* POSSIBILITY OF SUCH DAMAGE.
*/
-#define REDIS_VERSION "1.3.13"
+#define REDIS_VERSION "2.1.1"
#include "fmacros.h"
#include "config.h"
#include "lzf.h" /* LZF compression library */
#include "pqsort.h" /* Partial qsort for SORT+LIMIT */
#include "zipmap.h" /* Compact dictionary-alike data structure */
+#include "ziplist.h" /* Compact list data structure */
#include "sha1.h" /* SHA1 is used for DEBUG DIGEST */
#include "release.h" /* Release and/or git repository information */
/* Objects encoding. Some kind of objects like Strings and Hashes can be
* internally represented in multiple ways. The 'encoding' field of the object
* is set to one of this fields for this object. */
-#define REDIS_ENCODING_RAW 0 /* Raw representation */
-#define REDIS_ENCODING_INT 1 /* Encoded as integer */
-#define REDIS_ENCODING_ZIPMAP 2 /* Encoded as zipmap */
-#define REDIS_ENCODING_HT 3 /* Encoded as an hash table */
+#define REDIS_ENCODING_RAW 0 /* Raw representation */
+#define REDIS_ENCODING_INT 1 /* Encoded as integer */
+#define REDIS_ENCODING_HT 2 /* Encoded as hash table */
+#define REDIS_ENCODING_ZIPMAP 3 /* Encoded as zipmap */
+#define REDIS_ENCODING_LIST 4 /* Encoded as zipmap */
+#define REDIS_ENCODING_ZIPLIST 5 /* Encoded as ziplist */
static char* strencoding[] = {
"raw", "int", "zipmap", "hashtable"
#define REDIS_MULTI 8 /* This client is in a MULTI context */
#define REDIS_BLOCKED 16 /* The client is waiting in a blocking operation */
#define REDIS_IO_WAIT 32 /* The client is waiting for Virtual Memory I/O */
+#define REDIS_DIRTY_CAS 64 /* Watched keys modified. EXEC will fail. */
/* Slave replication state - slave side */
#define REDIS_REPL_NONE 0 /* No active replication */
typedef struct redisDb {
dict *dict; /* The keyspace for this DB */
dict *expires; /* Timeout of keys with a timeout set */
- dict *blockingkeys; /* Keys with clients waiting for data (BLPOP) */
+ dict *blocking_keys; /* Keys with clients waiting for data (BLPOP) */
dict *io_keys; /* Keys with clients waiting for VM I/O */
+ dict *watched_keys; /* WATCHED keys for MULTI/EXEC CAS */
int id;
} redisDb;
long repldboff; /* replication DB file offset */
off_t repldbsize; /* replication DB file size */
multiState mstate; /* MULTI/EXEC state */
- robj **blockingkeys; /* The key we are waiting to terminate a blocking
+ robj **blocking_keys; /* The key we are waiting to terminate a blocking
* operation such as BLPOP. Otherwise NULL. */
- int blockingkeysnum; /* Number of blocking keys */
+ int blocking_keys_num; /* Number of blocking keys */
time_t blockingto; /* Blocking operation timeout. If UNIX current time
* is >= blockingto then the operation timed out. */
list *io_keys; /* Keys this client is waiting to be loaded from the
* swap file in order to continue. */
+ list *watched_keys; /* Keys WATCHED for MULTI/EXEC CAS */
dict *pubsub_channels; /* channels a client is interested in (SUBSCRIBE) */
list *pubsub_patterns; /* patterns a client is interested in (SUBSCRIBE) */
} redisClient;
static int rewriteAppendOnlyFileBackground(void);
static int vmSwapObjectBlocking(robj *key, robj *val);
static int prepareForShutdown();
+static void touchWatchedKey(redisDb *db, robj *key);
+static void touchWatchedKeysOnFlush(int dbid);
+static void unwatchAllKeys(redisClient *c);
static void authCommand(redisClient *c);
static void pingCommand(redisClient *c);
static void psubscribeCommand(redisClient *c);
static void punsubscribeCommand(redisClient *c);
static void publishCommand(redisClient *c);
+static void watchCommand(redisClient *c);
+static void unwatchCommand(redisClient *c);
/*================================= Globals ================================= */
/* Global vars */
static struct redisServer server; /* server global state */
-static struct redisCommand cmdTable[] = {
+static struct redisCommand *commandTable;
+static struct redisCommand readonlyCommandTable[] = {
{"get",getCommand,2,REDIS_CMD_INLINE,NULL,1,1,1},
{"set",setCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,0,0,0},
{"setnx",setnxCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,0,0,0},
{"psubscribe",psubscribeCommand,-2,REDIS_CMD_INLINE,NULL,0,0,0},
{"punsubscribe",punsubscribeCommand,-1,REDIS_CMD_INLINE,NULL,0,0,0},
{"publish",publishCommand,3,REDIS_CMD_BULK|REDIS_CMD_FORCE_REPLICATION,NULL,0,0,0},
- {NULL,NULL,0,0,NULL,0,0,0}
+ {"watch",watchCommand,-2,REDIS_CMD_INLINE,NULL,0,0,0},
+ {"unwatch",unwatchCommand,1,REDIS_CMD_INLINE,NULL,0,0,0}
};
/*============================ Utility functions ============================ */
for (j = 0; j < server.dbnum; j++) {
server.db[j].dict = dictCreate(&dbDictType,NULL);
server.db[j].expires = dictCreate(&keyptrDictType,NULL);
- server.db[j].blockingkeys = dictCreate(&keylistDictType,NULL);
+ server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
+ server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
if (server.vm_enabled)
server.db[j].io_keys = dictCreate(&keylistDictType,NULL);
server.db[j].id = j;
if (c->flags & REDIS_BLOCKED)
unblockClientWaitingData(c);
+ /* UNWATCH all the keys */
+ unwatchAllKeys(c);
+ listRelease(c->watched_keys);
/* Unsubscribe from all the pubsub channels */
pubsubUnsubscribeAllChannels(c,0);
pubsubUnsubscribeAllPatterns(c,0);
ln = listSearchKey(server.clients,c);
redisAssert(ln != NULL);
listDelNode(server.clients,ln);
- /* Remove from the list of clients waiting for swapped keys */
+ /* Remove from the list of clients that are now ready to be restarted
+ * after waiting for swapped keys */
if (c->flags & REDIS_IO_WAIT && listLength(c->io_keys) == 0) {
ln = listSearchKey(server.io_ready_clients,c);
if (ln) {
server.vm_blocked_clients--;
}
}
+ /* Remove from the list of clients waiting for swapped keys */
while (server.vm_enabled && listLength(c->io_keys)) {
ln = listFirst(c->io_keys);
dontWaitForSwappedKey(c,ln->value);
}
}
+static int qsortRedisCommands(const void *r1, const void *r2) {
+ return strcasecmp(
+ ((struct redisCommand*)r1)->name,
