X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/841053366f12c27dcd225f83c8ae7312326be32e..ee5cfe385305071ed0ab293be3ad18b13f2f9c86:/redis.c diff --git a/redis.c b/redis.c index ab0b9be4..8658bc19 100644 --- a/redis.c +++ b/redis.c @@ -75,6 +75,7 @@ #include "zmalloc.h" /* total memory usage aware version of malloc/free */ #include "lzf.h" /* LZF compression library */ #include "pqsort.h" /* Partial qsort for SORT+LIMIT */ +#include "zipmap.h" /* Error codes */ #define REDIS_OK 0 @@ -118,9 +119,13 @@ #define REDIS_ZSET 3 #define REDIS_HASH 4 -/* Objects encoding */ +/* 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 */ /* Object types only used for dumping to disk */ #define REDIS_EXPIRETIME 253 @@ -221,6 +226,10 @@ #define APPENDFSYNC_ALWAYS 1 #define APPENDFSYNC_EVERYSEC 2 +/* Hashes related defaults */ +#define REDIS_HASH_MAX_ZIPMAP_ENTRIES 64 +#define REDIS_HASH_MAX_ZIPMAP_VALUE 512 + /* We can print the stacktrace, so our assert is defined this way: */ #define redisAssert(_e) ((_e)?(void)0 : (_redisAssert(#_e,__FILE__,__LINE__),_exit(1))) static void _redisAssert(char *estr, char *file, int line); @@ -386,6 +395,9 @@ struct redisServer { off_t vm_page_size; off_t vm_pages; unsigned long long vm_max_memory; + /* Hashes config */ + size_t hash_max_zipmap_entries; + size_t hash_max_zipmap_value; /* Virtual memory state */ FILE *vm_fp; int vm_fd; @@ -578,6 +590,7 @@ static void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mas static struct redisCommand *lookupCommand(char *name); static void call(redisClient *c, struct redisCommand *cmd); static void resetClient(redisClient *c); +static void convertToRealHash(robj *o); static void authCommand(redisClient *c); static void pingCommand(redisClient *c); @@ -661,6 +674,9 @@ static void brpopCommand(redisClient *c); static void appendCommand(redisClient *c); static void substrCommand(redisClient *c); static void zrankCommand(redisClient *c); +static void zrevrankCommand(redisClient *c); +static void hsetCommand(redisClient *c); +static void hgetCommand(redisClient *c); static void zremrangebyrankCommand(redisClient *c); /*================================= Globals ================================= */ @@ -717,6 +733,9 @@ static struct redisCommand cmdTable[] = { {"zcard",zcardCommand,2,REDIS_CMD_INLINE,1,1,1}, {"zscore",zscoreCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,1,1,1}, {"zrank",zrankCommand,3,REDIS_CMD_INLINE,1,1,1}, + {"zrevrank",zrevrankCommand,3,REDIS_CMD_INLINE,1,1,1}, + {"hset",hsetCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,1,1,1}, + {"hget",hgetCommand,3,REDIS_CMD_BULK,1,1,1}, {"incrby",incrbyCommand,3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,1,1,1}, {"decrby",decrbyCommand,3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,1,1,1}, {"getset",getsetCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,1,1,1}, @@ -1013,7 +1032,7 @@ static dictType zsetDictType = { }; /* Db->dict */ -static dictType hashDictType = { +static dictType dbDictType = { dictObjHash, /* hash function */ NULL, /* key dup */ NULL, /* val dup */ @@ -1032,6 +1051,16 @@ static dictType keyptrDictType = { NULL /* val destructor */ }; +/* Hash type hash table (note that small hashes are represented with