X-Git-Url: http://git.indexdata.com/?p=yaz-moved-to-github.git;a=blobdiff_plain;f=src%2Fcqltransform.c;h=133db946baeb2d465f676e2d461dc71a8c3d188c;hp=29ff07ad5a0541c17c8a655702bd06386bf96528;hb=e4368af2ede9079d33e7a8fd280d29b2bde1d1ad;hpb=4c176312acdc3444c9afc820f76a393e64668e52

diff --git a/src/cqltransform.c b/src/cqltransform.c
index 29ff07a..133db94 100644
--- a/src/cqltransform.c
+++ b/src/cqltransform.c
@@ -1,4 +1,4 @@
-/* $Id: cqltransform.c,v 1.13 2005-01-15 19:47:11 adam Exp $
+/* $Id: cqltransform.c,v 1.25 2006-10-25 09:58:19 adam Exp $
    Copyright (C) 1995-2005, Index Data ApS
    Index Data Aps
 
@@ -10,12 +10,24 @@ See the file LICENSE.
 /**
  * \file cqltransform.c
  * \brief Implements CQL transform (CQL to RPN conversion).
+ *
+ * Evaluation order of rules:
+ *
+ * always
+ * relation
+ * structure
+ * position
+ * truncation
+ * index
+ * relationModifier
  */
 
+#include <assert.h>
 #include <stdlib.h>
 #include <string.h>
 #include <yaz/cql.h>
 #include <yaz/xmalloc.h>
+#include <yaz/diagsrw.h>
 
 struct cql_prop_entry {
     char *pattern;
@@ -41,36 +53,48 @@ cql_transform_t cql_transform_open_FILE(FILE *f)
     {
         const char *cp_value_start;
         const char *cp_value_end;
+        const char *cp_pattern_start;
         const char *cp_pattern_end;
         const char *cp = line;
-        while (*cp && !strchr(" \t=\r\n#", *cp))
+
+        while (*cp && strchr(" \t", *cp))
+            cp++;
+        cp_pattern_start = cp;
+        
+        while (*cp && !strchr(" \t\r\n=#", *cp))
             cp++;
         cp_pattern_end = cp;
-        if (cp == line)
+        if (cp == cp_pattern_start)
             continue;
-        while (*cp && strchr(" \t\r\n", *cp))
+        while (*cp && strchr(" \t", *cp))
             cp++;
         if (*cp != '=')
-            continue;
+        {
+            *pp = 0;
+            cql_transform_close(ct);
+            return 0;
+        }
         cp++;
         while (*cp && strchr(" \t\r\n", *cp))
             cp++;
         cp_value_start = cp;
-        if (!(cp_value_end = strchr(cp, '#')))
+        cp_value_end = strchr(cp, '#');
+        if (!cp_value_end)
             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)->pattern = (char *) xmalloc(cp_pattern_end-cp_pattern_start + 1);
+        memcpy ((*pp)->pattern, cp_pattern_start,
+                cp_pattern_end-cp_pattern_start);
+        (*pp)->pattern[cp_pattern_end-cp_pattern_start] = '\0';
 
-        (*pp)->value = (char *) xmalloc (cp_value_end - cp_value_start + 1);
+        (*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)->value[cp_value_end - cp_value_start] = '\0';
         pp = &(*pp)->next;
     }
     *pp = 0;
@@ -109,7 +133,7 @@ cql_transform_t cql_transform_open_fname(const char *fname)
 
 static const char *cql_lookup_property(cql_transform_t ct,
                                        const char *pat1, const char *pat2,
-				       const char *pat3)
+                                       const char *pat3)
 {
     char pattern[120];
     struct cql_prop_entry *e;
@@ -123,21 +147,21 @@ static const char *cql_lookup_property(cql_transform_t ct,
     else if (pat1)
         sprintf (pattern, "%.39s", pat1);
     else
-	return 0;
+        return 0;
     
     for (e = ct->entry; e; e = e->next)
     {
-        if (!strcmp(e->pattern, pattern))
+        if (!cql_strcmp(e->pattern, pattern))
             return e->value;
     }
     return 0;
 }
 
 int cql_pr_attr_uri(cql_transform_t ct, const char *category,
-		   const char *uri, const char *val, const char *default_val,
-		   void (*pr)(const char *buf, void *client_data),
-		   void *client_data,
-		   int errcode)
+                   const char *uri, const char *val, const char *default_val,
+                   void (*pr)(const char *buf, void *client_data),
+                   void *client_data,
+                   int errcode)
 {
     const char *res = 0;
     const char *eval = val ? val : default_val;
@@ -145,24 +169,24 @@ int cql_pr_attr_uri(cql_transform_t ct, const char *category,
     
