X-Git-Url: http://git.indexdata.com/?p=yaz-moved-to-github.git;a=blobdiff_plain;f=retrieval%2Fd1_read.c;h=8ff1563c5be3985fbb0048e974f09c37bc5fd5ed;hp=e06f8f46fda8beb68d7044cc87b757796649bdf8;hb=82780038a8dfa288d03d6e9c8ef9df5f4e67fbfe;hpb=e147ddf129696b470aed3cfcdb405af84ee56ef6
diff --git a/retrieval/d1_read.c b/retrieval/d1_read.c
index e06f8f4..8ff1563 100644
--- a/retrieval/d1_read.c
+++ b/retrieval/d1_read.c
@@ -1,10 +1,45 @@
/*
- * Copyright (c) 1995-1998, Index Data.
+ * Copyright (c) 1995-2000, Index Data.
* See the file LICENSE for details.
* Sebastian Hammer, Adam Dickmeiss
*
* $Log: d1_read.c,v $
- * Revision 1.25 1999-04-20 09:56:48 adam
+ * Revision 1.35 2000-12-05 14:44:25 adam
+ * Readers skips sections.
+ *
+ * Revision 1.34 2000/12/05 10:06:23 adam
+ * Added support for null-data rules like .
+ *
+ * Revision 1.33 2000/11/29 14:22:47 adam
+ * Implemented XML/SGML attributes for data1 so that d1_read reads them
+ * and d1_write generates proper attributes for XML/SGML records. Added
+ * register locking for threaded version.
+ *
+ * Revision 1.32 2000/01/06 11:25:59 adam
+ * Added case to prevent warning.
+ *
+ * Revision 1.31 1999/12/21 14:16:20 ian
+ * Changed retrieval module to allow data1 trees with no associated absyn.
+ * Also added a simple interface for extracting values from data1 trees using
+ * a string based tagpath.
+ *
+ * Revision 1.30 1999/11/30 13:47:12 adam
+ * Improved installation. Moved header files to include/yaz.
+ *
+ * Revision 1.29 1999/10/21 12:06:29 adam
+ * Retrieval module no longer uses ctype.h - functions.
+ *
+ * Revision 1.28 1999/10/21 09:50:33 adam
+ * SGML reader uses own isspace - it doesn't do 8-bit on WIN32!
+ *
+ * Revision 1.27 1999/08/27 09:40:32 adam
+ * Renamed logf function to yaz_log. Removed VC++ project files.
+ *
+ * Revision 1.26 1999/07/13 13:23:48 adam
+ * Non-recursive version of data1_read_node. data1_read_nodex reads
+ * stream of bytes (instead of buffer in memory).
+ *
+ * Revision 1.25 1999/04/20 09:56:48 adam
* Added 'name' paramter to encoder/decoder routines (typedef Odr_fun).
* Modified all encoders/decoders to reflect this change.