zimpaps) */ +static dictType hashDictType = { + dictEncObjHash, /* hash function */ + NULL, /* key dup */ + NULL, /* val dup */ + dictEncObjKeyCompare, /* key compare */ + dictRedisObjectDestructor, /* key destructor */ + dictRedisObjectDestructor /* val destructor */ +}; + /* Keylist hash table type has unencoded redis objects as keys and * lists as values. It's used for blocking operations (BLPOP) and to * map swapped keys to a list of clients waiting for this keys to be loaded. */ @@ -1448,6 +1477,8 @@ static void initServerConfig() { server.vm_max_memory = 1024LL*1024*1024*1; /* 1 GB of RAM */ server.vm_max_threads = 4; server.vm_blocked_clients = 0; + server.hash_max_zipmap_entries = REDIS_HASH_MAX_ZIPMAP_ENTRIES; + server.hash_max_zipmap_value = REDIS_HASH_MAX_ZIPMAP_VALUE; resetServerSaveParams(); @@ -1495,7 +1526,7 @@ static void initServer() { exit(1); } for (j = 0; j < server.dbnum; j++) { - server.db[j].dict = dictCreate(&hashDictType,NULL); + server.db[j].dict = dictCreate(&dbDictType,NULL); server.db[j].expires = dictCreate(&keyptrDictType,NULL); server.db[j].blockingkeys = dictCreate(&keylistDictType,NULL); if (server.vm_enabled) @@ -1708,6 +1739,12 @@ static void loadServerConfig(char *filename) { server.vm_pages = strtoll(argv[1], NULL, 10); } else if (!strcasecmp(argv[0],"vm-max-threads") && argc == 2) { server.vm_max_threads = strtoll(argv[1], NULL, 10); + } else if (!strcasecmp(argv[0],"hash-max-zipmap-entries") && argc == 2){ + server.hash_max_zipmap_entries = strtol(argv[1], NULL, 10); + } else if (!strcasecmp(argv[0],"hash-max-zipmap-value") && argc == 2){ + server.hash_max_zipmap_value = strtol(argv[1], NULL, 10); + } else if (!strcasecmp(argv[0],"vm-max-threads") && argc == 2) { + server.vm_max_threads = strtoll(argv[1], NULL, 10); } else { err = "Bad directive or wrong number of arguments"; goto loaderr; } @@ -2545,6 +2582,16 @@ static robj *createSetObject(void) { return createObject(REDIS_SET,d); } +static robj *createHashObject(void) { + /* All the Hashes start as zipmaps. Will be automatically converted + * into hash tables if there are enough elements or big elements + * inside. */ + unsigned char *zm = zipmapNew(); + robj *o = createObject(REDIS_HASH,zm); + o->encoding = REDIS_ENCODING_ZIPMAP; + return o; +} + static robj *createZsetObject(void) { zset *zs = zmalloc(sizeof(*zs)); @@ -2576,7 +2623,17 @@ static void freeZsetObject(robj *o) { } static void freeHashObject(robj *o) { - dictRelease((dict*) o->ptr); + switch (o->encoding) { + case REDIS_ENCODING_HT: + dictRelease((dict*) o->ptr); + break; + case REDIS_ENCODING_ZIPMAP: + zfree(o->ptr); + break; + default: + redisAssert(0); + break; + } } static void incrRefCount(robj *o) { @@ -2881,7 +2938,7 @@ static int rdbSaveLen(FILE *fp, uint32_t len) { /* String objects in the form "2391" "-100" without any space and with a * range of values that can fit in an 8, 16 or 32 bit signed value can be * encoded as integers to save space */ -static int rdbTryIntegerEncoding(sds s, unsigned char *enc) { +static int rdbTryIntegerEncoding(char *s, size_t len, unsigned char *enc) { long long value; char *endptr, buf[32]; @@ -2892,7 +2949,7 @@ static int rdbTryIntegerEncoding(sds s, unsigned char *enc) { /* If the number converted