X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/5b4bff9c175a0189672d95eb953df5704c891d0c..dd94771578ca196db564ab3f51ae6ce6b185531c:/src/t_zset.c diff --git a/src/t_zset.c b/src/t_zset.c index 0fcd6ea3..4812709e 100644 --- a/src/t_zset.c +++ b/src/t_zset.c @@ -17,20 +17,14 @@ /* This skiplist implementation is almost a C translation of the original * algorithm described by William Pugh in "Skip Lists: A Probabilistic * Alternative to Balanced Trees", modified in three ways: - * a) this implementation allows for repeated values. + * a) this implementation allows for repeated scores. * b) the comparison is not just by key (our 'score') but by satellite data. * c) there is a back pointer, so it's a doubly linked list with the back * pointers being only at "level 1". This allows to traverse the list * from tail to head, useful for ZREVRANGE. */ zskiplistNode *zslCreateNode(int level, double score, robj *obj) { - zskiplistNode *zn = zmalloc(sizeof(*zn)); - - zn->forward = zmalloc(sizeof(zskiplistNode*) * level); - if (level > 1) - zn->span = zmalloc(sizeof(unsigned int) * (level - 1)); - else - zn->span = NULL; + zskiplistNode *zn = zmalloc(sizeof(*zn)+level*sizeof(struct zskiplistLevel)); zn->score = score; zn->obj = obj; return zn; @@ -45,11 +39,8 @@ zskiplist *zslCreate(void) { zsl->length = 0; zsl->header = zslCreateNode(ZSKIPLIST_MAXLEVEL,0,NULL); for (j = 0; j < ZSKIPLIST_MAXLEVEL; j++) { - zsl->header->forward[j] = NULL; - - /* span has space for ZSKIPLIST_MAXLEVEL-1 elements */ - if (j < ZSKIPLIST_MAXLEVEL-1) - zsl->header->span[j] = 0; + zsl->header->level[j].forward = NULL; + zsl->header->level[j].span = 0; } zsl->header->backward = NULL; zsl->tail = NULL; @@ -58,25 +49,25 @@ zskiplist *zslCreate(void) { void zslFreeNode(zskiplistNode *node) { decrRefCount(node->obj); - zfree(node->forward); - zfree(node->span); zfree(node); } void zslFree(zskiplist *zsl) { - zskiplistNode *node = zsl->header->forward[0], *next; + zskiplistNode *node = zsl->header->level[0].forward, *next; - zfree(zsl->header->forward); - zfree(zsl->header->span); zfree(zsl->header); while(node) { - next = node->forward[0]; + next = node->level[0].forward; zslFreeNode(node); node = next; } zfree(zsl); } +/* Returns a random level for the new skiplist node we are going to create. + * The return value of this function is between 1 and ZSKIPLIST_MAXLEVEL + * (both inclusive), with a powerlaw-alike distribution where higher + * levels are less likely to be returned. */ int zslRandomLevel(void) { int level = 1; while ((random()&0xFFFF) < (ZSKIPLIST_P * 0xFFFF)) @@ -84,22 +75,22 @@ int zslRandomLevel(void) { return (levelheader; for (i = zsl->level-1; i >= 0; i--) { /* store rank that is crossed to reach the insert position */ rank[i] = i == (zsl->level-1) ? 0 : rank[i+1]; - - while (x->forward[i] && - (x->forward[i]->score < score || - (x->forward[i]->score == score && - compareStringObjects(x->forward[i]->obj,obj) < 0))) { - rank[i] += i > 0 ? x->span[i-1] : 1; - x = x->forward[i]; + while (x->level[i].forward && + (x->level[i].forward->score < score || + (x->level[i].forward->score == score && + compareStringObjects(x->level[i].forward->obj,obj) < 0))) { + rank[i] += x->level[i].span; + x = x->level[i].forward; } update[i] = x; } @@ -112,56 +103,51 @@ void zslInsert(zskiplist *zsl, double score, robj *obj) { for (i = zsl->level; i < level; i++) { rank[i] = 0; update[i] = zsl->header; - update[i]->span[i-1] = zsl->length; + update[i]->level[i].span = zsl->length; } zsl->level = level; } x = zslCreateNode(level,score,obj); for (i = 0; i < level; i++) { - x->forward[i] = update[i]->forward[i]; - update[i]->forward[i] = x; + x->level[i].forward = update[i]->level[i].forward; + update[i]->level[i].forward = x; /* update span covered by update[i] as x is inserted here */ - if (i > 0) { - x->span[i-1] = update[i]->span[i-1] - (rank[0] - rank[i]); - update[i]->span[i-1] = (rank[0] - rank[i]) + 1; - } + x->level[i].span = update[i]->level[i].span - (rank[0] - rank[i]); + update[i]->level[i].span = (rank[0] - rank[i]) + 1; } /* increment span for untouched levels */ for (i = level; i < zsl->level; i++) { - update[i]->span[i-1]++; + update[i]->level[i].span++; } x->backward = (update[0] == zsl->header) ? NULL : update[0]; - if (x->forward[0]) - x->forward[0]->backward = x; + if (x->level[0].forward) + x->level[0].forward->backward = x; else zsl->tail = x; zsl->length++; + return x; } /* Internal function used by zslDelete, zslDeleteByScore and zslDeleteByRank */ void zslDeleteNode(zskiplist *zsl, zskiplistNode *x, zskiplistNode **update) { int i; for (i = 0; i < zsl->level; i++) { - if (update[i]->forward[i] == x) { - if (i > 0) { - update[i]->span[i-1] += x->span[i-1] - 1; - } - update[i]->forward[i] = x->forward[i]; + if (update[i]->level[i].forward == x) { + update[i]->level[i].span += x->level[i].span - 1; + update[i]->level[i].forward = x->level[i].forward; } else { - /* invariant: i > 0, because update[0]->forward[0] - * is always equal to x */ - update[i]->span[i-1] -= 1; + update[i]->level[i].span -= 1; } } - if (x->forward[0]) { - x->forward[0]->backward = x->backward; + if (x->level[0].forward) { + x->level[0].forward->backward = x->backward; } else { zsl->tail = x->backward; } - while(zsl->level > 1 && zsl->header->forward[zsl->level-1] == NULL) + while(zsl->level > 1 && zsl->header->level[zsl->level-1].forward == NULL) zsl->level--; zsl->length--; } @@ -173,16 +159,16 @@ int zslDelete(zskiplist *zsl, double score, robj *obj) { x = zsl->header; for (i = zsl->level-1; i >= 0; i--) { - while (x->forward[i] && - (x->forward[i]->score < score || - (x->forward[i]->score == score && - compareStringObjects(x->forward[i]->obj,obj) < 0))) - x = x->forward[i]; + while (x->level[i].forward && + (x->level[i].forward->score < score || + (x->level[i].forward->score == score && + compareStringObjects(x->level[i].forward->obj,obj) < 0))) + x = x->level[i].forward; update[i] = x; } /* We may have multiple elements with the same score, what we need * is to find the element with both the right score and object. */ - x = x->forward[0]; + x = x->level[0].forward; if (x && score == x->score && equalStringObjects(x->obj,obj)) { zslDeleteNode(zsl, x, update); zslFreeNode(x); @@ -193,33 +179,113 @@ int zslDelete(zskiplist *zsl, double score, robj *obj) { return 0; /* not found */ } +static int zslValueGteMin(double value, zrangespec *spec) { + return spec->minex ? (value > spec->min) : (value >= spec->min); +} + +static int zslValueLteMax(double value, zrangespec *spec) { + return spec->maxex ? (value < spec->max) : (value <= spec->max); +} + +/* Returns if there is a part of the zset is in range. */ +int zslIsInRange(zskiplist *zsl, zrangespec *range) { + zskiplistNode *x; + + /* Test for ranges that will always be empty. */ + if (range->min > range->max || + (range->min == range->max && (range->minex || range->maxex))) + return 0; + x = zsl->tail; + if (x == NULL || !zslValueGteMin(x->score,range)) + return 0; + x = zsl->header->level[0].forward; + if (x == NULL || !zslValueLteMax(x->score,range)) + return 0; + return 1; +} + +/* Find the first node that is contained in the specified range. + * Returns NULL when no element is contained in the range. */ +zskiplistNode *zslFirstInRange(zskiplist *zsl, zrangespec range) { + zskiplistNode *x; + int i; + + /* If everything is out of range, return early. */ + if (!zslIsInRange(zsl,&range)) return NULL; + + x = zsl->header; + for (i = zsl->level-1; i >= 0; i--) { + /* Go forward while *OUT* of range. */ + while (x->level[i].forward && + !zslValueGteMin(x->level[i].forward->score,&range)) + x = x->level[i].forward; + } + + /* This is an inner range, so the next node cannot be NULL. */ + x = x->level[0].forward; + redisAssert(x != NULL); + + /* Check if score <= max. */ + if (!zslValueLteMax(x->score,&range)) return NULL; + return x; +} + +/* Find the last node that is contained in the specified range. + * Returns NULL when no element is contained in the range. */ +zskiplistNode *zslLastInRange(zskiplist *zsl, zrangespec range) { + zskiplistNode *x; + int i; + + /* If everything is out of range, return early. */ + if (!zslIsInRange(zsl,&range)) return NULL; + + x = zsl->header; + for (i = zsl->level-1; i >= 0; i--) { + /* Go forward while *IN* range. */ + while (x->level[i].forward && + zslValueLteMax(x->level[i].forward->score,&range)) + x = x->level[i].forward; + } + + /* This is an inner range, so this node cannot be NULL. */ + redisAssert(x != NULL); + + /* Check if score >= min. */ + if (!zslValueGteMin(x->score,&range)) return NULL; + return x; +} + /* Delete all the elements with score between min and max from the skiplist. * Min and mx are inclusive, so a score >= min || score <= max is deleted. * Note that this function takes the reference to the hash table view of the * sorted set, in order to remove the elements from the hash table too. */ -unsigned long zslDeleteRangeByScore(zskiplist *zsl, double min, double max, dict *dict) { +unsigned long zslDeleteRangeByScore(zskiplist *zsl, zrangespec range, dict *dict) { zskiplistNode *update[ZSKIPLIST_MAXLEVEL], *x; unsigned long removed = 0; int i; x = zsl->header; for (i = zsl->level-1; i >= 0; i--) { - while (x->forward[i] && x->forward[i]->score < min) - x = x->forward[i]; + while (x->level[i].forward && (range.minex ? + x->level[i].forward->score <= range.min : + x->level[i].forward->score < range.min)) + x = x->level[i].forward; update[i] = x; } - /* We may have multiple elements with the same score, what we need - * is to find the element with both the right score and object. */ - x = x->forward[0]; - while (x && x->score <= max) { - zskiplistNode *next = x->forward[0]; - zslDeleteNode(zsl, x, update); + + /* Current node is the last with score < or <= min. */ + x = x->level[0].forward; + + /* Delete nodes while in range. */ + while (x && (range.maxex ? x->score < range.max : x->score <= range.max)) { + zskiplistNode *next = x->level[0].forward; + zslDeleteNode(zsl,x,update); dictDelete(dict,x->obj); zslFreeNode(x); removed++; x = next; } - return removed; /* not found */ + return removed; } /* Delete all the elements with rank between start and end from the skiplist. @@ -231,18 +297,18 @@ unsigned long zslDeleteRangeByRank(zskiplist *zsl, unsigned int start, unsigned x = zsl->header; for (i = zsl->level-1; i >= 0; i--) { - while (x->forward[i] && (traversed + (i > 0 ? x->span[i-1] : 1)) < start) { - traversed += i > 0 ? x->span[i-1] : 1; - x = x->forward[i]; + while (x->level[i].forward && (traversed + x->level[i].span) < start) { + traversed += x->level[i].span; + x = x->level[i].forward; } update[i] = x; } traversed++; - x = x->forward[0]; + x = x->level[0].forward; while (x && traversed <= end) { - zskiplistNode *next = x->forward[0]; - zslDeleteNode(zsl, x, update); + zskiplistNode *next = x->level[0].forward; + zslDeleteNode(zsl,x,update); dictDelete(dict,x->obj); zslFreeNode(x); removed++; @@ -252,39 +318,23 @@ unsigned long zslDeleteRangeByRank(zskiplist *zsl, unsigned int start, unsigned return removed; } -/* Find the first node having a score equal or greater than the specified one. - * Returns NULL if there is no match. */ -zskiplistNode *zslFirstWithScore(zskiplist *zsl, double score) { - zskiplistNode *x; - int i; - - x = zsl->header; - for (i = zsl->level-1; i >= 0; i--) { - while (x->forward[i] && x->forward[i]->score < score) - x = x->forward[i]; - } - /* We may have multiple elements with the same score, what we need - * is to find the element with both the right score and object. */ - return x->forward[0]; -} - /* Find the rank for an element by both score and key. * Returns 0 when the element cannot be found, rank otherwise. * Note that the rank is 1-based due to the span of zsl->header to the * first element. */ -unsigned long zslistTypeGetRank(zskiplist *zsl, double score, robj *o) { +unsigned long zslGetRank(zskiplist *zsl, double score, robj *o) { zskiplistNode *x; unsigned long rank = 0; int i; x = zsl->header; for (i = zsl->level-1; i >= 0; i--) { - while (x->forward[i] && - (x->forward[i]->score < score || - (x->forward[i]->score == score && - compareStringObjects(x->forward[i]->obj,o) <= 0))) { - rank += i > 0 ? x->span[i-1] : 1; - x = x->forward[i]; + while (x->level[i].forward && + (x->level[i].forward->score < score || + (x->level[i].forward->score == score && + compareStringObjects(x->level[i].forward->obj,o) <= 0))) { + rank += x->level[i].span; + x = x->level[i].forward; } /* x might be equal to zsl->header, so test if obj is non-NULL */ @@ -296,17 +346,17 @@ unsigned long zslistTypeGetRank(zskiplist *zsl, double score, robj *o) { } /* Finds an element by its rank. The rank argument needs to be 1-based. */ -zskiplistNode* zslistTypeGetElementByRank(zskiplist *zsl, unsigned long rank) { +zskiplistNode* zslGetElementByRank(zskiplist *zsl, unsigned long rank) { zskiplistNode *x; unsigned long traversed = 