X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=recctrl%2Frecgrs.c;h=55bdaa5df01f6f258cd1207bd430583ef5215ea4;hp=9a952c820720ea34561255073c26f5c40dec87cb;hb=ecb3935e78cd9bcfdebafdee0834cfb1060d7b5e;hpb=7ec81d893efe0407cda28aeca3cb232266dd713f

diff --git a/recctrl/recgrs.c b/recctrl/recgrs.c
index 9a952c8..55bdaa5 100644
--- a/recctrl/recgrs.c
+++ b/recctrl/recgrs.c
@@ -1,6 +1,6 @@
-/* $Id: recgrs.c,v 1.77 2003-04-24 22:29:52 pop Exp $
-   Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002,2003
-   Index Data Aps
+/* $Id: recgrs.c,v 1.109 2006-05-10 08:13:28 adam Exp $
+   Copyright (C) 1995-2005
+   Index Data ApS
 
 This file is part of the Zebra server.
 
@@ -23,187 +23,371 @@ Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
 #include <stdio.h>
 #include <assert.h>
 #include <sys/types.h>
-#ifndef WIN32
-#include <unistd.h>
-#endif
+#include <ctype.h>
 
 #include <yaz/log.h>
 #include <yaz/oid.h>
 
-#include <recctrl.h>
-#include "grsread.h"
+#include <d1_absyn.h>
+#include <idzebra/recgrs.h>
 
 #define GRS_MAX_WORD 512
 
-struct grs_handler {
-    RecTypeGrs type;
-    void *clientData;
-    int initFlag;
-    struct grs_handler *next;
-};
-
-struct grs_handlers {
-    struct grs_handler *handlers;
+struct source_parser {
+    int len;
+    const char *tok;
+    const char *src;
+    int lookahead;
+    NMEM nmem;
 };
 
-static int read_grs_type (struct grs_handlers *h,
-			  struct grs_read_info *p, const char *type,
-			  data1_node **root)
+static int sp_lex(struct source_parser *sp)
 {
-    struct grs_handler *gh = h->handlers;
-    const char *cp = strchr (type, '.');
+    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 (cp == NULL || cp == type)
+static int sp_range(struct source_parser *sp, data1_node *n, RecWord *wrd)
+{
+    int start, len;
+    RecWord tmp_w;
+    
+    /* ( */
+    sp_lex(sp);
+    if (sp->lookahead != '(')
+	return 0;
+    sp_lex(sp); /* skip ( */
+    
+    /* 1st arg: string */
+    if (!sp_expr(sp, n, wrd))
+	return 0;
+    
+    if (sp->lookahead != ',')
+	return 0;	
+    sp_lex(sp); /* skip , */
+    
+    /* 2nd arg: start */
+    if (!sp_expr(sp, n, &tmp_w))
+	return 0;
+    start = atoi_n(tmp_w.term_buf, tmp_w.term_len);
+    
+    if (sp->lookahead == ',')
     {
-        cp = strlen(type) + type;
-        *p->type = 0;
+	sp_lex(sp); /* skip , */
+	
+	/* 3rd arg: length */
+	if (!sp_expr(sp, n, &tmp_w))
+	    return 0;
+	len = atoi_n(tmp_w.term_buf, tmp_w.term_len);
     }
     else
-        strcpy (p->type, cp+1);
-    for (gh = h->handlers; gh; gh = gh->next)
+	len = wrd->term_len;
+    
+    /* ) */
+    if (sp->lookahead != ')')
+	return 0;	
+    sp_lex(sp);
+    
+    if (wrd->term_buf && wrd->term_len)
+    {
+	wrd->term_buf += start;
+	wrd->term_len -= start;
+	if (wrd->term_len > len)
+	    wrd->term_len = len;
+    }
+    return 1;
+}
+
+static int sp_first(struct source_parser *sp, data1_node *n, RecWord *wrd)
+{
+    char num_str[20];
+    int min_pos = -1;
+    sp_lex(sp);
+    if (sp->lookahead != '(')
+	return 0;
+    sp_lex(sp); /* skip ( */
+    if (!sp_expr(sp, n, wrd))
+	return 0;
+    while (sp->lookahead == ',')
     {
-        if (!memcmp (type, gh->type->type, cp-type))
+	RecWord search_w;
+	int i;
+	sp_lex(sp); /* skip , */
+	
+	if (!sp_expr(sp, n, &search_w))
+	    return 0;
+	for (i = 0; i<wrd->term_len; i++)
 	{
-	    if (!gh->initFlag)
+	    int j;
+	    for (j = 0; j<search_w.term_len && i+j < wrd->term_len; j++)
+		if (wrd->term_buf[i+j] != search_w.term_buf[j])
+		    break;
+	    if (j == search_w.term_len) /* match ? */
 	    {
-		gh->initFlag = 1;
-		gh->clientData = (*gh->type->init)();
+		if (min_pos == -1 || i < min_pos)
+		    min_pos = i;
+		break;
 	    }
-	    p->clientData = gh->clientData;
-            *root = (gh->type->read)(p);
-	    gh->clientData = p->clientData;
-	    return 0;
 	}
     }
+    if (sp->lookahead != ')')
+	return 0;
+    sp_lex(sp);
+    if (min_pos == -1)
+	min_pos = 0;  /* the default if not found */
+    sprintf(num_str, "%d", min_pos);
+    wrd->term_buf = nmem_strdup(sp->nmem, num_str);
+    wrd->term_len = strlen(wrd->term_buf);
     return 1;
 }
 