back into a string is not identical * then it's not possible to encode the string as integer */ - if (strlen(buf) != sdslen(s) || memcmp(buf,s,sdslen(s))) return 0; + if (strlen(buf) != len || memcmp(buf,s,len)) return 0; /* Finally check if it fits in our ranges */ if (value >= -(1<<7) && value <= (1<<7)-1) { @@ -2916,16 +2973,16 @@ static int rdbTryIntegerEncoding(sds s, unsigned char *enc) { } } -static int rdbSaveLzfStringObject(FILE *fp, robj *obj) { - unsigned int comprlen, outlen; +static int rdbSaveLzfStringObject(FILE *fp, unsigned char *s, size_t len) { + size_t comprlen, outlen; unsigned char byte; void *out; /* We require at least four bytes compression for this to be worth it */ - outlen = sdslen(obj->ptr)-4; - if (outlen <= 0) return 0; + if (len <= 4) return 0; + outlen = len-4; if ((out = zmalloc(outlen+1)) == NULL) return 0; - comprlen = lzf_compress(obj->ptr, sdslen(obj->ptr), out, outlen); + comprlen = lzf_compress(s, len, out, outlen); if (comprlen == 0) { zfree(out); return 0; @@ -2934,7 +2991,7 @@ static int rdbSaveLzfStringObject(FILE *fp, robj *obj) { byte = (REDIS_RDB_ENCVAL<<6)|REDIS_RDB_ENC_LZF; if (fwrite(&byte,1,1,fp) == 0) goto writeerr; if (rdbSaveLen(fp,comprlen) == -1) goto writeerr; - if (rdbSaveLen(fp,sdslen(obj->ptr)) == -1) goto writeerr; + if (rdbSaveLen(fp,len) == -1) goto writeerr; if (fwrite(out,comprlen,1,fp) == 0) goto writeerr; zfree(out); return comprlen; @@ -2946,16 +3003,13 @@ writeerr: /* Save a string objet as [len][data] on disk. If the object is a string * representation of an integer value we try to safe it in a special form */ -static int rdbSaveStringObjectRaw(FILE *fp, robj *obj) { - size_t len; +static int rdbSaveRawString(FILE *fp, unsigned char *s, size_t len) { int enclen; - len = sdslen(obj->ptr); - /* Try integer encoding */ if (len <= 11) { unsigned char buf[5]; - if ((enclen = rdbTryIntegerEncoding(obj->ptr,buf)) > 0) { + if ((enclen = rdbTryIntegerEncoding((char*)s,len,buf)) > 0) { if (fwrite(buf,enclen,1,fp) == 0) return -1; return 0; } @@ -2966,7 +3020,7 @@ static int rdbSaveStringObjectRaw(FILE *fp, robj *obj) { if (server.rdbcompression && len > 20) { int retval; - retval = rdbSaveLzfStringObject(fp,obj); + retval = rdbSaveLzfStringObject(fp,s,len); if (retval == -1) return -1; if (retval > 0) return 0; /* retval == 0 means data can't be compressed, save the old way */ @@ -2974,7 +3028,7 @@ static int rdbSaveStringObjectRaw(FILE *fp, robj *obj) { /* Store verbatim */ if (rdbSaveLen(fp,len) == -1) return -1; - if (len && fwrite(obj->ptr,len,1,fp) == 0) return -1; + if (len && fwrite(s,len,1,fp) == 0) return -1; return 0; } @@ -2989,10 +3043,10 @@ static int rdbSaveStringObject(FILE *fp, robj *obj) { * this in order to avoid bugs) */ if (obj->encoding != REDIS_ENCODING_RAW) { obj = getDecodedObject(obj); - retval = rdbSaveStringObjectRaw(fp,obj); + retval = rdbSaveRawString(fp,obj->ptr,sdslen(obj->ptr)); decrRefCount(obj); } else { - retval = rdbSaveStringObjectRaw(fp,obj); + retval = rdbSaveRawString(fp,obj->ptr,sdslen(obj->ptr)); } return retval; } @@ -3070,6 +3124,33 @@ static int rdbSaveObject(FILE *fp, robj *o) { if (rdbSaveDoubleValue(fp,*score) == -1) return -1; } dictReleaseIterator(di); + } else if (o->type == REDIS_HASH) { + /* Save a hash value */ + if (o->encoding == REDIS_ENCODING_ZIPMAP) { + unsigned char *p = zipmapRewind(o->ptr); + unsigned int count = zipmapLen(o->ptr); + unsigned char *key, *val; + unsigned int klen, vlen; + + if (rdbSaveLen(fp,count) == -1) return -1; + while((p = zipmapNext(p,&key,&klen,&val,&vlen)) != NULL) { + if (rdbSaveRawString(fp,key,klen) == -1) return -1; + if (rdbSaveRawString(fp,val,vlen) == -1) return -1; + } + } else { + dictIterator *di = dictGetIterator(o->ptr); + dictEntry *de; + + if (rdbSaveLen(fp,dictSize((dict*)o->ptr)) == -1) return -1; + while((de = dictNext(di)) != NULL) { + robj *key = dictGetEntryKey(de); + robj *val = dictGetEntryVal(de); + + if (rdbSaveStringObject(fp,key) == -1) return -1; + if (rdbSaveStringObject(fp,val) == -1) return -1; + } + dictReleaseIterator(di); + } } else { redisAssert(0 != 0); } @@ -3394,7 +3475,7 @@ static robj *rdbLoadObject(int type, FILE *fp) { } } else if (type == REDIS_ZSET) { /* Read list/set value */ - uint32_t zsetlen; + size_t zsetlen; zset *zs; if ((zsetlen = rdbLoadLen(fp,NULL)) == REDIS_RDB_LENERR) return NULL; @@ -3412,6 +3493,46 @@ static robj *rdbLoadObject(int type, FILE *fp) { zslInsert(zs->zsl,*score,ele); incrRefCount(ele); /* added to skiplist */ } + } else if (type == REDIS_HASH) { + size_t hashlen; + + if ((hashlen = rdbLoadLen(fp,NULL)) == REDIS_RDB_LENERR) return NULL; + o = createHashObject(); + /* Too many entries? Use an hash table. */ + if (hashlen > server.hash_max_zipmap_entries) + convertToRealHash(o); + /* Load every key/value, then set it into the zipmap or hash + * table, as needed. */ + while(hashlen--) { + robj *key, *val; + + if ((key = rdbLoadStringObject(fp)) == NULL) return NULL; + if ((val = rdbLoadStringObject(fp)) == NULL) return NULL; + /* If we are using a zipmap and there are too big values + * the object is converted to real hash table encoding. */ + if (o->encoding != REDIS_ENCODING_HT && + (sdslen(key->ptr) > server.hash_max_zipmap_value || + sdslen(val->ptr) > server.hash_max_zipmap_value)) + { + convertToRealHash(o); + } + + if (o->encoding == REDIS_ENCODING_ZIPMAP) { + unsigned char *zm = o->ptr; + + zm = zipmapSet(zm,key->ptr,sdslen(key->ptr), + val->ptr,sdslen(val->ptr),NULL); + o->ptr = zm; + decrRefCount(key); + decrRefCount(val); + } else { + tryObjectEncoding(key); + tryObjectEncoding(val); + dictAdd((dict*)o->ptr,key,val); + incrRefCount(key); + incrRefCount(val); + } + } } else { redisAssert(0 != 0); } @@ -3919,7 +4040,8 @@ static void typeCommand(redisClient *c) { case REDIS_LIST: type = "+list"; break; case REDIS_SET: type = "+set"; break; case REDIS_ZSET: type = "+zset"; break; - default: type = "unknown"; break; + case REDIS_HASH: type = "+hash"; break; + default: type = "+unknown"; break; } } addReplySds(c,sdsnew(type)); @@ -5575,7 +5697,7 @@ static void zscoreCommand(redisClient *c) { } } -static void zrankCommand(redisClient *c) { +static void zrankGenericCommand(redisClient *c, int reverse) { robj *o; o = lookupKeyRead(c->db,c->argv[1]); if (o == NULL) { @@ -5599,13 +5721,119 @@ static void zrankCommand(redisClient *c) { double *score = dictGetEntryVal(de); rank = zslGetRank(zsl, *score, c->argv[2]); if (rank) { - addReplyLong(c, rank-1); + if (reverse) { + addReplyLong(c, zsl->length - rank); + } else { + addReplyLong(c, rank-1); + } } else { addReply(c,shared.nullbulk); } } } +static void zrankCommand(redisClient *c) { + zrankGenericCommand(c, 0); +} + +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]); + + 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); + return; + } + } + if (o->encoding == REDIS_ENCODING_ZIPMAP) { + unsigned char *zm = o->ptr; + robj *valobj = getDecodedObject(c->argv[3]); + + zm = zipmapSet(zm,c->argv[2]->ptr,sdslen(c->argv[2]->ptr), + valobj->ptr,sdslen(valobj->ptr),&update); + decrRefCount(valobj); + o->ptr = zm; + } else { + if (dictAdd(o->ptr,c->argv[2],c->argv[3]) == DICT_OK) { + incrRefCount(c->argv[2]); + } else { + update = 1; + } + incrRefCount(c->argv[3]); + } + server.dirty++; + addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",update == 0)); +} + +static void hgetCommand(redisClient *c) { + robj *o = lookupKeyRead(c->db,c->argv[1]); + + if (o == NULL) { + addReply(c,shared.nullbulk); + return; + } else { + if (o->encoding == REDIS_ENCODING_ZIPMAP) { + unsigned char *zm = o->ptr; + unsigned char *val; + unsigned int vlen; + + if (zipmapGet(zm,c->argv[2]->ptr,sdslen(c->argv[2]->ptr), &val,&vlen)) { + addReplySds(c,sdscatprintf(sdsempty(),"$%u\r\n", vlen)); + addReplySds(c,sdsnewlen(val,vlen)); + addReply(c,shared.crlf); + return; + } else { + addReply(c,shared.nullbulk); + return; + } + } else { + struct dictEntry *de; + + de = dictFind(o->ptr,c->argv[2]); + if (de == NULL) { + addReply(c,shared.nullbulk); + } else { + robj *e = dictGetEntryVal(de); + + addReplyBulkLen(c,e); + addReply(c,e); + addReply(c,shared.crlf); + } + } + } +} + +static void convertToRealHash(robj *o) { + unsigned char *key, *val, *p, *zm = o->ptr; + unsigned int klen, vlen; + dict *dict = dictCreate(&hashDictType,NULL); + + assert(o->type == REDIS_HASH && o->encoding != REDIS_ENCODING_HT); + p = zipmapRewind(zm); + while((p = zipmapNext(p,&key,&klen,&val,&vlen)) != NULL) { + robj *keyobj, *valobj; + + keyobj = createStringObject((char*)key,klen); + valobj = createStringObject((char*)val,vlen); + tryObjectEncoding(keyobj); + tryObjectEncoding(valobj); + dictAdd(dict,keyobj,valobj); + } + o->encoding = REDIS_ENCODING_HT; + o->ptr = dict; + zfree(zm); +} + /* ========================= Non type-specific commands ==================== */ static void flushdbCommand(redisClient *c) { @@ -6746,7 +6974,7 @@ static void updateSlavesWaitingBgsave(int bgsaveerr) { static int syncWithMaster(void) { char buf[1024], tmpfile[256], authcmd[1024]; - int dumpsize; + long dumpsize; int fd = anetTcpConnect(NULL,server.masterhost,server.masterport); int dfd; @@ -6798,8 +7026,8 @@ static int syncWithMaster(void) { redisLog(REDIS_WARNING,"Bad protocol from MASTER, the first byte is not '$', are you sure the host and port are right?"); return REDIS_ERR; } - dumpsize = atoi(buf+1); - redisLog(REDIS_NOTICE,"Receiving %d bytes data dump from MASTER",dumpsize); + dumpsize = strtol(buf+1,NULL,10); + redisLog(REDIS_NOTICE,"Receiving %ld bytes data dump from MASTER",dumpsize); /* Read the bulk write data on a temp file */ snprintf(tmpfile,256,"temp-%d.%ld.rdb",(int)time(NULL),(long int)random()); dfd = open(tmpfile,O_CREAT|O_WRONLY,0644);