, size_t), void *fh,
474 char **buf = data1_get_read_buf (dh, &size);
480 *buf = (char *)xmalloc(*size = 4096);
484 if (rd + 4096 > *size && !(*buf =(char *)xrealloc(*buf, *size *= 2)))
486 if ((res = (*rf)(fh, *buf + rd, 4096)) <= 0)
491 return data1_read_node(dh, &bp, 0, &line, 0, m);
500 data1_node *data1_read_sgml (data1_handle dh, NMEM m, const char *buf)
502 const char *bp = buf;
504 return data1_read_node (dh, &bp, 0, &line, 0, m);