X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/4005fef12415087aed9bdfa4c1baadbb7a558bfb..55017f9da0c8ee8709029c4dbdbf4f1d4db4a181:/redis.c diff --git a/redis.c b/redis.c index 9e5bc94a..1bdd1421 100644 --- a/redis.c +++ b/redis.c @@ -91,7 +91,7 @@ #define REDIS_CONFIGLINE_MAX 1024 #define REDIS_OBJFREELIST_MAX 1000000 /* Max number of objects to cache */ #define REDIS_MAX_SYNC_TIME 60 /* Slave can't take more to sync */ -#define REDIS_EXPIRELOOKUPS_PER_CRON 10 /* try to expire 10 keys/loop */ +#define REDIS_EXPIRELOOKUPS_PER_CRON 10 /* lookup 10 expires per loop */ #define REDIS_MAX_WRITE_PER_EVENT (1024*64) #define REDIS_REQUEST_MAX_SIZE (1024*1024*256) /* max bytes in inline command */ @@ -342,8 +342,6 @@ struct redisServer { int port; int fd; redisDb *db; - dict *sharingpool; /* Poll used for object sharing */ - unsigned int sharingpoolsize; long long dirty; /* changes to DB from the last save */ list *clients; list *slaves, *monitors; @@ -381,6 +379,7 @@ struct redisServer { char *requirepass; int shareobjects; int rdbcompression; + int activerehashing; /* Replication related */ int isslave; char *masterauth; @@ -507,6 +506,7 @@ typedef struct zset { /* Our shared "common" objects */ +#define REDIS_SHARED_INTEGERS 10000 struct sharedObjectsStruct { robj *crlf, *ok, *err, *emptybulk, *czero, *cone, *pong, *space, *colon, *nullbulk, *nullmultibulk, *queued, @@ -515,7 +515,7 @@ struct sharedObjectsStruct { *select0, *select1, *select2, *select3, *select4, *select5, *select6, *select7, *select8, *select9, *messagebulk, *subscribebulk, *unsubscribebulk, *mbulk3, - *psubscribebulk, *punsubscribebulk; + *psubscribebulk, *punsubscribebulk, *integers[REDIS_SHARED_INTEGERS]; } shared; /* Global vars that are actally used as constants. The following double @@ -558,8 +558,7 @@ static robj *dupStringObject(robj *o); static void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc); static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc); static int syncWithMaster(void); -static robj *tryObjectSharing(robj *o); -static int tryObjectEncoding(robj *o); +static robj *tryObjectEncoding(robj *o); static robj *getDecodedObject(robj *o); static int removeExpire(redisDb *db, robj *key); static int expireIfNeeded(redisDb *db, robj *key); @@ -706,7 +705,10 @@ static void substrCommand(redisClient *c); static void zrankCommand(redisClient *c); static void zrevrankCommand(redisClient *c); static void hsetCommand(redisClient *c); +static void hsetnxCommand(redisClient *c); static void hgetCommand(redisClient *c); +static void hmsetCommand(redisClient *c); +static void hmgetCommand(redisClient *c); static void hdelCommand(redisClient *c); static void hlenCommand(redisClient *c); static void zremrangebyrankCommand(redisClient *c); @@ -782,8 +784,11 @@ static struct redisCommand cmdTable[] = { {"zrank",zrankCommand,3,REDIS_CMD_BULK,NULL,1,1,1}, {"zrevrank",zrevrankCommand,3,REDIS_CMD_BULK,NULL,1,1,1}, {"hset",hsetCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"hincrby",hincrbyCommand,4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,NULL,1,1,1}, + {"hsetnx",hsetnxCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1}, {"hget",hgetCommand,3,REDIS_CMD_BULK,NULL,1,1,1}, + {"hmset",hmsetCommand,-4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1}, + {"hmget",hmgetCommand,-3,REDIS_CMD_BULK,NULL,1,1,1}, + {"hincrby",hincrbyCommand,4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,NULL,1,1,1}, {"hdel",hdelCommand,3,REDIS_CMD_BULK,NULL,1,1,1}, {"hlen",hlenCommand,2,REDIS_CMD_INLINE,NULL,1,1,1}, {"hkeys",hkeysCommand,2,REDIS_CMD_INLINE,NULL,1,1,1}, @@ -1199,16 +1204,28 @@ static void tryResizeHashTables(void) { int j; for (j = 0; j < server.dbnum; j++) { - if (htNeedsResize(server.db[j].dict)) { - redisLog(REDIS_VERBOSE,"The hash table %d is too sparse, resize it...",j); + if (htNeedsResize(server.db[j].dict)) dictResize(server.db[j].dict); - redisLog(REDIS_VERBOSE,"Hash table %d resized.",j); - } if (htNeedsResize(server.db[j].expires)) dictResize(server.db[j].expires); } } +/* Our hash table implementation performs rehashing incrementally while + * we write/read from the hash table. Still if the server is idle, the hash + * table will use two tables for a long time. So we try to use 1 millisecond + * of CPU time at every serverCron() loop in order to rehash some key. */ +static void incrementallyRehash(void) { + int j; + + for (j = 0; j < server.dbnum; j++) { + if (dictIsRehashing(server.db[j].dict)) { + dictRehashMilliseconds(server.db[j].dict,1); + break; /* already used our millisecond for this loop... */ + } + } +} + /* A background saving child (BGSAVE) terminated its work. Handle this. */ void backgroundSaveDoneHandler(int statloc) { int exitcode = WEXITSTATUS(statloc); @@ -1223,7 +1240,7 @@ void backgroundSaveDoneHandler(int statloc) { redisLog(REDIS_WARNING, "Background saving error"); } else { redisLog(REDIS_WARNING, - "Background saving terminated by signal"); + "Background saving terminated by signal %d", WTERMSIG(statloc)); rdbRemoveTempFile(server.bgsavechildpid); } server.bgsavechildpid = -1; @@ -1284,7 +1301,8 @@ void backgroundRewriteDoneHandler(int statloc) { redisLog(REDIS_WARNING, "Background append only file rewriting error"); } else { redisLog(REDIS_WARNING, - "Background append only file rewriting terminated by signal"); + "Background append only file rewriting terminated by signal %d", + WTERMSIG(statloc)); } cleanup: sdsfree(server.bgrewritebuf); @@ -1293,6 +1311,19 @@ cleanup: server.bgrewritechildpid = -1; } +/* This function is called once a background process of some kind terminates, + * as we want to avoid resizing the hash tables when there is a child in order + * to play well with copy-on-write (otherwise when a resize happens lots of + * memory pages are copied). The goal of this function is to update the ability + * for dict.c to resize the hash tables accordingly to the fact we have o not + * running childs. */ +static void updateDictResizePolicy(void) { + if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1) + dictEnableResize(); + else + dictDisableResize(); +} + static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { int j, loops = server.cronloops++; REDIS_NOTUSED(eventLoop); @@ -1324,15 +1355,17 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD * if we resize the HT while there is the saving child at work actually * a lot of memory movements in the parent will cause a lot of pages * copied. */ - if (server.bgsavechildpid == -1 && !(loops % 10)) tryResizeHashTables(); + if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1) { + if (!(loops % 10)) tryResizeHashTables(); + if (server.activerehashing) incrementallyRehash(); + } /* Show information about connected clients */ if (!