*
@@ -139,13 +174,12 @@
*/
#include
-#include
#include
#include
-#include
-#include
-#include
+#include
+#include
+#include
/*
* get the tag which is the immediate parent of this node (this may mean
@@ -169,6 +203,41 @@ data1_node *data1_mk_node (data1_handle dh, NMEM m)
return r;
}
+data1_node *data1_mk_node_type (data1_handle dh, NMEM m, int type)
+{
+ data1_node *r;
+
+ r = data1_mk_node(dh, m);
+ r->which = type;
+ switch(type)
+ {
+ case DATA1N_tag:
+ r->u.tag.tag = 0;
+ r->u.tag.element = 0;
+ r->u.tag.no_data_requested = 0;
+ r->u.tag.node_selected = 0;
+ r->u.tag.make_variantlist = 0;
+ r->u.tag.get_bytes = -1;
+#if DATA1_USING_XATTR
+ r->u.tag.attributes = 0;
+#endif
+ break;
+ case DATA1N_root:
+ r->u.root.type = 0;
+ r->u.root.absyn = 0;
+ break;
+ case DATA1N_data:
+ r->u.data.data = 0;
+ r->u.data.len = 0;
+ r->u.data.what = 0;
+ r->u.data.formatted_text = 0;
+ break;
+ default:
+ logf (LOG_WARN, "data_mk_node_type. bad type = %d\n", type);
+ }
+ return r;
+}
+
void data1_free_tree (data1_handle dh, data1_node *t)
{
data1_node *p = t->child, *pn;
@@ -197,23 +266,17 @@ char *data1_insert_string (data1_handle dh, data1_node *res,
}
}
-
data1_node *data1_add_insert_taggeddata(data1_handle dh, data1_node *root,
data1_node *at,
const char *tagname, NMEM m,
int first_flag, int local_allowed)
{
data1_node *partag = get_parent_tag (dh, at);
- data1_node *tagn = data1_mk_node (dh, m);
+ data1_node *tagn = data1_mk_node_type (dh, m, DATA1N_tag);
data1_element *e = NULL;
data1_node *datn;
- tagn->which = DATA1N_tag;
tagn->u.tag.tag = data1_insert_string (dh, tagn, m, tagname);
- tagn->u.tag.node_selected = 0;
- tagn->u.tag.make_variantlist = 0;
- tagn->u.tag.no_data_requested = 0;
- tagn->u.tag.get_bytes = -1;
if (partag)
e = partag->u.tag.element;
@@ -221,11 +284,10 @@ data1_node *data1_add_insert_taggeddata(data1_handle dh, data1_node *root,
data1_getelementbytagname (dh, root->u.root.absyn, e, tagname);
if (!local_allowed && !tagn->u.tag.element)
return NULL;
- tagn->last_child = tagn->child = datn = data1_mk_node (dh, m);
+ tagn->last_child = tagn->child = datn = data1_mk_node_type (dh, m, DATA1N_data);
tagn->root = root;
datn->parent = tagn;
datn->root = root;
- datn->which = DATA1N_data;
datn->u.data.formatted_text = 0;
tagn->parent = at;
@@ -270,215 +332,348 @@ data1_node *data1_insert_taggeddata(data1_handle dh, data1_node *root,
return data1_add_insert_taggeddata (dh, root, at, tagname, m, 1, 0);
}
+#if DATA1_USING_XATTR
+data1_xattr *data1_read_xattr (data1_handle dh, NMEM m,
+ int (*get_byte)(void *fh), void *fh,
+ WRBUF wrbuf, int *ch)
+{
+ data1_xattr *p_first = 0;
+ data1_xattr **pp = &p_first;
+ int c = *ch;
+ for (;;)
+ {
+ data1_xattr *p;
+ int len;
+ while (c && d1_isspace(c))
+ c = (*get_byte)(fh);
+ if (!c || c == '>' || c == '/')
+ break;
+ *pp = p = nmem_malloc (m, sizeof(*p));
+ p->next = 0;
+ pp = &p->next;
+ p->value = 0;
+
+ wrbuf_rewind(wrbuf);
+ while (c && c != '=' && c != '>' && c != '/' && !d1_isspace(c))
+ {
+ wrbuf_putc (wrbuf, c);
+ c = (*get_byte)(fh);
+ }
+ wrbuf_putc (wrbuf, '\0');
+ len = wrbuf_len(wrbuf);
+ p->name = nmem_malloc (m, len);
+ strcpy (p->name, wrbuf_buf(wrbuf));
+ if (c == '=')
+ {
+ c = (*get_byte)(fh);
+ if (c == '"')
+ c = (*get_byte)(fh);
+ wrbuf_rewind(wrbuf);
+ while (c && c != '"' && c != '>' && c != '/')
+ {
+ wrbuf_putc (wrbuf, c);
+ c = (*get_byte)(fh);
+ }
+ wrbuf_putc (wrbuf, '\0');
+ len = wrbuf_len(wrbuf);
+ p->value = nmem_malloc (m, len);
+ strcpy (p->value, wrbuf_buf(wrbuf));
+ if (c == '"')
+ c = (*get_byte)(fh);
+ }
+ }
+ *ch = c;
+ return p_first;
+}
+#endif
+
/*
* Ugh. Sometimes functions just grow and grow on you. This one reads a
* 'node' and its children.