+ ((struct redisCommand*)r2)->name);
+}
+
+static void sortCommandTable() {
+ /* Copy and sort the read-only version of the command table */
+ commandTable = (struct redisCommand*)malloc(sizeof(readonlyCommandTable));
+ memcpy(commandTable,readonlyCommandTable,sizeof(readonlyCommandTable));
+ qsort(commandTable,
+ sizeof(readonlyCommandTable)/sizeof(struct redisCommand),
+ sizeof(struct redisCommand),qsortRedisCommands);
+}
+
static struct redisCommand *lookupCommand(char *name) {
- int j = 0;
- while(cmdTable[j].name != NULL) {
- if (!strcasecmp(name,cmdTable[j].name)) return &cmdTable[j];
- j++;
- }
- return NULL;
+ struct redisCommand tmp = {name,NULL,0,0,NULL,0,0,0};
+ return bsearch(
+ &tmp,
+ commandTable,
+ sizeof(readonlyCommandTable)/sizeof(struct redisCommand),
+ sizeof(struct redisCommand),
+ qsortRedisCommands);
}
/* resetClient prepare the client to process the next command */
}
/* Exec the command */
- if (c->flags & REDIS_MULTI && cmd->proc != execCommand && cmd->proc != discardCommand) {
+ if (c->flags & REDIS_MULTI &&
+ cmd->proc != execCommand && cmd->proc != discardCommand &&
+ cmd->proc != multiCommand && cmd->proc != watchCommand)
+ {
queueMultiCommand(c,cmd);
addReply(c,shared.queued);
} else {
c->reply = listCreate();
listSetFreeMethod(c->reply,decrRefCount);
listSetDupMethod(c->reply,dupClientReplyValue);
- c->blockingkeys = NULL;
- c->blockingkeysnum = 0;
+ c->blocking_keys = NULL;
+ c->blocking_keys_num = 0;
c->io_keys = listCreate();
+ c->watched_keys = listCreate();
listSetFreeMethod(c->io_keys,decrRefCount);
c->pubsub_channels = dictCreate(&setDictType,NULL);
c->pubsub_patterns = listCreate();
incrRefCount(shared.integers[value]);
o = shared.integers[value];
} else {
- o = createObject(REDIS_STRING, NULL);
if (value >= LONG_MIN && value <= LONG_MAX) {
+ o = createObject(REDIS_STRING, NULL);
o->encoding = REDIS_ENCODING_INT;
o->ptr = (void*)((long)value);
} else {
static robj *createListObject(void) {
list *l = listCreate();
-
+ robj *o = createObject(REDIS_LIST,l);
listSetFreeMethod(l,decrRefCount);
- return createObject(REDIS_LIST,l);
+ o->encoding = REDIS_ENCODING_LIST;
+ return o;
+}
+
+static robj *createZiplistObject(void) {
+ unsigned char *zl = ziplistNew();
+ robj *o = createObject(REDIS_LIST,zl);
+ o->encoding = REDIS_ENCODING_ZIPLIST;
+ return o;
}
static robj *createSetObject(void) {
}
static void freeListObject(robj *o) {
- listRelease((list*) o->ptr);
+ switch (o->encoding) {
+ case REDIS_ENCODING_LIST:
+ listRelease((list*) o->ptr);
+ break;
+ case REDIS_ENCODING_ZIPLIST:
+ zfree(o->ptr);
+ break;
+ default:
+ redisPanic("Unknown list encoding type");
+ }
}
static void freeSetObject(robj *o) {
static robj *lookupKeyWrite(redisDb *db, robj *key) {
deleteIfVolatile(db,key);
+ touchWatchedKey(db,key);
return lookupKey(db,key);
}
return 0;
}
+/* Save a long long value as either an encoded string or a string. */
+static int rdbSaveLongLongAsStringObject(FILE *fp, long long value) {
+ unsigned char buf[32];
+ int enclen = rdbEncodeInteger(value,buf);
+ if (enclen > 0) {
+ if (fwrite(buf,enclen,1,fp) == 0) return -1;
+ } else {
+ /* Encode as string */
+ enclen = ll2string((char*)buf,32,value);
+ redisAssert(enclen < 32);
+ if (rdbSaveLen(fp,enclen) == -1) return -1;
+ if (fwrite(buf,enclen,1,fp) == 0) return -1;
+ }
+ return 0;
+}
+
/* Like rdbSaveStringObjectRaw() but handle encoded objects */
static int rdbSaveStringObject(FILE *fp, robj *obj) {
- int retval;
-
/* Avoid to decode the object, then encode it again, if the
* object is alrady integer encoded. */
if (obj->encoding == REDIS_ENCODING_INT) {
- long val = (long) obj->ptr;
- unsigned char buf[5];
- int enclen;
-
- if ((enclen = rdbEncodeInteger(val,buf)) > 0) {
- if (fwrite(buf,enclen,1,fp) == 0) return -1;
- return 0;
- }
- /* otherwise... fall throught and continue with the usual
- * code path. */
- }
-
- /* Avoid incr/decr ref count business when possible.
- * This plays well with copy-on-write given that we are probably
- * in a child process (BGSAVE). Also this makes sure key objects
- * of swapped objects are not incRefCount-ed (an assert does not allow
- * this in order to avoid bugs) */
- if (obj->encoding != REDIS_ENCODING_RAW) {
- obj = getDecodedObject(obj);
- retval = rdbSaveRawString(fp,obj->ptr,sdslen(obj->ptr));
- decrRefCount(obj);
+ return rdbSaveLongLongAsStringObject(fp,(long)obj->ptr);
} else {
- retval = rdbSaveRawString(fp,obj->ptr,sdslen(obj->ptr));
+ redisAssert(obj->encoding == REDIS_ENCODING_RAW);
+ return rdbSaveRawString(fp,obj->ptr,sdslen(obj->ptr));
}
- return retval;
}
/* Save a double value. Doubles are saved as strings prefixed by an unsigned
if (rdbSaveStringObject(fp,o) == -1) return -1;
} else if (o->type == REDIS_LIST) {
/* Save a list value */
- list *list = o->ptr;
- listIter li;
- listNode *ln;
-
- if (rdbSaveLen(fp,listLength(list)) == -1) return -1;
- listRewind(list,&li);
- while((ln = listNext(&li))) {
- robj *eleobj = listNodeValue(ln);
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *p;
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vlong;
+
+ if (rdbSaveLen(fp,ziplistLen(o->ptr)) == -1) return -1;
+ p = ziplistIndex(o->ptr,0);
+ while(ziplistGet(p,&vstr,&vlen,&vlong)) {
+ if (vstr) {
+ if (rdbSaveRawString(fp,vstr,vlen) == -1)
+ return -1;
+ } else {
+ if (rdbSaveLongLongAsStringObject(fp,vlong) == -1)
+ return -1;
+ }
+ p = ziplistNext(o->ptr,p);
+ }
+ } else if (o->encoding == REDIS_ENCODING_LIST) {
+ list *list = o->ptr;
+ listIter li;
+ listNode *ln;
- if (rdbSaveStringObject(fp,eleobj) == -1) return -1;
+ if (rdbSaveLen(fp,listLength(list)) == -1) return -1;
+ listRewind(list,&li);
+ while((ln = listNext(&li))) {
+ robj *eleobj = listNodeValue(ln);
+ if (rdbSaveStringObject(fp,eleobj) == -1) return -1;
+ }
+ } else {
+ redisPanic("Unknown list encoding");
}
} else if (o->type == REDIS_SET) {
/* Save a set value */
/* Load a Redis object of the specified type from the specified file.