-static void grs_add_handler (struct grs_handlers *h, RecTypeGrs t)
+static int sp_expr(struct source_parser *sp, data1_node *n, RecWord *wrd)
 {
-    struct grs_handler *gh = (struct grs_handler *) xmalloc (sizeof(*gh));
-    gh->next = h->handlers;
-    h->handlers = gh;
-    gh->initFlag = 0;
-    gh->clientData = 0;
-    gh->type = t;
+    if (sp->lookahead != 't')
+	return 0;
+    if (sp->len == 4 && !memcmp(sp->tok, "data", sp->len))
+    {
+	if (n->which == DATA1N_data)
+	{
+	    wrd->term_buf = n->u.data.data;
+	    wrd->term_len = n->u.data.len;
+	}
+	sp_lex(sp);
+    }
+    else if (sp->len == 3 && !memcmp(sp->tok, "tag", sp->len))
+    {
+	if (n->which == DATA1N_tag)
+	{		
+	    wrd->term_buf = n->u.tag.tag;
+	    wrd->term_len = strlen(n->u.tag.tag);
+	}
+	sp_lex(sp);
+    }
+    else if (sp->len == 4 && !memcmp(sp->tok, "attr", sp->len))
+    {
+	RecWord tmp_w;
+	sp_lex(sp);
+	if (sp->lookahead != '(')
+	    return 0;
+	sp_lex(sp);
+
+	if (!sp_expr(sp, n, &tmp_w))
+	    return 0;
+	
+	wrd->term_buf = "";
+	wrd->term_len = 0;
+	if (n->which == DATA1N_tag)
+	{
+	    data1_xattr *p = n->u.tag.attributes;
+	    while (p && strlen(p->name) != tmp_w.term_len && 
+		   memcmp (p->name, tmp_w.term_buf, tmp_w.term_len))
+		p = p->next;
+	    if (p)
+	    {
+		wrd->term_buf = p->value;
+		wrd->term_len = strlen(p->value);
+	    }
+	}
+	if (sp->lookahead != ')')
+	    return 0;
+	sp_lex(sp);
+    }
+    else if (sp->len == 5 && !memcmp(sp->tok, "first", sp->len))
+    {
+	return sp_first(sp, n, wrd);
+    }
+    else if (sp->len == 5 && !memcmp(sp->tok, "range", sp->len))
+    {
+	return sp_range(sp, n, wrd);
+    }
+    else if (sp->len > 0 && isdigit(*(unsigned char *)sp->tok))
+    {
+	char *b;
+	wrd->term_len = sp->len;
+	b = nmem_malloc(sp->nmem, sp->len);
+	memcpy(b, sp->tok, sp->len);
+	wrd->term_buf = b;
+	sp_lex(sp);
+    }
+    else if (sp->len > 2 && sp->tok[0] == '\'' && sp->tok[sp->len-1] == '\'')
+    {
+	char *b;
+	wrd->term_len = sp->len - 2;
+	b = nmem_malloc(sp->nmem, wrd->term_len);
+	memcpy(b, sp->tok+1, wrd->term_len);
+	wrd->term_buf = b;
+	sp_lex(sp);
+    }
+    else 
+    {
+	wrd->term_buf = "";
+	wrd->term_len = 0;
+	sp_lex(sp);
+    }
+    return 1;
 }
 
-static void *grs_init(RecType recType)
+static struct source_parser *source_parser_create()
 {
-    struct grs_handlers *h = (struct grs_handlers *) xmalloc (sizeof(*h));
-    h->handlers = 0;
+    struct source_parser *sp = xmalloc(sizeof(*sp));
 
-    grs_add_handler (h, recTypeGrs_sgml);
-    grs_add_handler (h, recTypeGrs_regx);
-#if HAVE_TCL_H
-    grs_add_handler (h, recTypeGrs_tcl);
-#endif
-    grs_add_handler (h, recTypeGrs_marc);
-#if HAVE_EXPAT_H
-    grs_add_handler (h, recTypeGrs_xml);
-#endif
-#if HAVE_PERL
-    grs_add_handler (h, recTypeGrs_perl);
-#endif
-    return h;
+    sp->nmem = nmem_create();
+    return sp;
 }
 
-static void grs_destroy(void *clientData)
+static void source_parser_destroy(struct source_parser *sp)
 {
-    struct grs_handlers *h = (struct grs_handlers *) clientData;
-    struct grs_handler *gh = h->handlers, *gh_next;
-    while (gh)
-    {
-	gh_next = gh->next;
-	if (gh->initFlag)
-	    (*gh->type->destroy)(gh->clientData);
-	xfree (gh);
-	gh = gh_next;
+    if (!sp)
+	return;
+    nmem_destroy(sp->nmem);
+    xfree(sp);
+}
+    
+static int sp_parse(struct source_parser *sp, 
+		    data1_node *n, RecWord *wrd, const char *src)
+{
+    sp->len = 0;
+    sp->tok = 0;
+    sp->src = src;
+    sp->lookahead = 0;
+    nmem_reset(sp->nmem);
+
+    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(YLOG_WARN, 
+                         "  Only attributes (@) are supported in xelm xpath predicates");
+                    yaz_log(YLOG_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(YLOG_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(YLOG_WARN, 
+                                     "Only '=' relation is supported (%s)",p->u.relation.op);
+                                yaz_log(YLOG_WARN, "predicate %s ignored", p->u.relation.name);
+                                res = 1; break;
+                            } else {
+                                yaz_log(YLOG_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(YLOG_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(YLOG_WARN, "Unknown boolean relation %s, ignored",p->u.boolean.op);
+                return 1;
+            }
+        }
     }
-    xfree (h);
+    return 0;
 }
 
-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 != '@') {
-	  logf(LOG_WARN, 
-	       "  Only attributes (@) are supported in xelm xpath predicates");
-	  logf(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) {
-	  logf(LOG_DEBUG,"  - attribute %s <-> %s", attname, attr->name );
-
-	  if (!strcmp(attr->name, attname)) {
-	    if (p->u.relation.op[0]) {
-	      if (*p->u.relation.op != '=') {
-		logf(LOG_WARN, 
-		     "Only '=' relation is supported (%s)",p->u.relation.op);
-		logf(LOG_WARN, "predicate %s ignored", p->u.relation.name);
-		res = 1; break;
-	      } else {
-		logf(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;
+
+static int dfa_match_first(struct DFA_state **dfaar, const char *text)
+{
+    struct DFA_state *s = dfaar[0]; /* start state */
+    struct DFA_tran *t;
+    int i;
+    const char *p = text;
+    unsigned char c;
+    
+    for (c = *p++, t = s->trans, i = s->tran_no; --i >= 0; t++)
+    {
+	if (c >= t->ch[0] && c <= t->ch[1])
+	{
+	    while (i >= 0)
+	    {
+		/* move to next state and return if we get a match */
+		s = dfaar[t->to];
+		if (s->rule_no)
+		    return 1;
+		/* next char */
+		if (!c)
+		    return 0;
+		c = *p++;
+		for (t = s->trans, i = s->tran_no; --i >= 0; t++)
+		    if (c >= t->ch[0] && c <= t->ch[1])
+			break;
 	    }
-	  }
 	}
-	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 {
-	logf(LOG_WARN, "Unknown boolean relation %s, ignored",p->u.boolean.op);
-	return (1);
-      }
     }
-  }
-  return 0;
+    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.
+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
+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???)
+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
+pop, 2003-01-17
 