     if (uri)
     {
-	struct cql_prop_entry *e;
-	
-	for (e = ct->entry; e; e = e->next)
-	    if (!memcmp(e->pattern, "set.", 4) && e->value &&
-		!strcmp(e->value, uri))
-	    {
-		prefix = e->pattern+4;
-		break;
-	    }
-	/* must have a prefix now - if not it's an error */
+        struct cql_prop_entry *e;
+        
+        for (e = ct->entry; e; e = e->next)
+            if (!memcmp(e->pattern, "set.", 4) && e->value &&
+                !strcmp(e->value, uri))
+            {
+                prefix = e->pattern+4;
+                break;
+            }
+        /* must have a prefix now - if not it's an error */
     }
 
     if (!uri || prefix)
     {
-	if (!res)
-	    res = cql_lookup_property(ct, category, prefix, eval);
-	if (!res)
-	    res = cql_lookup_property(ct, category, prefix, "*");
+        if (!res)
+            res = cql_lookup_property(ct, category, prefix, eval);
+        if (!res)
+            res = cql_lookup_property(ct, category, prefix, "*");
     }
     if (res)
     {
@@ -190,25 +214,112 @@ int cql_pr_attr_uri(cql_transform_t ct, const char *category,
     if (errcode && !ct->error)
     {
         ct->error = errcode;
-	if (val)
-	    ct->addinfo = xstrdup(val);
-	else
-	    ct->addinfo = 0;
+        if (val)
+            ct->addinfo = xstrdup(val);
+        else
+            ct->addinfo = 0;
     }
     return 0;
 }
 
 int cql_pr_attr(cql_transform_t ct, const char *category,
-		const char *val, const char *default_val,
-		void (*pr)(const char *buf, void *client_data),
-		void *client_data,
-		int errcode)
+                const char *val, const char *default_val,
+                void (*pr)(const char *buf, void *client_data),
+                void *client_data,
+                int errcode)
 {
     return cql_pr_attr_uri(ct, category, 0 /* uri */,
-			   val, default_val, pr, client_data, errcode);
+                           val, default_val, pr, client_data, errcode);
+}
+
+
+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 != 0) {
+        char *name = mods->u.st.index;
+        char *term = mods->u.st.term;
+        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 = 40; /* Unsupported proximity 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 = 42; /* Unsupported proximity unit */
+                ct->addinfo = xstrdup(term);
+                return 0;
+            }
+        } else {
+            ct->error = 46;     /* Unsupported 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().
@@ -220,9 +331,9 @@ static const char *wcchar(const char *term, int length)
     char *whichp;
 
     for (whichp = "*?"; *whichp != '\0'; whichp++) {
-	current = (const char *) memchr(term, *whichp, length);
-	if (current != 0 && (best == 0 || current < best))
-	    best = current;
+        current = (const char *) memchr(term, *whichp, length);
+        if (current != 0 && (best == 0 || current < best))
+            best = current;
     }
 
     return best;
@@ -230,11 +341,16 @@ static const char *wcchar(const char *term, int length)
 
 
 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;
+    const char *ns = cn->u.st.index_uri;
+
+    assert(cn->which == CQL_NODE_ST);
+
     if (length > 0)
     {
         if (length > 1 && term[0] == '^' && term[length-1] == '^')
@@ -249,7 +365,7 @@ void emit_term(cql_transform_t ct,
             cql_pr_attr(ct, "position", "first", 0,
                         pr, client_data, 32);
             term++;
-	    length--;
+            length--;
         }
         else if (term[length-1] == '^')
         {
@@ -266,70 +382,95 @@ void emit_term(cql_transform_t ct,
 
     if (length > 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.
-	 */
+        /* 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 &&
+            wcchar(term+1, length-2) == 0 &&
             cql_pr_attr(ct, "truncation", "both", 0,
-			pr, client_data, 0)) {
-	    term++;
-	    length -= 2;
+                        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)) {
-	    term++;
-	    length--;
+                 wcchar(term+1, length-1) == 0 &&
+                 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)) {
-	    length--;
+                 wcchar(term, length-1) == 0 &&
+                 cql_pr_attr(ct, "truncation", "right", 0,
+                             pr, client_data, 0)) {
+            length--;
         }
         else if (wcchar(term, length))
         {
-	    /* We have one or more wildcard characters, but not in a
-	     * way that can be dealt with using only the standard
-	     * left-, right- and both-truncation attributes.  We need
-	     * to translate the pattern into a Z39.58-type pattern,
-	     * which has been supported in BIB-1 since 1996.  If
-	     * there's no configuration element for "truncation.z3958"
-	     * we indicate this as error 28 "Masking character not
-	     * supported".
-	     */
-	    int i;
-	    char *mem;
+            /* We have one or more wildcard characters, but not in a
+             * way that can be dealt with using only the standard
+             * left-, right- and both-truncation attributes.  We need
+             * to translate the pattern into a Z39.58-type pattern,
+             * which has been supported in BIB-1 since 1996.  If
+             * there's no configuration element for "truncation.z3958"
+             * we indicate this as error 28 "Masking character not
+             * supported".
+             */
+            int i;
+            char *mem;
             cql_pr_attr(ct, "truncation", "z3958", 0,
                         pr, client_data, 28);
-	    mem = (char *) xmalloc(length+1);
+            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];
-	    }
-	    mem[length] = '\0';
-	    term = mem;
+                if (term[i] == '*')      mem[i] = '?';
+                else if (term[i] == '?') mem[i] = '#';
+                else                     mem[i] = term[i];
+            }
+            mem[length] = '\0';
+            term = mem;
         }
         else {
-	    /* No masking characters.  Use "truncation.none" if given. */
+            /* No masking characters.  Use "truncation.none" if given. */
             cql_pr_attr(ct, "truncation", "none", 0,
                         pr, client_data, 0);
         }
     }
+    if (ns) {
+        cql_pr_attr_uri(ct, "index", ns,
+                        cn->u.st.index, "serverChoice",
+                        pr, client_data, 16);
+    }
+    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);
+        }
+    }
 