* On success a newly allocated object is returned, otherwise NULL. */
static robj *rdbLoadObject(int type, FILE *fp) {
- robj *o;
+ robj *o, *ele, *dec;
+ size_t len;
redisLog(REDIS_DEBUG,"LOADING OBJECT %d (at %d)\n",type,ftell(fp));
if (type == REDIS_STRING) {
/* Read string value */
if ((o = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
o = tryObjectEncoding(o);
- } else if (type == REDIS_LIST || type == REDIS_SET) {
- /* Read list/set value */
- uint32_t listlen;
+ } else if (type == REDIS_LIST) {
+ /* Read list value */
+ if ((len = rdbLoadLen(fp,NULL)) == REDIS_RDB_LENERR) return NULL;
+
+ o = createZiplistObject();
+
+ /* Load every single element of the list */
+ while(len--) {
+ if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
- if ((listlen = rdbLoadLen(fp,NULL)) == REDIS_RDB_LENERR) return NULL;
- o = (type == REDIS_LIST) ? createListObject() : createSetObject();
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ dec = getDecodedObject(ele);
+ o->ptr = ziplistPush(o->ptr,dec->ptr,sdslen(dec->ptr),REDIS_TAIL);
+ decrRefCount(dec);
+ decrRefCount(ele);
+ } else {
+ ele = tryObjectEncoding(ele);
+ listAddNodeTail(o->ptr,ele);
+ incrRefCount(ele);
+ }
+ }
+ } else if (type == REDIS_SET) {
+ /* Read list/set value */
+ if ((len = rdbLoadLen(fp,NULL)) == REDIS_RDB_LENERR) return NULL;
+ o = createSetObject();
/* It's faster to expand the dict to the right size asap in order
* to avoid rehashing */
- if (type == REDIS_SET && listlen > DICT_HT_INITIAL_SIZE)
- dictExpand(o->ptr,listlen);
+ if (len > DICT_HT_INITIAL_SIZE)
+ dictExpand(o->ptr,len);
/* Load every single element of the list/set */
- while(listlen--) {
- robj *ele;
-
+ while(len--) {
if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
ele = tryObjectEncoding(ele);
- if (type == REDIS_LIST) {
- listAddNodeTail((list*)o->ptr,ele);
- } else {
- dictAdd((dict*)o->ptr,ele,NULL);
- }
+ dictAdd((dict*)o->ptr,ele,NULL);
}
} else if (type == REDIS_ZSET) {
/* Read list/set value */
}
}
+ touchWatchedKey(c->db,key);
if (nx) deleteIfVolatile(c->db,key);
retval = dictAdd(c->db->dict,key,val);
if (retval == DICT_ERR) {
for (j = 1; j < c->argc; j++) {
if (deleteKey(c->db,c->argv[j])) {
+ touchWatchedKey(c->db,c->argv[j]);
server.dirty++;
deleted++;
}
incrRefCount(c->argv[2]);
}
deleteKey(c->db,c->argv[1]);
+ touchWatchedKey(c->db,c->argv[2]);
server.dirty++;
addReply(c,nx ? shared.cone : shared.ok);
}
}
/* =================================== Lists ================================ */
-static void pushGenericCommand(redisClient *c, int where) {
- robj *lobj;
- list *list;
+static void lPush(robj *subject, robj *value, int where) {
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST) {
+ int pos = (where == REDIS_HEAD) ? ZIPLIST_HEAD : ZIPLIST_TAIL;
+ value = getDecodedObject(value);
+ subject->ptr = ziplistPush(subject->ptr,value->ptr,sdslen(value->ptr),pos);
+ decrRefCount(value);
+ } else if (subject->encoding == REDIS_ENCODING_LIST) {
+ if (where == REDIS_HEAD) {
+ listAddNodeHead(subject->ptr,value);
+ } else {
+ listAddNodeTail(subject->ptr,value);
+ }
+ incrRefCount(value);
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+}
- lobj = lookupKeyWrite(c->db,c->argv[1]);
+static robj *lPop(robj *subject, int where) {
+ robj *value = NULL;
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *p;
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vlong;
+ int pos = (where == REDIS_HEAD) ? 0 : -1;
+ p = ziplistIndex(subject->ptr,pos);
+ if (ziplistGet(p,&vstr,&vlen,&vlong)) {
+ if (vstr) {
+ value = createStringObject((char*)vstr,vlen);
+ } else {
+ value = createStringObjectFromLongLong(vlong);
+ }
+ /* We only need to delete an element when it exists */
+ subject->ptr = ziplistDelete(subject->ptr,&p);
+ }
+ } else if (subject->encoding == REDIS_ENCODING_LIST) {
+ list *list = subject->ptr;
+ listNode *ln;
+ if (where == REDIS_HEAD) {
+ ln = listFirst(list);
+ } else {
+ ln = listLast(list);
+ }
+ if (ln != NULL) {
+ value = listNodeValue(ln);
+ incrRefCount(value);
+ listDelNode(list,ln);
+ }
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+ return value;
+}
+
+static unsigned long lLength(robj *subject) {
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST) {
+ return ziplistLen(subject->ptr);
+ } else if (subject->encoding == REDIS_ENCODING_LIST) {
+ return listLength((list*)subject->ptr);
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+}
+
+/* Structure to hold set iteration abstraction. */
+typedef struct {
+ robj *subject;
+ unsigned char encoding;
+ unsigned char direction; /* Iteration direction */
+ unsigned char *zi;
+ listNode *ln;
+} lIterator;
+
+/* Structure for an entry while iterating over a list. */
+typedef struct {
+ lIterator *li;
+ unsigned char *zi; /* Entry in ziplist */
+ listNode *ln; /* Entry in linked list */
+} lEntry;
+
+/* Initialize an iterator at the specified index. */
+static lIterator *lInitIterator(robj *subject, int index, unsigned char direction) {
+ lIterator *li = zmalloc(sizeof(lIterator));
+ li->subject = subject;
+ li->encoding = subject->encoding;
+ li->direction = direction;
+ if (li->encoding == REDIS_ENCODING_ZIPLIST) {
+ li->zi = ziplistIndex(subject->ptr,index);
+ } else if (li->encoding == REDIS_ENCODING_LIST) {
+ li->ln = listIndex(subject->ptr,index);
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+ return li;
+}
+
+/* Clean up the iterator. */
+static void lReleaseIterator(lIterator *li) {
+ zfree(li);
+}
+
+/* Stores pointer to current the entry in the provided entry structure
+ * and advances the position of the iterator. Returns 1 when the current
+ * entry is in fact an entry, 0 otherwise. */
+static int lNext(lIterator *li, lEntry *entry) {
+ entry->li = li;
+ if (li->encoding == REDIS_ENCODING_ZIPLIST) {
+ entry->zi = li->zi;
+ if (entry->zi != NULL) {
+ if (li->direction == REDIS_TAIL)
+ li->zi = ziplistNext(li->subject->ptr,li->zi);
+ else
+ li->zi = ziplistPrev(li->subject->ptr,li->zi);
+ return 1;
+ }
+ } else if (li->encoding == REDIS_ENCODING_LIST) {
+ entry->ln = li->ln;
+ if (entry->ln != NULL) {
+ if (li->direction == REDIS_TAIL)
+ li->ln = li->ln->next;
+ else
+ li->ln = li->ln->prev;
+ return 1;
+ }
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+ return 0;
+}
+
+/* Return entry or NULL at the current position of the iterator. */
+static robj *lGet(lEntry *entry) {
+ lIterator *li = entry->li;
+ robj *value = NULL;
+ if (li->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vlong;
+ redisAssert(entry->zi != NULL);
+ if (ziplistGet(entry->zi,&vstr,&vlen,&vlong)) {
+ if (vstr) {
+ value = createStringObject((char*)vstr,vlen);
+ } else {
+ value = createStringObjectFromLongLong(vlong);
+ }
+ }
+ } else if (li->encoding == REDIS_ENCODING_LIST) {
+ redisAssert(entry->ln != NULL);
+ value = listNodeValue(entry->ln);
+ incrRefCount(value);
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+ return value;
+}
+
+/* Compare the given object with the entry at the current position. */
+static int lEqual(lEntry *entry, robj *o) {
+ lIterator *li = entry->li;
+ if (li->encoding == REDIS_ENCODING_ZIPLIST) {
+ redisAssert(o->encoding == REDIS_ENCODING_RAW);
+ return ziplistCompare(entry->zi,o->ptr,sdslen(o->ptr));
+ } else if (li->encoding == REDIS_ENCODING_LIST) {
+ return equalStringObjects(o,listNodeValue(entry->ln));
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+}
+
+/* Delete the element pointed to. */
+static void lDelete(lEntry *entry) {
+ lIterator *li = entry->li;
+ if (li->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *p = entry->zi;
+ li->subject->ptr = ziplistDelete(li->subject->ptr,&p);
+
+ /* Update position of the iterator depending on the direction */
+ if (li->direction == REDIS_TAIL)
+ li->zi = p;
+ else
+ li->zi = ziplistPrev(li->subject->ptr,p);
+ } else if (entry->li->encoding == REDIS_ENCODING_LIST) {
+ listNode *next;
+ if (li->direction == REDIS_TAIL)
+ next = entry->ln->next;
+ else
+ next = entry->ln->prev;
+ listDelNode(li->subject->ptr,entry->ln);
+ li->ln = next;
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+}
+
+static void pushGenericCommand(redisClient *c, int where) {
+ robj *lobj = lookupKeyWrite(c->db,c->argv[1]);
if (lobj == NULL) {
