Reformat a bit
[yaz-moved-to-github.git] / src / cqltransform.c
index a203f61..3163775 100644 (file)
-/* $Id: cqltransform.c,v 1.17 2005-12-21 00:04:14 mike Exp $
-   Copyright (C) 1995-2005, Index Data ApS
-   Index Data Aps
-
-This file is part of the YAZ toolkit.
-
-See the file LICENSE.
-*/
-
+/* This file is part of the YAZ toolkit.
+ * Copyright (C) 1995-2011 Index Data
+ * See the file LICENSE for details.
+ */
 /**
  * \file cqltransform.c
  * \brief Implements CQL transform (CQL to RPN conversion).
+ *
+ * Evaluation order of rules:
+ *
+ * always
+ * relation
+ * structure
+ * position
+ * truncation
+ * index
+ * relationModifier
  */
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
 
+#include <assert.h>
 #include <stdlib.h>
 #include <string.h>
-#include <yaz/cql.h>
+#include <yaz/rpn2cql.h>
 #include <yaz/xmalloc.h>
+#include <yaz/diagsrw.h>
+#include <yaz/tokenizer.h>
+#include <yaz/wrbuf.h>
+#include <yaz/z-core.h>
+#include <yaz/matchstr.h>
+#include <yaz/oid_db.h>
+#include <yaz/log.h>
 
 struct cql_prop_entry {
     char *pattern;
     char *value;
+    Z_AttributeList attr_list;
     struct cql_prop_entry *next;
 };
 
 struct cql_transform_t_ {
     struct cql_prop_entry *entry;
+    yaz_tok_cfg_t tok_cfg;
     int error;
     char *addinfo;
+    WRBUF w;
+    NMEM nmem;
 };
 