(loops % 50)) { - redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use, %d shared objects", + redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use", listLength(server.clients)-listLength(server.slaves), listLength(server.slaves), - zmalloc_used_memory(), - dictSize(server.sharingpool)); + zmalloc_used_memory()); } /* Close connections of timedout clients */ @@ -1350,6 +1383,7 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD } else { backgroundRewriteDoneHandler(statloc); } + updateDictResizePolicy(); } } else { /* If there is not a background saving in progress check if @@ -1469,6 +1503,8 @@ static void beforeSleep(struct aeEventLoop *eventLoop) { } static void createSharedObjects(void) { + int j; + shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n")); shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n")); shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n")); @@ -1509,6 +1545,10 @@ static void createSharedObjects(void) { shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17); shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19); shared.mbulk3 = createStringObject("*3\r\n",4); + for (j = 0; j < REDIS_SHARED_INTEGERS; j++) { + shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j); + shared.integers[j]->encoding = REDIS_ENCODING_INT; + } } static void appendServerSaveParams(time_t seconds, int changes) { @@ -1545,7 +1585,7 @@ static void initServerConfig() { server.requirepass = NULL; server.shareobjects = 0; server.rdbcompression = 1; - server.sharingpoolsize = 1024; + server.activerehashing = 1; server.maxclients = 0; server.blpop_blocked_clients = 0; server.maxmemory = 0; @@ -1598,7 +1638,6 @@ static void initServer() { createSharedObjects(); server.el = aeCreateEventLoop(); server.db = zmalloc(sizeof(redisDb)*server.dbnum); - server.sharingpool = dictCreate(&setDictType,NULL); server.fd = anetTcpServer(server.neterr, server.port, server.bindaddr); if (server.fd == -1) { redisLog(REDIS_WARNING, "Opening TCP port: %s", server.neterr); @@ -1669,15 +1708,12 @@ static void loadServerConfig(char *filename) { char buf[REDIS_CONFIGLINE_MAX+1], *err = NULL; int linenum = 0; sds line = NULL; - char *errormsg = "Fatal error, can't open config file '%s'"; - char *errorbuf = zmalloc(sizeof(char)*(strlen(errormsg)+strlen(filename))); - sprintf(errorbuf, errormsg, filename); if (filename[0] == '-' && filename[1] == '\0') fp = stdin; else { if ((fp = fopen(filename,"r")) == NULL) { - redisLog(REDIS_WARNING, errorbuf); + redisLog(REDIS_WARNING, "Fatal error, can't open config file '%s'", filename); exit(1); } } @@ -1783,10 +1819,9 @@ static void loadServerConfig(char *filename) { if ((server.rdbcompression = yesnotoi(argv[1])) == -1) { err = "argument must be 'yes' or 'no'"; goto loaderr; } - } else if (!strcasecmp(argv[0],"shareobjectspoolsize") && argc == 2) { - server.sharingpoolsize = atoi(argv[1]); - if (server.sharingpoolsize < 1) { - err = "invalid object sharing pool size"; goto loaderr; + } else if (!strcasecmp(argv[0],"activerehashing") && argc == 2) { + if ((server.activerehashing = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; } } else if (!strcasecmp(argv[0],"daemonize") && argc == 2) { if ((server.daemonize = yesnotoi(argv[1])) == -1) { @@ -1964,7 +1999,7 @@ static void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) /* Use writev() if we have enough buffers to send */ if (!server.glueoutputbuf && - listLength(c->reply) > REDIS_WRITEV_THRESHOLD && + listLength(c->reply) > REDIS_WRITEV_THRESHOLD && !(c->flags & REDIS_MASTER)) { sendReplyToClientWritev(el, fd, privdata, mask); @@ -2042,7 +2077,7 @@ static void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int o = listNodeValue(node); objlen = sdslen(o->ptr); - if (totwritten + objlen - offset > REDIS_MAX_WRITE_PER_EVENT) + if (totwritten + objlen - offset > REDIS_MAX_WRITE_PER_EVENT) break; if(ion == REDIS_WRITEV_IOVEC_COUNT) @@ -2090,7 +2125,7 @@ static void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int } } - if (totwritten > 0) + if (totwritten > 0) c->lastinteraction = time(NULL); if (listLength(c->reply) == 0) { @@ -2267,15 +2302,9 @@ static int processCommand(redisClient *c) { return 1; } } - /* Let's try to share objects on the command arguments vector */ - if (server.shareobjects) { - int j; - for(j = 1; j < c->argc; j++) - c->argv[j] = tryObjectSharing(c->argv[j]); - } /* Let's try to encode the bulk object to save space. */ if (cmd->flags & REDIS_CMD_BULK) - tryObjectEncoding(c->argv[c->argc-1]); + c->argv[c->argc-1] = tryObjectEncoding(c->argv[c->argc-1]); /* Check if the user is authenticated */ if (server.requirepass && !c->authenticated && cmd->proc != authCommand) { @@ -2294,7 +2323,8 @@ static int processCommand(redisClient *c) { } /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */ - if (dictSize(c->pubsub_channels) > 0 && + if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0) + && cmd->proc != subscribeCommand && cmd->proc != unsubscribeCommand && cmd->proc != psubscribeCommand && cmd->proc != punsubscribeCommand) { addReplySds(c,sdsnew("-ERR only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context\r\n")); @@ -2406,7 +2436,7 @@ again: if (p) { sds query, *argv; int argc, j; - + query = c->querybuf; c->querybuf = sdsempty(); querylen = 1+(p-(query)); @@ -2601,6 +2631,21 @@ static void addReplyLong(redisClient *c, long l) { addReplySds(c,sdsnewlen(buf,len)); } +static void addReplyLongLong(redisClient *c, long long ll) { + char buf[128]; + size_t len; + + if (ll == 0) { + addReply(c,shared.czero); + return; + } else if (ll == 1) { + addReply(c,shared.cone); + return; + } + len = snprintf(buf,sizeof(buf),":%lld\r\n",ll); + addReplySds(c,sdsnewlen(buf,len)); +} + static void addReplyUlong(redisClient *c, unsigned long ul) { char buf[128]; size_t len; @@ -2800,7 +2845,6 @@ static void freeHashObject(robj *o) { } static void incrRefCount(robj *o) { - redisAssert(!server.vm_enabled || o->storage == REDIS_VM_MEMORY); o->refcount++; } @@ -2812,9 +2856,6 @@ static void decrRefCount(void *obj) { if (server.vm_enabled && (o->storage == REDIS_VM_SWAPPED || o->storage == REDIS_VM_LOADING)) { - if (o->storage == REDIS_VM_SWAPPED || o->storage == REDIS_VM_LOADING) { - redisAssert(o->refcount == 1); - } if (o->storage == REDIS_VM_LOADING) vmCancelThreadedIOJob(obj); redisAssert(o->type == REDIS_STRING); freeStringObject(o); @@ -2927,51 +2968,6 @@ static int deleteKey(redisDb *db, robj *key) { return retval == DICT_OK; } -/* Try to share an object against the shared objects pool */ -static robj *tryObjectSharing(robj *o) { - struct dictEntry *de; - unsigned long c; - - if (o == NULL || server.shareobjects == 0) return o; - - redisAssert(o->type == REDIS_STRING); - de = dictFind(server.sharingpool,o); - if (de) { - robj *shared = dictGetEntryKey(de); - - c = ((unsigned long) dictGetEntryVal(de))+1; - dictGetEntryVal(de) = (void*) c; - incrRefCount(shared); - decrRefCount(o); - return shared; - } else { - /* Here we are using a stream algorihtm: Every time an object is - * shared we increment its count, everytime there is a miss we - * recrement the counter of a random object. If this object reaches - * zero we remove the object and put the current object instead. */ - if (dictSize(server.sharingpool) >= - server.sharingpoolsize) { - de = dictGetRandomKey(server.sharingpool); - redisAssert(de != NULL); - c = ((unsigned long) dictGetEntryVal(de))-1; - dictGetEntryVal(de) = (void*) c; - if (c == 0) { - dictDelete(server.sharingpool,de->key); - } - } else { - c = 0; /* If the pool is empty we want to add this object */ - } - if (c == 0) { - int retval; - - retval = dictAdd(server.sharingpool,o,(void*)1); - redisAssert(retval == DICT_OK); - incrRefCount(o); - } - return o; - } -} - /* Check if the nul-terminated string 's' can be represented by a long * (that is, is a number that fits into long without any other space or * character before or after the digits). @@ -2982,7 +2978,7 @@ static int isStringRepresentableAsLong(sds s, long *longval) { char buf[32], *endptr; long value; int slen; - + value = strtol(s, &endptr, 10); if (endptr[0] != '\0') return REDIS_ERR; slen = snprintf(buf,32,"%ld",value); @@ -2995,36 +2991,42 @@ static