if (handleClientsWaitingListPush(c,c->argv[1],c->argv[2])) {
addReply(c,shared.cone);
return;
}
- lobj = createListObject();
- list = lobj->ptr;
- if (where == REDIS_HEAD) {
- listAddNodeHead(list,c->argv[2]);
- } else {
- listAddNodeTail(list,c->argv[2]);
- }
+ lobj = createZiplistObject();
dictAdd(c->db->dict,c->argv[1],lobj);
incrRefCount(c->argv[1]);
- incrRefCount(c->argv[2]);
} else {
if (lobj->type != REDIS_LIST) {
addReply(c,shared.wrongtypeerr);
addReply(c,shared.cone);
return;
}
- list = lobj->ptr;
- if (where == REDIS_HEAD) {
- listAddNodeHead(list,c->argv[2]);
- } else {
- listAddNodeTail(list,c->argv[2]);
- }
- incrRefCount(c->argv[2]);
}
+ lPush(lobj,c->argv[2],where);
+ addReplyLongLong(c,lLength(lobj));
server.dirty++;
- addReplyLongLong(c,listLength(list));
}
static void lpushCommand(redisClient *c) {
}
static void llenCommand(redisClient *c) {
- robj *o;
- list *l;
-
- if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL ||
- checkType(c,o,REDIS_LIST)) return;
-
- l = o->ptr;
- addReplyUlong(c,listLength(l));
+ robj *o = lookupKeyReadOrReply(c,c->argv[1],shared.czero);
+ if (o == NULL || checkType(c,o,REDIS_LIST)) return;
+ addReplyUlong(c,lLength(o));
}
static void lindexCommand(redisClient *c) {
- robj *o;
+ robj *o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk);
+ if (o == NULL || checkType(c,o,REDIS_LIST)) return;
int index = atoi(c->argv[2]->ptr);
- list *list;
- listNode *ln;
-
- if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL ||
- checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
+ robj *value = NULL;
- ln = listIndex(list, index);
- if (ln == NULL) {
- addReply(c,shared.nullbulk);
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *p;
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vlong;
+ p = ziplistIndex(o->ptr,index);
+ if (ziplistGet(p,&vstr,&vlen,&vlong)) {
+ if (vstr) {
+ value = createStringObject((char*)vstr,vlen);
+ } else {
+ value = createStringObjectFromLongLong(vlong);
+ }
+ addReplyBulk(c,value);
+ decrRefCount(value);
+ } else {
+ addReply(c,shared.nullbulk);
+ }
+ } else if (o->encoding == REDIS_ENCODING_LIST) {
+ listNode *ln = listIndex(o->ptr,index);
+ if (ln != NULL) {
+ value = listNodeValue(ln);
+ addReplyBulk(c,value);
+ } else {
+ addReply(c,shared.nullbulk);
+ }
} else {
- robj *ele = listNodeValue(ln);
- addReplyBulk(c,ele);
+ redisPanic("Unknown list encoding");
}
}
static void lsetCommand(redisClient *c) {
- robj *o;
+ robj *o = lookupKeyWriteOrReply(c,c->argv[1],shared.nokeyerr);
+ if (o == NULL || checkType(c,o,REDIS_LIST)) return;
int index = atoi(c->argv[2]->ptr);
- list *list;
- listNode *ln;
-
- if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.nokeyerr)) == NULL ||
- checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
+ robj *value = c->argv[3];
- ln = listIndex(list, index);
- if (ln == NULL) {
- addReply(c,shared.outofrangeerr);
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *p, *zl = o->ptr;
+ p = ziplistIndex(zl,index);
+ if (p == NULL) {
+ addReply(c,shared.outofrangeerr);
+ } else {
+ o->ptr = ziplistDelete(o->ptr,&p);
+ value = getDecodedObject(value);
+ o->ptr = ziplistInsert(o->ptr,p,value->ptr,sdslen(value->ptr));
+ decrRefCount(value);
+ addReply(c,shared.ok);
+ server.dirty++;
+ }
+ } else if (o->encoding == REDIS_ENCODING_LIST) {
+ listNode *ln = listIndex(o->ptr,index);
+ if (ln == NULL) {
+ addReply(c,shared.outofrangeerr);
+ } else {
+ decrRefCount((robj*)listNodeValue(ln));
+ listNodeValue(ln) = value;
+ incrRefCount(value);
+ addReply(c,shared.ok);
+ server.dirty++;
+ }
} else {
- robj *ele = listNodeValue(ln);
-
- decrRefCount(ele);
- listNodeValue(ln) = c->argv[3];
- incrRefCount(c->argv[3]);
- addReply(c,shared.ok);
- server.dirty++;
+ redisPanic("Unknown list encoding");
}
}
static void popGenericCommand(redisClient *c, int where) {
- robj *o;
- list *list;
- listNode *ln;
-
- if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.nullbulk)) == NULL ||
- checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
+ robj *o = lookupKeyWriteOrReply(c,c->argv[1],shared.nullbulk);
+ if (o == NULL || checkType(c,o,REDIS_LIST)) return;
- if (where == REDIS_HEAD)
- ln = listFirst(list);
- else
- ln = listLast(list);
-
- if (ln == NULL) {
+ robj *value = lPop(o,where);
+ if (value == NULL) {
addReply(c,shared.nullbulk);
} else {
- robj *ele = listNodeValue(ln);
- addReplyBulk(c,ele);
- listDelNode(list,ln);
- if (listLength(list) == 0) deleteKey(c->db,c->argv[1]);
+ addReplyBulk(c,value);
+ decrRefCount(value);
+ if (lLength(o) == 0) deleteKey(c->db,c->argv[1]);
server.dirty++;
}
}
}
static void lrangeCommand(redisClient *c) {
- robj *o;
+ robj *o, *value;
int start = atoi(c->argv[2]->ptr);
int end = atoi(c->argv[3]->ptr);
int llen;
int rangelen, j;
- list *list;
- listNode *ln;
- robj *ele;
+ lEntry entry;
if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptymultibulk)) == NULL
|| checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
- llen = listLength(list);
+ llen = lLength(o);
/* convert negative indexes */
if (start < 0) start = llen+start;
rangelen = (end-start)+1;
/* Return the result in form of a multi-bulk reply */
- ln = listIndex(list, start);
addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",rangelen));
+ lIterator *li = lInitIterator(o,start,REDIS_TAIL);
for (j = 0; j < rangelen; j++) {
- ele = listNodeValue(ln);
- addReplyBulk(c,ele);
- ln = ln->next;
+ redisAssert(lNext(li,&entry));
+ value = lGet(&entry);
+ addReplyBulk(c,value);
+ decrRefCount(value);
}
+ lReleaseIterator(li);
}
static void ltrimCommand(redisClient *c) {
if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.ok)) == NULL ||
checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
- llen = listLength(list);
+ llen = lLength(o);
/* convert negative indexes */
if (start < 0) start = llen+start;
}
/* Remove list elements to perform the trim */
- for (j = 0; j < ltrim; j++) {
- ln = listFirst(list);
- listDelNode(list,ln);
- }
- for (j = 0; j < rtrim; j++) {
- ln = listLast(list);
- listDelNode(list,ln);
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ o->ptr = ziplistDeleteRange(o->ptr,0,ltrim);
+ o->ptr = ziplistDeleteRange(o->ptr,-rtrim,rtrim);
+ } else if (o->encoding == REDIS_ENCODING_LIST) {
+ list = o->ptr;
+ for (j = 0; j < ltrim; j++) {
+ ln = listFirst(list);
+ listDelNode(list,ln);
+ }
+ for (j = 0; j < rtrim; j++) {
+ ln = listLast(list);
+ listDelNode(list,ln);
+ }
+ } else {
+ redisPanic("Unknown list encoding");
}
- if (listLength(list) == 0) deleteKey(c->db,c->argv[1]);
+ if (lLength(o) == 0) deleteKey(c->db,c->argv[1]);
server.dirty++;
addReply(c,shared.ok);
}
static void lremCommand(redisClient *c) {
- robj *o;
- list *list;
- listNode *ln, *next;
+ robj *subject, *obj = c->argv[3];
int toremove = atoi(c->argv[2]->ptr);
int removed = 0;
- int fromtail = 0;
+ lEntry entry;
- if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL ||
- checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
+ subject = lookupKeyWriteOrReply(c,c->argv[1],shared.czero);
+ if (subject == NULL || checkType(c,subject,REDIS_LIST)) return;
+ /* Make sure obj is raw when we're dealing with a ziplist */
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST)
+ obj = getDecodedObject(obj);
+
+ lIterator *li;
if (toremove < 0) {
toremove = -toremove;
- fromtail = 1;
+ li = lInitIterator(subject,-1,REDIS_HEAD);
+ } else {
+ li = lInitIterator(subject,0,REDIS_TAIL);
}
- ln = fromtail ? list->tail : list->head;
- while (ln) {
- robj *ele = listNodeValue(ln);
- next = fromtail ? ln->prev : ln->next;
- if (equalStringObjects(ele,c->argv[3])) {
- listDelNode(list,ln);
+ while (lNext(li,&entry)) {
+ if (lEqual(&entry,obj)) {
+ lDelete(&entry);
server.dirty++;
removed++;
if (toremove && removed == toremove) break;
}
- ln = next;
}
- if (listLength(list) == 0) deleteKey(c->db,c->argv[1]);
+ lReleaseIterator(li);
+
+ /* Clean up raw encoded object */
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST)
+ decrRefCount(obj);
+
+ if (lLength(subject) == 0) deleteKey(c->db,c->argv[1]);
addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",removed));
}
* as well. This command was originally proposed by Ezra Zygmuntowicz.