+
+cql_transform_t cql_transform_create(void)
+{
+    cql_transform_t ct = (cql_transform_t) xmalloc(sizeof(*ct));
+    ct->tok_cfg = yaz_tok_cfg_create();
+    ct->w = wrbuf_alloc();
+    ct->error = 0;
+    ct->addinfo = 0;
+    ct->entry = 0;
+    ct->nmem = nmem_create();
+    return ct;
+}
+
+static int cql_transform_parse_tok_line(cql_transform_t ct,
+                                        const char *pattern,
+                                        yaz_tok_parse_t tp)
+{
+    int ae_num = 0;
+    Z_AttributeElement *ae[20];
+    int ret = 0; /* 0=OK, != 0 FAIL */
+    int t;
+    t = yaz_tok_move(tp);
+    
+    while (t == YAZ_TOK_STRING && ae_num < 20)
+    {
+        WRBUF type_str = wrbuf_alloc();
+        WRBUF set_str = 0;
+        Z_AttributeElement *elem = 0;
+        const char *value_str = 0;
+        /* attset type=value  OR  type=value */
+        
+        elem = (Z_AttributeElement *) nmem_malloc(ct->nmem, sizeof(*elem));
+        elem->attributeSet = 0;
+        ae[ae_num] = elem;
+        wrbuf_puts(ct->w, yaz_tok_parse_string(tp));
+        wrbuf_puts(type_str, yaz_tok_parse_string(tp));
+        t = yaz_tok_move(tp);
+        if (t == YAZ_TOK_EOF)
+        {
+            wrbuf_destroy(type_str);
+            if (set_str)
+                wrbuf_destroy(set_str);                
+            break;
+        }
+        if (t == YAZ_TOK_STRING)  
+        {  
+            wrbuf_puts(ct->w, " ");
+            wrbuf_puts(ct->w, yaz_tok_parse_string(tp));
+            set_str = type_str;
+            
+            elem->attributeSet =
+                yaz_string_to_oid_nmem(yaz_oid_std(), CLASS_ATTSET,
+                                       wrbuf_cstr(set_str), ct->nmem);
+            
+            type_str = wrbuf_alloc();
+            wrbuf_puts(type_str, yaz_tok_parse_string(tp));
+            t = yaz_tok_move(tp);
+        }
+        elem->attributeType = nmem_intdup(ct->nmem, 0);
+        if (sscanf(wrbuf_cstr(type_str), ODR_INT_PRINTF, elem->attributeType)
+            != 1)
+        {
+            wrbuf_destroy(type_str);
+            if (set_str)
+                wrbuf_destroy(set_str);                
+            yaz_log(YLOG_WARN, "Expected numeric attribute type");
+            ret = -1;
+            break;
+        }
+
+        wrbuf_destroy(type_str);
+        if (set_str)
+            wrbuf_destroy(set_str);                
+        
+        if (t != '=')
+        {
+            yaz_log(YLOG_WARN, "Expected = after after attribute type");
+            ret = -1;
+            break;
+        }
+        t = yaz_tok_move(tp);
+        if (t != YAZ_TOK_STRING) /* value */
+        {
+            yaz_log(YLOG_WARN, "Missing attribute value");
+            ret = -1;
+            break;
+        }
+        value_str = yaz_tok_parse_string(tp);
+        if (yaz_isdigit(*value_str))
+        {
+            elem->which = Z_AttributeValue_numeric;
+            elem->value.numeric =
+                nmem_intdup(ct->nmem, atoi(value_str));
+        }
+        else
+        {
+            Z_ComplexAttribute *ca = (Z_ComplexAttribute *)
+                nmem_malloc(ct->nmem, sizeof(*ca));
+            elem->which = Z_AttributeValue_complex;
+            elem->value.complex = ca;
+            ca->num_list = 1;
+            ca->list = (Z_StringOrNumeric **)
+                nmem_malloc(ct->nmem, sizeof(Z_StringOrNumeric *));
+            ca->list[0] = (Z_StringOrNumeric *)
+                nmem_malloc(ct->nmem, sizeof(Z_StringOrNumeric));
+            ca->list[0]->which = Z_StringOrNumeric_string;
+            ca->list[0]->u.string = nmem_strdup(ct->nmem, value_str);
+            ca->num_semanticAction = 0;
+            ca->semanticAction = 0;
+        }
+        wrbuf_puts(ct->w, "=");
+        wrbuf_puts(ct->w, yaz_tok_parse_string(tp));
+        t = yaz_tok_move(tp);
+        wrbuf_puts(ct->w, " ");
+        ae_num++;
+    }
+    if (ret == 0) /* OK? */
+    {
+        struct cql_prop_entry **pp = &ct->entry;
+        while (*pp)
+            pp = &(*pp)->next;
+        *pp = (struct cql_prop_entry *) xmalloc(sizeof(**pp));
+        (*pp)->pattern = xstrdup(pattern);
+        (*pp)->value = xstrdup(wrbuf_cstr(ct->w));
+
+        (*pp)->attr_list.num_attributes = ae_num;
+        if (ae_num == 0)
+            (*pp)->attr_list.attributes = 0;
+        else
+        {
+            (*pp)->attr_list.attributes = (Z_AttributeElement **)
+                nmem_malloc(ct->nmem,
+                            ae_num * sizeof(Z_AttributeElement *));
+            memcpy((*pp)->attr_list.attributes, ae, 
+                   ae_num * sizeof(Z_AttributeElement *));
+        }
+        (*pp)->next = 0;
+
+        if (0)
+        {
+            ODR pr = odr_createmem(ODR_PRINT);
+            Z_AttributeList *alp = &(*pp)->attr_list;
+            odr_setprint(pr, yaz_log_file());
+            z_AttributeList(pr, &alp, 0, 0);
+            odr_setprint(pr, 0);
+            odr_destroy(pr);
+        }
+    }
+    return ret;
+}
+
+int cql_transform_define_pattern(cql_transform_t ct, const char *pattern,
+                                 const char *value)
+{
+    int r;
+    yaz_tok_parse_t tp = yaz_tok_parse_buf(ct->tok_cfg, value);
+    yaz_tok_cfg_single_tokens(ct->tok_cfg, "=");
+    r = cql_transform_parse_tok_line(ct, pattern, tp);
+    yaz_tok_parse_destroy(tp);
+    return r;
+}
+    
 cql_transform_t cql_transform_open_FILE(FILE *f)
 {
+    cql_transform_t ct = cql_transform_create();
     char line[1024];
-    cql_transform_t ct = (cql_transform_t) xmalloc (sizeof(*ct));
-    struct cql_prop_entry **pp = &ct->entry;
 