int isStringRepresentableAsLong(sds s, long *longval) { } /* Try to encode a string object in order to save space */ -static int tryObjectEncoding(robj *o) { +static robj *tryObjectEncoding(robj *o) { long value; sds s = o->ptr; if (o->encoding != REDIS_ENCODING_RAW) - return REDIS_ERR; /* Already encoded */ + return o; /* Already encoded */ - /* It's not save to encode shared objects: shared objects can be shared + /* It's not safe to encode shared objects: shared objects can be shared * everywhere in the "object space" of Redis. Encoded objects can only * appear as "values" (and not, for instance, as keys) */ - if (o->refcount > 1) return REDIS_ERR; + if (o->refcount > 1) return o; /* Currently we try to encode only strings */ redisAssert(o->type == REDIS_STRING); /* Check if we can represent this string as a long integer */ - if (isStringRepresentableAsLong(s,&value) == REDIS_ERR) return REDIS_ERR; + if (isStringRepresentableAsLong(s,&value) == REDIS_ERR) return o; /* Ok, this object can be encoded */ - o->encoding = REDIS_ENCODING_INT; - sdsfree(o->ptr); - o->ptr = (void*) value; - return REDIS_OK; + if (value >= 0 && value < REDIS_SHARED_INTEGERS) { + decrRefCount(o); + incrRefCount(shared.integers[value]); + return shared.integers[value]; + } else { + o->encoding = REDIS_ENCODING_INT; + sdsfree(o->ptr); + o->ptr = (void*) value; + return o; + } } /* Get a decoded version of an encoded object (returned as a new object). * If the object is already raw-encoded just increment the ref count. */ static robj *getDecodedObject(robj *o) { robj *dec; - + if (o->encoding == REDIS_ENCODING_RAW) { incrRefCount(o); return o; @@ -3082,6 +3084,77 @@ static size_t stringObjectLen(robj *o) { } } +static int getDoubleFromObject(redisClient *c, robj *o, double *value) { + double parsedValue; + char *eptr = NULL; + + if (o && o->type != REDIS_STRING) { + addReplySds(c,sdsnew("-ERR value is not a double\r\n")); + return REDIS_ERR; + } + + if (o == NULL) + parsedValue = 0; + else if (o->encoding == REDIS_ENCODING_RAW) + parsedValue = strtod(o->ptr, &eptr); + else if (o->encoding == REDIS_ENCODING_INT) + parsedValue = (long)o->ptr; + else + redisAssert(1 != 1); + + if (eptr != NULL && *eptr != '\0') { + addReplySds(c,sdsnew("-ERR value is not a double\r\n")); + return REDIS_ERR; + } + + *value = parsedValue; + + return REDIS_OK; +} + +static int getLongLongFromObject(redisClient *c, robj *o, long long *value) { + long long parsedValue; + char *eptr = NULL; + + if (o && o->type != REDIS_STRING) { + addReplySds(c,sdsnew("-ERR value is not an integer\r\n")); + return REDIS_ERR; + } + + if (o == NULL) + parsedValue = 0; + else if (o->encoding == REDIS_ENCODING_RAW) + parsedValue = strtoll(o->ptr, &eptr, 10); + else if (o->encoding == REDIS_ENCODING_INT) + parsedValue = (long)o->ptr; + else + redisAssert(1 != 1); + + if (eptr != NULL && *eptr != '\0') { + addReplySds(c,sdsnew("-ERR value is not an integer\r\n")); + return REDIS_ERR; + } + + *value = parsedValue; + + return REDIS_OK; +} + +static int getLongFromObject(redisClient *c, robj *o, long *value) { + long long actualValue; + + if (getLongLongFromObject(c, o, &actualValue) != REDIS_OK) return REDIS_ERR; + + if (actualValue < LONG_MIN || actualValue > LONG_MAX) { + addReplySds(c,sdsnew("-ERR value is out of range\r\n")); + return REDIS_ERR; + } + + *value = actualValue; + + return REDIS_OK; +} + /*============================ RDB saving/loading =========================== */ static int rdbSaveType(FILE *fp, unsigned char type) { @@ -3354,7 +3427,7 @@ static off_t rdbSavedObjectLen(robj *o, FILE *fp) { /* Return the number of pages required to save this object in the swap file */ static off_t rdbSavedObjectPages(robj *o, FILE *fp) { off_t bytes = rdbSavedObjectLen(o,fp); - + return (bytes+(server.vm_page_size-1))/server.vm_page_size; } @@ -3437,7 +3510,7 @@ static int rdbSave(char *filename) { fflush(fp); fsync(fileno(fp)); fclose(fp); - + /* Use RENAME to make sure the DB file is changed atomically only * if the generate DB file is ok. */ if (rename(tmpfile,filename) == -1) { @@ -3481,6 +3554,7 @@ static int rdbSaveBackground(char *filename) { } redisLog(REDIS_NOTICE,"Background saving started by pid %d",childpid); server.bgsavechildpid = childpid; + updateDictResizePolicy(); return REDIS_OK; } return REDIS_OK; /* unreached */ @@ -3590,9 +3664,9 @@ static robj *rdbLoadStringObject(FILE*fp) { case REDIS_RDB_ENC_INT8: case REDIS_RDB_ENC_INT16: case REDIS_RDB_ENC_INT32: - return tryObjectSharing(rdbLoadIntegerObject(fp,len)); + return rdbLoadIntegerObject(fp,len); case REDIS_RDB_ENC_LZF: - return tryObjectSharing(rdbLoadLzfStringObject(fp)); + return rdbLoadLzfStringObject(fp); default: redisAssert(0); } @@ -3604,7 +3678,7 @@ static robj *rdbLoadStringObject(FILE*fp) { sdsfree(val); return NULL; } - return tryObjectSharing(createObject(REDIS_STRING,val)); + return createObject(REDIS_STRING,val); } /* For information about double serialization check rdbSaveDoubleValue() */ @@ -3634,7 +3708,7 @@ static robj *rdbLoadObject(int type, FILE *fp) { if (type == REDIS_STRING) { /* Read string value */ if ((o = rdbLoadStringObject(fp)) == NULL) return NULL; - tryObjectEncoding(o); + o = tryObjectEncoding(o); } else if (type == REDIS_LIST || type == REDIS_SET) { /* Read list/set value */ uint32_t listlen; @@ -3650,7 +3724,7 @@ static robj *rdbLoadObject(int type, FILE *fp) { robj *ele; if ((ele = rdbLoadStringObject(fp)) == NULL) return NULL; - tryObjectEncoding(ele); + ele = tryObjectEncoding(ele); if (type == REDIS_LIST) { listAddNodeTail((list*)o->ptr,ele); } else { @@ -3671,7 +3745,7 @@ static robj *rdbLoadObject(int type, FILE *fp) { double *score = zmalloc(sizeof(double)); if ((ele = rdbLoadStringObject(fp)) == NULL) return NULL; - tryObjectEncoding(ele); + ele = tryObjectEncoding(ele); if (rdbLoadDoubleValue(fp,score) == -1) return NULL; dictAdd(zs->dict,ele,score); zslInsert(zs->zsl,*score,ele); @@ -3710,8 +3784,8 @@ static robj *rdbLoadObject(int type, FILE *fp) { decrRefCount(key); decrRefCount(val); } else { - tryObjectEncoding(key); - tryObjectEncoding(val); + key = tryObjectEncoding(key); + val = tryObjectEncoding(val); dictAdd((dict*)o->ptr,key,val); } } @@ -3866,7 +3940,7 @@ static void setnxCommand(redisClient *c) { static int getGenericCommand(redisClient *c) { robj *o; - + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL) return REDIS_OK; @@ -3897,7 +3971,7 @@ static void getsetCommand(redisClient *c) { static void mgetCommand(redisClient *c) { int j; - + addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",c->argc-1)); for (j = 1; j < c->argc; j++) { robj *o = lookupKeyRead(c->db,c->argv[j]); @@ -3937,7 +4011,7 @@ static void msetGenericCommand(redisClient *c, int nx) { for (j = 1; j < c->argc; j += 2) { int retval; - tryObjectEncoding(c->argv[j+1]); + c->argv[j+1] = tryObjectEncoding(c->argv[j+1]); retval = dictAdd(c->db->dict,c->argv[j],c->argv[j+1]); if (retval == DICT_ERR) { dictReplace(c->db->dict,c->argv[j],c->argv[j+1]); @@ -3964,28 +4038,14 @@ static void incrDecrCommand(redisClient *c, long long incr) { long long value; int retval; robj *o; - + o = lookupKeyWrite(c->db,c->argv[1]); - if (o == NULL) { - value = 0; - } else { - if (o->type != REDIS_STRING) { - value = 0; - } else { - char *eptr; - if (o->encoding == REDIS_ENCODING_RAW) - value = strtoll(o->ptr, &eptr, 10); - else if (o->encoding == REDIS_ENCODING_INT) - value = (long)o->ptr; - else - redisAssert(1 != 1); - } - } + if (getLongLongFromObject(c, o, &value) != REDIS_OK) return; value += incr; o = createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",value)); - tryObjectEncoding(o); + o = tryObjectEncoding(o); retval = dictAdd(c->db->dict,c->argv[1],o); if (retval == DICT_ERR) { dictReplace(c->db->dict,c->argv[1],o); @@ -4008,12 +4068,18 @@ static void decrCommand(redisClient *c) { } static void incrbyCommand(redisClient *c) { - long long incr = strtoll(c->argv[2]->ptr, NULL, 10); + long long incr; + + if (getLongLongFromObject(c, c->argv[2], &incr) != REDIS_OK) return; + incrDecrCommand(c,incr); } static void decrbyCommand(redisClient *c) { - long long incr = strtoll(c->argv[2]->ptr, NULL, 10); + long long incr; + + if (getLongLongFromObject(c, c->argv[2], &incr) != REDIS_OK) return; + incrDecrCommand(c,-incr); } @@ -4031,7 +4097,7 @@ static void appendCommand(redisClient *c) { totlen = stringObjectLen(c->argv[2]); } else { dictEntry *de; - + de = dictFind(c->db->dict,c->argv[1]); assert(de != NULL); @@ -4120,7 +4186,7 @@ static void existsCommand(redisClient *c) { static void selectCommand(redisClient *c) { int id = atoi(c->argv[1]->ptr); - + if (selectDb(c,id) == REDIS_ERR) { addReplySds(c,sdsnew("-ERR invalid DB index\r\n")); } else { @@ -4130,7 +4196,7 @@ static void selectCommand(redisClient *c) { static void randomkeyCommand(redisClient *c) { dictEntry *de; - + while(1) { de = dictGetRandomKey(c->db->dict); if (!de || expireIfNeeded(c->db,dictGetEntryKey(de)) == 0) break; @@ -4258,7 +4324,7 @@ static void shutdownCommand(redisClient *c) { * in the next cron() Redis will be notified that the background * saving aborted, handling special stuff like slaves pending for * synchronization... */ - redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit"); + redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit"); addReplySds(c, sdsnew("-ERR can't quit, problems saving the DB\r\n")); } @@ -4402,7 +4468,7 @@ static void llenCommand(redisClient *c) { if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,o,REDIS_LIST)) return; - + l = o->ptr; addReplyUlong(c,listLength(l)); } @@ -4493,8 +4559,8 @@ static void lrangeCommand(redisClient *c) { listNode *ln; robj *ele; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullmultibulk)) == NULL || - checkType(c,o,REDIS_LIST)) return; + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptymultibulk)) == NULL + || checkType(c,o,REDIS_LIST)) return; list = o->ptr; llen = listLength(list); @@ -4758,7 +4824,7 @@ static void scardCommand(redisClient *c) { if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,o,REDIS_SET)) return; - + s = o->ptr; addReplyUlong(c,dictSize(s)); } @@ -4827,7 +4893,7 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, unsigned long server.dirty++; addReply(c,shared.czero); } else { - addReply(c,shared.nullmultibulk); + addReply(c,shared.emptymultibulk); } return; } @@ -5049,7 +5115,7 @@ static zskiplistNode *zslCreateNode(int level, double score, robj *obj) { static zskiplist *zslCreate(void) { int j; zskiplist *zsl; - + zsl = zmalloc(sizeof(*zsl)); zsl->level = 1; zsl->length = 0; @@ -5091,7 +5157,7 @@ static int zslRandomLevel(void) { int level = 1; while ((random()&0xFFFF) < (ZSKIPLIST_P * 0xFFFF)) level += 1; - return level; + return (leveldict,ele); redisAssert(de != NULL); @@ -5411,14 +5477,16 @@ static void zaddGenericCommand(redisClient *c, robj *key, robj *ele, double scor static void zaddCommand(redisClient *c) { double scoreval; - scoreval = strtod(c->argv[2]->ptr,NULL); + if (getDoubleFromObject(c, c->argv[2], &scoreval) != REDIS_OK) return; + zaddGenericCommand(c,c->argv[1],c->argv[3],scoreval,0); } static void zincrbyCommand(redisClient *c) { double scoreval; - scoreval = strtod(c->argv[2]->ptr,NULL); + if (getDoubleFromObject(c, c->argv[2], &scoreval) != REDIS_OK) return; + zaddGenericCommand(c,c->argv[1],c->argv[3],scoreval,1); } @@ -5452,12 +5520,15 @@ static void zremCommand(redisClient *c) { } static void zremrangebyscoreCommand(redisClient *c) { - double min = strtod(c->argv[2]->ptr,NULL); - double max = strtod(c->argv[3]->ptr,NULL); + double min; + double max; long deleted; robj *zsetobj; zset *zs; + if ((getDoubleFromObject(c, c->argv[2], &min) != REDIS_OK) || + (getDoubleFromObject(c, c->argv[3], &max) != REDIS_OK)) return; + if ((zsetobj = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,zsetobj,REDIS_ZSET)) return; @@ -5470,13 +5541,16 @@ static void zremrangebyscoreCommand(redisClient *c) { } static void zremrangebyrankCommand(redisClient *c) { - int start = atoi(c->argv[2]->ptr); - int end = atoi(c->argv[3]->ptr); + long start; + long end; int llen; long deleted; robj *zsetobj; zset *zs; + if ((getLongFromObject(c, c->argv[2], &start) != REDIS_OK) || + (getLongFromObject(c, c->argv[3], &end) != REDIS_OK)) return; + if ((zsetobj = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,zsetobj,REDIS_ZSET)) return; zs = zsetobj->ptr; @@ -5583,7 +5657,8 @@ static void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) { if (remaining >= (zsetnum + 1) && !strcasecmp(c->argv[j]->ptr,"weights")) { j++; remaining--; for (i = 0; i < zsetnum; i++, j++, remaining--) { - src[i].weight = strtod(c->argv[j]->ptr, NULL); + if (getDoubleFromObject(c, c->argv[j], &src[i].weight) != REDIS_OK) + return; } } else if (remaining >= 2 && !strcasecmp(c->argv[j]->ptr,"aggregate")) { j++; remaining--; @@ -5705,8 +5780,8 @@ static void zinterCommand(redisClient *c) { static void zrangeGenericCommand(redisClient *c, int reverse) { robj *o; - int start = atoi(c->argv[2]->ptr); - int end = atoi(c->argv[3]->ptr); + long start; + long end; int withscores = 0; int llen; int rangelen, j; @@ -5715,6 +5790,9 @@ static void zrangeGenericCommand(redisClient *c, int reverse) { zskiplistNode *ln; robj *ele; + if ((getLongFromObject(c, c->argv[2], &start) != REDIS_OK) || + (getLongFromObject(c, c->argv[3], &end) != REDIS_OK)) return; + if (c->argc == 5 && !strcasecmp(c->argv[4]->ptr,"withscores")) { withscores = 1; } else if (c->argc >= 5) { @@ -5722,8 +5800,8 @@ static void zrangeGenericCommand(redisClient *c, int reverse) { return; } - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullmultibulk)) == NULL || - checkType(c,o,REDIS_ZSET)) return; + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptymultibulk)) == NULL + || checkType(c,o,REDIS_ZSET)) return; zsetobj = o->ptr; zsl = zsetobj->zsl; llen = zsl->length; @@ -5829,7 +5907,7 @@ static void genericZrangebyscoreCommand(redisClient *c, int justcount) { /* Ok, lookup the key and get the range */ o = lookupKeyRead(c->db,c->argv[1]); if (o == NULL) { - addReply(c,justcount ? shared.czero : shared.nullmultibulk); + addReply(c,justcount ? shared.czero : shared.emptymultibulk); } else { if (o->type != REDIS_ZSET) { addReply(c,shared.wrongtypeerr); @@ -5966,295 +6044,378 @@ static void zrevrankCommand(redisClient *c) { zrankGenericCommand(c, 1); } -/* =================================== Hashes =============================== */ -static void hsetCommand(redisClient *c) { - int update = 0; - robj *o = lookupKeyWrite(c->db,c->argv[1]); +/* ========================= Hashes utility functions ======================= */ +#define REDIS_HASH_KEY 1 +#define REDIS_HASH_VALUE 2 - if (o == NULL) { - o = createHashObject(); - dictAdd(c->db->dict,c->argv[1],o); - incrRefCount(c->argv[1]); - } else { - if (o->type != REDIS_HASH) { - addReply(c,shared.wrongtypeerr); +/* Check the length of a number of objects to see if we need to convert a + * zipmap to a real hash. Note