-/* $Id: cqltransform.c,v 1.17 2005-12-21 00:04:14 mike 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
/**
* \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;
{
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;
}
+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().
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] == '^')
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);
}
(*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)
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,
void *client_data)
{
const char *ns;
+ struct cql_node *mods;
if (!cn)
return;
(*pr)("\" ", client_data);
return ;
}
- cql_pr_attr_uri(ct, "index", ns,
- cn->u.st.index, "serverChoice",
- pr, client_data, 16);
}
else
{
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
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"))
}
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;
(*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);
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;