-    ct->error = 0;
-    ct->addinfo = 0;
+    yaz_tok_cfg_single_tokens(ct->tok_cfg, "=");
+
     while (fgets(line, sizeof(line)-1, f))
     {
-        const char *cp_value_start;
-        const char *cp_value_end;
-        const char *cp_pattern_end;
-        const char *cp = line;
-        while (*cp && !strchr(" \t=\r\n#", *cp))
-            cp++;
-        cp_pattern_end = cp;
-        if (cp == line)
-            continue;
-        while (*cp && strchr(" \t\r\n", *cp))
-            cp++;
-        if (*cp != '=')
-            continue;
-        cp++;
-        while (*cp && strchr(" \t\r\n", *cp))
-            cp++;
-        cp_value_start = cp;
-        if (!(cp_value_end = strchr(cp, '#')))
-            cp_value_end = strlen(line) + line;
-
-        if (cp_value_end != cp_value_start &&
-            strchr(" \t\r\n", cp_value_end[-1]))
-            cp_value_end--;
-        *pp = (struct cql_prop_entry *) xmalloc (sizeof(**pp));
-        (*pp)->pattern = (char *) xmalloc (cp_pattern_end - line + 1);
-        memcpy ((*pp)->pattern, line, cp_pattern_end - line);
-        (*pp)->pattern[cp_pattern_end-line] = 0;
-
-        (*pp)->value = (char *) xmalloc (cp_value_end - cp_value_start + 1);
-        if (cp_value_start != cp_value_end)
-            memcpy ((*pp)->value, cp_value_start, cp_value_end-cp_value_start);
-        (*pp)->value[cp_value_end - cp_value_start] = 0;
-        pp = &(*pp)->next;
+        yaz_tok_parse_t tp = yaz_tok_parse_buf(ct->tok_cfg, line);
+        int t;
+        wrbuf_rewind(ct->w);
+        t = yaz_tok_move(tp);
+        if (t == YAZ_TOK_STRING)
+        {
+            char * pattern = xstrdup(yaz_tok_parse_string(tp));
+            t = yaz_tok_move(tp);
+            if (t != '=')
+            {
+                yaz_tok_parse_destroy(tp);
+                cql_transform_close(ct);
+                return 0;
+            }
+            if (cql_transform_parse_tok_line(ct, pattern, tp))
+            {
+                yaz_tok_parse_destroy(tp);
+                cql_transform_close(ct);
+                return 0;
+            }
+            xfree(pattern);
+        }
+        else if (t != YAZ_TOK_EOF)
+        {
+            yaz_tok_parse_destroy(tp);
+            cql_transform_close(ct);
+            return 0;
+        }
+        yaz_tok_parse_destroy(tp);
     }
-    *pp = 0;
     return ct;
 }
 
@@ -86,14 +262,16 @@ void cql_transform_close(cql_transform_t ct)
     while (pe)
     {
         struct cql_prop_entry *pe_next = pe->next;
-        xfree (pe->pattern);
-        xfree (pe->value);
-        xfree (pe);
+        xfree(pe->pattern);
+        xfree(pe->value);
+        xfree(pe);
         pe = pe_next;
     }
-    if (ct->addinfo)
-        xfree (ct->addinfo);
-    xfree (ct);
+    xfree(ct->addinfo);
+    yaz_tok_cfg_destroy(ct->tok_cfg);
+    wrbuf_destroy(ct->w);
+    nmem_destroy(ct->nmem);
+    xfree(ct);
 }
 
 cql_transform_t cql_transform_open_fname(const char *fname)
@@ -107,6 +285,78 @@ cql_transform_t cql_transform_open_fname(const char *fname)
     return ct;
 }
 
+#if 0
+struct Z_AttributeElement {
+       Z_AttributeSetId *attributeSet; /* OPT */
+       int *attributeType;
+       int which;
+       union {
+               int *numeric;
+               Z_ComplexAttribute *complex;
+#define Z_AttributeValue_numeric 1
+#define Z_AttributeValue_complex 2
+       } value;
+};
+#endif
+
+static int compare_attr(Z_AttributeElement *a, Z_AttributeElement *b)
+{
+    ODR odr_a = odr_createmem(ODR_ENCODE);
+    ODR odr_b = odr_createmem(ODR_ENCODE);
+    int len_a, len_b;
+    char *buf_a, *buf_b;
+    int ret;
+
+    z_AttributeElement(odr_a, &a, 0, 0);
+    z_AttributeElement(odr_b, &b, 0, 0);
+    
+    buf_a = odr_getbuf(odr_a, &len_a, 0);
+    buf_b = odr_getbuf(odr_b, &len_b, 0);
+
+    ret = yaz_memcmp(buf_a, buf_b, len_a, len_b);
+
+    odr_destroy(odr_a);
+    odr_destroy(odr_b);
+    return ret;
+}
+
+const char *cql_lookup_reverse(cql_transform_t ct, 
+                               const char *category,
+                               Z_AttributeList *attributes)
+{
+    struct cql_prop_entry *e;
+    size_t clen = strlen(category);
+    for (e = ct->entry; e; e = e->next)
+    {
+        if (!strncmp(e->pattern, category, clen))
+        {
+            /* category matches.. See if attributes in pattern value
+               are all listed in actual attributes */
+            int i;
+            for (i = 0; i < e->attr_list.num_attributes; i++)
+            {
+                /* entry attribute */
+                Z_AttributeElement *e_ae = e->attr_list.attributes[i];
+                int j;
+                for (j = 0; j < attributes->num_attributes; j++)
+                {
+                    /* actual attribute */
+                    Z_AttributeElement *a_ae = attributes->attributes[j];
+                    int r = compare_attr(e_ae, a_ae);
+                    if (r == 0)
+                        break;
+                }
+                if (j == attributes->num_attributes)
+                    break; /* i was not found at all.. try next pattern */
+                    
+            }
+            if (i == e->attr_list.num_attributes)
+                return e->pattern + clen;
+        }
+    }
+    return 0;
+}
+                                      
 static const char *cql_lookup_property(cql_transform_t ct,
                                        const char *pat1, const char *pat2,
                                        const char *pat3)
@@ -115,13 +365,13 @@ static const char *cql_lookup_property(cql_transform_t ct,
     struct cql_prop_entry *e;
 