*/
static void rpoplpushcommand(redisClient *c) {
- robj *sobj;
- list *srclist;
- listNode *ln;
-
+ robj *sobj, *value;
if ((sobj = lookupKeyWriteOrReply(c,c->argv[1],shared.nullbulk)) == NULL ||
checkType(c,sobj,REDIS_LIST)) return;
- srclist = sobj->ptr;
- ln = listLast(srclist);
- if (ln == NULL) {
+ if (lLength(sobj) == 0) {
addReply(c,shared.nullbulk);
} else {
robj *dobj = lookupKeyWrite(c->db,c->argv[2]);
- robj *ele = listNodeValue(ln);
- list *dstlist;
-
- if (dobj && dobj->type != REDIS_LIST) {
- addReply(c,shared.wrongtypeerr);
- return;
- }
+ if (dobj && checkType(c,dobj,REDIS_LIST)) return;
+ value = lPop(sobj,REDIS_TAIL);
/* Add the element to the target list (unless it's directly
* passed to some BLPOP-ing client */
- if (!handleClientsWaitingListPush(c,c->argv[2],ele)) {
- if (dobj == NULL) {
- /* Create the list if the key does not exist */
- dobj = createListObject();
+ if (!handleClientsWaitingListPush(c,c->argv[2],value)) {
+ /* Create the list if the key does not exist */
+ if (!dobj) {
+ dobj = createZiplistObject();
dictAdd(c->db->dict,c->argv[2],dobj);
incrRefCount(c->argv[2]);
}
- dstlist = dobj->ptr;
- listAddNodeHead(dstlist,ele);
- incrRefCount(ele);
+ lPush(dobj,value,REDIS_HEAD);
}
/* Send the element to the client as reply as well */
- addReplyBulk(c,ele);
+ addReplyBulk(c,value);
+
+ /* lPop returns an object with its refcount incremented */
+ decrRefCount(value);
- /* Finally remove the element from the source list */
- listDelNode(srclist,ln);
- if (listLength(srclist) == 0) deleteKey(c->db,c->argv[1]);
+ /* Delete the source list when it is empty */
+ if (lLength(sobj) == 0) deleteKey(c->db,c->argv[1]);
server.dirty++;
}
}
zskiplistNode *zn = zmalloc(sizeof(*zn));
zn->forward = zmalloc(sizeof(zskiplistNode*) * level);
- if (level > 0)
+ if (level > 1)
zn->span = zmalloc(sizeof(unsigned int) * (level - 1));
+ else
+ zn->span = NULL;
zn->score = score;
zn->obj = obj;
return zn;
zset *zs;
double *score;
+ if (isnan(scoreval)) {
+ addReplySds(c,sdsnew("-ERR provide score is Not A Number (nan)\r\n"));
+ return;
+ }
+
zsetobj = lookupKeyWrite(c->db,key);
if (zsetobj == NULL) {
zsetobj = createZsetObject();
} else {
*score = scoreval;
}
+ if (isnan(*score)) {
+ addReplySds(c,
+ sdsnew("-ERR resulting score is Not A Number (nan)\r\n"));
+ zfree(score);
+ /* Note that we don't need to check if the zset may be empty and
+ * should be removed here, as we can only obtain Nan as score if
+ * there was already an element in the sorted set. */
+ return;
+ }
} else {
*score = scoreval;
}
#define REDIS_AGGR_SUM 1
#define REDIS_AGGR_MIN 2
#define REDIS_AGGR_MAX 3
+#define zunionInterDictValue(_e) (dictGetEntryVal(_e) == NULL ? 1.0 : *(double*)dictGetEntryVal(_e))
inline static void zunionInterAggregate(double *target, double val, int aggregate) {
if (aggregate == REDIS_AGGR_SUM) {
}
static void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) {
- int i, j, zsetnum;
+ int i, j, setnum;
int aggregate = REDIS_AGGR_SUM;
zsetopsrc *src;
robj *dstobj;
dictIterator *di;
dictEntry *de;
- /* expect zsetnum input keys to be given */
- zsetnum = atoi(c->argv[2]->ptr);
- if (zsetnum < 1) {
+ /* expect setnum input keys to be given */
+ setnum = atoi(c->argv[2]->ptr);
+ if (setnum < 1) {
addReplySds(c,sdsnew("-ERR at least 1 input key is needed for ZUNIONSTORE/ZINTERSTORE\r\n"));
return;
}
/* test if the expected number of keys would overflow */
- if (3+zsetnum > c->argc) {
+ if (3+setnum > c->argc) {
addReply(c,shared.syntaxerr);
return;
}
/* read keys to be used for input */
- src = zmalloc(sizeof(zsetopsrc) * zsetnum);
- for (i = 0, j = 3; i < zsetnum; i++, j++) {
- robj *zsetobj = lookupKeyWrite(c->db,c->argv[j]);
- if (!zsetobj) {
+ src = zmalloc(sizeof(zsetopsrc) * setnum);
+ for (i = 0, j = 3; i < setnum; i++, j++) {
+ robj *obj = lookupKeyWrite(c->db,c->argv[j]);
+ if (!obj) {
src[i].dict = NULL;
} else {
- if (zsetobj->type != REDIS_ZSET) {
+ if (obj->type == REDIS_ZSET) {
+ src[i].dict = ((zset*)obj->ptr)->dict;
+ } else if (obj->type == REDIS_SET) {
+ src[i].dict = (obj->ptr);
+ } else {
zfree(src);
addReply(c,shared.wrongtypeerr);
return;
}
- src[i].dict = ((zset*)zsetobj->ptr)->dict;
}
/* default all weights to 1 */
int remaining = c->argc - j;
while (remaining) {
- if (remaining >= (zsetnum + 1) && !strcasecmp(c->argv[j]->ptr,"weights")) {
+ if (remaining >= (setnum + 1) && !strcasecmp(c->argv[j]->ptr,"weights")) {
j++; remaining--;
- for (i = 0; i < zsetnum; i++, j++, remaining--) {
+ for (i = 0; i < setnum; i++, j++, remaining--) {
if (getDoubleFromObjectOrReply(c, c->argv[j], &src[i].weight, NULL) != REDIS_OK)
return;
}
/* sort sets from the smallest to largest, this will improve our
* algorithm's performance */
- qsort(src,zsetnum,sizeof(zsetopsrc), qsortCompareZsetopsrcByCardinality);
+ qsort(src,setnum,sizeof(zsetopsrc),qsortCompareZsetopsrcByCardinality);
dstobj = createZsetObject();
dstzset = dstobj->ptr;
di = dictGetIterator(src[0].dict);
while((de = dictNext(di)) != NULL) {
double *score = zmalloc(sizeof(double)), value;
- *score = src[0].weight * (*(double*)dictGetEntryVal(de));
+ *score = src[0].weight * zunionInterDictValue(de);
- for (j = 1; j < zsetnum; j++) {
+ for (j = 1; j < setnum; j++) {
dictEntry *other = dictFind(src[j].dict,dictGetEntryKey(de));
if (other) {
- value = src[j].weight * (*(double*)dictGetEntryVal(other));
+ value = src[j].weight * zunionInterDictValue(other);
zunionInterAggregate(score, value, aggregate);
} else {
break;
}
/* skip entry when not present in every source dict */
- if (j != zsetnum) {
+ if (j != setnum) {
zfree(score);
} else {
robj *o = dictGetEntryKey(de);
dictReleaseIterator(di);
}
} else if (op == REDIS_OP_UNION) {
- for (i = 0; i < zsetnum; i++) {
+ for (i = 0; i < setnum; i++) {
if (!src[i].dict) continue;
di = dictGetIterator(src[i].dict);