*/
-data1_node *data1_read_node (data1_handle dh, const char **buf,
- data1_node *parent, int *line,
- data1_absyn *absyn, NMEM m)
+data1_node *data1_read_nodex (data1_handle dh, NMEM m,
+ int (*get_byte)(void *fh), void *fh, WRBUF wrbuf)
{
+ data1_absyn *absyn = 0;
+ data1_node *d1_stack[256];
data1_node *res;
+ int c;
+ int level = 0;
+ int line = 1;
- while (**buf && isspace(**buf))
+ d1_stack[level] = 0;
+ c = (*get_byte)(fh);
+ while (1)
{
- if (**buf == '\n')
- (*line)++;
- (*buf)++;
- }
- if (!**buf)
- return 0;
-
- if (**buf == '<') /* beginning of tag */
- {
- char tag[64];
- char args[256];
- size_t i;
- const char *t = (*buf) + 1;
- data1_node **pp;
- data1_element *elem = 0;
-
- for (i = 0; *t && *t != '>' && !isspace(*t); t++)
- if (i < (sizeof(tag)-1))
- tag[i++] = *t;
- tag[i] = '\0';
- while (isspace(*t))
- t++;
- for (i = 0; *t && *t != '>'; t++)
- if (i < (sizeof(args)-1))
- args[i++] = *t;
- args[i] = '\0';
- if (*t != '>' && !isspace(*t))
+ data1_node *parent = level ? d1_stack[level-1] : 0;
+ while (c != '\0' && d1_isspace(c))
{
- logf(LOG_WARN, "d1: %d: Malformed tag", *line);
- return 0;
+ if (c == '\n')
+ line++;
+ c = (*get_byte)(fh);
}
- /*
- * if end-tag, see if we terminate parent. If so, consume and return.
- * Else, return.
- */
- if (*tag == '/')
- {
- if (parent && (!*(tag +1) ||
- (parent->which == DATA1N_root &&
- !strcmp(tag + 1,parent->u.root.type)) ||
- (parent->which == DATA1N_tag &&
- !strcmp(tag + 1, parent->u.tag.tag))))
- *buf = t + 1;
- return 0;
- }
- if (!absyn) /* parent node - what are we? */
+ if (c == '\0')
+ break;
+
+ if (c == '<') /* beginning of tag */
{
- if (!(absyn = data1_get_absyn (dh, tag)))
+#if DATA1_USING_XATTR
+ data1_xattr *xattr;
+#endif
+ char tag[64];
+ char args[256];
+ int null_tag = 0;
+ int end_tag = 0;
+ size_t i = 0;
+
+ c = (*get_byte)(fh);
+ if (c == '/')
{
- logf(LOG_WARN, "Unable to acquire abstract syntax for '%s'",
- tag);
- return 0;
+ end_tag = 1;
+ c = (*get_byte)(fh);
}
- res = data1_mk_node (dh, m);
- res->which = DATA1N_root;
- res->u.root.type = data1_insert_string (dh, res, m, tag);
- res->u.root.absyn = absyn;
- res->root = res;
- *buf = t + 1;
- }
- else if (!strcmp(tag, "var"))
- {
- char tclass[DATA1_MAX_SYMBOL], type[DATA1_MAX_SYMBOL];
- data1_vartype *tp;
- int val_offset;
- data1_node *p;
-
- if (sscanf(args, "%s %s %n", tclass, type, &val_offset) != 2)
+ else if (c == '!') /* tags/comments that we don't deal with yet */
{
- logf(LOG_WARN, "Malformed variant triple at '%s'", tag);
- return 0;
+ while (c && c != '>')
+ c = (*get_byte)(fh);
+ if (c)
+ c = (*get_byte)(fh);
+ continue;
}
- if (!(tp =
- data1_getvartypebyct(dh, parent->root->u.root.absyn->varset,
- tclass, type)))
- return 0;
-
- /*
- * If we're the first variant in this group, create a parent var,
- * and insert it before the current variant.