     if (pat1 && pat2 && pat3)
-        sprintf (pattern, "%.39s.%.39s.%.39s", pat1, pat2, pat3);
+        sprintf(pattern, "%.39s.%.39s.%.39s", pat1, pat2, pat3);
     else if (pat1 && pat2)
-        sprintf (pattern, "%.39s.%.39s", pat1, pat2);
+        sprintf(pattern, "%.39s.%.39s", pat1, pat2);
     else if (pat1 && pat3)
-        sprintf (pattern, "%.39s.%.39s", pat1, pat3);
+        sprintf(pattern, "%.39s.%.39s", pat1, pat3);
     else if (pat1)
-        sprintf (pattern, "%.39s", pat1);
+        sprintf(pattern, "%.39s", pat1);
     else
         return 0;
     
@@ -161,6 +411,18 @@ int cql_pr_attr_uri(cql_transform_t ct, const char *category,
     {
         if (!res)
             res = cql_lookup_property(ct, category, prefix, eval);
+        /* we have some aliases for some relations unfortunately.. */
+        if (!res && !prefix && !strcmp(category, "relation"))
+        {
+            if (!strcmp(val, "=="))
+                res = cql_lookup_property(ct, category, prefix, "exact");
+            if (!strcmp(val, "="))
+                res = cql_lookup_property(ct, category, prefix, "eq");
+            if (!strcmp(val, "<="))
+                res = cql_lookup_property(ct, category, prefix, "le");
+            if (!strcmp(val, ">="))
+                res = cql_lookup_property(ct, category, prefix, "ge");
+        }
         if (!res)
             res = cql_lookup_property(ct, category, prefix, "*");
     }
@@ -171,14 +433,27 @@ int cql_pr_attr_uri(cql_transform_t ct, const char *category,
         const char *cp0 = res, *cp1;
         while ((cp1 = strchr(cp0, '=')))
         {
+            int i;
             while (*cp1 && *cp1 != ' ')
                 cp1++;
-            if (cp1 - cp0 >= sizeof(buf))
+            if (cp1 - cp0 >= (ptrdiff_t) sizeof(buf))
                 break;
-            memcpy (buf, cp0, cp1 - cp0);
+            memcpy(buf, cp0, cp1 - cp0);
             buf[cp1-cp0] = 0;
             (*pr)("@attr ", client_data);
-            (*pr)(buf, client_data);
+
+            for (i = 0; buf[i]; i++)
+            {
+                if (buf[i] == '*')
+                    (*pr)(eval, client_data);
+                else
+                {
+                    char tmp[2];
+                    tmp[0] = buf[i];
+                    tmp[1] = '\0';
+                    (*pr)(tmp, client_data);
+                }
+            }
             (*pr)(" ", client_data);
             cp0 = cp1;
             while (*cp0 == ' ')
@@ -209,90 +484,206 @@ int cql_pr_attr(cql_transform_t ct, const char *category,
 }
 
