summaryrefslogtreecommitdiff
path: root/src/object.c
diff options
context:
space:
mode:
authorantirez <antirez@gmail.com>2013-07-12 11:56:52 +0200
committerantirez <antirez@gmail.com>2013-07-12 11:56:52 +0200
commit81e55ec0f399d5bb2b75e5b2ffa5a62d60a31bc2 (patch)
treec44f65dd6be94e6d27fe16ac28b1084fccc9e187 /src/object.c
parent2ed64e1f7a1724ee57e92a1a502f32134ac4d99a (diff)
downloadredis-81e55ec0f399d5bb2b75e5b2ffa5a62d60a31bc2.tar.gz
Fixed compareStringObject() and introduced collateStringObject().
compareStringObject was not always giving the same result when comparing two exact strings, but encoded as integers or as sds strings, since it switched to strcmp() when at least one of the strings were not sds encoded. For instance the two strings "123" and "123\x00456", where the first string was integer encoded, would result into the old implementation of compareStringObject() to return 0 as if the strings were equal, while instead the second string is "greater" than the first in a binary comparison. The same compasion, but with "123" encoded as sds string, would instead return a value < 0, as it is correct. It is not impossible that the above caused some obscure bug, since the comparison was not always deterministic, and compareStringObject() is used in the implementation of skiplists, hash tables, and so forth. At the same time, collateStringObject() was introduced by this commit, so that can be used by SORT command to return sorted strings usign collation instead of binary comparison. See next commit.
Diffstat (limited to 'src/object.c')
-rw-r--r--src/object.c41
1 files changed, 33 insertions, 8 deletions
diff --git a/src/object.c b/src/object.c
index 2554656a3..58668da5b 100644
--- a/src/object.c
+++ b/src/object.c
@@ -332,35 +332,60 @@ robj *getDecodedObject(robj *o) {
}
}
-/* Compare two string objects via strcmp() or alike.
+/* Compare two string objects via strcmp() or strcoll() depending on flags.
* Note that the objects may be integer-encoded. In such a case we
* use ll2string() to get a string representation of the numbers on the stack
* and compare the strings, it's much faster than calling getDecodedObject().
*
- * Important note: if objects are not integer encoded, but binary-safe strings,
- * sdscmp() from sds.c will apply memcmp() so this function ca be considered
- * binary safe. */
-int compareStringObjects(robj *a, robj *b) {
+ * Important note: when REDIS_COMPARE_BINARY is used a binary-safe comparison
+ * is used. */
+
+#define REDIS_COMPARE_BINARY (1<<0)
+#define REDIS_COMPARE_COLL (1<<1)
+
+int compareStringObjectsWithFlags(robj *a, robj *b, int flags) {
redisAssertWithInfo(NULL,a,a->type == REDIS_STRING && b->type == REDIS_STRING);
char bufa[128], bufb[128], *astr, *bstr;
+ size_t alen, blen, minlen;
int bothsds = 1;
if (a == b) return 0;
if (a->encoding != REDIS_ENCODING_RAW) {
- ll2string(bufa,sizeof(bufa),(long) a->ptr);
+ alen = ll2string(bufa,sizeof(bufa),(long) a->ptr);
astr = bufa;
bothsds = 0;
} else {
astr = a->ptr;
+ alen = sdslen(astr);
}
if (b->encoding != REDIS_ENCODING_RAW) {
- ll2string(bufb,sizeof(bufb),(long) b->ptr);
+ blen = ll2string(bufb,sizeof(bufb),(long) b->ptr);
bstr = bufb;
bothsds = 0;
} else {
bstr = b->ptr;
+ blen = sdslen(bstr);
+ }
+ if (flags & REDIS_COMPARE_COLL) {
+ return strcoll(astr,bstr);
+ } else {
+ int cmp;
+
+ minlen = (alen < blen) ? alen : blen;
+ cmp = memcmp(astr,bstr,minlen);
+ if (cmp == 0) return alen-blen;
+ return cmp;
}
- return bothsds ? sdscmp(astr,bstr) : strcmp(astr,bstr);
+}
+
+/* Wrapper for compareStringObjectsWithFlags() using binary comparison. */
+int compareStringObjects(robj *a, robj *b) {
+ return compareStringObjectsWithFlags(a,b,REDIS_COMPARE_BINARY);
+}
+
+/* Wrapper for compareStringObjectsWithFlags() using collation. */
+int collateStringObjects(robj *a, robj *b) {
+ return compareStringObjectsWithFlags(a,b,REDIS_COMPARE_COLL);
}
/* Equal string objects return 1 if the two objects are the same from the