0; int i; x = zsl->header; for (i = zsl->level-1; i >= 0; i--) { - while (x->forward[i] && (traversed + (i>0 ? x->span[i-1] : 1)) <= rank) + while (x->level[i].forward && (traversed + x->level[i].span) <= rank) { - traversed += i > 0 ? x->span[i-1] : 1; - x = x->forward[i]; + traversed += x->level[i].span; + x = x->level[i].forward; } if (traversed == rank) { return x; @@ -315,189 +365,704 @@ zskiplistNode* zslistTypeGetElementByRank(zskiplist *zsl, unsigned long rank) { return NULL; } +/* Populate the rangespec according to the objects min and max. */ +static int zslParseRange(robj *min, robj *max, zrangespec *spec) { + char *eptr; + spec->minex = spec->maxex = 0; + + /* Parse the min-max interval. If one of the values is prefixed + * by the "(" character, it's considered "open". For instance + * ZRANGEBYSCORE zset (1.5 (2.5 will match min < x < max + * ZRANGEBYSCORE zset 1.5 2.5 will instead match min <= x <= max */ + if (min->encoding == REDIS_ENCODING_INT) { + spec->min = (long)min->ptr; + } else { + if (((char*)min->ptr)[0] == '(') { + spec->min = strtod((char*)min->ptr+1,&eptr); + if (eptr[0] != '\0' || isnan(spec->min)) return REDIS_ERR; + spec->minex = 1; + } else { + spec->min = strtod((char*)min->ptr,&eptr); + if (eptr[0] != '\0' || isnan(spec->min)) return REDIS_ERR; + } + } + if (max->encoding == REDIS_ENCODING_INT) { + spec->max = (long)max->ptr; + } else { + if (((char*)max->ptr)[0] == '(') { + spec->max = strtod((char*)max->ptr+1,&eptr); + if (eptr[0] != '\0' || isnan(spec->max)) return REDIS_ERR; + spec->maxex = 1; + } else { + spec->max = strtod((char*)max->ptr,&eptr); + if (eptr[0] != '\0' || isnan(spec->max)) return REDIS_ERR; + } + } + + return REDIS_OK; +} + /*----------------------------------------------------------------------------- - * Sorted set commands + * Ziplist-backed sorted set API *----------------------------------------------------------------------------*/ -/* This generic command implements both ZADD and ZINCRBY. - * scoreval is the score if the operation is a ZADD (doincrement == 0) or - * the increment if the operation is a ZINCRBY (doincrement == 1). */ -void zaddGenericCommand(redisClient *c, robj *key, robj *ele, double scoreval, int doincrement) { - robj *zsetobj; - zset *zs; - double *score; +double zzlGetScore(unsigned char *sptr) { + unsigned char *vstr; + unsigned int vlen; + long long vlong; + char buf[128]; + double score; - if (isnan(scoreval)) { - addReplySds(c,sdsnew("-ERR provide score is Not A Number (nan)\r\n")); - return; + redisAssert(sptr != NULL); + redisAssert(ziplistGet(sptr,&vstr,&vlen,&vlong)); + + if (vstr) { + memcpy(buf,vstr,vlen); + buf[vlen] = '\0'; + score = strtod(buf,NULL); + } else { + score = vlong; } - zsetobj = lookupKeyWrite(c->db,key); - if (zsetobj == NULL) { - zsetobj = createZsetObject(); - dbAdd(c->db,key,zsetobj); + return score; +} + +/* Compare element in sorted set with given element. */ +int zzlCompareElements(unsigned char *eptr, unsigned char *cstr, unsigned int clen) { + unsigned char *vstr; + unsigned int vlen; + long long vlong; + unsigned char vbuf[32]; + int minlen, cmp; + + redisAssert(ziplistGet(eptr,&vstr,&vlen,&vlong)); + if (vstr == NULL) { + /* Store string representation of long long in buf. */ + vlen = ll2string((char*)vbuf,sizeof(vbuf),vlong); + vstr = vbuf; + } + + minlen = (vlen < clen) ? vlen : clen; + cmp = memcmp(vstr,cstr,minlen); + if (cmp == 0) return vlen-clen; + return cmp; +} + +unsigned int zzlLength(unsigned char *zl) { + return ziplistLen(zl)/2; +} + +/* Move to next entry based on the values in eptr and sptr. Both are set to + * NULL when there is no next entry. */ +void zzlNext(unsigned char *zl, unsigned char **eptr, unsigned char **sptr) { + unsigned char *_eptr, *_sptr; + redisAssert(*eptr != NULL && *sptr != NULL); + + _eptr = ziplistNext(zl,*sptr); + if (_eptr != NULL) { + _sptr = ziplistNext(zl,_eptr); + redisAssert(_sptr != NULL); } else { - if (zsetobj->type != REDIS_ZSET) { - addReply(c,shared.wrongtypeerr); - return; - } + /* No next entry. */ + _sptr = NULL; } - zs = zsetobj->ptr; - /* Ok now since we implement both ZADD and ZINCRBY here the code - * needs to handle the two different conditions. It's all about setting - * '*score', that is, the new score to set, to the right value. */ - score = zmalloc(sizeof(double)); - if (doincrement) { - dictEntry *de; + *eptr = _eptr; + *sptr = _sptr; +} - /* Read the old score. If the element was not present starts from 0 */ - de = dictFind(zs->dict,ele); - if (de) { - double *oldscore = dictGetEntryVal(de); - *score = *oldscore + scoreval; - } 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; - } +/* Move to the previous entry based on the values in eptr and sptr. Both are + * set to NULL when there is no next entry. */ +void zzlPrev(unsigned char *zl, unsigned char **eptr, unsigned char **sptr) { + unsigned char *_eptr, *_sptr; + redisAssert(*eptr != NULL && *sptr != NULL); + + _sptr = ziplistPrev(zl,*eptr); + if (_sptr != NULL) { + _eptr = ziplistPrev(zl,_sptr); + redisAssert(_eptr != NULL); } else { - *score = scoreval; + /* No previous entry. */ + _eptr = NULL; } - /* What follows is a simple remove and re-insert operation that is common - * to both ZADD and ZINCRBY... */ - if (dictAdd(zs->dict,ele,score) == DICT_OK) { - /* case 1: New element */ - incrRefCount(ele); /* added to hash */ - zslInsert(zs->zsl,*score,ele); - incrRefCount(ele); /* added to skiplist */ - touchWatchedKey(c->db,c->argv[1]); - server.dirty++; - if (doincrement) - addReplyDouble(c,*score); + *eptr = _eptr; + *sptr = _sptr; +} + +/* Returns if there is a part of the zset is in range. Should only be used + * internally by zzlFirstInRange and zzlLastInRange. */ +int zzlIsInRange(unsigned char *zl, zrangespec *range) { + unsigned char *p; + double score; + + /* Test for ranges that will always be empty. */ + if (range->min > range->max || + (range->min == range->max && (range->minex || range->maxex))) + return 0; + + p = ziplistIndex(zl,-1); /* Last score. */ + if (p == NULL) return 0; /* Empty sorted set */ + score = zzlGetScore(p); + if (!zslValueGteMin(score,range)) + return 0; + + p = ziplistIndex(zl,1); /* First score. */ + redisAssert(p != NULL); + score = zzlGetScore(p); + if (!zslValueLteMax(score,range)) + return 0; + + return 1; +} + +/* Find pointer to the first element contained in the specified range. + * Returns NULL when no element is contained in the range. */ +unsigned char *zzlFirstInRange(unsigned char *zl, zrangespec range) { + unsigned char *eptr = ziplistIndex(zl,0), *sptr; + double score; + + /* If everything is out of range, return early. */ + if (!zzlIsInRange(zl,&range)) return NULL; + + while (eptr != NULL) { + sptr = ziplistNext(zl,eptr); + redisAssert(sptr != NULL); + + score = zzlGetScore(sptr); + if (zslValueGteMin(score,&range)) { + /* Check if score <= max. */ + if (zslValueLteMax(score,&range)) + return eptr; + return NULL; + } + + /* Move to next element. */ + eptr = ziplistNext(zl,sptr); + } + + return NULL; +} + +/* Find pointer to the last element contained in the specified range. + * Returns NULL when no element is contained in the range. */ +unsigned char *zzlLastInRange(unsigned char *zl, zrangespec range) { + unsigned char *eptr = ziplistIndex(zl,-2), *sptr; + double score; + + /* If everything is out of range, return early. */ + if (!zzlIsInRange(zl,&range)) return NULL; + + while (eptr != NULL) { + sptr = ziplistNext(zl,eptr); + redisAssert(sptr != NULL); + + score = zzlGetScore(sptr); + if (zslValueLteMax(score,&range)) { + /* Check if score >= min. */ + if (zslValueGteMin(score,&range)) + return eptr; + return NULL; + } + + /* Move to previous element by moving to the score of previous element. + * When this returns NULL, we know there also is no element. */ + sptr = ziplistPrev(zl,eptr); + if (sptr != NULL) + redisAssert((eptr = ziplistPrev(zl,sptr)) != NULL); else - addReply(c,shared.cone); + eptr = NULL; + } + + return NULL; +} + +unsigned char *zzlFind(unsigned char *zl, robj *ele, double *score) { + unsigned char *eptr = ziplistIndex(zl,0), *sptr; + + ele = getDecodedObject(ele); + while (eptr != NULL) { + sptr = ziplistNext(zl,eptr); + redisAssertWithInfo(NULL,ele,sptr != NULL); + + if (ziplistCompare(eptr,ele->ptr,sdslen(ele->ptr))) { + /* Matching element, pull out score. */ + if (score != NULL) *score = zzlGetScore(sptr); + decrRefCount(ele); + return eptr; + } + + /* Move to next element. */ + eptr = ziplistNext(zl,sptr); + } + + decrRefCount(ele); + return NULL; +} + +/* Delete (element,score) pair from ziplist. Use local copy of eptr because we + * don't want to modify the one given as argument. */ +unsigned char *zzlDelete(unsigned char *zl, unsigned char *eptr) { + unsigned char *p = eptr; + + /* TODO: add function to ziplist API to delete N elements from offset. */ + zl = ziplistDelete(zl,&p); + zl = ziplistDelete(zl,&p); + return zl; +} + +unsigned char *zzlInsertAt(unsigned char *zl, unsigned char *eptr, robj *ele, double score) { + unsigned char *sptr; + char scorebuf[128]; + int scorelen; + size_t offset; + + redisAssertWithInfo(NULL,ele,ele->encoding == REDIS_ENCODING_RAW); + scorelen = d2string(scorebuf,sizeof(scorebuf),score); + if (eptr == NULL) { + zl = ziplistPush(zl,ele->ptr,sdslen(ele->ptr),ZIPLIST_TAIL); + zl = ziplistPush(zl,(unsigned char*)scorebuf,scorelen,ZIPLIST_TAIL); } else { - dictEntry *de; - double *oldscore; + /* Keep offset relative to zl, as it might be re-allocated. */ + offset = eptr-zl; + zl = ziplistInsert(zl,eptr,ele->ptr,sdslen(ele->ptr)); + eptr = zl+offset; + + /* Insert score after the element. */ + redisAssertWithInfo(NULL,ele,(sptr = ziplistNext(zl,eptr)) != NULL); + zl = ziplistInsert(zl,sptr,(unsigned char*)scorebuf,scorelen); + } - /* case 2: Score update operation */ - de = dictFind(zs->dict,ele); - redisAssert(de != NULL); - oldscore = dictGetEntryVal(de); - if (*score != *oldscore) { - int deleted; - - /* Remove and insert the element in the skip list with new score */ - deleted = zslDelete(zs->zsl,*oldscore,ele); - redisAssert(deleted != 0); - zslInsert(zs->zsl,*score,ele); - incrRefCount(ele); - /* Update the score in the hash table */ - dictReplace(zs->dict,ele,score); - touchWatchedKey(c->db,c->argv[1]); - server.dirty++; - } else { - zfree(score); + return zl; +} + +/* Insert (element,score) pair in ziplist. This function assumes the element is + * not yet present in the list. */ +unsigned char *zzlInsert(unsigned char *zl, robj *ele, double score) { + unsigned char *eptr = ziplistIndex(zl,0), *sptr; + double s; + + ele = getDecodedObject(ele); + while (eptr != NULL) { + sptr = ziplistNext(zl,eptr); + redisAssertWithInfo(NULL,ele,sptr != NULL); + s = zzlGetScore(sptr); + + if (s > score) { + /* First element with score larger than score for element to be + * inserted. This means we should take its spot in the list to + * maintain ordering. */ + zl = zzlInsertAt(zl,eptr,ele,score); + break; + } else if (s == score) { + /* Ensure lexicographical ordering for elements. */ + if (zzlCompareElements(eptr,ele->ptr,sdslen(ele->ptr)) > 0) { + zl = zzlInsertAt(zl,eptr,ele,score); + break; + } } - if (doincrement) - addReplyDouble(c,*score); - else - addReply(c,shared.czero); + + /* Move to next element. */ + eptr = ziplistNext(zl,sptr); } + + /* Push on tail of list when it was not yet inserted. */ + if (eptr == NULL) + zl = zzlInsertAt(zl,NULL,ele,score); + + decrRefCount(ele); + return zl; } -void zaddCommand(redisClient *c) { - double scoreval; +unsigned char *zzlDeleteRangeByScore(unsigned char *zl, zrangespec range, unsigned long *deleted) { + unsigned char *eptr, *sptr; + double score; + unsigned long num = 0; + + if (deleted != NULL) *deleted = 0; + + eptr = zzlFirstInRange(zl,range); + if (eptr == NULL) return zl; + + /* When the tail of the ziplist is deleted, eptr will point to the sentinel + * byte and ziplistNext will return NULL. */ + while ((sptr = ziplistNext(zl,eptr)) != NULL) { + score = zzlGetScore(sptr); + if (zslValueLteMax(score,&range)) { + /* Delete both the element and the score. */ + zl = ziplistDelete(zl,&eptr); + zl = ziplistDelete(zl,&eptr); + num++; + } else { + /* No longer in range. */ + break; + } + } - if (getDoubleFromObjectOrReply(c, c->argv[2], &scoreval, NULL) != REDIS_OK) return; - zaddGenericCommand(c,c->argv[1],c->argv[3],scoreval,0); + if (deleted != NULL) *deleted = num; + return zl; } -void zincrbyCommand(redisClient *c) { - double scoreval; +/* Delete all the elements with rank between start and end from the skiplist. + * Start and end are inclusive. Note that start and end need to be 1-based */ +unsigned char *zzlDeleteRangeByRank(unsigned char *zl, unsigned int start, unsigned int end, unsigned long *deleted) { + unsigned int num = (end-start)+1; + if (deleted) *deleted = num; + zl = ziplistDeleteRange(zl,2*(start-1),2*num); + return zl; +} - if (getDoubleFromObjectOrReply(c, c->argv[2], &scoreval, NULL) != REDIS_OK) return; - zaddGenericCommand(c,c->argv[1],c->argv[3],scoreval,1); +/*----------------------------------------------------------------------------- + * Common sorted set API + *----------------------------------------------------------------------------*/ + +unsigned int zsetLength(robj *zobj) { + int length = -1; + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + length = zzlLength(zobj->ptr); + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + length = ((zset*)zobj->ptr)->zsl->length; + } else { + redisPanic("Unknown sorted set encoding"); + } + return length; } -void zremCommand(redisClient *c) { - robj *zsetobj; +void zsetConvert(robj *zobj, int encoding) { zset *zs; - dictEntry *de; - double *oldscore; - int deleted; + zskiplistNode *node, *next; + robj *ele; + double score; + + if (zobj->encoding == encoding) return; + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + unsigned char *zl = zobj->ptr; + unsigned char *eptr, *sptr; + unsigned char *vstr; + unsigned int vlen; + long long vlong; + + if (encoding != REDIS_ENCODING_SKIPLIST) + redisPanic("Unknown target encoding"); + + zs = zmalloc(sizeof(*zs)); + zs->dict = dictCreate(&zsetDictType,NULL); + zs->zsl = zslCreate(); + + eptr = ziplistIndex(zl,0); + redisAssertWithInfo(NULL,zobj,eptr != NULL); + sptr = ziplistNext(zl,eptr); + redisAssertWithInfo(NULL,zobj,sptr != NULL); + + while (eptr != NULL) { + score = zzlGetScore(sptr); + redisAssertWithInfo(NULL,zobj,ziplistGet(eptr,&vstr,&vlen,&vlong)); + if (vstr == NULL) + ele = createStringObjectFromLongLong(vlong); + else + ele = createStringObject((char*)vstr,vlen); + + /* Has incremented refcount since it was just created. */ + node = zslInsert(zs->zsl,score,ele); + redisAssertWithInfo(NULL,zobj,dictAdd(zs->dict,ele,&node->score) == DICT_OK); + incrRefCount(ele); /* Added to dictionary. */ + zzlNext(zl,&eptr,&sptr); + } - if ((zsetobj = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || - checkType(c,zsetobj,REDIS_ZSET)) return; + zfree(zobj->ptr); + zobj->ptr = zs; + zobj->encoding = REDIS_ENCODING_SKIPLIST; + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + unsigned char *zl = ziplistNew(); + + if (encoding != REDIS_ENCODING_ZIPLIST) + redisPanic("Unknown target encoding"); + + /* Approach similar to zslFree(), since we want to free the skiplist at + * the same time as creating the ziplist. */ + zs = zobj->ptr; + dictRelease(zs->dict); + node = zs->zsl->header->level[0].forward; + zfree(zs->zsl->header); + zfree(zs->zsl); + + while (node) { + ele = getDecodedObject(node->obj); + zl = zzlInsertAt(zl,NULL,ele,node->score); + decrRefCount(ele); + + next = node->level[0].forward; + zslFreeNode(node); + node = next; + } - zs = zsetobj->ptr; - de = dictFind(zs->dict,c->argv[2]); - if (de == NULL) { - addReply(c,shared.czero); + zfree(zs); + zobj->ptr = zl; + zobj->encoding = REDIS_ENCODING_ZIPLIST; + } else { + redisPanic("Unknown sorted set encoding"); + } +} + +/*----------------------------------------------------------------------------- + * Sorted set commands + *----------------------------------------------------------------------------*/ + +/* This generic command implements both ZADD and ZINCRBY. */ +void zaddGenericCommand(redisClient *c, int incr) { + static char *nanerr = "resulting score is not a number (NaN)"; + robj *key = c->argv[1]; + robj *ele; + robj *zobj; + robj *curobj; + double score = 0, *scores, curscore = 0.0; + int j, elements = (c->argc-2)/2; + int added = 0; + + if (c->argc % 2) { + addReply(c,shared.syntaxerr); return; } - /* Delete from the skiplist */ - oldscore = dictGetEntryVal(de); - deleted = zslDelete(zs->zsl,*oldscore,c->argv[2]); - redisAssert(deleted != 0); - /* Delete from the hash table */ - dictDelete(zs->dict,c->argv[2]); - if (htNeedsResize(zs->dict)) dictResize(zs->dict); - if (dictSize(zs->dict) == 0) dbDelete(c->db,c->argv[1]); - touchWatchedKey(c->db,c->argv[1]); - server.dirty++; - addReply(c,shared.cone); + /* Start parsing all the scores, we need to emit any syntax error + * before executing additions to the sorted set, as the command should + * either execute fully or nothing at all. */ + scores = zmalloc(sizeof(double)*elements); + for (j = 0; j < elements; j++) { + if (getDoubleFromObjectOrReply(c,c->argv[2+j*2],&scores[j],NULL) + != REDIS_OK) + { + zfree(scores); + return; + } + } + + /* Lookup the key and create the sorted set if does not exist. */ + zobj = lookupKeyWrite(c->db,key); + if (zobj == NULL) { + if (server.zset_max_ziplist_entries == 0 || + server.zset_max_ziplist_value < sdslen(c->argv[3]->ptr)) + { + zobj = createZsetObject(); + } else { + zobj = createZsetZiplistObject(); + } + dbAdd(c->db,key,zobj); + } else { + if (zobj->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + zfree(scores); + return; + } + } + + for (j = 0; j < elements; j++) { + score = scores[j]; + + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + unsigned char *eptr; + + /* Prefer non-encoded element when dealing with ziplists. */ + ele = c->argv[3+j*2]; + if ((eptr = zzlFind(zobj->ptr,ele,&curscore)) != NULL) { + if (incr) { + score += curscore; + if (isnan(score)) { + addReplyError(c,nanerr); + /* Don't need to check if the sorted set is empty + * because we know it has at least one element. */ + zfree(scores); + return; + } + } + + /* Remove and re-insert when score changed. */ + if (score != curscore) { + zobj->ptr = zzlDelete(zobj->ptr,eptr); + zobj->ptr = zzlInsert(zobj->ptr,ele,score); + + signalModifiedKey(c->db,key); + server.dirty++; + } + } else { + /* Optimize: check if the element is too large or the list + * becomes too long *before* executing zzlInsert. */ + zobj->ptr = zzlInsert(zobj->ptr,ele,score); + if (zzlLength(zobj->ptr) > server.zset_max_ziplist_entries) + zsetConvert(zobj,REDIS_ENCODING_SKIPLIST); + if (sdslen(ele->ptr) > server.zset_max_ziplist_value) + zsetConvert(zobj,REDIS_ENCODING_SKIPLIST); + + signalModifiedKey(c->db,key); + server.dirty++; + if (!incr) added++; + } + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + zset *zs = zobj->ptr; + zskiplistNode *znode; + dictEntry *de; + + ele = c->argv[3+j*2] = tryObjectEncoding(c->argv[3+j*2]); + de = dictFind(zs->dict,ele); + if (de != NULL) { + curobj = dictGetKey(de); + curscore = *(double*)dictGetVal(de); + + if (incr) { + score += curscore; + if (isnan(score)) { + addReplyError(c,nanerr); + /* Don't need to check if the sorted set is empty + * because we know it has at least