+struct source_parser {
+ int len;
+ const char *tok;
+ const char *src;
+ int lookahead;
+};
+
+static int sp_lex(struct source_parser *sp)
+{
+ while (*sp->src == ' ')
+ (sp->src)++;
+ sp->tok = sp->src;
+ sp->len = 0;
+ while (*sp->src && !strchr("<>();,-: ", *sp->src))
+ {
+ sp->src++;
+ sp->len++;
+ }
+ if (sp->len)
+ sp->lookahead = 't';
+ else
+ {
+ sp->lookahead = *sp->src;
+ if (*sp->src)
+ sp->src++;
+ }
+ return sp->lookahead;
+}
+
+
+static int sp_expr(struct source_parser *sp, data1_node *n, RecWord *wrd)
+{
+ if (sp->lookahead != 't')
+ return 0;
+ if (sp->len == 4 && !memcmp(sp->tok, "data", sp->len))
+ {
+ if (n->which == DATA1N_data)
+ {
+ wrd->string = n->u.data.data;
+ wrd->length = n->u.data.len;
+ }
+ sp_lex(sp);
+ }
+ else if (sp->len == 3 && !memcmp(sp->tok, "tag", sp->len))
+ {
+ if (n->which == DATA1N_tag)
+ {
+ wrd->string = n->u.tag.tag;
+ wrd->length = strlen(n->u.tag.tag);
+ }
+ sp_lex(sp);
+ }
+ else if (sp->len == 4 && !memcmp(sp->tok, "attr", sp->len))
+ {
+ sp_lex(sp);
+ if (sp->lookahead != '(')
+ return 0;
+ sp_lex(sp);
+ if (sp->lookahead != 't')
+ return 0;
+
+ if (n->which == DATA1N_tag)
+ {
+ data1_xattr *p = n->u.tag.attributes;
+ while (p && strlen(p->name) != sp->len &&
+ memcmp (p->name, sp->tok, sp->len))
+ p = p->next;
+ if (p)
+ {
+ wrd->string = p->value;
+ wrd->length = strlen(p->value);
+ }
+ }
+ sp_lex(sp);
+ if (sp->lookahead != ')')
+ return 0;
+ sp_lex(sp);
+ }
+ else if (sp->len == 5 && !memcmp(sp->tok, "range", sp->len))
+ {
+ int start, len;
+ sp_lex(sp);
+ if (sp->lookahead != '(')
+ return 0;
+
+ sp_lex(sp);
+ sp_expr(sp, n, wrd);
+ if (sp->lookahead != ',')
+ return 0;
+
+ sp_lex(sp);
+ if (sp->lookahead != 't')
+ return 0;
+ start = atoi_n(sp->tok, sp->len);
+
+ sp_lex(sp);
+ if (sp->lookahead != ',')
+ return 0;
+
+ sp_lex(sp);
+ if (sp->lookahead != 't')
+ return 0;
+ len = atoi_n(sp->tok, sp->len);
+
+ sp_lex(sp);
+ if (sp->lookahead != ')')
+ return 0;
+
+ sp_lex(sp);
+ if (wrd->string && wrd->length)
+ {
+ wrd->string += start;
+ wrd->length -= start;
+ if (wrd->length > len)
+ wrd->length = len;
+ }
+ }
+ return 1;
+}
+
+static int sp_parse(data1_node *n, RecWord *wrd, const char *src)
+{
+ struct source_parser sp;
+ sp.len = 0;
+ sp.tok = 0;
+ sp.src = src;
+ sp.lookahead = 0;
+ sp_lex(&sp);
+
+ return sp_expr(&sp, n, wrd);
+}
+
+int d1_check_xpath_predicate(data1_node *n, struct xpath_predicate *p)
+{
+ int res = 1;
+ char *attname;
+ data1_xattr *attr;
+
+ if (!p) {
+ return 1;
+ } else {
+ if (p->which == XPATH_PREDICATE_RELATION) {
+ if (p->u.relation.name[0]) {
+ if (*p->u.relation.name != '@') {
+ yaz_log(LOG_WARN,
+ " Only attributes (@) are supported in xelm xpath predicates");
+ yaz_log(LOG_WARN, "predicate %s ignored", p->u.relation.name);
+ return (1);
+ }
+ attname = p->u.relation.name + 1;
+ res = 0;
+ /* looking for the attribute with a specified name */
+ for (attr = n->u.tag.attributes; attr; attr = attr->next) {
+ yaz_log(LOG_DEBUG," - attribute %s <-> %s", attname, attr->name );
+
+ if (!strcmp(attr->name, attname)) {
+ if (p->u.relation.op[0]) {
+ if (*p->u.relation.op != '=') {
+ yaz_log(LOG_WARN,
+ "Only '=' relation is supported (%s)",p->u.relation.op);
+ yaz_log(LOG_WARN, "predicate %s ignored", p->u.relation.name);
+ res = 1; break;
+ } else {
+ yaz_log(LOG_DEBUG," - value %s <-> %s",
+ p->u.relation.value, attr->value );
+ if (!strcmp(attr->value, p->u.relation.value)) {