 
+static void cql_pr_int(int val,
+                       void (*pr)(const char *buf, void *client_data),
+                       void *client_data)
+{
+    char buf[21];              /* enough characters to 2^64 */
+    sprintf(buf, "%d", val);
+    (*pr)(buf, client_data);
+    (*pr)(" ", client_data);
+}
+
+
+static int cql_pr_prox(cql_transform_t ct, struct cql_node *mods,
+                       void (*pr)(const char *buf, void *client_data),
+                       void *client_data)
+{
+    int exclusion = 0;
+    int distance;               /* to be filled in later depending on unit */
+    int distance_defined = 0;
+    int ordered = 0;
+    int proxrel = 2;            /* less than or equal */
+    int unit = 2;               /* word */
+
+    while (mods)
+    {
+        const char *name = mods->u.st.index;
+        const char *term = mods->u.st.term;
+        const char *relation = mods->u.st.relation;
+
+        if (!strcmp(name, "distance")) {
+            distance = strtol(term, (char**) 0, 0);
+            distance_defined = 1;
+            if (!strcmp(relation, "="))
+                proxrel = 3;
+            else if (!strcmp(relation, ">"))
+                proxrel = 5;
+            else if (!strcmp(relation, "<"))
+                proxrel = 1;
+            else if (!strcmp(relation, ">=")) 
+                proxrel = 4;
+            else if (!strcmp(relation, "<="))
+                proxrel = 2;
+            else if (!strcmp(relation, "<>"))
+                proxrel = 6;
+            else 
+            {
+                ct->error = YAZ_SRW_UNSUPP_PROX_RELATION;
+                ct->addinfo = xstrdup(relation);
+                return 0;
+            }
+        } 
+        else if (!strcmp(name, "ordered"))
+            ordered = 1;
+        else if (!strcmp(name, "unordered"))
+            ordered = 0;
+        else if (!strcmp(name, "unit"))
+        {
+            if (!strcmp(term, "word"))
+                unit = 2;
+            else if (!strcmp(term, "sentence"))
+                unit = 3;
+            else if (!strcmp(term, "paragraph"))
+                unit = 4;
+            else if (!strcmp(term, "element"))
+                unit = 8;
+            else 
+            {
+                ct->error = YAZ_SRW_UNSUPP_PROX_UNIT;
+                ct->addinfo = xstrdup(term);
+                return 0;
+            }
+        } 
+        else 
+        {
+            ct->error = YAZ_SRW_UNSUPP_BOOLEAN_MODIFIER;
+            ct->addinfo = xstrdup(name);
+            return 0;
+        }
+        mods = mods->u.st.modifiers;
+    }
+
+    if (!distance_defined)
+        distance = (unit == 2) ? 1 : 0;
+
+    cql_pr_int(exclusion, pr, client_data);
+    cql_pr_int(distance, pr, client_data);
+    cql_pr_int(ordered, pr, client_data);
+    cql_pr_int(proxrel, pr, client_data);
+    (*pr)("k ", client_data);
+    cql_pr_int(unit, pr, client_data);
+
+    return 1;
+}
+
 /* Returns location of first wildcard character in the `length'
  * characters starting at `term', or a null pointer of there are
  * none -- like memchr().
  */
-static const char *wcchar(const char *term, int length)
+static const char *wcchar(int start, const char *term, int length)
 {
-    const char *best = 0;
-    const char *current;
-    char *whichp;
-
-    for (whichp = "*?"; *whichp != '\0'; whichp++) {
-        current = (const char *) memchr(term, *whichp, length);
-        if (current != 0 && (best == 0 || current < best))
-            best = current;
+    while (length > 0)
+    {
+        if (start || term[-1] != '\\')
+            if (strchr("*?", *term))
+                return term;
+        term++;
+        length--;
+        start = 0;
+    }
+    return 0;
+}
+
+
+/* ### checks for CQL relation-name rather than Type-1 attribute */
+static int has_modifier(struct cql_node *cn, const char *name) {
+    struct cql_node *mod;
+    for (mod = cn->u.st.modifiers; mod != 0; mod = mod->u.st.modifiers) {
+        if (!strcmp(mod->u.st.index, name))
+            return 1;
     }
 
-    return best;
+    return 0;
 }
 
 
-void emit_term(cql_transform_t ct,
-               const char *term, int length,
-               void (*pr)(const char *buf, void *client_data),
-               void *client_data)
+static void emit_term(cql_transform_t ct,
+                      struct cql_node *cn,
+                      const char *term, int length,
+                      void (*pr)(const char *buf, void *client_data),
+                      void *client_data)
 {
     int i;
-    if (length > 0)
+    const char *ns = cn->u.st.index_uri;
+    int process_term = !has_modifier(cn, "regexp");
+    char *z3958_mem = 0;
+
+    assert(cn->which == CQL_NODE_ST);
+
+    if (process_term && length > 0)
     {
         if (length > 1 && term[0] == '^' && term[length-1] == '^')
         {
             cql_pr_attr(ct, "position", "firstAndLast", 0,
-                        pr, client_data, 32);
+                        pr, client_data, YAZ_SRW_ANCHORING_CHAR_IN_UNSUPP_POSITION);
             term++;
             length -= 2;
         }
         else if (term[0] == '^')
         {
             cql_pr_attr(ct, "position", "first", 0,
-                        pr, client_data, 32);
+                        pr, client_data, YAZ_SRW_ANCHORING_CHAR_IN_UNSUPP_POSITION);
             term++;
             length--;
         }
         else if (term[length-1] == '^')
         {
             cql_pr_attr(ct, "position", "last", 0,
-                        pr, client_data, 32);
+                        pr, client_data, YAZ_SRW_ANCHORING_CHAR_IN_UNSUPP_POSITION);
             length--;
         }
         else
         {
             cql_pr_attr(ct, "position", "any", 0,
-                        pr, client_data, 32);
+                        pr, client_data, YAZ_SRW_ANCHORING_CHAR_IN_UNSUPP_POSITION);
         }
     }
 