- */
+*/
 
 data1_termlist *xpath_termlist_by_tagpath(char *tagpath, data1_node *n)
 {
@@ -212,81 +396,52 @@ data1_termlist *xpath_termlist_by_tagpath(char *tagpath, data1_node *n)
     data1_node *nn;
 #ifdef ENHANCED_XELM 
     struct xpath_location_step *xp;
-
 #endif
-    char *pexpr = xmalloc(strlen(tagpath)+2);
+    char *pexpr = xmalloc(strlen(tagpath)+5);
     int ok = 0;
     
-    sprintf (pexpr, "%s\n", tagpath);
-    logf(LOG_DEBUG,"Checking tagpath %s",tagpath);
-    while (xpe) 
+    sprintf (pexpr, "/%s\n", tagpath);
+    for (; xpe; xpe = xpe->next)
     {
-        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;
-
-	logf(LOG_DEBUG," - xpath %s",xpe->xpath_expr);
-
-        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);
-	}
-        pexpr--;
+	int i;
+	ok = dfa_match_first(xpe->dfa->states, 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--) {
-	    
-	    logf(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)) {
-	      logf(LOG_DEBUG,"  Predicates didn't match");
-	      ok = 0;
-	      break;
-	    }
-
-	    if (nn->which == DATA1N_tag) {
-	      nn = nn->parent;
-	    }
-	  }
+            /* we have to check the perdicates up to the root node */
+            xp = xpe->xpath;
+            
+            /* find the first tag up in the node structure */
+            for (nn = n; 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(YLOG_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(YLOG_DEBUG, "  Predicates didn't match");
+                    ok = 0;
+                    break;
+                }
+                
+                if (nn->which == DATA1N_tag)
+                    nn = nn->parent;
+            }
 #endif
-	  if (ok) {
-	    break;
-	  }
+            if (ok)
+                break;
 	}
-        xpe = xpe->next;
     } 
-
+    
     xfree(pexpr);
     
     if (ok) {
-      logf(LOG_DEBUG,"Got it");
+	yaz_log(YLOG_DEBUG, "Got it");
         return xpe->termlists;
     } else {
         return NULL;
@@ -314,162 +469,215 @@ data1_termlist *xpath_termlist_by_tagpath(char *tagpath, data1_node *n)
 /* 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) {
-
-  /*
-  char comb[512];
-  
-  wrd->attrSet = VAL_IDXPATH;
-  wrd->attrUse = 3;
-  wrd->reg_type = '0';
-  wrd->string = name;
-  wrd->length = strlen(name);
-                    
-  wrd->seqno--;
-  (*p->tokenAdd)(wrd);
-
-  if (value && 
-      strlen(name) + strlen(value) < sizeof(comb)-2) {
-    strcpy (comb, name);
-    strcat (comb, "=");
-    strcat (comb, value);
-    
-    wrd->attrUse = 3;
-    wrd->reg_type = '0';
-    wrd->string = comb;
-    wrd->length = strlen(comb);
-    wrd->seqno--;
-    
+			      RecWord *wrd)
+{
+#if NATTR
+    wrd->index_name = ZEBRA_XPATH_ELM_BEGIN;
+#else
+    wrd->attrSet = VAL_IDXPATH;
+    wrd->attrUse = 1;
+#endif
+    wrd->index_type = '0';
+    wrd->term_buf = tag_path;
+    wrd->term_len = strlen(tag_path);
     (*p->tokenAdd)(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);
+    
+    if (value) {
+#if NATTR
+	wrd->index_name = ZEBRA_XPATH_ATTR;
+#else
+        wrd->attrUse = 1015;
+#endif
+        wrd->index_type = 'w';
+        wrd->term_buf = value;
+        wrd->term_len = strlen(value);
+        (*p->tokenAdd)(wrd);
+    }
+#if NATTR
+    wrd->index_name = ZEBRA_XPATH_ELM_END;
+#else
+    wrd->attrUse = 2;
+#endif
+    wrd->index_type = '0';
+    wrd->term_buf = tag_path;
+    wrd->term_len = strlen(tag_path);
     (*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)
+static void mk_tag_path_full(char *tag_path_full, size_t max, data1_node *n)
 {
-    int i;
-    char tag_path_full[1024];
     size_t flen = 0;
     data1_node *nn;
+
+    /* 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 > (max - 2))
+		break;
+	    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;
+}
+	
+
+static void index_xpath(struct source_parser *sp, data1_node *n,
+			struct recExtractCtrl *p,
+			int level, RecWord *wrd,
+#if NATTR
+			char *xpath_index,
+			int xpath_is_start
+#else
+			int use
+#endif
+    )
+{
+    int i;
+    char tag_path_full[1024];
     int termlist_only = 1;
+    data1_termlist *tl;
+    int xpdone = 0;
+#if NATTR
+#else
+    int xpath_is_start = 0;
+    if (use == 1)
+	xpath_is_start = 1;
+#endif
 
+#if NATTR
+    yaz_log(YLOG_DEBUG, "index_xpath level=%d xpath_index=%s",
+	    level, xpath_index);
+#else
+    yaz_log(YLOG_DEBUG, "index_xpath level=%d use=%d", level, use);
+#endif
     if ((!n->root->u.root.absyn) ||
 	(n->root->u.root.absyn->enable_xpath_indexing)) {
-      termlist_only = 0;
+	termlist_only = 0;
     }
 
     switch (n->which)
     {
     case DATA1N_data:
-        wrd->string = n->u.data.data;
-        wrd->length = n->u.data.len;
-        if (p->flagShowRecords)
-        {
-            printf("%*s data=", (level + 1) * 4, "");
-            for (i = 0; i<wrd->length && i < 8; i++)
-                fputc (wrd->string[i], stdout);
-            printf("\n");
-        }  
-        else  {
-            data1_termlist *tl;
-            int xpdone = 0;
-            flen = 0;
-            
-	    //	    logf (LOG_LOG,"Index value, %s",wrd->string);
-
-            /* 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;
-            }
+        wrd->term_buf = n->u.data.data;
+        wrd->term_len = n->u.data.len;
+        xpdone = 0;
 