that we only check string encoded objects + * as their string length can be queried in constant time. */ +static void hashTryConversion(robj *subject, robj **argv, int start, int end) { + int i; + if (subject->encoding != REDIS_ENCODING_ZIPMAP) return; + + for (i = start; i <= end; i++) { + if (argv[i]->encoding == REDIS_ENCODING_RAW && + sdslen(argv[i]->ptr) > server.hash_max_zipmap_value) + { + convertToRealHash(subject); return; } } - /* We want to convert the zipmap into an hash table right now if the - * entry to be added is too big. Note that we check if the object - * is integer encoded before to try fetching the length in the test below. - * This is because integers are small, but currently stringObjectLen() - * performs a slow conversion: not worth it. */ - if (o->encoding == REDIS_ENCODING_ZIPMAP && - ((c->argv[2]->encoding == REDIS_ENCODING_RAW && - sdslen(c->argv[2]->ptr) > server.hash_max_zipmap_value) || - (c->argv[3]->encoding == REDIS_ENCODING_RAW && - sdslen(c->argv[3]->ptr) > server.hash_max_zipmap_value))) - { - convertToRealHash(o); +} + +/* Get the value from a hash identified by key. Returns either a string + * object or NULL if the value cannot be found. The refcount of the object + * is always increased by 1 when the value was found. */ +static robj *hashGet(robj *o, robj *key) { + robj *value = NULL; + if (o->encoding == REDIS_ENCODING_ZIPMAP) { + unsigned char *v; + unsigned int vlen; + key = getDecodedObject(key); + if (zipmapGet(o->ptr,key->ptr,sdslen(key->ptr),&v,&vlen)) { + value = createStringObject((char*)v,vlen); + } + decrRefCount(key); + } else { + dictEntry *de = dictFind(o->ptr,key); + if (de != NULL) { + value = dictGetEntryVal(de); + incrRefCount(value); + } } + return value; +} +/* Test if the key exists in the given hash. Returns 1 if the key + * exists and 0 when it doesn't. */ +static int hashExists(robj *o, robj *key) { if (o->encoding == REDIS_ENCODING_ZIPMAP) { - unsigned char *zm = o->ptr; - robj *valobj = getDecodedObject(c->argv[3]); + key = getDecodedObject(key); + if (zipmapExists(o->ptr,key->ptr,sdslen(key->ptr))) { + decrRefCount(key); + return 1; + } + decrRefCount(key); + } else { + if (dictFind(o->ptr,key) != NULL) { + return 1; + } + } + return 0; +} - zm = zipmapSet(zm,c->argv[2]->ptr,sdslen(c->argv[2]->ptr), - valobj->ptr,sdslen(valobj->ptr),&update); - decrRefCount(valobj); - o->ptr = zm; +/* Add an element, discard the old if the key already exists. + * Return 0 on insert and 1 on update. */ +static int hashSet(robj *o, robj *key, robj *value) { + int update = 0; + if (o->encoding == REDIS_ENCODING_ZIPMAP) { + key = getDecodedObject(key); + value = getDecodedObject(value); + o->ptr = zipmapSet(o->ptr, + key->ptr,sdslen(key->ptr), + value->ptr,sdslen(value->ptr), &update); + decrRefCount(key); + decrRefCount(value); - /* And here there is the second check for hash conversion. */ - if (zipmapLen(zm) > server.hash_max_zipmap_entries) + /* Check if the zipmap needs to be upgraded to a real hash table */ + if (zipmapLen(o->ptr) > server.hash_max_zipmap_entries) convertToRealHash(o); } else { - tryObjectEncoding(c->argv[2]); - /* note that c->argv[3] is already encoded, as the latest arg - * of a bulk command is always integer encoded if possible. */ - if (dictReplace(o->ptr,c->argv[2],c->argv[3])) { - incrRefCount(c->argv[2]); + value = tryObjectEncoding(value); + if (dictReplace(o->ptr,key,value)) { + /* Insert */ + incrRefCount(key); } else { + /* Update */ update = 1; } - incrRefCount(c->argv[3]); + incrRefCount(value); } - server.dirty++; - addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",update == 0)); + return update; } -static void hincrbyCommand(redisClient *c) { - long long value = 0, incr = 0; - robj *o = lookupKeyWrite(c->db,c->argv[1]); +/* Delete an element from a hash. + * Return 1 on deleted and 0 on not found. */ +static int hashDelete(robj *o, robj *key) { + int deleted = 0; + if (o->encoding == REDIS_ENCODING_ZIPMAP) { + key = getDecodedObject(key); + o->ptr = zipmapDel(o->ptr,key->ptr,sdslen(key->ptr), &deleted); + decrRefCount(key); + } else { + deleted = dictDelete((dict*)o->ptr,key) == DICT_OK; + /* Always check if the dictionary needs a resize after a delete. */ + if (deleted && htNeedsResize(o->ptr)) dictResize(o->ptr); + } + return deleted; +} + +/* Return the number of elements in a hash. */ +static unsigned long hashLength(robj *o) { + return (o->encoding == REDIS_ENCODING_ZIPMAP) ? + zipmapLen((unsigned char*)o->ptr) : dictSize((dict*)o->ptr); +} + +/* Structure to hold hash iteration abstration. Note that iteration over + * hashes involves both fields and values. Because it is possible that + * not both are required, store pointers in the iterator to avoid + * unnecessary memory allocation for fields/values. */ +typedef struct { + int encoding; + unsigned char *zi; + unsigned char *zk, *zv; + unsigned int zklen, zvlen; + + dictIterator *di; + dictEntry *de; +} hashIterator; + +static hashIterator *hashInitIterator(robj *subject) { + hashIterator *hi = zmalloc(sizeof(hashIterator)); + hi->encoding = subject->encoding; + if (hi->encoding == REDIS_ENCODING_ZIPMAP) { + hi->zi = zipmapRewind(subject->ptr); + } else if (hi->encoding == REDIS_ENCODING_HT) { + hi->di = dictGetIterator(subject->ptr); + } else { + redisAssert(NULL); + } + return hi; +} +static void hashReleaseIterator(hashIterator *hi) { + if (hi->encoding == REDIS_ENCODING_HT) { + dictReleaseIterator(hi->di); + } + zfree(hi); +} + +/* Move to the next entry in the hash. Return REDIS_OK when the next entry + * could be found and REDIS_ERR when the iterator reaches the end. */ +static int hashNext(hashIterator *hi) { + if (hi->encoding == REDIS_ENCODING_ZIPMAP) { + if ((hi->zi = zipmapNext(hi->zi, &hi->zk, &hi->zklen, + &hi->zv, &hi->zvlen)) == NULL) return REDIS_ERR; + } else { + if ((hi->de = dictNext(hi->di)) == NULL) return REDIS_ERR; + } + return REDIS_OK; +} + +/* Get key or value object at current iteration position. + * This increases the refcount of the field object by 1. */ +static robj *hashCurrent(hashIterator *hi, int what) { + robj *o; + if (hi->encoding == REDIS_ENCODING_ZIPMAP) { + if (what & REDIS_HASH_KEY) { + o = createStringObject((char*)hi->zk,hi->zklen); + } else { + o = createStringObject((char*)hi->zv,hi->zvlen); + } + } else { + if (what & REDIS_HASH_KEY) { + o = dictGetEntryKey(hi->de); + } else { + o = dictGetEntryVal(hi->de); + } + incrRefCount(o); + } + return o; +} + +static robj *hashLookupWriteOrCreate(redisClient *c, robj *key) { + robj *o = lookupKeyWrite(c->db,key); if (o == NULL) { o = createHashObject(); - dictAdd(c->db->dict,c->argv[1],o); - incrRefCount(c->argv[1]); + dictAdd(c->db->dict,key,o); + incrRefCount(key); } else { if (o->type != REDIS_HASH) { addReply(c,shared.wrongtypeerr); - return; + return NULL; } } + return o; +} - robj *o_incr = getDecodedObject(c->argv[3]); - incr = strtoll(o_incr->ptr, NULL, 10); - decrRefCount(o_incr); +/* ============================= Hash commands ============================== */ +static void hsetCommand(redisClient *c) { + int update; + robj *o; - if (o->encoding == REDIS_ENCODING_ZIPMAP) { - unsigned char *zm = o->ptr; - unsigned char *zval; - unsigned int zvlen; - - /* Find value if already present in hash */ - if (zipmapGet(zm,c->argv[2]->ptr,sdslen(c->argv[2]->ptr), - &zval,&zvlen)) { - /* strtoll needs the char* to have a trailing \0, but - * the zipmap doesn't include them. */ - sds szval = sdsnewlen(zval, zvlen); - value = strtoll(szval,NULL,10); - sdsfree(szval); - } - - value += incr; - sds svalue = sdscatprintf(sdsempty(),"%lld",value); - zm = zipmapSet(zm,c->argv[2]->ptr,sdslen(c->argv[2]->ptr), - (unsigned char*)svalue,sdslen(svalue),NULL); - sdsfree(svalue); - o->ptr = zm; - - /* Check if the zipmap needs to be converted. */ - if (zipmapLen(zm) > server.hash_max_zipmap_entries) - convertToRealHash(o); + if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return; + hashTryConversion(o,c->argv,2,3); + update = hashSet(o,c->argv[2],c->argv[3]); + addReply(c, update ? shared.czero : shared.cone); + server.dirty++; +} + +static void hsetnxCommand(redisClient *c) { + robj *o; + if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return; + hashTryConversion(o,c->argv,2,3); + + if (hashExists(o, c->argv[2])) { + addReply(c, shared.czero); } else { - robj *hval; - dictEntry *de; + hashSet(o,c->argv[2],c->argv[3]); + addReply(c, shared.cone); + server.dirty++; + } +} - /* Find value if already present in hash */ - de = dictFind(o->ptr,c->argv[2]); - if (de != NULL) { - hval = dictGetEntryVal(de); - if (hval->encoding == REDIS_ENCODING_RAW) - value = strtoll(hval->ptr,NULL,10); - else if (hval->encoding == REDIS_ENCODING_INT) - value = (long)hval->ptr; - else - redisAssert(1 != 1); - } +static void hmsetCommand(redisClient *c) { + int i; + robj *o; - value += incr; - hval = createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",value)); - tryObjectEncoding(hval); - if (dictReplace(o->ptr,c->argv[2],hval)) { - incrRefCount(c->argv[2]); - } + if ((c->argc % 2) == 1) { + addReplySds(c,sdsnew("-ERR wrong number of arguments for HMSET\r\n")); + return; } + if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return; + hashTryConversion(o,c->argv,2,c->argc-1); + for (i = 2; i < c->argc; i += 2) { + hashSet(o,c->argv[i],c->argv[i+1]); + } + addReply(c, shared.ok); +} + +static void hincrbyCommand(redisClient *c) { + long long value, incr; + robj *o, *current, *new; + + if (getLongLongFromObject(c,c->argv[3],&incr) != REDIS_OK) return; + if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return; + if ((current = hashGet(o,c->argv[2])) != NULL) { + if (current->encoding == REDIS_ENCODING_RAW) + value = strtoll(current->ptr,NULL,10); + else if (current->encoding == REDIS_ENCODING_INT) + value = (long)current->ptr; + else + redisAssert(1 != 1); + decrRefCount(current); + } else { + value = 0; + } + + value += incr; + new = createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",value)); + hashSet(o,c->argv[2],new); + decrRefCount(new); + addReplyLongLong(c,value); server.dirty++; - addReplyLong(c, value); } static void hgetCommand(redisClient *c) { - robj *o; - + robj *o, *value; if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL || checkType(c,o,REDIS_HASH)) return; - if (o->encoding == REDIS_ENCODING_ZIPMAP) { - unsigned char *zm = o->ptr; - unsigned char *val; - unsigned int vlen; - robj *field; - - field = getDecodedObject(c->argv[2]); - if (zipmapGet(zm,field->ptr,sdslen(field->ptr), &val,&vlen)) { - addReplySds(c,sdscatprintf(sdsempty(),"$%u\r\n", vlen)); - addReplySds(c,sdsnewlen(val,vlen)); - addReply(c,shared.crlf); - decrRefCount(field); - return; - } else { - addReply(c,shared.nullbulk); - decrRefCount(field); - return; - } + if ((value = hashGet(o,c->argv[2])) != NULL) { + addReplyBulk(c,value); + decrRefCount(value); } else { - struct dictEntry *de; + addReply(c,shared.nullbulk); + } +} - de = dictFind(o->ptr,c->argv[2]); - if (de == NULL) { - addReply(c,shared.nullbulk); - } else { - robj *e = dictGetEntryVal(de); +static void hmgetCommand(redisClient *c) { + int i; + robj *o, *value; + o = lookupKeyRead(c->db,c->argv[1]); + if (o != NULL && o->type != REDIS_HASH) { + addReply(c,shared.wrongtypeerr); + } - addReplyBulk(c,e); + /* Note the check for o != NULL happens inside the loop. This is + * done because objects that cannot be found are considered to be + * an empty hash. The reply should then be a series of NULLs. */ + addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",c->argc-2)); + for (i = 2; i < c->argc; i++) { + if (o != NULL && (value = hashGet(o,c->argv[i])) != NULL) { + addReplyBulk(c,value); + decrRefCount(value); + } else { + addReply(c,shared.nullbulk); } } } static void hdelCommand(redisClient *c) { robj *o; - int deleted = 0; - if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,o,REDIS_HASH)) return; - if (o->encoding == REDIS_ENCODING_ZIPMAP) { - robj *field = getDecodedObject(c->argv[2]); - - o->ptr = zipmapDel((unsigned char*) o->ptr, - (unsigned char*) field->ptr, - sdslen(field->ptr), &deleted); - decrRefCount(field); - if (zipmapLen((unsigned char*) o->ptr) == 0) - deleteKey(c->db,c->argv[1]); + if (hashDelete(o,c->argv[2])) { + if (hashLength(o) == 0) deleteKey(c->db,c->argv[1]); + addReply(c,shared.cone); + server.dirty++; } else { - deleted = dictDelete((dict*)o->ptr,c->argv[2]) == DICT_OK; - if (htNeedsResize(o->ptr)) dictResize(o->ptr); - if (dictSize((dict*)o->ptr) == 0) deleteKey(c->db,c->argv[1]); + addReply(c,shared.czero); } - if (deleted) server.dirty++; - addReply(c,deleted ? shared.cone : shared.czero); } static void hlenCommand(redisClient *c) { robj *o; - unsigned long len; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,o,REDIS_HASH)) return; - len = (o->encoding == REDIS_ENCODING_ZIPMAP) ? - zipmapLen((unsigned char*)o->ptr) : dictSize((dict*)o->ptr); - addReplyUlong(c,len); + addReplyUlong(c,hashLength(o)); } -#define REDIS_GETALL_KEYS 1 -#define REDIS_GETALL_VALS 2 static void genericHgetallCommand(redisClient *c, int flags) { - robj *o, *lenobj; + robj *o, *lenobj, *obj; unsigned long count = 0; + hashIterator *hi; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullmultibulk)) == NULL + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptymultibulk)) == NULL || checkType(c,o,REDIS_HASH)) return; lenobj = createObject(REDIS_STRING,NULL); addReply(c,lenobj); decrRefCount(lenobj); - if (o->encoding == REDIS_ENCODING_ZIPMAP) { - unsigned char *p = zipmapRewind(o->ptr); - unsigned char *field, *val; - unsigned int flen, vlen; - - while((p = zipmapNext(p,&field,&flen,&val,&vlen)) != NULL) { - robj *aux; - - if (flags & REDIS_GETALL_KEYS) { - aux = createStringObject((char*)field,flen); - addReplyBulk(c,aux); - decrRefCount(aux); - count++; - } - if (flags & REDIS_GETALL_VALS) { - aux = createStringObject((char*)val,vlen); - addReplyBulk(c,aux); - decrRefCount(aux); - count++; - } + hi = hashInitIterator(o); + while (hashNext(hi) != REDIS_ERR) { + if (flags & REDIS_HASH_KEY) { + obj = hashCurrent(hi,REDIS_HASH_KEY); + addReplyBulk(c,obj); + decrRefCount(obj); + count++; } - } else { - dictIterator *di = dictGetIterator(o->ptr); - dictEntry *de; - - while((de = dictNext(di)) != NULL) { - robj *fieldobj = dictGetEntryKey(de); - robj *valobj = dictGetEntryVal(de); - - if (flags & REDIS_GETALL_KEYS) { - addReplyBulk(c,fieldobj); - count++; - } - if (flags & REDIS_GETALL_VALS) { - addReplyBulk(c,valobj); - count++; - } + if (flags & REDIS_HASH_VALUE) { + obj = hashCurrent(hi,REDIS_HASH_VALUE); + addReplyBulk(c,obj); + decrRefCount(obj); + count++; } - dictReleaseIterator(di); } + hashReleaseIterator(hi); + lenobj->ptr = sdscatprintf(sdsempty(),"*%lu\r\n",count); } static void hkeysCommand(redisClient *c) { - genericHgetallCommand(c,REDIS_GETALL_KEYS); + genericHgetallCommand(c,REDIS_HASH_KEY); } static void hvalsCommand(redisClient *c) { - genericHgetallCommand(c,REDIS_GETALL_VALS); + genericHgetallCommand(c,REDIS_HASH_VALUE); } static