-    if (length > 0)
+    if (process_term && length > 0)
     {
+        const char *first_wc = wcchar(1, term, length);
+        const char *second_wc = first_wc ?
+            wcchar(0, first_wc+1, length-(first_wc-term)-1) : 0;
+
         /* Check for well-known globbing patterns that represent
          * simple truncation attributes as expected by, for example,
          * Bath-compliant server.  If we find such a pattern but
          * there's no mapping for it, that's fine: we just use a
          * general pattern-matching attribute.
          */
-        if (length > 1 && term[0] == '*' && term[length-1] == '*' &&
-            wcchar(term+1, length-2) == 0 &&
-            cql_pr_attr(ct, "truncation", "both", 0,
-                        pr, client_data, 0)) {
+        if (first_wc == term && second_wc == term + length-1 
+            && *first_wc == '*' && *second_wc == '*' 
+            && cql_pr_attr(ct, "truncation", "both", 0, pr, client_data, 0)) 
+        {
             term++;
             length -= 2;
         }
-        else if (term[0] == '*' &&
-                 wcchar(term+1, length-1) == 0 &&
-                 cql_pr_attr(ct, "truncation", "left", 0,
-                             pr, client_data, 0)) {
+        else if (first_wc == term && second_wc == 0 && *first_wc == '*'
+                 && cql_pr_attr(ct, "truncation", "left", 0,
+                                pr, client_data, 0))
+        {
             term++;
             length--;
         }
-        else if (term[length-1] == '*' &&
-                 wcchar(term, length-1) == 0 &&
-                 cql_pr_attr(ct, "truncation", "right", 0,
-                             pr, client_data, 0)) {
+        else if (first_wc == term + length-1 && second_wc == 0
+                 && *first_wc == '*'
+                 && cql_pr_attr(ct, "truncation", "right", 0, 
+                                pr, client_data, 0))
+        {
             length--;
         }
-        else if (wcchar(term, length))
+        else if (first_wc)
         {
             /* We have one or more wildcard characters, but not in a
              * way that can be dealt with using only the standard
@@ -304,17 +695,22 @@ void emit_term(cql_transform_t ct,
              * supported".
              */
             int i;
-            char *mem;
             cql_pr_attr(ct, "truncation", "z3958", 0,
-                        pr, client_data, 28);
-            mem = (char *) xmalloc(length+1);
-            for (i = 0; i < length; i++) {
-                if (term[i] == '*')      mem[i] = '?';
-                else if (term[i] == '?') mem[i] = '#';
-                else                     mem[i] = term[i];
+                        pr, client_data, YAZ_SRW_MASKING_CHAR_UNSUPP);
+            z3958_mem = (char *) xmalloc(length+1);
+            for (i = 0; i < length; i++)
+            {
+                if (i > 0 && term[i-1] == '\\')
+                    z3958_mem[i] = term[i];
+                else if (term[i] == '*')
+                    z3958_mem[i] = '?';
+                else if (term[i] == '?')
+                    z3958_mem[i] = '#';
+                else
+                    z3958_mem[i] = term[i];
             }
-            mem[length] = '\0';
-            term = mem;
+            z3958_mem[length] = '\0';
+            term = z3958_mem;
         }
         else {
             /* No masking characters.  Use "truncation.none" if given. */
@@ -322,23 +718,70 @@ void emit_term(cql_transform_t ct,
                         pr, client_data, 0);
         }
     }
+    if (ns) {
+        cql_pr_attr_uri(ct, "index", ns,
+                        cn->u.st.index, "serverChoice",
+                        pr, client_data, YAZ_SRW_UNSUPP_INDEX);
+    }
+    if (cn->u.st.modifiers)
+    {
+        struct cql_node *mod = cn->u.st.modifiers;
+        for (; mod; mod = mod->u.st.modifiers)
+        {
+            cql_pr_attr(ct, "relationModifier", mod->u.st.index, 0,
+                        pr, client_data, YAZ_SRW_UNSUPP_RELATION_MODIFIER);
+        }
+    }
 