if (dictFind(dstzset->dict,dictGetEntryKey(de)) != NULL) continue;
double *score = zmalloc(sizeof(double)), value;
- *score = src[i].weight * (*(double*)dictGetEntryVal(de));
+ *score = src[i].weight * zunionInterDictValue(de);
/* because the zsets are sorted by size, its only possible
* for sets at larger indices to hold this entry */
- for (j = (i+1); j < zsetnum; j++) {
+ for (j = (i+1); j < setnum; j++) {
dictEntry *other = dictFind(src[j].dict,dictGetEntryKey(de));
if (other) {
- value = src[j].weight * (*(double*)dictGetEntryVal(other));
+ value = src[j].weight * zunionInterDictValue(other);
zunionInterAggregate(score, value, aggregate);
}
}
static void flushdbCommand(redisClient *c) {
server.dirty += dictSize(c->db->dict);
+ touchWatchedKeysOnFlush(c->db->id);
dictEmpty(c->db->dict);
dictEmpty(c->db->expires);
addReply(c,shared.ok);
}
static void flushallCommand(redisClient *c) {
+ touchWatchedKeysOnFlush(-1);
server.dirty += emptyDb();
addReply(c,shared.ok);
if (server.bgsavechildpid != -1) {
* is optimized for speed and a bit less for readability */
static void sortCommand(redisClient *c) {
list *operations;
- int outputlen = 0;
+ unsigned int outputlen = 0;
int desc = 0, alpha = 0;
int limit_start = 0, limit_count = -1, start, end;
int j, dontsort = 0, vectorlen;
/* Load the sorting vector with all the objects to sort */
switch(sortval->type) {
- case REDIS_LIST: vectorlen = listLength((list*)sortval->ptr); break;
+ case REDIS_LIST: vectorlen = lLength(sortval); break;
case REDIS_SET: vectorlen = dictSize((dict*)sortval->ptr); break;
case REDIS_ZSET: vectorlen = dictSize(((zset*)sortval->ptr)->dict); break;
default: vectorlen = 0; redisPanic("Bad SORT type"); /* Avoid GCC warning */
j = 0;
if (sortval->type == REDIS_LIST) {
- list *list = sortval->ptr;
- listNode *ln;
- listIter li;
-
- listRewind(list,&li);
- while((ln = listNext(&li))) {
- robj *ele = ln->value;
- vector[j].obj = ele;
+ lIterator *li = lInitIterator(sortval,0,REDIS_TAIL);
+ lEntry entry;
+ while(lNext(li,&entry)) {
+ vector[j].obj = lGet(&entry);
vector[j].u.score = 0;
vector[j].u.cmpobj = NULL;
j++;
}
+ lReleaseIterator(li);
} else {
dict *set;
dictIterator *di;
}
}
} else {
- robj *listObject = createListObject();
- list *listPtr = (list*) listObject->ptr;
+ robj *sobj = createZiplistObject();
/* STORE option specified, set the sorting result as a List object */
for (j = start; j <= end; j++) {
listIter li;
if (!getop) {
- listAddNodeTail(listPtr,vector[j].obj);
- incrRefCount(vector[j].obj);
- }
- listRewind(operations,&li);
- while((ln = listNext(&li))) {
- redisSortOperation *sop = ln->value;
- robj *val = lookupKeyByPattern(c->db,sop->pattern,
- vector[j].obj);
+ lPush(sobj,vector[j].obj,REDIS_TAIL);
+ } else {
+ listRewind(operations,&li);
+ while((ln = listNext(&li))) {
+ redisSortOperation *sop = ln->value;
+ robj *val = lookupKeyByPattern(c->db,sop->pattern,
+ vector[j].obj);
- if (sop->type == REDIS_SORT_GET) {
- if (!val) {
- listAddNodeTail(listPtr,createStringObject("",0));
+ if (sop->type == REDIS_SORT_GET) {
+ if (!val) val = createStringObject("",0);
+
+ /* lPush does an incrRefCount, so we should take care
+ * care of the incremented refcount caused by either
+ * lookupKeyByPattern or createStringObject("",0) */
+ lPush(sobj,val,REDIS_TAIL);
+ decrRefCount(val);
} else {
- /* We should do a incrRefCount on val because it is
- * added to the list, but also a decrRefCount because
- * it is returned by lookupKeyByPattern. This results
- * in doing nothing at all. */
- listAddNodeTail(listPtr,val);
+ /* always fails */
+ redisAssert(sop->type == REDIS_SORT_GET);
}
- } else {
- redisAssert(sop->type == REDIS_SORT_GET); /* always fails */
}
}
}
- if (dictReplace(c->db->dict,storekey,listObject)) {
+ if (dictReplace(c->db->dict,storekey,sobj)) {
incrRefCount(storekey);
}
/* Note: we add 1 because the DB is dirty anyway since even if the
}
/* Cleanup */
+ if (sortval->type == REDIS_LIST)
+ for (j = 0; j < vectorlen; j++)
+ decrRefCount(vector[j].obj);
decrRefCount(sortval);
listRelease(operations);
for (j = 0; j < vectorlen; j++) {
}
static void multiCommand(redisClient *c) {
+ if (c->flags & REDIS_MULTI) {
+ addReplySds(c,sdsnew("-ERR MULTI calls can not be nested\r\n"));
+ return;
+ }
c->flags |= REDIS_MULTI;
addReply(c,shared.ok);
}
return;
}
+ /* Check if we need to abort the EXEC if some WATCHed key was touched.
+ * A failed EXEC will return a multi bulk nil object. */
+ if (c->flags & REDIS_DIRTY_CAS) {
+ freeClientMultiState(c);
+ initClientMultiState(c);
+ c->flags &= ~(REDIS_MULTI|REDIS_DIRTY_CAS);
+ unwatchAllKeys(c);
+ addReply(c,shared.nullmultibulk);
+ return;
+ }
+
/* Replicate a MULTI request now that we are sure the block is executed.
* This way we'll deliver the MULTI/..../EXEC block as a whole and
* both the AOF and the replication link will have the same consistency
execCommandReplicateMulti(c);
/* Exec all the queued commands */
+ unwatchAllKeys(c); /* Unwatch ASAP otherwise we'll waste CPU cycles */
orig_argv = c->argv;
orig_argc = c->argc;
addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",c->mstate.count));
c->argc = orig_argc;
freeClientMultiState(c);
initClientMultiState(c);
- c->flags &= (~REDIS_MULTI);
+ c->flags &= ~(REDIS_MULTI|REDIS_DIRTY_CAS);
/* Make sure the EXEC command is always replicated / AOF, since we
* always send the MULTI command (we can't know beforehand if the
* next operations will contain at least a modification to the DB). */
* empty we need to block. In order to do so we remove the notification for
* new data to read in the client socket (so that we'll not serve new
* requests if the blocking request is not served). Also we put the client
- * in a dictionary (db->blockingkeys) mapping keys to a list of clients
+ * in a dictionary (db->blocking_keys) mapping keys to a list of clients
* blocking for this keys.