+ res = 1; break;
+ }
+ }
+ } else {
+ /* attribute exists, no value specified */
+ res = 1; break;
+ }
+ }
+ }
+ yaz_log(LOG_DEBUG, "return %d", res);
+ return res;
+ } else {
+ return 1;
+ }
+ }
+ else if (p->which == XPATH_PREDICATE_BOOLEAN) {
+ if (!strcmp(p->u.boolean.op,"and")) {
+ return d1_check_xpath_predicate(n, p->u.boolean.left)
+ && d1_check_xpath_predicate(n, p->u.boolean.right);
+ }
+ else if (!strcmp(p->u.boolean.op,"or")) {
+ return (d1_check_xpath_predicate(n, p->u.boolean.left)
+ || d1_check_xpath_predicate(n, p->u.boolean.right));
+ } else {
+ yaz_log(LOG_WARN, "Unknown boolean relation %s, ignored",p->u.boolean.op);
+ return 1;
+ }
+ }
+ }
+ return 0;
+}
+
+
+/* *ostrich*
+
+New function, looking for xpath "element" definitions in abs, by
+tagpath, using a kind of ugly regxp search.The DFA was built while
+parsing abs, so here we just go trough them and try to match
+against the given tagpath. The first matching entry is returned.
+
+pop, 2002-12-13
+
+Added support for enhanced xelm. Now [] predicates are considered
+as well, when selecting indexing rules... (why the hell it's called
+termlist???)
+
+pop, 2003-01-17
+
+*/
+
+data1_termlist *xpath_termlist_by_tagpath(char *tagpath, data1_node *n)
+{
+ data1_absyn *abs = n->root->u.root.absyn;
+ data1_xpelement *xpe = abs->xp_elements;
+ data1_node *nn;
+#ifdef ENHANCED_XELM
+ struct xpath_location_step *xp;
+
+#endif
+ char *pexpr = xmalloc(strlen(tagpath)+2);
+ int ok = 0;
+
+ sprintf (pexpr, "%s\n", tagpath);
+ yaz_log(LOG_DEBUG,"Checking tagpath %s",tagpath);
+ while (xpe)
+ {
+ struct DFA_state **dfaar = xpe->dfa->states;
+ struct DFA_state *s=dfaar[0];
+ struct DFA_tran *t;
+ const char *p;
+ int i;
+ unsigned char c;
+ int start_line = 1;
+
+ c = *pexpr++; t = s->trans; i = s->tran_no;
+ if ((c >= t->ch[0] && c <= t->ch[1]) || (!t->ch[0])) {
+ p = pexpr;
+ do {
+ if ((s = dfaar[t->to])->rule_no &&
+ (start_line || s->rule_nno)) {
+ ok = 1;
+ break;
+ }
+ for (t=s->trans, i=s->tran_no; --i >= 0; t++) {
+ if ((unsigned) *p >= t->ch[0] && (unsigned) *p <= t->ch[1])
+ break;
+ }
+ p++;
+ } while (i >= 0);
+ }
+ if (ok)
+ yaz_log(LOG_DEBUG," xpath match %s",xpe->xpath_expr);
+ else
+ yaz_log(LOG_DEBUG," xpath no match %s",xpe->xpath_expr);
+
+ pexpr--;
+ if (ok) {
+#ifdef ENHANCED_XELM
+ /* we have to check the perdicates up to the root node */
+ xp = xpe->xpath;
+
+ /* find the first tag up in the node structure */
+ nn = n; while (nn && nn->which != DATA1N_tag) {
+ nn = nn->parent;
+ }
+
+ /* go from inside out in the node structure, while going
+ backwards trough xpath location steps ... */
+ for (i=xpe->xpath_len - 1; i>0; i--) {
+
+ yaz_log(LOG_DEBUG,"Checking step %d: %s on tag %s",
+ i,xp[i].part,nn->u.tag.tag);
+
+ if (!d1_check_xpath_predicate(nn, xp[i].predicate)) {
+ yaz_log(LOG_DEBUG," Predicates didn't match");
+ ok = 0;
+ break;
+ }
+
+ if (nn->which == DATA1N_tag) {
+ nn = nn->parent;
+ }
+ }
+#endif
+ if (ok) {
+ break;
+ }
+ }
+ xpe = xpe->next;
+ }
+
+ xfree(pexpr);
+
+ if (ok) {
+ yaz_log(LOG_DEBUG,"Got it");
+ return xpe->termlists;
+ } else {
+ return NULL;
+ }
+}
+
+/* use
+ 1 start element (tag)
+ 2 end element
+ 3 start attr (and attr-exact)
+ 4 end attr
+
+ 1016 cdata
+ 1015 attr data
+
+ *ostrich*
+
+ Now, if there is a matching xelm described in abs, for the
+ indexed element or the attribute, then the data is handled according
+ to those definitions...
