X-Git-Url: http://git.indexdata.com/?p=yaz-moved-to-github.git;a=blobdiff_plain;f=src%2Fcqltransform.c;h=3f54faa53d3de4f96a32338cf6c17bb3e1f80693;hp=a203f61add8ed24993fc05dc5607024441f0b43a;hb=3b792849c676d96918850e2f1e0af87a75278501;hpb=ccb96652ac7afe10d0cd7db9f0e348001c6b3f82

diff --git a/src/cqltransform.c b/src/cqltransform.c
index a203f61..3f54faa 100644
--- a/src/cqltransform.c
+++ b/src/cqltransform.c
@@ -1,5 +1,5 @@
-/* $Id: cqltransform.c,v 1.17 2005-12-21 00:04:14 mike Exp $
-   Copyright (C) 1995-2005, Index Data ApS
+/* $Id: cqltransform.c,v 1.29 2007-10-31 21:58:07 adam Exp $
+   Copyright (C) 1995-2007, Index Data ApS
    Index Data Aps
 
 This file is part of the YAZ toolkit.
@@ -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;
@@ -171,6 +195,7 @@ 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))
@@ -178,7 +203,19 @@ int cql_pr_attr_uri(cql_transform_t ct, const char *category,
             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 == ' ')
@@ -209,33 +246,138 @@ 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 != 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().
  */
-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,
+               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] == '^')
         {
@@ -264,35 +406,40 @@ void emit_term(cql_transform_t ct,
         }
     }
 
-    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 +451,22 @@ 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];
+            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. */
@@ -322,16 +474,42 @@ 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, 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);
+    xfree(z3958_mem);
 }
 
 void emit_wordlist(cql_transform_t ct,
@@ -354,7 +532,7 @@ 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 +542,7 @@ 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 +551,7 @@ void cql_transform_r(cql_transform_t ct,
                      void *client_data)
 {
     const char *ns;
+    struct cql_node *mods;
 
     if (!cn)
         return;
@@ -390,9 +569,6 @@ 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
         {
@@ -402,6 +578,7 @@ void cql_transform_r(cql_transform_t ct,
                 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);
@@ -414,15 +591,6 @@ void cql_transform_r(cql_transform_t ct,
         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, "structure", cn->u.st.relation, 0,
                     pr, client_data, 24);
         if (cn->u.st.relation && !cql_strcmp(cn->u.st.relation, "all"))
@@ -435,7 +603,7 @@ void cql_transform_r(cql_transform_t ct,
         }
         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 +611,20 @@ 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);
@@ -494,6 +673,16 @@ 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;