* - If a PUSH operation against a key with blocked clients waiting is
* performed, we serve the first in the list: basically instead to push
list *l;
int j;
- c->blockingkeys = zmalloc(sizeof(robj*)*numkeys);
- c->blockingkeysnum = numkeys;
+ c->blocking_keys = zmalloc(sizeof(robj*)*numkeys);
+ c->blocking_keys_num = numkeys;
c->blockingto = timeout;
for (j = 0; j < numkeys; j++) {
/* Add the key in the client structure, to map clients -> keys */
- c->blockingkeys[j] = keys[j];
+ c->blocking_keys[j] = keys[j];
incrRefCount(keys[j]);
/* And in the other "side", to map keys -> clients */
- de = dictFind(c->db->blockingkeys,keys[j]);
+ de = dictFind(c->db->blocking_keys,keys[j]);
if (de == NULL) {
int retval;
/* For every key we take a list of clients blocked for it */
l = listCreate();
- retval = dictAdd(c->db->blockingkeys,keys[j],l);
+ retval = dictAdd(c->db->blocking_keys,keys[j],l);
incrRefCount(keys[j]);
assert(retval == DICT_OK);
} else {
list *l;
int j;
- assert(c->blockingkeys != NULL);
+ assert(c->blocking_keys != NULL);
/* The client may wait for multiple keys, so unblock it for every key. */
- for (j = 0; j < c->blockingkeysnum; j++) {
+ for (j = 0; j < c->blocking_keys_num; j++) {
/* Remove this client from the list of clients waiting for this key. */
- de = dictFind(c->db->blockingkeys,c->blockingkeys[j]);
+ de = dictFind(c->db->blocking_keys,c->blocking_keys[j]);
assert(de != NULL);
l = dictGetEntryVal(de);
listDelNode(l,listSearchKey(l,c));
/* If the list is empty we need to remove it to avoid wasting memory */
if (listLength(l) == 0)
- dictDelete(c->db->blockingkeys,c->blockingkeys[j]);
- decrRefCount(c->blockingkeys[j]);
+ dictDelete(c->db->blocking_keys,c->blocking_keys[j]);
+ decrRefCount(c->blocking_keys[j]);
}
/* Cleanup the client structure */
- zfree(c->blockingkeys);
- c->blockingkeys = NULL;
+ zfree(c->blocking_keys);
+ c->blocking_keys = NULL;
c->flags &= (~REDIS_BLOCKED);
server.blpop_blocked_clients--;
/* We want to process data if there is some command waiting
list *l;
listNode *ln;
- de = dictFind(c->db->blockingkeys,key);
+ de = dictFind(c->db->blocking_keys,key);
if (de == NULL) return 0;
l = dictGetEntryVal(de);
ln = listFirst(l);
exit(1);
}
-/* Write an object into a file in the bulk format $<count>\r\n<payload>\r\n */
-static int fwriteBulkObject(FILE *fp, robj *obj) {
- char buf[128];
- int decrrc = 0;
-
- /* Avoid the incr/decr ref count business if possible to help
- * copy-on-write (we are often in a child process when this function
- * is called).
- * Also makes sure that key objects don't get incrRefCount-ed when VM
- * is enabled */
- if (obj->encoding != REDIS_ENCODING_RAW) {
- obj = getDecodedObject(obj);
- decrrc = 1;
- }
- snprintf(buf,sizeof(buf),"$%ld\r\n",(long)sdslen(obj->ptr));
- if (fwrite(buf,strlen(buf),1,fp) == 0) goto err;
- if (sdslen(obj->ptr) && fwrite(obj->ptr,sdslen(obj->ptr),1,fp) == 0)
- goto err;
- if (fwrite("\r\n",2,1,fp) == 0) goto err;
- if (decrrc) decrRefCount(obj);
- return 1;
-err:
- if (decrrc) decrRefCount(obj);
- return 0;
-}
-
/* Write binary-safe string into a file in the bulkformat
* $<count>\r\n<payload>\r\n */
static int fwriteBulkString(FILE *fp, char *s, unsigned long len) {
- char buf[128];
-
- snprintf(buf,sizeof(buf),"$%ld\r\n",(unsigned long)len);
- if (fwrite(buf,strlen(buf),1,fp) == 0) return 0;
- if (len && fwrite(s,len,1,fp) == 0) return 0;
+ char cbuf[128];
+ int clen;
+ cbuf[0] = '$';
+ clen = 1+ll2string(cbuf+1,sizeof(cbuf)-1,len);
+ cbuf[clen++] = '\r';
+ cbuf[clen++] = '\n';
+ if (fwrite(cbuf,clen,1,fp) == 0) return 0;
+ if (len > 0 && fwrite(s,len,1,fp) == 0) return 0;
if (fwrite("\r\n",2,1,fp) == 0) return 0;
return 1;
}
}
/* Write a long value in bulk format $<count>\r\n<payload>\r\n */
-static int fwriteBulkLong(FILE *fp, long l) {
- char buf[128], lbuf[128];
-
- snprintf(lbuf,sizeof(lbuf),"%ld\r\n",l);
- snprintf(buf,sizeof(buf),"$%lu\r\n",(unsigned long)strlen(lbuf)-2);
- if (fwrite(buf,strlen(buf),1,fp) == 0) return 0;
- if (fwrite(lbuf,strlen(lbuf),1,fp) == 0) return 0;
+static int fwriteBulkLongLong(FILE *fp, long long l) {
+ char bbuf[128], lbuf[128];
+ unsigned int blen, llen;
+ llen = ll2string(lbuf,32,l);
+ blen = snprintf(bbuf,sizeof(bbuf),"$%u\r\n%s\r\n",llen,lbuf);
+ if (fwrite(bbuf,blen,1,fp) == 0) return 0;
return 1;
}
+/* Delegate writing an object to writing a bulk string or bulk long long. */
+static int fwriteBulkObject(FILE *fp, robj *obj) {
+ /* Avoid using getDecodedObject to help copy-on-write (we are often
+ * in a child process when this function is called). */
+ if (obj->encoding == REDIS_ENCODING_INT) {
+ return fwriteBulkLongLong(fp,(long)obj->ptr);
+ } else if (obj->encoding == REDIS_ENCODING_RAW) {
+ return fwriteBulkString(fp,obj->ptr,sdslen(obj->ptr));
+ } else {
+ redisPanic("Unknown string encoding");
+ }
+}
+
/* Write a sequence of commands able to fully rebuild the dataset into
* "filename". Used both by REWRITEAOF and BGREWRITEAOF. */
static int rewriteAppendOnlyFile(char *filename) {
/* SELECT the new DB */
if (fwrite(selectcmd,sizeof(selectcmd)-1,1,fp) == 0) goto werr;
- if (fwriteBulkLong(fp,j) == 0) goto werr;
+ if (fwriteBulkLongLong(fp,j) == 0) goto werr;
/* Iterate this DB writing every entry */
while((de = dictNext(di)) != NULL) {
if (fwriteBulkObject(fp,o) == 0) goto werr;
} else if (o->type == REDIS_LIST) {
/* Emit the RPUSHes needed to rebuild the list */
- list *list = o->ptr;
- listNode *ln;
- listIter li;
+ char cmd[]="*3\r\n$5\r\nRPUSH\r\n";
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *zl = o->ptr;
+ unsigned char *p = ziplistIndex(zl,0);
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vlong;
+
+ while(ziplistGet(p,&vstr,&vlen,&vlong)) {
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulkObject(fp,key) == 0) goto werr;
+ if (vstr) {
+ if (fwriteBulkString(fp,(char*)vstr,vlen) == 0)
+ goto werr;
+ } else {
+ if (fwriteBulkLongLong(fp,vlong) == 0)
+ goto werr;
+ }
+ p = ziplistNext(zl,p);
+ }
+ } else if (o->encoding == REDIS_ENCODING_LIST) {
+ list *list = o->ptr;
+ listNode *ln;
+ listIter li;
- listRewind(list,&li);
- while((ln = listNext(&li))) {
- char cmd[]="*3\r\n$5\r\nRPUSH\r\n";
- robj *eleobj = listNodeValue(ln);
+ listRewind(list,&li);
+ while((ln = listNext(&li))) {
+ robj *eleobj = listNodeValue(ln);
- if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
- if (fwriteBulkObject(fp,key) == 0) goto werr;
- if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulkObject(fp,key) == 0) goto werr;
+ if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
+ }
+ } else {
+ redisPanic("Unknown list encoding");
}
} else if (o->type == REDIS_SET) {
/* Emit the SADDs needed to rebuild the set */
if (expiretime < now) continue;
if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
if (fwriteBulkObject(fp,key) == 0) goto werr;
- if (fwriteBulkLong(fp,expiretime) == 0) goto werr;
+ if (fwriteBulkLongLong(fp,expiretime) == 0) goto werr;
}
if (swapped) decrRefCount(o);
}
addReplyLongLong(c,receivers);
}
+/* ===================== WATCH (CAS alike for MULTI/EXEC) ===================
+ *
+ * The implementation uses a per-DB hash table mapping keys to list of clients
+ * WATCHing those keys, so that given a key that is going to be modified
+ * we can mark all the associated clients as dirty.