-            tag_path_full[flen] = 0;
-            
-            /* If we have a matching termlist... */
-            if ((tl = xpath_termlist_by_tagpath(tag_path_full, n))) {
-                for (; tl; tl = tl->next) {
-                    wrd->reg_type = *tl->structure;
-                    /* this is the ! case, so structure is for the xpath index */
-                    if (!tl->att) {
-                        wrd->attrSet = VAL_IDXPATH;
-                        wrd->attrUse = use;
-                        (*p->tokenAdd)(wrd);
-                        xpdone = 1;
-                        /* this is just the old fashioned attribute based index */
-                    } else {
-                        wrd->attrSet = (int) (tl->att->parent->reference);
-                        wrd->attrUse = tl->att->locals->local;
-                        (*p->tokenAdd)(wrd);
-                    }
-                }
-            }
-            /* xpath indexing is done, if there was no termlist given, 
-               or no ! in the termlist, and default indexing is enabled... */
-            if ((!xpdone) && (!termlist_only)) {
-                wrd->attrSet = VAL_IDXPATH;
-                wrd->attrUse = use;
-                wrd->reg_type = 'w';
-                (*p->tokenAdd)(wrd);
-            }
+	mk_tag_path_full(tag_path_full, sizeof(tag_path_full), n);
+	
+	/* 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->index_type = *tl->structure;
+		memcpy (&wrd_tl, wrd, sizeof(*wrd));
+		if (tl->source)
+		    sp_parse(sp, n, &wrd_tl, tl->source);
+#if NATTR
+		if (!tl->index_name)
+#else
+		if (!tl->att)
+#endif
+		{
+		    /* this is the ! case, so structure is for the xpath index */
+#if NATTR
+		    wrd_tl.index_name = xpath_index;
+#else
+		    wrd_tl.attrSet = VAL_IDXPATH;
+		    wrd_tl.attrUse = use;
+#endif
+		    if (p->flagShowRecords)
+		    {
+			int i;
+		        printf("%*sXPath index", (level + 1) * 4, "");
+			printf (" XData:\"");
+			for (i = 0; i<wrd_tl.term_len && i < 40; i++)
+			    fputc (wrd_tl.term_buf[i], stdout);
+			fputc ('"', stdout);
+			if (wrd_tl.term_len > 40)
+			    printf (" ...");
+			fputc ('\n', stdout);
+		    }
+		    else
+			(*p->tokenAdd)(&wrd_tl);
+		    xpdone = 1;
+		} else {
+		    /* this is just the old fashioned attribute based index */
+#if NATTR
+		    wrd_tl.index_name = tl->index_name;
+#else
+		    wrd_tl.attrSet = (int) (tl->att->parent->reference);
+		    wrd_tl.attrUse = tl->att->locals->local;
+#endif
+		    if (p->flagShowRecords)
+		    {
+			int i;
+			printf("%*sIdx: [%s]", (level + 1) * 4, "",
+			       tl->structure);
+#if NATTR
+			printf("%s %s", tl->index_name, tl->source);
+#else
+			printf("%s:%s [%d] %s",
+			       tl->att->parent->name,
+			       tl->att->name, tl->att->value,
+			       tl->source);
+#endif
+			printf (" XData:\"");
+			for (i = 0; i<wrd_tl.term_len && i < 40; i++)
+			    fputc (wrd_tl.term_buf[i], stdout);
+			fputc ('"', stdout);
+			if (wrd_tl.term_len > 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)
+	{
+#if NATTR
+	    wrd->index_name = xpath_index;
+#else
+	    wrd->attrSet = VAL_IDXPATH;
+	    wrd->attrUse = use;
+#endif
+	    wrd->index_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;
-        }
-
+	mk_tag_path_full(tag_path_full, sizeof(tag_path_full), n);
 