- */
- if (parent->which != DATA1N_variant)
+ while (c && c != '>' && c != '/' && !d1_isspace(c))
{
- res = data1_mk_node (dh, m);
- res->which = DATA1N_variant;
- res->u.variant.type = 0;
- res->u.variant.value = 0;
- res->root = parent->root;
+ if (i < (sizeof(tag)-1))
+ tag[i++] = c;
+ c = (*get_byte)(fh);
+ }
+ tag[i] = '\0';
+#if DATA1_USING_XATTR
+ xattr = data1_read_xattr (dh, m, get_byte, fh, wrbuf, &c);
+ args[0] = '\0';
+#else
+ while (d1_isspace(c))
+ c = (*get_byte)(fh);
+ for (i = 0; c && c != '>' && c != '/'; c = (*get_byte)(fh))
+ if (i < (sizeof(args)-1))
+ args[i++] = c;
+ args[i] = '\0';
+#endif
+ if (c == '/')
+ { /* or */
+ null_tag = 1;
+ c = (*get_byte)(fh);
+ }
+ if (c != '>')
+ {
+ yaz_log(LOG_WARN, "d1: %d: Malformed tag", line);
+ return 0;
}
else
+ c = (*get_byte)(fh);
+
+ /* End tag? */
+ if (end_tag)
+ {
+ if (*tag == '\0')
+ --level; /* */
+ else
+ { /* */
+ int i = level;
+ while (i > 0)
+ {
+ parent = d1_stack[--i];
+ if ((parent->which == DATA1N_root &&
+ !strcmp(tag, parent->u.root.type)) ||
+ (parent->which == DATA1N_tag &&
+ !strcmp(tag, parent->u.tag.tag)))
+ {
+ level = i;
+ break;
+ }
+ }
+ if (i != level)
+ {
+ yaz_log (LOG_WARN, "%d: no begin tag for %s",
+ line, tag);
+ break;
+ }
+ }
+ if (level == 0)
+ return d1_stack[0];
+ continue;
+ }
+ if (level == 0) /* root ? */
+ {
+ if (!(absyn = data1_get_absyn (dh, tag)))
+ {
+ yaz_log(LOG_WARN, "Unable to acquire abstract syntax " "for '%s'", tag);
+ /* It's now OK for a record not to have an absyn */
+ }
+ res = data1_mk_node_type (dh, m, DATA1N_root);
+ res->u.root.type = data1_insert_string (dh, res, m, tag);
+ res->u.root.absyn = absyn;
+ res->root = res;
+ }
+ else if (!strcmp(tag, "var"))
{
+ char tclass[DATA1_MAX_SYMBOL], type[DATA1_MAX_SYMBOL];
+ data1_vartype *tp;
+ int val_offset;
+
+ if (sscanf(args, "%s %s %n", tclass, type, &val_offset) != 2)
+ {
+ yaz_log(LOG_WARN, "Malformed variant triple at '%s'", tag);
+ continue;
+ }
+ if (!(tp =
+ data1_getvartypebyct(dh,
+ parent->root->u.root.absyn->varset,
+ tclass, type)))
+ continue;
/*
- * now determine if one of our ancestor triples is of same type.
- * If so, we break here. This will make the parser unwind until
- * we become a sibling (alternate variant) to the aforementioned
- * triple. It stinks that we re-parse these tags on every
- * iteration of this. This is a function in need of a rewrite.
+ * If we're the first variant in this group, create a parent
+ * variant, and insert it before the current variant.