+
+ modified by pop, 2002-12-13
+*/
+
+/* add xpath index for an attribute */
+static void index_xpath_attr (char *tag_path, char *name, char *value,
+ char *structure, struct recExtractCtrl *p,
+ RecWord *wrd)
+{
+ wrd->attrSet = VAL_IDXPATH;
+ wrd->attrUse = 1;
+ wrd->reg_type = '0';
+ wrd->string = tag_path;
+ wrd->length = strlen(tag_path);
+ (*p->tokenAdd)(wrd);
+
+ if (value) {
+ wrd->attrUse = 1015;
+ wrd->reg_type = 'w';
+ wrd->string = value;
+ wrd->length = strlen(value);
+ (*p->tokenAdd)(wrd);
+ }
+
+ wrd->attrUse = 2;
+ wrd->reg_type = '0';
+ wrd->string = tag_path;
+ wrd->length = strlen(tag_path);
+ (*p->tokenAdd)(wrd);
+}
+
+
+static void index_xpath (data1_node *n, struct recExtractCtrl *p,
+ int level, RecWord *wrd, int use)
+{
+ int i;
+ char tag_path_full[1024];
+ size_t flen = 0;
+ data1_node *nn;
+ int termlist_only = 1;
+ data1_termlist *tl;
+ int xpdone = 0;
+
+ yaz_log(LOG_DEBUG, "index_xpath level=%d use=%d", level, use);
+ if ((!n->root->u.root.absyn) ||
+ (n->root->u.root.absyn->enable_xpath_indexing)) {
+ termlist_only = 0;
+ }
+
+ switch (n->which)
+ {
+ case DATA1N_data:
+ wrd->string = n->u.data.data;
+ wrd->length = n->u.data.len;
+ xpdone = 0;
+ flen = 0;
+
+ /* we have to fetch the whole path to the data tag */
+ for (nn = n; nn; nn = nn->parent) {
+ if (nn->which == DATA1N_tag) {
+ size_t tlen = strlen(nn->u.tag.tag);
+ if (tlen + flen > (sizeof(tag_path_full)-2)) return;
+ memcpy (tag_path_full + flen, nn->u.tag.tag, tlen);
+ flen += tlen;
+ tag_path_full[flen++] = '/';
+ }
+ else if (nn->which == DATA1N_root) break;
+ }
+
+ tag_path_full[flen] = 0;
+
+ /* If we have a matching termlist... */
+ if (n->root->u.root.absyn &&
+ (tl = xpath_termlist_by_tagpath(tag_path_full, n)))
+ {
+ for (; tl; tl = tl->next)
+ {
+ /* need to copy recword because it may be changed */
+ RecWord wrd_tl;
+ wrd->reg_type = *tl->structure;
+ /* this is the ! case, so structure is for the xpath index */
+ memcpy (&wrd_tl, wrd, sizeof(*wrd));
+ if (tl->source)
+ sp_parse(n, &wrd_tl, tl->source);
+ if (!tl->att) {
+ wrd_tl.attrSet = VAL_IDXPATH;
+ wrd_tl.attrUse = use;
+ if (p->flagShowRecords)
+ {
+ int i;
+ printf("%*sXPath index", (level + 1) * 4, "");
+ printf (" XData:\"");
+ for (i = 0; i<wrd_tl.length && i < 40; i++)
+ fputc (wrd_tl.string[i], stdout);
+ fputc ('"', stdout);
+ if (wrd_tl.length > 40)
+ printf (" ...");
+ fputc ('\n', stdout);
+ }
+ else
+ (*p->tokenAdd)(&wrd_tl);
+ xpdone = 1;
+ } else {
+ /* this is just the old fashioned attribute based index */
+ wrd_tl.attrSet = (int) (tl->att->parent->reference);
+ wrd_tl.attrUse = tl->att->locals->local;
+ if (p->flagShowRecords)
+ {
+ int i;
+ printf("%*sIdx: [%s]", (level + 1) * 4, "",
+ tl->structure);
+ printf("%s:%s [%d] %s",
+ tl->att->parent->name,
+ tl->att->name, tl->att->value,
+ tl->source);
+ printf (" XData:\"");