-        wrd->reg_type = '0';
-        wrd->string = tag_path_full;
-        wrd->length = flen;
+        wrd->index_type = '0';
+        wrd->term_buf = tag_path_full;
+        wrd->term_len = strlen(tag_path_full);
+#if NATTR
+	wrd->index_name = xpath_index;
+#else
         wrd->attrSet = VAL_IDXPATH;
         wrd->attrUse = use;
+#endif
         if (p->flagShowRecords)
         {
             printf("%*s tag=", (level + 1) * 4, "");
-            for (i = 0; i<wrd->length && i < 40; i++)
-                fputc (wrd->string[i], stdout);
+            for (i = 0; i<wrd->term_len && i < 40; i++)
+                fputc (wrd->term_buf[i], stdout);
             if (i == 40)
                 printf (" ..");
             printf("\n");
@@ -479,131 +687,171 @@ static void index_xpath (data1_node *n, struct recExtractCtrl *p,
             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;} }
-	      }
+            
+            /* 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 NATTR
+			if (!tl->index_name)
+#else
+			if (!tl->att) 
+#endif
+			    do_xpindex = 1;
+		    }
+                }
 	    }
 	    if (do_xpindex) {
-	      (*p->tokenAdd)(wrd);   /* index element pag (AKA tag path) */
+                (*p->tokenAdd)(wrd);   /* index element pag (AKA tag path) */
 	    }
-
-            if (use == 1) /* only for the starting tag... */
+            
+            if (xpath_is_start == 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++;
-	      }
-	    }
+                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]; 
+                    
+                    /* this could be cached as well */
+                    sprintf (attr_tag_path_full, "@%s/%s",
+                             xp->name, 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 NATTR
+			    if (!tl->index_name)
+				do_xpindex = 1;
+#else
+                            if (!tl->att)
+                                do_xpindex = 1;
+#endif
+                        }
+                    }
+                    
+                    if (do_xpindex) {
+                        
+                        /* attribute  (no value) */
+                        wrd->index_type = '0';
+#if NATTR
+			wrd->index_name = ZEBRA_XPATH_ATTR;
+#else
+                        wrd->attrUse = 3;
+#endif
+                        wrd->term_buf = xp->name;
+                        wrd->term_len = 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);
+
+#if NATTR
+                            wrd->index_name = ZEBRA_XPATH_ATTR;
+#else
+                            wrd->attrUse = 3;
+#endif
+                            wrd->index_type = '0';
+                            wrd->term_buf = comb;
+                            wrd->term_len = 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 xpdone = 0;
+                    
+                    sprintf (attr_tag_path_full, "@%s/%s",
+                             xp->name, tag_path_full);
+                    
+                    if ((tl = tll[i]))
+                    {
+                        /* If there is a termlist given (=xelm directive) */
+                        for (; tl; tl = tl->next)
+                        {
+#if NATTR
+			    if (!tl->index_name)
+#else
+                            if (!tl->att) 
+#endif
+			    {
+                                /* add xpath index for the attribute */
+                                index_xpath_attr (attr_tag_path_full, xp->name,
+                                                  xp->value, tl->structure,
+                                                  p, wrd);
+                                xpdone = 1;
+                            } else {
+                                /* index attribute value (only path/@attr) */
+                                if (xp->value) 
+				{
+#if NATTR
+				    wrd->index_name = tl->index_name;
+#else
+                                    wrd->attrSet = (int) 
+                                        (tl->att->parent->reference);
+                                    wrd->attrUse = tl->att->locals->local;
+#endif
+                                    wrd->index_type = *tl->structure;
+                                    wrd->term_buf = xp->value;
+                                    wrd->term_len = 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,
+static void index_termlist (struct source_parser *sp, data1_node *par,
+			    data1_node *n,
                             struct recExtractCtrl *p, int level, RecWord *wrd)
 {
     data1_termlist *tlist = 0;
@@ -622,68 +870,55 @@ static void index_termlist (data1_node *par, data1_node *n,
         return;
     if (par->u.tag.element->tag)
         dtype = par->u.tag.element->tag->kind;
-    
+
     for (; tlist; tlist = tlist->next)
     {
-
-	char xattr[512];
 	/* consider source */
-	wrd->string = 0;
+	wrd->term_buf = 0;
+	assert(tlist->source);
+	sp_parse(sp, n, wrd, tlist->source);
 
-	if (!strcmp (tlist->source, "data") && n->which == DATA1N_data)
-	{
-	    wrd->string = n->u.data.data;
-	    wrd->length = n->u.data.len;
-	}
-	else if (!strcmp (tlist->source, "tag") && n->which == DATA1N_tag)
-        {
-	    wrd->string = n->u.tag.tag;
-	    wrd->length = strlen(n->u.tag.tag);
-	}
-	else if (sscanf (tlist->source, "attr(%511[^)])", xattr) == 1 &&
-	    n->which == DATA1N_tag)
-	{
-	    data1_xattr *p = n->u.tag.attributes;
-	    while (p && strcmp (p->name, xattr))
-		p = p->next;
-	    if (p)
-	    {
-		wrd->string = p->value;
-		wrd->length = strlen(p->value);
-	    }
-	}
-	if (wrd->string)
+	if (wrd->term_buf && wrd->term_len)
 	{
 	    if (p->flagShowRecords)
 	    {
 		int i;
 		printf("%*sIdx: [%s]", (level + 1) * 4, "",
 		       tlist->structure);
+#if NATTR
+		printf("%s %s", tlist->index_name, tlist->source);
+#else
 		printf("%s:%s [%d] %s",
 		       tlist->att->parent->name,
 		       tlist->att->name, tlist->att->value,
 		       tlist->source);
-		printf (" data=\"");
-		for (i = 0; i<wrd->length && i < 40; i++)
-		    fputc (wrd->string[i], stdout);
+#endif
+		printf (" XData:\"");
+		for (i = 0; i<wrd->term_len && i < 40; i++)
+		    fputc (wrd->term_buf[i], stdout);
 		fputc ('"', stdout);
-		if (wrd->length > 40)
+		if (wrd->term_len > 40)
 		    printf (" ...");
 		fputc ('\n', stdout);
 	    }
 	    else
 	    {
-		wrd->reg_type = *tlist->structure;
+		wrd->index_type = *tlist->structure;
+#if NATTR
+		wrd->index_name = tlist->index_name;
+#else
 		wrd->attrSet = (int) (tlist->att->parent->reference);
 		wrd->attrUse = tlist->att->locals->local;
+#endif
 		(*p->tokenAdd)(wrd);
 	    }
 	}
     }
 }
 
-static int dumpkeys(data1_node *n, struct recExtractCtrl *p, int level,
-                    RecWord *wrd)
+static int dumpkeys_r(struct source_parser *sp,
+		      data1_node *n, struct recExtractCtrl *p, int level,
+		      RecWord *wrd)
 {
     for (; n; n = n->next)
     {
@@ -726,14 +961,20 @@ static int dumpkeys(data1_node *n, struct recExtractCtrl *p, int level,
 
 	if (n->which == DATA1N_tag)
 	{
-            index_termlist (n, n, p, level, wrd);
+            index_termlist(sp, n, n, p, level, wrd);
             /* index start tag */
-            assert (n->root->u.root.absyn);
-      	    index_xpath (n, p, level, wrd, 1);
+#if NATTR
+	    if (n->root->u.root.absyn)
+      	        index_xpath(sp, n, p, level, wrd, ZEBRA_XPATH_ELM_BEGIN, 
+			    1 /* is start */);
+#else
+	    if (n->root->u.root.absyn)
+      	        index_xpath(sp, n, p, level, wrd, 1);
+#endif
  	}
 
 	if (n->child)
-	    if (dumpkeys(n->child, p, level + 1, wrd) < 0)
+	    if (dumpkeys_r(sp, n->child, p, level + 1, wrd) < 0)
 		return -1;
 
 
@@ -746,8 +987,8 @@ static int dumpkeys(data1_node *n, struct recExtractCtrl *p, int level,
 		printf("%*s", level * 4, "");
 		printf("Data: ");
 		if (n->u.data.len > 256)
-		    printf("'%.240s ... %.6s'\n", n->u.data.data,
-			   n->u.data.data + n->u.data.len-6);
+		    printf("'%.170s ... %.70s'\n", n->u.data.data,
+			   n->u.data.data + n->u.data.len-70);
 		else if (n->u.data.len > 0)
 		    printf("'%.*s'\n", n->u.data.len, n->u.data.data);
 		else
@@ -755,15 +996,25 @@ static int dumpkeys(data1_node *n, struct recExtractCtrl *p, int level,
 	    }
 
 	    if (par)
-		index_termlist (par, n, p, level, wrd);
+		index_termlist(sp, par, n, p, level, wrd);
 