     (*pr)("\"", client_data);
     for (i = 0; i<length; i++)
     {
-        char buf[2];
-        buf[0] = term[i];
-        buf[1] = 0;
-        (*pr)(buf, client_data);
+        /* pr(int) each character */
+        char buf[3];
+        const char *cp;
+
+        buf[1] = term[i];
+        buf[2] = 0;
+        /* do we have to escape this char? */
+        if (buf[1] == '"')
+        {
+            buf[0] = '\\';
+            cp = buf;
+        }
+        else
+            cp = buf+1;
+        (*pr)(cp, client_data);
     }
     (*pr)("\" ", client_data);
 }
@@ -354,7 +495,7 @@ 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 +505,7 @@ 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,
@@ -373,69 +514,59 @@ void cql_transform_r(cql_transform_t ct,
                      void *client_data)
 {
     const char *ns;
+    struct cql_node *mods;
 
     if (!cn)
         return;
     switch (cn->which)
     {
     case CQL_NODE_ST:
-	ns = cn->u.st.index_uri;
+        ns = cn->u.st.index_uri;
         if (ns)
         {
             if (!strcmp(ns, cql_uri())
-		&& cn->u.st.index && !strcmp(cn->u.st.index, "resultSet"))
+                && cn->u.st.index && !cql_strcmp(cn->u.st.index, "resultSet"))
             {
                 (*pr)("@set \"", client_data);
                 (*pr)(cn->u.st.term, client_data);
                 (*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->addinfo = 0;
-	    }
-	}
-        if (cn->u.st.relation && !strcmp(cn->u.st.relation, "="))
+        else
+        {
+            if (!ct->error)
+            {
+                ct->error = 15;
+                ct->addinfo = 0;
+            }
+        }
+        cql_pr_attr(ct, "always", 0, 0, pr, client_data, 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 && !strcmp(cn->u.st.relation, "<="))
+        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 && !strcmp(cn->u.st.relation, ">="))
+        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.term, 0,
-                            pr, client_data, 20);
-            }
-        }
         cql_pr_attr(ct, "structure", cn->u.st.relation, 0,
                     pr, client_data, 24);
-        if (cn->u.st.relation && !strcmp(cn->u.st.relation, "all"))
+        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 && !strcmp(cn->u.st.relation, "any"))
+        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),
+            emit_term(ct, cn, cn->u.st.term, strlen(cn->u.st.term),
                       pr, client_data);
         }
         break;
@@ -443,9 +574,24 @@ void cql_transform_r(cql_transform_t ct,
         (*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 = 46; /* SRW diag: "Unsupported 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;
+
+    default:
+        fprintf(stderr, "Fatal: impossible CQL node-type %d\n", cn->which);
+        abort();
     }
 }
 
@@ -464,10 +610,10 @@ int cql_transform(cql_transform_t ct,
 
     for (e = ct->entry; e ; e = e->next)
     {
-        if (!memcmp(e->pattern, "set.", 4))
-	    cql_apply_prefix(nmem, cn, e->pattern+4, e->value);
-        else if (!strcmp(e->pattern, "set"))
-	    cql_apply_prefix(nmem, cn, 0, e->value);
+        if (!cql_strncmp(e->pattern, "set.", 4))
+            cql_apply_prefix(nmem, cn, e->pattern+4, e->value);
+        else if (!cql_strcmp(e->pattern, "set"))
+            cql_apply_prefix(nmem, cn, 0, e->value);
     }
     cql_transform_r (ct, cn, pr, client_data);
     nmem_destroy(nmem);
@@ -490,6 +636,16 @@ 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;
@@ -500,3 +656,11 @@ int cql_transform_error(cql_transform_t ct, const char **addinfo)
     *addinfo = ct->addinfo;
     return ct->error;
 }
+/*
+ * Local variables:
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ * vim: shiftwidth=4 tabstop=8 expandtab
+ */
+