one element. */ + zfree(scores); + return; + } + } + + /* Remove and re-insert when score changed. We can safely + * delete the key object from the skiplist, since the + * dictionary still has a reference to it. */ + if (score != curscore) { + redisAssertWithInfo(c,curobj,zslDelete(zs->zsl,curscore,curobj)); + znode = zslInsert(zs->zsl,score,curobj); + incrRefCount(curobj); /* Re-inserted in skiplist. */ + dictGetVal(de) = &znode->score; /* Update score ptr. */ + + signalModifiedKey(c->db,key); + server.dirty++; + } + } else { + znode = zslInsert(zs->zsl,score,ele); + incrRefCount(ele); /* Inserted in skiplist. */ + redisAssertWithInfo(c,NULL,dictAdd(zs->dict,ele,&znode->score) == DICT_OK); + incrRefCount(ele); /* Added to dictionary. */ + + signalModifiedKey(c->db,key); + server.dirty++; + if (!incr) added++; + } + } else { + redisPanic("Unknown sorted set encoding"); + } + } + zfree(scores); + if (incr) /* ZINCRBY */ + addReplyDouble(c,score); + else /* ZADD */ + addReplyLongLong(c,added); } -void zremrangebyscoreCommand(redisClient *c) { - double min; - double max; - long deleted; - robj *zsetobj; - zset *zs; +void zaddCommand(redisClient *c) { + zaddGenericCommand(c,0); +} - if ((getDoubleFromObjectOrReply(c, c->argv[2], &min, NULL) != REDIS_OK) || - (getDoubleFromObjectOrReply(c, c->argv[3], &max, NULL) != REDIS_OK)) return; +void zincrbyCommand(redisClient *c) { + zaddGenericCommand(c,1); +} + +void zremCommand(redisClient *c) { + robj *key = c->argv[1]; + robj *zobj; + int deleted = 0, j; + + if ((zobj = lookupKeyWriteOrReply(c,key,shared.czero)) == NULL || + checkType(c,zobj,REDIS_ZSET)) return; + + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + unsigned char *eptr; + + for (j = 2; j < c->argc; j++) { + if ((eptr = zzlFind(zobj->ptr,c->argv[j],NULL)) != NULL) { + deleted++; + zobj->ptr = zzlDelete(zobj->ptr,eptr); + if (zzlLength(zobj->ptr) == 0) { + dbDelete(c->db,key); + break; + } + } + } + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + zset *zs = zobj->ptr; + dictEntry *de; + double score; + + for (j = 2; j < c->argc; j++) { + de = dictFind(zs->dict,c->argv[j]); + if (de != NULL) { + deleted++; + + /* Delete from the skiplist */ + score = *(double*)dictGetVal(de); + redisAssertWithInfo(c,c->argv[j],zslDelete(zs->zsl,score,c->argv[j])); + + /* Delete from the hash table */ + dictDelete(zs->dict,c->argv[j]); + if (htNeedsResize(zs->dict)) dictResize(zs->dict); + if (dictSize(zs->dict) == 0) { + dbDelete(c->db,key); + break; + } + } + } + } else { + redisPanic("Unknown sorted set encoding"); + } + + if (deleted) { + signalModifiedKey(c->db,key); + server.dirty += deleted; + } + addReplyLongLong(c,deleted); +} + +void zremrangebyscoreCommand(redisClient *c) { + robj *key = c->argv[1]; + robj *zobj; + zrangespec range; + unsigned long deleted; + + /* Parse the range arguments. */ + if (zslParseRange(c->argv[2],c->argv[3],&range) != REDIS_OK) { + addReplyError(c,"min or max is not a float"); + return; + } - if ((zsetobj = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || - checkType(c,zsetobj,REDIS_ZSET)) return; + if ((zobj = lookupKeyWriteOrReply(c,key,shared.czero)) == NULL || + checkType(c,zobj,REDIS_ZSET)) return; + + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + zobj->ptr = zzlDeleteRangeByScore(zobj->ptr,range,&deleted); + if (zzlLength(zobj->ptr) == 0) dbDelete(c->db,key); + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + zset *zs = zobj->ptr; + deleted = zslDeleteRangeByScore(zs->zsl,range,zs->dict); + if (htNeedsResize(zs->dict)) dictResize(zs->dict); + if (dictSize(zs->dict) == 0) dbDelete(c->db,key); + } else { + redisPanic("Unknown sorted set encoding"); + } - zs = zsetobj->ptr; - deleted = zslDeleteRangeByScore(zs->zsl,min,max,zs->dict); - if (htNeedsResize(zs->dict)) dictResize(zs->dict); - if (dictSize(zs->dict) == 0) dbDelete(c->db,c->argv[1]); - if (deleted) touchWatchedKey(c->db,c->argv[1]); + if (deleted) signalModifiedKey(c->db,key); server.dirty += deleted; addReplyLongLong(c,deleted); } void zremrangebyrankCommand(redisClient *c) { + robj *key = c->argv[1]; + robj *zobj; long start; long end; int llen; - long deleted; - robj *zsetobj; - zset *zs; + unsigned long deleted; if ((getLongFromObjectOrReply(c, c->argv[2], &start, NULL) != REDIS_OK) || (getLongFromObjectOrReply(c, c->argv[3], &end, NULL) != REDIS_OK)) return; - if ((zsetobj = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || - checkType(c,zsetobj,REDIS_ZSET)) return; - zs = zsetobj->ptr; - llen = zs->zsl->length; + if ((zobj = lookupKeyWriteOrReply(c,key,shared.czero)) == NULL || + checkType(c,zobj,REDIS_ZSET)) return; - /* convert negative indexes */ + /* Sanitize indexes. */ + llen = zsetLength(zobj); if (start < 0) start = llen+start; if (end < 0) end = llen+end; if (start < 0) start = 0; @@ -510,37 +1075,365 @@ void zremrangebyrankCommand(redisClient *c) { } if (end >= llen) end = llen-1; - /* increment start and end because zsl*Rank functions - * use 1-based rank */ - deleted = zslDeleteRangeByRank(zs->zsl,start+1,end+1,zs->dict); - if (htNeedsResize(zs->dict)) dictResize(zs->dict); - if (dictSize(zs->dict) == 0) dbDelete(c->db,c->argv[1]); - if (deleted) touchWatchedKey(c->db,c->argv[1]); + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + /* Correct for 1-based rank. */ + zobj->ptr = zzlDeleteRangeByRank(zobj->ptr,start+1,end+1,&deleted); + if (zzlLength(zobj->ptr) == 0) dbDelete(c->db,key); + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + zset *zs = zobj->ptr; + + /* Correct for 1-based rank. */ + deleted = zslDeleteRangeByRank(zs->zsl,start+1,end+1,zs->dict); + if (htNeedsResize(zs->dict)) dictResize(zs->dict); + if (dictSize(zs->dict) == 0) dbDelete(c->db,key); + } else { + redisPanic("Unknown sorted set encoding"); + } + + if (deleted) signalModifiedKey(c->db,key); server.dirty += deleted; - addReplyLongLong(c, deleted); + addReplyLongLong(c,deleted); } typedef struct { - dict *dict; + robj *subject; + int type; /* Set, sorted set */ + int encoding; double weight; + + union { + /* Set iterators. */ + union _iterset { + struct { + intset *is; + int ii; + } is; + struct { + dict *dict; + dictIterator *di; + dictEntry *de; + } ht; + } set; + + /* Sorted set iterators. */ + union _iterzset { + struct { + unsigned char *zl; + unsigned char *eptr, *sptr; + } zl; + struct { + zset *zs; + zskiplistNode *node; + } sl; + } zset; + } iter; } zsetopsrc; -int qsortCompareZsetopsrcByCardinality(const void *s1, const