-	    index_xpath (n, p, level, wrd, 1016);
+#if NATTR
+	    index_xpath(sp, n, p, level, wrd, ZEBRA_XPATH_CDATA, 
+			0 /* is start */);
+#else
+	    index_xpath(sp, n, p, level, wrd, 1016);
+#endif
  	}
 
 	if (n->which == DATA1N_tag)
 	{
             /* index end tag */
-	    index_xpath (n, p, level, wrd, 2);
+#if NATTR
+	    index_xpath(sp, n, p, level, wrd, ZEBRA_XPATH_ELM_END, 
+			0 /* is start */);
+#else
+	    index_xpath(sp, n, p, level, wrd, 2);
+#endif
 	}
 
 	if (p->flagShowRecords && n->which == DATA1N_root)
@@ -774,6 +1025,14 @@ static int dumpkeys(data1_node *n, struct recExtractCtrl *p, int level,
     return 0;
 }
 
+static int dumpkeys(data1_node *n, struct recExtractCtrl *p, RecWord *wrd)
+{
+    struct source_parser *sp = source_parser_create();
+    int r = dumpkeys_r(sp, n, p, 0, wrd);
+    source_parser_destroy(sp);
+    return r;
+}
+
 int grs_extract_tree(struct recExtractCtrl *p, data1_node *n)
 {
     oident oe;
@@ -791,11 +1050,12 @@ int grs_extract_tree(struct recExtractCtrl *p, data1_node *n)
     }
     (*p->init)(p, &wrd);
 
-    return dumpkeys(n, p, 0, &wrd);
+    return dumpkeys(n, p, &wrd);
 }
 
-static int grs_extract_sub(struct grs_handlers *h, struct recExtractCtrl *p,
-			   NMEM mem)
+static int grs_extract_sub(void *clientData, struct recExtractCtrl *p,
+			   NMEM mem,
+			   data1_node *(*grs_read)(struct grs_read_info *))
 {
     data1_node *n;
     struct grs_read_info gri;
@@ -811,9 +1071,9 @@ static int grs_extract_sub(struct grs_handlers *h, struct recExtractCtrl *p,
     gri.offset = p->offset;
     gri.mem = mem;
     gri.dh = p->dh;
+    gri.clientData = clientData;
 
-    if (read_grs_type (h, &gri, p->subType, &n))
-	return RECCTRL_EXTRACT_ERROR_NO_SUCH_FILTER;
+    n = (*grs_read)(&gri);
     if (!n)
         return RECCTRL_EXTRACT_EOF;
     oe.proto = PROTO_Z3950;
@@ -828,6 +1088,7 @@ static int grs_extract_sub(struct grs_handlers *h, struct recExtractCtrl *p,
         if ((oid_ent_to_oid (&oe, oidtmp)))
             (*p->schemaAdd)(p, oidtmp);
     }
+    data1_concat_text(p->dh, mem, n);
 
     /* ensure our data1 tree is UTF-8 */
     data1_iconv (p->dh, mem, n, "UTF-8", data1_get_encoding(p->dh, n));
@@ -837,7 +1098,7 @@ static int grs_extract_sub(struct grs_handlers *h, struct recExtractCtrl *p,
 #endif
 
     (*p->init)(p, &wrd);
-    if (dumpkeys(n, p, 0, &wrd) < 0)
+    if (dumpkeys(n, p, &wrd) < 0)
     {
 	data1_free_tree(p->dh, n);
 	return RECCTRL_EXTRACT_ERROR_GENERIC;
@@ -846,13 +1107,12 @@ static int grs_extract_sub(struct grs_handlers *h, struct recExtractCtrl *p,
     return RECCTRL_EXTRACT_OK;
 }
 