+ for (i = 0; i<wrd_tl.length && i < 40; i++)
+ fputc (wrd_tl.string[i], stdout);
+ fputc ('"', stdout);
+ if (wrd_tl.length > 40)
+ printf (" ...");
+ fputc ('\n', stdout);
+ }
+ else
+ (*p->tokenAdd)(&wrd_tl);
+ }
+ }
+ }
+ /* xpath indexing is done, if there was no termlist given,
+ or no ! in the termlist, and default indexing is enabled... */
+ if (!p->flagShowRecords && !xpdone && !termlist_only)
+ {
+ wrd->attrSet = VAL_IDXPATH;
+ wrd->attrUse = use;
+ wrd->reg_type = 'w';
+ (*p->tokenAdd)(wrd);
+ }
+ break;
+ case DATA1N_tag:
+ flen = 0;
+ for (nn = n; nn; nn = nn->parent)
+ {
+ if (nn->which == DATA1N_tag)
+ {
+ size_t tlen = strlen(nn->u.tag.tag);
+ if (tlen + flen > (sizeof(tag_path_full)-2))
+ return;
+ memcpy (tag_path_full + flen, nn->u.tag.tag, tlen);
+ flen += tlen;
+ tag_path_full[flen++] = '/';
+ }
+ else if (nn->which == DATA1N_root)
+ break;
+ }
+
+
+ wrd->reg_type = '0';
+ wrd->string = tag_path_full;
+ wrd->length = flen;
+ wrd->attrSet = VAL_IDXPATH;
+ wrd->attrUse = use;
+ if (p->flagShowRecords)
+ {
+ printf("%*s tag=", (level + 1) * 4, "");
+ for (i = 0; i<wrd->length && i < 40; i++)
+ fputc (wrd->string[i], stdout);
+ if (i == 40)
+ printf (" ..");
+ printf("\n");
+ }
+ else
+ {
+ data1_xattr *xp;
+ data1_termlist *tl;
+ int do_xpindex;
+
+ tag_path_full[flen] = 0;
+
+ /* Add tag start/end xpath index, only when there is a ! in the apropriate xelm
+ directive, or default xpath indexing is enabled */
+ if (!(do_xpindex = 1 - termlist_only)) {
+ if ((tl = xpath_termlist_by_tagpath(tag_path_full, n))) {
+ for (; tl; tl = tl->next) { if (!tl->att) {do_xpindex = 1;} }
+ }
+ }
+ if (do_xpindex) {
+ (*p->tokenAdd)(wrd); /* index element pag (AKA tag path) */
+ }
+
+ if (use == 1) /* only for the starting tag... */
+ {
+#define MAX_ATTR_COUNT 50
+ data1_termlist *tll[MAX_ATTR_COUNT];
+
+ int i = 0;
+
+ /* get termlists for attributes, and find out, if we have to do xpath indexing */
+ for (xp = n->u.tag.attributes; xp; xp = xp->next) {
+ i++;
+ }
+
+ i = 0;
+ for (xp = n->u.tag.attributes; xp; xp = xp->next) {
+ char comb[512];
+ int do_xpindex = 1 - termlist_only;
+ data1_termlist *tl;
+ char attr_tag_path_full[1024];
+ int int_len = flen;
+
+ /* this could be cached as well */
+ sprintf (attr_tag_path_full, "@%s/%.*s",
+ xp->name, int_len, tag_path_full);
+
+ tll[i] = xpath_termlist_by_tagpath(attr_tag_path_full,n);
+
+ /* if there is a ! in the xelm termlist, or default indexing is on,
+ proceed with xpath idx */
+ if ((tl = tll[i]))
+ {
+ for (; tl; tl = tl->next)
+ {
+ if (!tl->att)
+ do_xpindex = 1;
+ }
+ }
+
+ if (do_xpindex) {
+
+ /* attribute (no value) */
+ wrd->reg_type = '0';
+ wrd->attrUse = 3;
+ wrd->string = xp->name;
+ wrd->length = strlen(xp->name);
+
+ wrd->seqno--;
+ (*p->tokenAdd)(wrd);
+
+ if (xp->value &&
+ strlen(xp->name) + strlen(xp->value) < sizeof(comb)-2) {