*/
- for (p = parent; p->which == DATA1N_variant; p = p->parent)
- if (p->u.variant.type == tp)
- return 0;
- res = data1_mk_node (dh, m);
- res->which = DATA1N_variant;
+ if (parent->which != DATA1N_variant)
+ {
+ res = data1_mk_node (dh, m);
+ res->which = DATA1N_variant;
+ res->u.variant.type = 0;
+ res->u.variant.value = 0;
+ }
+ else
+ {
+ /*
+ * now determine if one of our ancestor triples is of
+ * same type. If so, we break here.
+ */
+ int i;
+ for (i = level-1; d1_stack[i]->which==DATA1N_variant; --i)
+ if (d1_stack[i]->u.variant.type == tp)
+ {
+ level = i;
+ break;
+ }
+ res = data1_mk_node (dh, m);
+ res->which = DATA1N_variant;
+ res->u.variant.type = tp;
+ res->u.variant.value =
+ data1_insert_string (dh, res, m, args + val_offset);
+ }
+ }
+ else /* tag.. acquire our element in the abstract syntax */
+ {
+ data1_node *partag = get_parent_tag (dh, parent);
+ data1_element *elem, *e = 0;
+ int localtag = 0;
+
+ if (parent->which == DATA1N_variant)
+ return 0;
+ if (partag)
+ if (!(e = partag->u.tag.element))
+ localtag = 1; /* our parent is a local tag */
+
+ elem = data1_getelementbytagname(dh, absyn, e, tag);
+ res = data1_mk_node_type (dh, m, DATA1N_tag);
+ res->u.tag.tag = data1_insert_string (dh, res, m, tag);
+ res->u.tag.element = elem;
+#if DATA1_USING_XATTR
+ res->u.tag.attributes = xattr;
+#endif
+ }
+ if (parent)
+ {
+ parent->last_child = res;
res->root = parent->root;
- res->u.variant.type = tp;
- res->u.variant.value =
- data1_insert_string (dh, res, m, args + val_offset);
- *buf = t + 1;
}
+ res->parent = parent;
+ if (d1_stack[level])
+ d1_stack[level]->next = res;
+ else if (parent)
+ parent->child = res;
+ d1_stack[level] = res;
+ d1_stack[level+1] = 0;
+ if (!null_tag)
+ ++level;
}
- else /* tag.. acquire our element in the abstract syntax */
+ else /* != '<'... this is a body of text */
{
- data1_node *partag = get_parent_tag (dh, parent);
- data1_element *e = 0;
- int localtag = 0;
-
- if (parent->which == DATA1N_variant)
- return 0;
- if (partag)
- if (!(e = partag->u.tag.element))
- localtag = 1; /* our parent is a local tag */
-
- elem = data1_getelementbytagname(dh, absyn, e, tag);
- res = data1_mk_node (dh, m);
- res->which = DATA1N_tag;
- res->u.tag.tag = data1_insert_string (dh, res, m, tag);
- res->u.tag.element = elem;
- res->u.tag.node_selected = 0;
- res->u.tag.make_variantlist = 0;
- res->u.tag.no_data_requested = 0;
- res->u.tag.get_bytes = -1;
+ const char *src;
+ char *dst;
+ int len, prev_char = 0;
+
+ if (level == 0)
+ {
+ c = (*get_byte)(fh);
+ continue;
+ }
+ res = data1_mk_node_type (dh, m, DATA1N_data);
+ res->parent = parent;
+ res->u.data.what = DATA1I_text;
+ res->u.data.formatted_text = 0;
res->root = parent->root;
- *buf = t + 1;
- }
-
- res->parent = parent;
- pp = &res->child;
- /*
- * Read child nodes.