-static int grs_extract(void *clientData, struct recExtractCtrl *p)
+int zebra_grs_extract(void *clientData, struct recExtractCtrl *p,
+		      data1_node *(*grs_read)(struct grs_read_info *))
 {
     int ret;
     NMEM mem = nmem_create ();
-    struct grs_handlers *h = (struct grs_handlers *) clientData;
-
-    ret = grs_extract_sub(h, p, mem);
+    ret = grs_extract_sub(clientData, p, mem, grs_read);
     nmem_destroy(mem);
     return ret;
 }
@@ -860,7 +1120,8 @@ static int grs_extract(void *clientData, struct recExtractCtrl *p)
 /*
  * Return: -1: Nothing done. 0: Ok. >0: Bib-1 diagnostic.
  */
-static int process_comp(data1_handle dh, data1_node *n, Z_RecordComposition *c)
+static int process_comp(data1_handle dh, data1_node *n, Z_RecordComposition *c,
+			char **addinfo, ODR o)
 {
     data1_esetname *eset;
     Z_Espec1 *espec = 0;
@@ -874,10 +1135,11 @@ static int process_comp(data1_handle dh, data1_node *n, Z_RecordComposition *c)
 	if (!(eset = data1_getesetbyname(dh, n->u.root.absyn,
 					 c->u.simple->u.generic)))
 	{
-	    logf(LOG_LOG, "Unknown esetname '%s'", c->u.simple->u.generic);
+	    yaz_log(YLOG_LOG, "Unknown esetname '%s'", c->u.simple->u.generic);
+	    *addinfo = odr_strdup(o, c->u.simple->u.generic);
 	    return 25; /* invalid esetname */
 	}
-	logf(LOG_DEBUG, "Esetname '%s' in simple compspec",
+	yaz_log(YLOG_DEBUG, "Esetname '%s' in simple compspec",
 	     c->u.simple->u.generic);
 	espec = eset->spec;
 	break;
@@ -894,23 +1156,24 @@ static int process_comp(data1_handle dh, data1_node *n, Z_RecordComposition *c)
 			  data1_getesetbyname(dh, n->u.root.absyn,
 					      p->u.elementSetName)))
 		    {
-			logf(LOG_LOG, "Unknown esetname '%s'",
+			yaz_log(YLOG_DEBUG, "Unknown esetname '%s'",
 			     p->u.elementSetName);
+			*addinfo = odr_strdup(o, p->u.elementSetName);
 			return 25; /* invalid esetname */
 		    }
-		    logf(LOG_DEBUG, "Esetname '%s' in complex compspec",
+		    yaz_log(YLOG_DEBUG, "Esetname '%s' in complex compspec",
 			 p->u.elementSetName);
 		    espec = eset->spec;
 		    break;
 		case Z_ElementSpec_externalSpec:
 		    if (p->u.externalSpec->which == Z_External_espec1)
 		    {
-			logf(LOG_DEBUG, "Got Espec-1");
+			yaz_log(YLOG_DEBUG, "Got Espec-1");
 			espec = p->u.externalSpec-> u.espec1;
 		    }
 		    else
 		    {
-			logf(LOG_LOG, "Unknown external espec.");
+			yaz_log(YLOG_LOG, "Unknown external espec.");
 			return 25; /* bad. what is proper diagnostic? */
 		    }
 		    break;
@@ -922,12 +1185,12 @@ static int process_comp(data1_handle dh, data1_node *n, Z_RecordComposition *c)
     }
     if (espec)
     {
-        logf (LOG_DEBUG, "Element: Espec-1 match");
+        yaz_log(YLOG_DEBUG, "Element: Espec-1 match");
 	return data1_doespec1(dh, n, espec);
     }
     else
     {
-	logf (LOG_DEBUG, "Element: all match");
+	yaz_log(YLOG_DEBUG, "Element: all match");
 	return -1;
     }
 }
@@ -971,7 +1234,7 @@ static void zebra_xml_metadata (struct recRetrieveCtrl *p, data1_node *top,
         data1_mk_tag_data_int (p->dh, n, "score", p->score, mem);
     }
     data1_mk_text (p->dh, mem, i4, n);
-    data1_mk_tag_data_int (p->dh, n, "localnumber", p->localno, mem);
+    data1_mk_tag_data_zint (p->dh, n, "localnumber", p->localno, mem);
     if (p->fname)
     {
         data1_mk_text (p->dh, mem, i4, n);
@@ -980,7 +1243,8 @@ static void zebra_xml_metadata (struct recRetrieveCtrl *p, data1_node *top,
     data1_mk_text (p->dh, mem, i2, n);
 }
 
-static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
+int zebra_grs_retrieve(void *clientData, struct recRetrieveCtrl *p,
+		       data1_node *(*grs_read)(struct grs_read_info *))
 {
     data1_node *node = 0, *onode = 0, *top;
     data1_node *dnew;
@@ -989,7 +1253,7 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
     NMEM mem;
     struct grs_read_info gri;
     const char *tagname;
-    struct grs_handlers *h = (struct grs_handlers *) clientData;
+
     int requested_schema = VAL_NONE;
     data1_marctab *marctab;
     int dummy;
@@ -1003,29 +1267,24 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
     gri.offset = 0;
     gri.mem = mem;
     gri.dh = p->dh;
+    gri.clientData = clientData;
 
-    logf (LOG_DEBUG, "grs_retrieve");
-    if (read_grs_type (h, &gri, p->subType, &node))
-    {
-	p->diagnostic = 14;
-        nmem_destroy (mem);
-	return 0;
-    }
+    yaz_log(YLOG_DEBUG, "grs_retrieve");
+    node = (*grs_read)(&gri);
     if (!node)
     {
 	p->diagnostic = 14;
         nmem_destroy (mem);
 	return 0;
     }
-    /* ensure our data1 tree is UTF-8 */
-    data1_iconv (p->dh, mem, node, "UTF-8", data1_get_encoding(p->dh, node));
+    data1_concat_text(p->dh, mem, node);
 
 #if 0
     data1_pr_tree (p->dh, node, stdout);
 #endif
     top = data1_get_root_tag (p->dh, node);
 
-    logf (LOG_DEBUG, "grs_retrieve: size");
+    yaz_log(YLOG_DEBUG, "grs_retrieve: size");
     tagname = data1_systag_lookup(node->u.root.absyn, "size", "size");
     if (tagname &&
         (dnew = data1_mk_tag_data_wd(p->dh, top, tagname, mem)))
@@ -1040,7 +1299,7 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
     if (tagname && p->score >= 0 &&
 	(dnew = data1_mk_tag_data_wd(p->dh, top, tagname, mem)))
     {
-        logf (LOG_DEBUG, "grs_retrieve: %s", tagname);
+        yaz_log(YLOG_DEBUG, "grs_retrieve: %s", tagname);
 	dnew->u.data.what = DATA1I_num;
 	dnew->u.data.data = dnew->lbuf;
 	sprintf(dnew->u.data.data, "%d", p->score);
@@ -1052,17 +1311,20 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
     if (tagname && p->localno > 0 &&
         (dnew = data1_mk_tag_data_wd(p->dh, top, tagname, mem)))
     {
-        logf (LOG_DEBUG, "grs_retrieve: %s", tagname);
+        yaz_log(YLOG_DEBUG, "grs_retrieve: %s", tagname);
 	dnew->u.data.what = DATA1I_text;
 	dnew->u.data.data = dnew->lbuf;
         