     (*pr)("\"", client_data);
-    for (i = 0; i<length; i++)
+    for (i = 0; i < length; i++)
     {
+        /* pr(int) each character */
+        /* we do not need to deal with \-sequences because the
+           CQL and PQF terms have same \-format, bug #1988 */
         char buf[2];
+
         buf[0] = term[i];
-        buf[1] = 0;
+        buf[1] = '\0';
         (*pr)(buf, client_data);
     }
     (*pr)("\" ", client_data);
+    xfree(z3958_mem);
 }
 
-void emit_wordlist(cql_transform_t ct,
-                   struct cql_node *cn,
-                   void (*pr)(const char *buf, void *client_data),
-                   void *client_data,
-                   const char *op)
+static void emit_terms(cql_transform_t ct,
+                       struct cql_node *cn,
+                       void (*pr)(const char *buf, void *client_data),
+                       void *client_data,
+                       const char *op)
+{
+    struct cql_node *ne = cn->u.st.extra_terms;
+    if (ne)
+    {
+        (*pr)("@", client_data);
+        (*pr)(op, client_data);
+        (*pr)(" ", client_data);
+    }
+    emit_term(ct, cn, cn->u.st.term, strlen(cn->u.st.term),
+              pr, client_data);
+    for (; ne; ne = ne->u.st.extra_terms)
+    {
+        if (ne->u.st.extra_terms)
+        {
+            (*pr)("@", client_data);
+            (*pr)(op, client_data);
+            (*pr)(" ", client_data);
+        }            
+        emit_term(ct, cn, ne->u.st.term, strlen(ne->u.st.term),
+                  pr, client_data);
+    }
+}
+
+static void emit_wordlist(cql_transform_t ct,
+                          struct cql_node *cn,
+                          void (*pr)(const char *buf, void *client_data),
+                          void *client_data,
+                          const char *op)
 {
     const char *cp0 = cn->u.st.term;
     const char *cp1;
@@ -354,7 +797,7 @@ void emit_wordlist(cql_transform_t ct,
             (*pr)("@", client_data);
             (*pr)(op, client_data);
             (*pr)(" ", client_data);
-            emit_term(ct, last_term, last_length, pr, client_data);
+            emit_term(ct, cn, last_term, last_length, pr, client_data);
         }
         last_term = cp0;
         if (cp1)
@@ -364,7 +807,7 @@ void emit_wordlist(cql_transform_t ct,
         cp0 = cp1;
     }
     if (last_term)
-        emit_term(ct, last_term, last_length, pr, client_data);
+        emit_term(ct, cn, last_term, last_length, pr, client_data);
 }
 
 void cql_transform_r(cql_transform_t ct,
@@ -373,6 +816,7 @@ void cql_transform_r(cql_transform_t ct,
                      void *client_data)
 {
     const char *ns;
+    struct cql_node *mods;
 