+
+ /* attribute value exact */
+ strcpy (comb, xp->name);
+ strcat (comb, "=");
+ strcat (comb, xp->value);
+
+ wrd->attrUse = 3;
+ wrd->reg_type = '0';
+ wrd->string = comb;
+ wrd->length = strlen(comb);
+ wrd->seqno--;
+
+ (*p->tokenAdd)(wrd);
+ }
+ }
+ i++;
+ }
+
+ i = 0;
+ for (xp = n->u.tag.attributes; xp; xp = xp->next) {
+ data1_termlist *tl;
+ char attr_tag_path_full[1024];
+ int int_len = flen;
+ int xpdone = 0;
+
+ sprintf (attr_tag_path_full, "@%s/%.*s",
+ xp->name, int_len, tag_path_full);
+
+ if ((tl = tll[i]))
+ {
+ /* If there is a termlist given (=xelm directive) */
+ for (; tl; tl = tl->next)
+ {
+ if (!tl->att) {
+ /* add xpath index for the attribute */
+ index_xpath_attr (attr_tag_path_full, xp->name,
+ xp->value, tl->structure,
+ p, wrd);
+ xpdone = 1;
+ } else {
+ /* add attribute based index for the attribute */
+ if (xp->value) {
+ wrd->attrSet = (int)
+ (tl->att->parent->reference);
+ wrd->attrUse = tl->att->locals->local;
+ wrd->reg_type = *tl->structure;
+ wrd->string = xp->value;
+ wrd->length = strlen(xp->value);
+ (*p->tokenAdd)(wrd);
+ }
+ }
+ }
+ }
+ /* if there was no termlist for the given path,
+ or the termlist didn't have a ! element, index
+ the attribute as "w" */
+ if ((!xpdone) && (!termlist_only))
+ {
+ index_xpath_attr (attr_tag_path_full, xp->name,
+ xp->value, "w", p, wrd);
+ }
+ i++;
+ }
+ }
+ }
+ }
+}
+
+static void index_termlist (data1_node *par, data1_node *n,
+ struct recExtractCtrl *p, int level, RecWord *wrd)
+{
+ data1_termlist *tlist = 0;
+ data1_datatype dtype = DATA1K_string;
+
+ /*
+ * cycle up towards the root until we find a tag with an att..
+ * this has the effect of indexing locally defined tags with
+ * the attribute of their ancestor in the record.
+ */
+
+ while (!par->u.tag.element)
+ if (!par->parent || !(par=get_parent_tag(p->dh, par->parent)))
+ break;
+ if (!par || !(tlist = par->u.tag.element->termlists))
+ return;
+ if (par->u.tag.element->tag)
+ dtype = par->u.tag.element->tag->kind;
+
+ for (; tlist; tlist = tlist->next)
+ {
+ /* consider source */
+ wrd->string = 0;
+ assert(tlist->source);
+ sp_parse(n, wrd, tlist->source);
+
+ if (wrd->string)
+ {
+ if (p->flagShowRecords)
+ {
+ int i;
+ printf("%*sIdx: [%s]", (level + 1) * 4, "",
+ tlist->structure);
+ printf("%s:%s [%d] %s",
+ tlist->att->parent->name,
+ tlist->att->name, tlist->att->value,
+ tlist->source);
+ printf (" XData:\"");
+ for (i = 0; i<wrd->length && i < 40; i++)
+ fputc (wrd->string[i], stdout);
+ fputc ('"', stdout);
+ if (wrd->length > 40)
+ printf (" ...");
+ fputc ('\n', stdout);
+ }
+ else
+ {
+ wrd->reg_type = *tlist->structure;
+ wrd->attrSet = (int) (tlist->att->parent->reference);
+ wrd->attrUse = tlist->att->locals->local;
+ (*p->tokenAdd)(wrd);
+ }
+ }
+ }
+}
+
+static int dumpkeys(data1_node *n, struct recExtractCtrl *p, int level,
+ RecWord *wrd)