- */
- while ((*pp = data1_read_node(dh, buf, res, line, absyn, m)))
- {
- res->last_child = *pp;
- pp = &(*pp)->next;
- }
- }
- else /* != '<'... this is a body of text */
- {
- const char *src;
- char *dst;
- int len, prev_char = 0;
-
- if (!parent)
- return 0;
-
- res = data1_mk_node(dh, m);
- res->parent = parent;
- res->which = DATA1N_data;
- res->u.data.what = DATA1I_text;
- res->u.data.formatted_text = 0;
- res->root = parent->root;
-
- /* determine length of "data" */
- src = strchr (*buf, '<');
- if (src)
- len = src - *buf;
- else
- len = strlen (*buf);
+ parent->last_child = res;
+ if (d1_stack[level])
+ d1_stack[level]->next = res;
+ else
+ parent->child = res;
+ d1_stack[level] = res;
+
+ wrbuf_rewind(wrbuf);
- /* use local buffer of nmem if too large */
- if (len >= DATA1_LOCALDATA)
- res->u.data.data = (char*) nmem_malloc (m, len);
- else
- res->u.data.data = res->lbuf;
+ while (c && c != '<')
+ {
+ wrbuf_putc (wrbuf, c);
+ c = (*get_byte)(fh);
+ }
+ len = wrbuf_len(wrbuf);
- /* read "data" and transfer while removing white space */
- dst = res->u.data.data;
- for (src = *buf; --len >= 0; src++)
- {
- if (*src == '\n')
- (*line)++;
- if (isspace (*src))
- prev_char = ' ';
+ /* use local buffer of nmem if too large */
+ if (len >= DATA1_LOCALDATA)
+ res->u.data.data = (char*) nmem_malloc (m, len);
else
+ res->u.data.data = res->lbuf;
+
+ /* read "data" and transfer while removing white space */
+ dst = res->u.data.data;
+ for (src = wrbuf_buf(wrbuf); --len >= 0; src++)
{
- if (prev_char)
+ if (*src == '\n')
+ line++;
+ if (d1_isspace (*src))
+ prev_char = ' ';
+ else
{
- *dst++ = prev_char;
- prev_char = 0;
+ if (prev_char)
+ {
+ *dst++ = prev_char;
+ prev_char = 0;
+ }
+ *dst++ = *src;
}
- *dst++ = *src;
}
+ res->u.data.len = dst - res->u.data.data;
}
- *buf = src;
- res->u.data.len = dst - res->u.data.data;
}
- return res;
+ return 0;
+}
+
+int getc_mem (void *fh)
+{
+ const char **p = (const char **) fh;
+ if (**p)
+ return *(*p)++;
+ return 0;
+}
+
+data1_node *data1_read_node (data1_handle dh, const char **buf, NMEM m)
+{
+ WRBUF wrbuf = wrbuf_alloc();
+ data1_node *node;
+
+ node = data1_read_nodex(dh, m, getc_mem, (void *) (buf), wrbuf);
+ wrbuf_free (wrbuf, 1);
+ return node;
}
/*
@@ -492,21 +687,21 @@ data1_node *data1_read_record(data1_handle dh,
char **buf = data1_get_read_buf (dh, &size);
const char *bp;
int rd = 0, res;
- int line = 0;
if (!*buf)
*buf = (char *)xmalloc(*size = 4096);
for (;;)
{
- if (rd + 4096 > *size && !(*buf =(char *)xrealloc(*buf, *size *= 2)))
+ if (rd + 2048 >= *size && !(*buf =(char *)xrealloc(*buf, *size *= 2)))
abort();
- if ((res = (*rf)(fh, *buf + rd, 4096)) <= 0)
+ if ((res = (*rf)(fh, *buf + rd, 2048)) <= 0)
{
if (!res)
{
bp = *buf;
- return data1_read_node(dh, &bp, 0, &line, 0, m);
+ (*buf)[rd] = '\0';
+ return data1_read_node(dh, &bp, m);
}
else
return 0;
@@ -518,6 +713,6 @@ data1_node *data1_read_record(data1_handle dh,
data1_node *data1_read_sgml (data1_handle dh, NMEM m, const char *buf)
{
const char *bp = buf;
- int line = 0;
- return data1_read_node (dh, &bp, 0, &line, 0, m);
+ return data1_read_node (dh, &bp, m);
}
+