+ *
+ * Also every client contains a list of WATCHed keys so that's possible to
+ * un-watch such keys when the client is freed or when UNWATCH is called. */
+
+/* In the client->watched_keys list we need to use watchedKey structures
+ * as in order to identify a key in Redis we need both the key name and the
+ * DB */
+typedef struct watchedKey {
+ robj *key;
+ redisDb *db;
+} watchedKey;
+
+/* Watch for the specified key */
+static void watchForKey(redisClient *c, robj *key) {
+ list *clients = NULL;
+ listIter li;
+ listNode *ln;
+ watchedKey *wk;
+
+ /* Check if we are already watching for this key */
+ listRewind(c->watched_keys,&li);
+ while((ln = listNext(&li))) {
+ wk = listNodeValue(ln);
+ if (wk->db == c->db && equalStringObjects(key,wk->key))
+ return; /* Key already watched */
+ }
+ /* This key is not already watched in this DB. Let's add it */
+ clients = dictFetchValue(c->db->watched_keys,key);
+ if (!clients) {
+ clients = listCreate();
+ dictAdd(c->db->watched_keys,key,clients);
+ incrRefCount(key);
+ }
+ listAddNodeTail(clients,c);
+ /* Add the new key to the lits of keys watched by this client */
+ wk = zmalloc(sizeof(*wk));
+ wk->key = key;
+ wk->db = c->db;
+ incrRefCount(key);
+ listAddNodeTail(c->watched_keys,wk);
+}
+
+/* Unwatch all the keys watched by this client. To clean the EXEC dirty
+ * flag is up to the caller. */
+static void unwatchAllKeys(redisClient *c) {
+ listIter li;
+ listNode *ln;
+
+ if (listLength(c->watched_keys) == 0) return;
+ listRewind(c->watched_keys,&li);
+ while((ln = listNext(&li))) {
+ list *clients;
+ watchedKey *wk;
+
+ /* Lookup the watched key -> clients list and remove the client
+ * from the list */
+ wk = listNodeValue(ln);
+ clients = dictFetchValue(wk->db->watched_keys, wk->key);
+ assert(clients != NULL);
+ listDelNode(clients,listSearchKey(clients,c));
+ /* Kill the entry at all if this was the only client */
+ if (listLength(clients) == 0)
+ dictDelete(wk->db->watched_keys, wk->key);
+ /* Remove this watched key from the client->watched list */
+ listDelNode(c->watched_keys,ln);
+ decrRefCount(wk->key);
+ zfree(wk);
+ }
+}
+
+/* "Touch" a key, so that if this key is being WATCHed by some client the
+ * next EXEC will fail. */
+static void touchWatchedKey(redisDb *db, robj *key) {
+ list *clients;
+ listIter li;
+ listNode *ln;
+
+ if (dictSize(db->watched_keys) == 0) return;
+ clients = dictFetchValue(db->watched_keys, key);
+ if (!clients) return;
+
+ /* Mark all the clients watching this key as REDIS_DIRTY_CAS */
+ /* Check if we are already watching for this key */
+ listRewind(clients,&li);
+ while((ln = listNext(&li))) {
+ redisClient *c = listNodeValue(ln);
+
+ c->flags |= REDIS_DIRTY_CAS;
+ }
+}
+
+/* On FLUSHDB or FLUSHALL all the watched keys that are present before the
+ * flush but will be deleted as effect of the flushing operation should
+ * be touched. "dbid" is the DB that's getting the flush. -1 if it is
+ * a FLUSHALL operation (all the DBs flushed). */
+static void touchWatchedKeysOnFlush(int dbid) {
+ listIter li1, li2;
+ listNode *ln;
+
+ /* For every client, check all the waited keys */
+ listRewind(server.clients,&li1);
+ while((ln = listNext(&li1))) {
+ redisClient *c = listNodeValue(ln);
+ listRewind(c->watched_keys,&li2);
+ while((ln = listNext(&li2))) {
+ watchedKey *wk = listNodeValue(ln);
+
+ /* For every watched key matching the specified DB, if the
+ * key exists, mark the client as dirty, as the key will be
+ * removed. */
+ if (dbid == -1 || wk->db->id == dbid) {
+ if (dictFind(wk->db->dict, wk->key) != NULL)
+ c->flags |= REDIS_DIRTY_CAS;
+ }
+ }
+ }
+}
+
+static void watchCommand(redisClient *c) {
+ int j;
+
+ if (c->flags & REDIS_MULTI) {
+ addReplySds(c,sdsnew("-ERR WATCH inside MULTI is not allowed\r\n"));
+ return;
+ }
+ for (j = 1; j < c->argc; j++)
+ watchForKey(c,c->argv[j]);
+ addReply(c,shared.ok);
+}
+
+static void unwatchCommand(redisClient *c) {
+ unwatchAllKeys(c);
+ c->flags &= (~REDIS_DIRTY_CAS);
+ addReply(c,shared.ok);
+}
+
/* ================================= Debugging ============================== */
/* Compute the sha1 of string at 's' with 'len' bytes long.
if (o->type == REDIS_STRING) {
mixObjectDigest(digest,o);
} else if (o->type == REDIS_LIST) {
- list *list = o->ptr;
- listNode *ln;
- listIter li;
-
- listRewind(list,&li);
- while((ln = listNext(&li))) {
- robj *eleobj = listNodeValue(ln);
-
+ lIterator *li = lInitIterator(o,0,REDIS_TAIL);
+ lEntry entry;
+ while(lNext(li,&entry)) {
+ robj *eleobj = lGet(&entry);
mixObjectDigest(digest,eleobj);
+ decrRefCount(eleobj);
}
+ lReleaseIterator(li);
} else if (o->type == REDIS_SET) {
dict *set = o->ptr;
dictIterator *di = dictGetIterator(set);
}
static void version() {
- printf("Redis server version %s\n", REDIS_VERSION);
+ printf("Redis server version %s (%s:%d)\n", REDIS_VERSION,
+ REDIS_GIT_SHA1, atoi(REDIS_GIT_DIRTY) > 0);
exit(0);
}
time_t start;
initServerConfig();
+ sortCommandTable();
if (argc == 2) {
if (strcmp(argv[1], "-v") == 0 ||
strcmp(argv[1], "--version") == 0) version();
projects / redis.git / blobdiff