void hgetallCommand(redisClient *c) { - genericHgetallCommand(c,REDIS_GETALL_KEYS|REDIS_GETALL_VALS); + genericHgetallCommand(c,REDIS_HASH_KEY|REDIS_HASH_VALUE); } static void hexistsCommand(redisClient *c) { robj *o; - int exists = 0; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,o,REDIS_HASH)) return; - if (o->encoding == REDIS_ENCODING_ZIPMAP) { - robj *field; - unsigned char *zm = o->ptr; - - field = getDecodedObject(c->argv[2]); - exists = zipmapExists(zm,field->ptr,sdslen(field->ptr)); - decrRefCount(field); - } else { - exists = dictFind(o->ptr,c->argv[2]) != NULL; - } - addReply(c,exists ? shared.cone : shared.czero); + addReply(c, hashExists(o,c->argv[2]) ? shared.cone : shared.czero); } static void convertToRealHash(robj *o) { @@ -6269,8 +6430,8 @@ static void convertToRealHash(robj *o) { keyobj = createStringObject((char*)key,klen); valobj = createStringObject((char*)val,vlen); - tryObjectEncoding(keyobj); - tryObjectEncoding(valobj); + keyobj = tryObjectEncoding(keyobj); + valobj = tryObjectEncoding(valobj); dictAdd(dict,keyobj,valobj); } o->encoding = REDIS_ENCODING_HT; @@ -6306,23 +6467,26 @@ static redisSortOperation *createSortOperation(int type, robj *pattern) { } /* Return the value associated to the key with a name obtained - * substituting the first occurence of '*' in 'pattern' with 'subst' */ + * substituting the first occurence of '*' in 'pattern' with 'subst'. + * The returned object will always have its refcount increased by 1 + * when it is non-NULL. */ static robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { - char *p; + char *p, *f; sds spat, ssub; - robj keyobj; - int prefixlen, sublen, postfixlen; + robj keyobj, fieldobj, *o; + int prefixlen, sublen, postfixlen, fieldlen; /* Expoit the internal sds representation to create a sds string allocated on the stack in order to make this function faster */ struct { long len; long free; char buf[REDIS_SORTKEY_MAX+1]; - } keyname; + } keyname, fieldname; /* If the pattern is "#" return the substitution object itself in order * to implement the "SORT ... GET #" feature. */ spat = pattern->ptr; if (spat[0] == '#' && spat[1] == '\0') { + incrRefCount(subst); return subst; } @@ -6339,20 +6503,47 @@ static robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { return NULL; } + /* Find out if we're dealing with a hash dereference. */ + if ((f = strstr(p+1, "->")) != NULL) { + fieldlen = sdslen(spat)-(f-spat); + /* this also copies \0 character */ + memcpy(fieldname.buf,f+2,fieldlen-1); + fieldname.len = fieldlen-2; + } else { + fieldlen = 0; + } + prefixlen = p-spat; sublen = sdslen(ssub); - postfixlen = sdslen(spat)-(prefixlen+1); + postfixlen = sdslen(spat)-(prefixlen+1)-fieldlen; memcpy(keyname.buf,spat,prefixlen); memcpy(keyname.buf+prefixlen,ssub,sublen); memcpy(keyname.buf+prefixlen+sublen,p+1,postfixlen); keyname.buf[prefixlen+sublen+postfixlen] = '\0'; keyname.len = prefixlen+sublen+postfixlen; - - initStaticStringObject(keyobj,((char*)&keyname)+(sizeof(long)*2)) decrRefCount(subst); - /* printf("lookup '%s' => %p\n", keyname.buf,de); */ - return lookupKeyRead(db,&keyobj); + /* Lookup substituted key */ + initStaticStringObject(keyobj,((char*)&keyname)+(sizeof(long)*2)); + o = lookupKeyRead(db,&keyobj); + if (o == NULL) return NULL; + + if (fieldlen > 0) { + if (o->type != REDIS_HASH || fieldname.len < 1) return NULL; + + /* Retrieve value from hash by the field name. This operation + * already increases the refcount of the returned object. */ + initStaticStringObject(fieldobj,((char*)&fieldname)+(sizeof(long)*2)); + o = hashGet(o, &fieldobj); + } else { + if (o->type != REDIS_STRING) return NULL; + + /* Every object that this function returns needs to have its refcount + * increased. sortCommand decreases it again. */ + incrRefCount(o); + } + + return o; } /* sortCompare() is used by qsort in sortCommand(). Given that qsort_r with @@ -6387,14 +6578,9 @@ static int sortCompare(const void *s1, const void *s2) { cmp = strcoll(so1->u.cmpobj->ptr,so2->u.cmpobj->ptr); } } else { - /* Compare elements directly */ - robj *dec1, *dec2; - - dec1 = getDecodedObject(so1->obj); - dec2 = getDecodedObject(so2->obj); - cmp = strcoll(dec1->ptr,dec2->ptr); - decrRefCount(dec1); - decrRefCount(dec2); + /* Compare elements directly. Note that these objects already + * need to be non-encoded (see sortCommand). */ + cmp = strcoll(so1->obj->ptr,so2->obj->ptr); } } return server.sort_desc ? -cmp : cmp; @@ -6415,7 +6601,7 @@ static void sortCommand(redisClient *c) { /* Lookup the key to sort. It must be of the right types */ sortval = lookupKeyRead(c->db,c->argv[1]); if (sortval == NULL) { - addReply(c,shared.nullmultibulk); + addReply(c,shared.emptymultibulk); return; } if (sortval->type != REDIS_SET && sortval->type != REDIS_LIST && @@ -6521,38 +6707,36 @@ static void sortCommand(redisClient *c) { /* Now it's time to load the right scores in the sorting vector */ if (dontsort == 0) { for (j = 0; j < vectorlen; j++) { + robj *byval; if (sortby) { - robj *byval; - + /* lookup value to sort by */ byval = lookupKeyByPattern(c->db,sortby,vector[j].obj); - if (!byval || byval->type != REDIS_STRING) continue; - if (alpha) { - vector[j].u.cmpobj = getDecodedObject(byval); - } else { - if (byval->encoding == REDIS_ENCODING_RAW) { - vector[j].u.score = strtod(byval->ptr,NULL); - } else { - /* Don't need to decode the object if it's - * integer-encoded (the only encoding supported) so - * far. We can just cast it */ - if (byval->encoding == REDIS_ENCODING_INT) { - vector[j].u.score = (long)byval->ptr; - } else - redisAssert(1 != 1); - } - } + if (!byval) continue; } else { - if (!alpha) { - if (vector[j].obj->encoding == REDIS_ENCODING_RAW) - vector[j].u.score = strtod(vector[j].obj->ptr,NULL); - else { - if (vector[j].obj->encoding == REDIS_ENCODING_INT) - vector[j].u.score = (long) vector[j].obj->ptr; - else - redisAssert(1 != 1); - } + /* use object itself to sort by */ + byval = vector[j].obj; + } + + if (alpha) { + vector[j].u.cmpobj = getDecodedObject(byval); + } else { + if (byval->encoding == REDIS_ENCODING_RAW) { + vector[j].u.score = strtod(byval->ptr,NULL); + } else if (byval->encoding == REDIS_ENCODING_INT) { + /* Don't need to decode the object if it's + * integer-encoded (the only encoding supported) so + * far. We can just cast it */ + vector[j].u.score = (long)byval->ptr; + } else { + redisAssert(1 != 1); } } + + /* when the object was retrieved using lookupKeyByPattern, + * its refcount needs to be decreased. */ + if (sortby) { + decrRefCount(byval); + } } } @@ -6594,10 +6778,11 @@ static void sortCommand(redisClient *c) { vector[j].obj); if (sop->type == REDIS_SORT_GET) { - if (!val || val->type != REDIS_STRING) { + if (!val) { addReply(c,shared.nullbulk); } else { addReplyBulk(c,val); + decrRefCount(val); } } else { redisAssert(sop->type == REDIS_SORT_GET); /* always fails */ @@ -6624,11 +6809,14 @@ static void sortCommand(redisClient *c) { vector[j].obj); if (sop->type == REDIS_SORT_GET) { - if (!val || val->type != REDIS_STRING) { + if (!val) { listAddNodeTail(listPtr,createStringObject("",0)); } 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); - incrRefCount(val); } } else { redisAssert(sop->type == REDIS_SORT_GET); /* always fails */ @@ -6649,7 +6837,7 @@ static void sortCommand(redisClient *c) { decrRefCount(sortval); listRelease(operations); for (j = 0; j < vectorlen; j++) { - if (sortby && alpha && vector[j].u.cmpobj) + if (alpha && vector[j].u.cmpobj) decrRefCount(vector[j].u.cmpobj); } zfree(vector); @@ -6872,15 +7060,20 @@ static int deleteIfVolatile(redisDb *db, robj *key) { return dictDelete(db->dict,key) == DICT_OK; } -static void expireGenericCommand(redisClient *c, robj *key, time_t seconds) { +static void expireGenericCommand(redisClient *c, robj *key, robj *param, long offset) { dictEntry *de; + time_t seconds; + + if (getLongFromObject(c, param, &seconds) != REDIS_OK) return; + + seconds -= offset; de = dictFind(c->db->dict,key); if (de == NULL) { addReply(c,shared.czero); return; } - if (seconds < 0) { + if (seconds <= 0) { if (deleteKey(c->db,key)) server.dirty++; addReply(c, shared.cone); return; @@ -6897,11 +7090,11 @@ static void expireGenericCommand(redisClient *c, robj *key, time_t seconds) { } static void expireCommand(redisClient *c) { - expireGenericCommand(c,c->argv[1],strtol(c->argv[2]->ptr,NULL,10)); + expireGenericCommand(c,c->argv[1],c->argv[2],0); } static void expireatCommand(redisClient *c) { - expireGenericCommand(c,c->argv[1],strtol(c->argv[2]->ptr,NULL,10)-time(NULL)); + expireGenericCommand(c,c->argv[1],c->argv[2],time(NULL)); } static void ttlCommand(redisClient *c) { @@ -7145,7 +7338,7 @@ static void blockingPopGenericCommand(redisClient *c, int where) { * non-blocking POP operation */ robj *argv[2], **orig_argv; int orig_argc; - + /* We need to alter the command arguments before to call * popGenericCommand() as the command takes a single key. */ orig_argv = c->argv; @@ -7382,7 +7575,7 @@ static void updateSlavesWaitingBgsave(int bgsaveerr) { slave->replstate = REDIS_REPL_WAIT_BGSAVE_END; } else if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) { struct redis_stat buf; - + if (bgsaveerr != REDIS_OK) { freeClient(slave); redisLog(REDIS_WARNING,"SYNC failed. BGSAVE child returned an error"); @@ -7780,14 +7973,9 @@ int loadAppendOnlyFile(char *filename) { redisLog(REDIS_WARNING,"Unknown command '%s' reading the append only file", argv[0]->ptr); exit(1); } - /* Try object sharing and encoding */ - if (server.shareobjects) { - int j; - for(j = 1; j < argc; j++) - argv[j] = tryObjectSharing(argv[j]); - } + /* Try object encoding */ if (cmd->flags & REDIS_CMD_BULK) - tryObjectEncoding(argv[argc-1]); + argv[argc-1] = tryObjectEncoding(argv[argc-1]); /* Run the command in the context of a fake client */ fakeClient->argc = argc; fakeClient->argv = argv; @@ -8044,7 +8232,7 @@ static int rewriteAppendOnlyFile(char *filename) { fflush(fp); fsync(fileno(fp)); fclose(fp); - + /* Use RENAME to make sure the DB file is changed atomically only * if the generate DB file is ok. */ if (rename(tmpfile,filename) == -1) { @@ -8103,6 +8291,7 @@ static int rewriteAppendOnlyFileBackground(void) { redisLog(REDIS_NOTICE, "Background append only file rewriting started by pid %d",childpid); server.bgrewritechildpid = childpid; + updateDictResizePolicy(); /* We set appendseldb to -1 in order to force the next call to the * feedAppendOnlyFile() to issue a SELECT command, so the differences * accumulated by the parent into server.bgrewritebuf will start @@ -8161,7 +8350,7 @@ static void aofRemoveTempFile(pid_t childpid) { static void expandVmSwapFilename(void) { char *p = strstr(server.vm_swap_file,"%p"); sds new; - + if (!p) return; new = sdsempty(); *p = '\0'; @@ -8286,7 +8475,7 @@ static int vmFreePage(off_t page) { } /* Find N contiguous free pages storing the first page of the cluster in *first. - * Returns REDIS_OK if it was able to find N contiguous pages, otherwise + * Returns REDIS_OK if it was able to find N contiguous pages, otherwise * REDIS_ERR is returned. * * This function uses a simple algorithm: we try to allocate @@ -8297,7 +8486,7 @@ static int vmFreePage(off_t page) { * we try to find less populated places doing a forward jump of * REDIS_VM_MAX_RANDOM_JUMP, then we start scanning again a few pages * without hurry, and then we jump again and so forth... - * + * * This function can be improved using a free list to avoid to guess * too much, since we could collect data about freed pages. * @@ -8670,7 +8859,10 @@ static void freeIOJob(iojob *j) { j->type == REDIS_IOJOB_DO_SWAP || j->type == REDIS_IOJOB_LOAD) && j->val != NULL) decrRefCount(j->val); - decrRefCount(j->key); + /* We don't decrRefCount the j->key field as we did't incremented + * the count creating IO Jobs. This is because the key field here is + * just used as an indentifier and if a key is removed the Job should + * never be touched again. */ zfree(j); } @@ -8843,7 +9035,7 @@ again: iojob *job = ln->value; if (job->canceled) continue; /* Skip this, already canceled. */ - if (compareStringObjects(job->key,o) == 0) { + if (job->key == o) { redisLog(REDIS_DEBUG,"*** CANCELED %p (%s) (type %d) (LIST ID %d)\n", (void*)job, (char*)o->ptr, job->type, i); /* Mark the pages as free since the swap didn't happened @@ -8948,7 +9140,7 @@ static void *IOThreadEntryPoint(void *arg) { listDelNode(server.io_processing,ln); listAddNodeTail(server.io_processed,j); unlockThreadedIO(); - + /* Signal the main thread there is new stuff to process */ assert(write(server.io_ready_pipe_write,"x",1) == 1); } @@ -9026,14 +9218,14 @@ static void queueIOJob(iojob *j) { static int vmSwapObjectThreaded(robj *key, robj *val, redisDb *db) { iojob *j; - + assert(key->storage == REDIS_VM_MEMORY); assert(key->refcount == 1); j = zmalloc(sizeof(*j)); j->type = REDIS_IOJOB_PREPARE_SWAP; j->db = db; - j->key = dupStringObject(key); + j->key = key; j->val = val; incrRefCount(val); j->canceled = 0; @@ -9070,7 +9262,7 @@ static int waitForSwappedKey(redisClient *c, robj *key) { vmCancelThreadedIOJob(o); return 0; } - + /* OK: the key is either swapped, or being loaded just now. */ /* Add the key to the list of keys this client is waiting for. @@ -9101,7 +9293,7 @@ static int waitForSwappedKey(redisClient *c, robj *key) { j = zmalloc(sizeof(*j)); j->type = REDIS_IOJOB_LOAD; j->db = c->db; - j->key = dupStringObject(key); + j->key = o; j->key->vtype = o->vtype; j->page = o->vm.page; j->val = NULL; @@ -9620,6 +9812,9 @@ static void debugCommand(redisClient *c) { (void*)key, key->refcount, (unsigned long long) key->vm.page, (unsigned long long) key->vm.usedpages)); } + } else if (!strcasecmp(c->argv[1]->ptr,"swapin") && c->argc == 3) { + lookupKeyRead(c->db,c->argv[2]); + addReply(c,shared.ok); } else if (!strcasecmp(c->argv[1]->ptr,"swapout") && c->argc == 3) { dictEntry *de = dictFind(c->db->dict,c->argv[2]); robj *key, *val; @@ -9651,7 +9846,7 @@ static void debugCommand(redisClient *c) { } } else { addReplySds(c,sdsnew( - "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT |SWAPOUT |RELOAD]\r\n")); + "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT |SWAPIN |SWAPOUT |RELOAD]\r\n")); } } @@ -9780,7 +9975,7 @@ static void *getMcontextEip(ucontext_t *uc) { return (void*) uc->uc_mcontext->__ss.__rip; #else return (void*) uc->uc_mcontext->__ss.__eip; - #endif + #endif #elif defined(__i386__) || defined(__X86_64__) || defined(__x86_64__) return (void*) uc->uc_mcontext.gregs[REG_EIP]; /* Linux 32/64 bit */ #elif defined(__ia64__) /* Linux IA64 */ @@ -9805,7 +10000,7 @@ static void segvHandler(int sig, siginfo_t *info, void *secret) { redisLog(REDIS_WARNING, "%s",infostring); /* It's not safe to sdsfree() the returned string under memory * corruption conditions. Let it leak as we are going to abort */ - + trace_size = backtrace(trace, 100); /* overwrite sigaction with caller's address */ if (getMcontextEip(uc) != NULL) {