-	sprintf(dnew->u.data.data, "%d", p->localno);
+	sprintf(dnew->u.data.data, ZINT_FORMAT, p->localno);
 	dnew->u.data.len = strlen(dnew->u.data.data);
     }
+
+    if (p->input_format == VAL_TEXT_XML)
+       zebra_xml_metadata (p, top, mem);
+
 #if 0
     data1_pr_tree (p->dh, node, stdout);
 #endif
-#if YAZ_VERSIONL >= 0x010903L
     if (p->comp && p->comp->which == Z_RecordComp_complex &&
 	p->comp->u.complex->generic &&
         p->comp->u.complex->generic->which == Z_Schema_oid &&
@@ -1072,22 +1334,12 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
 	if (oe)
 	    requested_schema = oe->value;
     }
-#else
-    if (p->comp && p->comp->which == Z_RecordComp_complex &&
-	p->comp->u.complex->generic && p->comp->u.complex->generic->schema)
-    {
-	oident *oe = oid_getentbyoid (p->comp->u.complex->generic->schema);
-	if (oe)
-	    requested_schema = oe->value;
-    }
-#endif
-
     /* If schema has been specified, map if possible, then check that
      * we got the right one 
      */
     if (requested_schema != VAL_NONE)
     {
-	logf (LOG_DEBUG, "grs_retrieve: schema mapping");
+	yaz_log(YLOG_DEBUG, "grs_retrieve: schema mapping");
 	for (map = node->u.root.absyn->maptabs; map; map = map->next)
 	{
 	    if (map->target_absyn_ref == requested_schema)
@@ -1115,7 +1367,7 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
      * the overlap of schema and formatting which is inherent in the MARC
      * family)
      */
-    yaz_log (LOG_DEBUG, "grs_retrieve: syntax mapping");
+    yaz_log(YLOG_DEBUG, "grs_retrieve: syntax mapping");
     if (node->u.root.absyn)
         for (map = node->u.root.absyn->maptabs; map; map = map->next)
         {
@@ -1131,7 +1383,7 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
                 break;
             }
         }
-    yaz_log (LOG_DEBUG, "grs_retrieve: schemaIdentifier");
+    yaz_log(YLOG_DEBUG, "grs_retrieve: schemaIdentifier");
     if (node->u.root.absyn &&
 	node->u.root.absyn->reference != VAL_NONE &&
 	p->input_format == VAL_GRS1)
@@ -1169,8 +1421,9 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
 	}
     }
 
-    logf (LOG_DEBUG, "grs_retrieve: element spec");
-    if (p->comp && (res = process_comp(p->dh, node, p->comp)) > 0)
+    yaz_log(YLOG_DEBUG, "grs_retrieve: element spec");
+    if (p->comp && (res = process_comp(p->dh, node, p->comp, &p->addinfo,
+				       p->odr)) > 0)
     {
 	p->diagnostic = res;
 	if (onode)
@@ -1185,19 +1438,18 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
 #if 0
     data1_pr_tree (p->dh, node, stdout);
 #endif
-    logf (LOG_DEBUG, "grs_retrieve: transfer syntax mapping");
+    yaz_log(YLOG_DEBUG, "grs_retrieve: transfer syntax mapping");
     switch (p->output_format = (p->input_format != VAL_NONE ?
 				p->input_format : VAL_SUTRS))
     {
     case VAL_TEXT_XML:
-        zebra_xml_metadata (p, top, mem);
-
 #if 0
         data1_pr_tree (p->dh, node, stdout);
 #endif
-
-        if (p->encoding)
-            data1_iconv (p->dh, mem, node, p->encoding, "UTF-8");
+	/* default output encoding for XML is UTF-8 */
+	data1_iconv (p->dh, mem, node,
+		     p->encoding ? p->encoding : "UTF-8",
+		     data1_get_encoding(p->dh, node));
 
 	if (!(p->rec_buf = data1_nodetoidsgml(p->dh, node, selected,
 					      &p->rec_len)))
@@ -1210,6 +1462,7 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
 	}
 	break;
     case VAL_GRS1:
+	data1_iconv (p->dh, mem, node, "UTF-8", data1_get_encoding(p->dh, node));
 	dummy = 0;
 	if (!(p->rec_buf = data1_nodetogr(p->dh, node, selected,
 					  p->odr, &dummy)))
@@ -1218,6 +1471,9 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
 	    p->rec_len = (size_t) (-1);
 	break;
     case VAL_EXPLAIN:
+	/* ensure our data1 tree is UTF-8 */
+	data1_iconv (p->dh, mem, node, "UTF-8", data1_get_encoding(p->dh, node));
+	
 	if (!(p->rec_buf = data1_nodetoexplain(p->dh, node, selected,
 					       p->odr)))
 	    p->diagnostic = 238;
@@ -1225,6 +1481,8 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
 	    p->rec_len = (size_t) (-1);
 	break;
     case VAL_SUMMARY:
+	/* ensure our data1 tree is UTF-8 */
+	data1_iconv (p->dh, mem, node, "UTF-8", data1_get_encoding(p->dh, node));
 	if (!(p->rec_buf = data1_nodetosummary(p->dh, node, selected,
 					       p->odr)))
 	    p->diagnostic = 238;
@@ -1232,8 +1490,9 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
 	    p->rec_len = (size_t) (-1);
 	break;
     case VAL_SUTRS:
-        if (p->encoding)
-            data1_iconv (p->dh, mem, node, p->encoding, "UTF-8");
+	if (p->encoding)
+            data1_iconv (p->dh, mem, node, p->encoding,
+			 data1_get_encoding(p->dh, node));
 	if (!(p->rec_buf = data1_nodetobuf(p->dh, node, selected,
 					   &p->rec_len)))
 	    p->diagnostic = 238;
@@ -1245,6 +1504,9 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
 	}
 	break;
     case VAL_SOIF:
+	if (p->encoding)
+            data1_iconv (p->dh, mem, node, p->encoding,
+			 data1_get_encoding(p->dh, node));
 	if (!(p->rec_buf = data1_nodetosoif(p->dh, node, selected,
 					    &p->rec_len)))
 	    p->diagnostic = 238;
@@ -1270,8 +1532,9 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
 	    p->diagnostic = 238;
 	    break;
 	}
-        if (p->encoding)
-            data1_iconv (p->dh, mem, node, p->encoding, "UTF-8");
+	if (p->encoding)
+            data1_iconv (p->dh, mem, node, p->encoding,
+			 data1_get_encoding(p->dh, node));
 	if (!(p->rec_buf = data1_nodetomarc(p->dh, marctab, node,
 					selected, &p->rec_len)))
 	    p->diagnostic = 238;
@@ -1290,13 +1553,11 @@ static int grs_retrieve(void *clientData, struct recRetrieveCtrl *p)
     return 0;
 }
 
-static struct recType grs_type =
-{
-    "grs",
-    grs_init,
-    grs_destroy,
-    grs_extract,
-    grs_retrieve
-};
+/*
+ * Local variables:
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ * vim: shiftwidth=4 tabstop=8 expandtab
+ */
 
-RecType recTypeGrs = &grs_type;