     if (!cn)
         return;
@@ -390,71 +834,58 @@ void cql_transform_r(cql_transform_t ct,
                 (*pr)("\" ", client_data);
                 return ;
             }
-            cql_pr_attr_uri(ct, "index", ns,
-                            cn->u.st.index, "serverChoice",
-                            pr, client_data, 16);
         }
         else
         {
             if (!ct->error)
             {
-                ct->error = 15;
+                ct->error = YAZ_SRW_UNSUPP_CONTEXT_SET;
                 ct->addinfo = 0;
             }
         }
-        if (cn->u.st.relation && !cql_strcmp(cn->u.st.relation, "="))
-            cql_pr_attr(ct, "relation", "eq", "scr",
-                        pr, client_data, 19);
-        else if (cn->u.st.relation && !cql_strcmp(cn->u.st.relation, "<="))
-            cql_pr_attr(ct, "relation", "le", "scr",
-                        pr, client_data, 19);
-        else if (cn->u.st.relation && !cql_strcmp(cn->u.st.relation, ">="))
-            cql_pr_attr(ct, "relation", "ge", "scr",
-                        pr, client_data, 19);
-        else
-            cql_pr_attr(ct, "relation", cn->u.st.relation, "eq",
-                        pr, client_data, 19);
-        if (cn->u.st.modifiers)
-        {
-            struct cql_node *mod = cn->u.st.modifiers;
-            for (; mod; mod = mod->u.st.modifiers)
-            {
-                cql_pr_attr(ct, "relationModifier", mod->u.st.index, 0,
-                            pr, client_data, 20);
-            }
-        }
+        cql_pr_attr(ct, "always", 0, 0, pr, client_data, 0);
+        cql_pr_attr(ct, "relation", cn->u.st.relation, 0, pr, client_data,
+                    YAZ_SRW_UNSUPP_RELATION);
         cql_pr_attr(ct, "structure", cn->u.st.relation, 0,
-                    pr, client_data, 24);
+                    pr, client_data, YAZ_SRW_UNSUPP_COMBI_OF_RELATION_AND_TERM);
         if (cn->u.st.relation && !cql_strcmp(cn->u.st.relation, "all"))
-        {
             emit_wordlist(ct, cn, pr, client_data, "and");
-        }
         else if (cn->u.st.relation && !cql_strcmp(cn->u.st.relation, "any"))
-        {
             emit_wordlist(ct, cn, pr, client_data, "or");
-        }
         else
-        {
-            emit_term(ct, cn->u.st.term, strlen(cn->u.st.term),
-                      pr, client_data);
-        }
+            emit_terms(ct, cn, pr, client_data, "and");
         break;
     case CQL_NODE_BOOL:
         (*pr)("@", client_data);
         (*pr)(cn->u.boolean.value, client_data);
         (*pr)(" ", client_data);
+        mods = cn->u.boolean.modifiers;
+        if (!strcmp(cn->u.boolean.value, "prox")) 
+        {
+            if (!cql_pr_prox(ct, mods, pr, client_data))
+                return;
+        } 
+        else if (mods)
+        {
+            /* Boolean modifiers other than on proximity not supported */
+            ct->error = YAZ_SRW_UNSUPP_BOOLEAN_MODIFIER;
+            ct->addinfo = xstrdup(mods->u.st.index);
+            return;
+        }
 
         cql_transform_r(ct, cn->u.boolean.left, pr, client_data);
         cql_transform_r(ct, cn->u.boolean.right, pr, client_data);
-
+        break;
+    case CQL_NODE_SORT:
+        cql_transform_r(ct, cn->u.sort.search, pr, client_data);
+        break;
     default:
         fprintf(stderr, "Fatal: impossible CQL node-type %d\n", cn->which);
         abort();
     }
 }
 
-int cql_transform(cql_transform_t ct,
-                  struct cql_node *cn,
+int cql_transform(cql_transform_t ct, struct cql_node *cn,
                   void (*pr)(const char *buf, void *client_data),
                   void *client_data)
 {
@@ -462,8 +893,7 @@ int cql_transform(cql_transform_t ct,
     NMEM nmem = nmem_create();
 
     ct->error = 0;
-    if (ct->addinfo)
-        xfree (ct->addinfo);
+    xfree(ct->addinfo);
     ct->addinfo = 0;
 
     for (e = ct->entry; e ; e = e->next)
@@ -473,7 +903,7 @@ int cql_transform(cql_transform_t ct,
         else if (!cql_strcmp(e->pattern, "set"))
             cql_apply_prefix(nmem, cn, 0, e->value);
     }
-    cql_transform_r (ct, cn, pr, client_data);
+    cql_transform_r(ct, cn, pr, client_data);
     nmem_destroy(nmem);
     return ct->error;
 }
@@ -494,6 +924,16 @@ int cql_transform_buf(cql_transform_t ct, struct cql_node *cn,
     info.max = max;
     info.buf = out;
     r = cql_transform(ct, cn, cql_buf_write_handler, &info);
+    if (info.off < 0) {
+        /* Attempt to write past end of buffer.  For some reason, this
+           SRW diagnostic is deprecated, but it's so perfect for our
+           purposes that it would be stupid not to use it. */
+        char numbuf[30];
+        ct->error = YAZ_SRW_TOO_MANY_CHARS_IN_QUERY;
+        sprintf(numbuf, "%ld", (long) info.max);
+        ct->addinfo = xstrdup(numbuf);
+        return -1;
+    }
     if (info.off >= 0)
         info.buf[info.off] = '\0';
     return r;
@@ -504,9 +944,18 @@ int cql_transform_error(cql_transform_t ct, const char **addinfo)
     *addinfo = ct->addinfo;
     return ct->error;
 }
+
+void cql_transform_set_error(cql_transform_t ct, int error, const char *addinfo)
+{
+    xfree(ct->addinfo);
+    ct->addinfo = addinfo ? xstrdup(addinfo) : 0;
+    ct->error = error;
+}
+
 /*
  * Local variables:
  * c-basic-offset: 4
+ * c-file-style: "Stroustrup"
  * indent-tabs-mode: nil
  * End:
  * vim: shiftwidth=4 tabstop=8 expandtab