void *s2) { - zsetopsrc *d1 = (void*) s1, *d2 = (void*) s2; - unsigned long size1, size2; - size1 = d1->dict ? dictSize(d1->dict) : 0; - size2 = d2->dict ? dictSize(d2->dict) : 0; - return size1 - size2; + +/* Use dirty flags for pointers that need to be cleaned up in the next + * iteration over the zsetopval. The dirty flag for the long long value is + * special, since long long values don't need cleanup. Instead, it means that + * we already checked that "ell" holds a long long, or tried to convert another + * representation into a long long value. When this was successful, + * OPVAL_VALID_LL is set as well. */ +#define OPVAL_DIRTY_ROBJ 1 +#define OPVAL_DIRTY_LL 2 +#define OPVAL_VALID_LL 4 + +/* Store value retrieved from the iterator. */ +typedef struct { + int flags; + unsigned char _buf[32]; /* Private buffer. */ + robj *ele; + unsigned char *estr; + unsigned int elen; + long long ell; + double score; +} zsetopval; + +typedef union _iterset iterset; +typedef union _iterzset iterzset; + +void zuiInitIterator(zsetopsrc *op) { + if (op->subject == NULL) + return; + + if (op->type == REDIS_SET) { + iterset *it = &op->iter.set; + if (op->encoding == REDIS_ENCODING_INTSET) { + it->is.is = op->subject->ptr; + it->is.ii = 0; + } else if (op->encoding == REDIS_ENCODING_HT) { + it->ht.dict = op->subject->ptr; + it->ht.di = dictGetIterator(op->subject->ptr); + it->ht.de = dictNext(it->ht.di); + } else { + redisPanic("Unknown set encoding"); + } + } else if (op->type == REDIS_ZSET) { + iterzset *it = &op->iter.zset; + if (op->encoding == REDIS_ENCODING_ZIPLIST) { + it->zl.zl = op->subject->ptr; + it->zl.eptr = ziplistIndex(it->zl.zl,0); + if (it->zl.eptr != NULL) { + it->zl.sptr = ziplistNext(it->zl.zl,it->zl.eptr); + redisAssert(it->zl.sptr != NULL); + } + } else if (op->encoding == REDIS_ENCODING_SKIPLIST) { + it->sl.zs = op->subject->ptr; + it->sl.node = it->sl.zs->zsl->header->level[0].forward; + } else { + redisPanic("Unknown sorted set encoding"); + } + } else { + redisPanic("Unsupported type"); + } +} + +void zuiClearIterator(zsetopsrc *op) { + if (op->subject == NULL) + return; + + if (op->type == REDIS_SET) { + iterset *it = &op->iter.set; + if (op->encoding == REDIS_ENCODING_INTSET) { + REDIS_NOTUSED(it); /* skip */ + } else if (op->encoding == REDIS_ENCODING_HT) { + dictReleaseIterator(it->ht.di); + } else { + redisPanic("Unknown set encoding"); + } + } else if (op->type == REDIS_ZSET) { + iterzset *it = &op->iter.zset; + if (op->encoding == REDIS_ENCODING_ZIPLIST) { + REDIS_NOTUSED(it); /* skip */ + } else if (op->encoding == REDIS_ENCODING_SKIPLIST) { + REDIS_NOTUSED(it); /* skip */ + } else { + redisPanic("Unknown sorted set encoding"); + } + } else { + redisPanic("Unsupported type"); + } +} + +int zuiLength(zsetopsrc *op) { + if (op->subject == NULL) + return 0; + + if (op->type == REDIS_SET) { + iterset *it = &op->iter.set; + if (op->encoding == REDIS_ENCODING_INTSET) { + return intsetLen(it->is.is); + } else if (op->encoding == REDIS_ENCODING_HT) { + return dictSize(it->ht.dict); + } else { + redisPanic("Unknown set encoding"); + } + } else if (op->type == REDIS_ZSET) { + iterzset *it = &op->iter.zset; + if (op->encoding == REDIS_ENCODING_ZIPLIST) { + return zzlLength(it->zl.zl); + } else if (op->encoding == REDIS_ENCODING_SKIPLIST) { + return it->sl.zs->zsl->length; + } else { + redisPanic("Unknown sorted set encoding"); + } + } else { + redisPanic("Unsupported type"); + } +} + +/* Check if the current value is valid. If so, store it in the passed structure + * and move to the next element. If not valid, this means we have reached the + * end of the structure and can abort. */ +int zuiNext(zsetopsrc *op, zsetopval *val) { + if (op->subject == NULL) + return 0; + + if (val->flags & OPVAL_DIRTY_ROBJ) + decrRefCount(val->ele); + + memset(val,0,sizeof(zsetopval)); + + if (op->type == REDIS_SET) { + iterset *it = &op->iter.set; + if (op->encoding == REDIS_ENCODING_INTSET) { + int64_t ell; + + if (!intsetGet(it->is.is,it->is.ii,&ell)) + return 0; + val->ell = ell; + val->score = 1.0; + + /* Move to next element. */ + it->is.ii++; + } else if (op->encoding == REDIS_ENCODING_HT) { + if (it->ht.de == NULL) + return 0; + val->ele = dictGetKey(it->ht.de); + val->score = 1.0; + + /* Move to next element. */ + it->ht.de = dictNext(it->ht.di); + } else { + redisPanic("Unknown set encoding"); + } + } else if (op->type == REDIS_ZSET) { + iterzset *it = &op->iter.zset; + if (op->encoding == REDIS_ENCODING_ZIPLIST) { + /* No need to check both, but better be explicit. */ + if (it->zl.eptr == NULL || it->zl.sptr == NULL) + return 0; + redisAssert(ziplistGet(it->zl.eptr,&val->estr,&val->elen,&val->ell)); + val->score = zzlGetScore(it->zl.sptr); + + /* Move to next element. */ + zzlNext(it->zl.zl,&it->zl.eptr,&it->zl.sptr); + } else if (op->encoding == REDIS_ENCODING_SKIPLIST) { + if (it->sl.node == NULL) + return 0; + val->ele = it->sl.node->obj; + val->score = it->sl.node->score; + + /* Move to next element. */ + it->sl.node = it->sl.node->level[0].forward; + } else { + redisPanic("Unknown sorted set encoding"); + } + } else { + redisPanic("Unsupported type"); + } + return 1; +} + +int zuiLongLongFromValue(zsetopval *val) { + if (!(val->flags & OPVAL_DIRTY_LL)) { + val->flags |= OPVAL_DIRTY_LL; + + if (val->ele != NULL) { + if (val->ele->encoding == REDIS_ENCODING_INT) { + val->ell = (long)val->ele->ptr; + val->flags |= OPVAL_VALID_LL; + } else if (val->ele->encoding == REDIS_ENCODING_RAW) { + if (string2ll(val->ele->ptr,sdslen(val->ele->ptr),&val->ell)) + val->flags |= OPVAL_VALID_LL; + } else { + redisPanic("Unsupported element encoding"); + } + } else if (val->estr != NULL) { + if (string2ll((char*)val->estr,val->elen,&val->ell)) + val->flags |= OPVAL_VALID_LL; + } else { + /* The long long was already set, flag as valid. */ + val->flags |= OPVAL_VALID_LL; + } + } + return val->flags & OPVAL_VALID_LL; +} + +robj *zuiObjectFromValue(zsetopval *val) { + if (val->ele == NULL) { + if (val->estr != NULL) { + val->ele = createStringObject((char*)val->estr,val->elen); + } else { + val->ele = createStringObjectFromLongLong(val->ell); + } + val->flags |= OPVAL_DIRTY_ROBJ; + } + return val->ele; +} + +int zuiBufferFromValue(zsetopval *val) { + if (val->estr == NULL) { + if (val->ele != NULL) { + if (val->ele->encoding == REDIS_ENCODING_INT) { + val->elen = ll2string((char*)val->_buf,sizeof(val->_buf),(long)val->ele->ptr); + val->estr = val->_buf; + } else if (val->ele->encoding == REDIS_ENCODING_RAW) { + val->elen = sdslen(val->ele->ptr); + val->estr = val->ele->ptr; + } else { + redisPanic("Unsupported element encoding"); + } + } else { + val->elen = ll2string((char*)val->_buf,sizeof(val->_buf),val->ell); + val->estr = val->_buf; + } + } + return 1; +} + +/* Find value pointed to by val in the source pointer to by op. When found, + * return 1 and store its score in target. Return 0 otherwise. */ +int zuiFind(zsetopsrc *op, zsetopval *val, double *score) { + if (op->subject == NULL) + return 0; + + if (op->type == REDIS_SET) { + iterset *it = &op->iter.set; + + if (op->encoding == REDIS_ENCODING_INTSET) { + if (zuiLongLongFromValue(val) && intsetFind(it->is.is,val->ell)) { + *score = 1.0; + return 1; + } else { + return 0; + } + } else if (op->encoding == REDIS_ENCODING_HT) { + zuiObjectFromValue(val); + if (dictFind(it->ht.dict,val->ele) != NULL) { + *score = 1.0; + return 1; + } else { + return 0; + } + } else { + redisPanic("Unknown set encoding"); + } + } else if (op->type == REDIS_ZSET) { + iterzset *it = &op->iter.zset; + zuiObjectFromValue(val); + + if (op->encoding == REDIS_ENCODING_ZIPLIST) { + if (zzlFind(it->zl.zl,val->ele,score) != NULL) { + /* Score is already set by zzlFind. */ + return 1; + } else { + return 0; + } + } else if (op->encoding == REDIS_ENCODING_SKIPLIST) { + dictEntry *de; + if ((de = dictFind(it->sl.zs->dict,val->ele)) != NULL) { + *score = *(double*)dictGetVal(de); + return 1; + } else { + return 0; + } + } else { + redisPanic("Unknown sorted set encoding"); + } + } else { + redisPanic("Unsupported type"); + } +} + +int zuiCompareByCardinality(const void *s1, const void *s2) { + return zuiLength((zsetopsrc*)s1) - zuiLength((zsetopsrc*)s2); } #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)) +#define zunionInterDictValue(_e) (dictGetVal(_e) == NULL ? 1.0 : *(double*)dictGetVal(_e)) inline static void zunionInterAggregate(double *target, double val, int aggregate) { if (aggregate == REDIS_AGGR_SUM) { *target = *target + val; + /* The result of adding two doubles is NaN when one variable + * is +inf and the other is -inf. When these numbers are added, + * we maintain the convention of the result being 0.0. */ + if (isnan(*target)) *target = 0.0; } else if (aggregate == REDIS_AGGR_MIN) { *target = val < *target ? val : *target; } else if (aggregate == REDIS_AGGR_MAX) { @@ -552,18 +1445,25 @@ inline static void zunionInterAggregate(double *target, double val, int aggregat } void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) { - int i, j, setnum; + int i, j; + long setnum; int aggregate = REDIS_AGGR_SUM; zsetopsrc *src; + zsetopval zval; + robj *tmp; + unsigned int maxelelen = 0; robj *dstobj; zset *dstzset; - dictIterator *di; - dictEntry *de; + zskiplistNode *znode; + int touched = 0; /* expect setnum input keys to be given */ - setnum = atoi(c->argv[2]->ptr); + if ((getLongFromObjectOrReply(c, c->argv[2], &setnum, NULL) != REDIS_OK)) + return; + if (setnum < 1) { - addReplySds(c,sdsnew("-ERR at least 1 input key is needed for ZUNIONSTORE/ZINTERSTORE\r\n")); + addReplyError(c, + "at least 1 input key is needed for ZUNIONSTORE/ZINTERSTORE"); return; } @@ -574,24 +1474,24 @@ void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) { } /* read keys to be used for input */ - src = zmalloc(sizeof(zsetopsrc) * setnum); + src = zcalloc(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 (obj->type == REDIS_ZSET) { - src[i].dict = ((zset*)obj->ptr)->dict; - } else if (obj->type == REDIS_SET) { - src[i].dict = (obj->ptr); - } else { + if (obj != NULL) { + if (obj->type != REDIS_ZSET && obj->type != REDIS_SET) { zfree(src); addReply(c,shared.wrongtypeerr); return; } + + src[i].subject = obj; + src[i].type = obj->type; + src[i].encoding = obj->encoding; + } else { + src[i].subject = NULL; } - /* default all weights to 1 */ + /* Default all weights to 1. */ src[i].weight = 1.0; } @@ -603,8 +1503,12 @@ void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) { if (remaining >= (setnum + 1) && !strcasecmp(c->argv[j]->ptr,"weights")) { j++; remaining--; for (i = 0; i < setnum; i++, j++, remaining--) { - if (getDoubleFromObjectOrReply(c, c->argv[j], &src[i].weight, NULL) != REDIS_OK) + if (getDoubleFromObjectOrReply(c,c->argv[j],&src[i].weight, + "weight value is not a float") != REDIS_OK) + { + zfree(src); return; + } } } else if (remaining >= 2 && !strcasecmp(c->argv[j]->ptr,"aggregate")) { j++; remaining--; @@ -628,90 +1532,122 @@ void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) { } } + for (i = 0; i < setnum; i++) + zuiInitIterator(&src[i]); + /* sort sets from the smallest to largest, this will improve our * algorithm's performance */ - qsort(src,setnum,sizeof(zsetopsrc),qsortCompareZsetopsrcByCardinality); + qsort(src,setnum,sizeof(zsetopsrc),zuiCompareByCardinality); dstobj = createZsetObject(); dstzset = dstobj->ptr; + memset(&zval, 0, sizeof(zval)); if (op == REDIS_OP_INTER) { - /* skip going over all entries if the smallest zset is NULL or empty */ - if (src[0].dict && dictSize(src[0].dict) > 0) { - /* precondition: as src[0].dict is non-empty and the zsets are ordered - * from small to large, all src[i > 0].dict are non-empty too */ - di = dictGetIterator(src[0].dict); - while((de = dictNext(di)) != NULL) { - double *score = zmalloc(sizeof(double)), value; - *score = src[0].weight * zunionInterDictValue(de); + /* Skip everything if the smallest input is empty. */ + if (zuiLength(&src[0]) > 0) { + /* Precondition: as src[0] is non-empty and the inputs are ordered + * by size, all src[i > 0] are non-empty too. */ + while (zuiNext(&src[0],&zval)) { + double score, value; + + score = src[0].weight * zval.score; + if (isnan(score)) score = 0; for (j = 1; j < setnum; j++) { - dictEntry *other = dictFind(src[j].dict,dictGetEntryKey(de)); - if (other) { - value = src[j].weight * zunionInterDictValue(other); - zunionInterAggregate(score, value, aggregate); + /* It is not safe to access the zset we are + * iterating, so explicitly check for equal object. */ + if (src[j].subject == src[0].subject) { + value = zval.score*src[j].weight; + zunionInterAggregate(&score,value,aggregate); + } else if (zuiFind(&src[j],&zval,&value)) { + value *= src[j].weight; + zunionInterAggregate(&score,value,aggregate); } else { break; } } - /* skip entry when not present in every source dict */ - if (j != setnum) { - zfree(score); - } else { - robj *o = dictGetEntryKey(de); - dictAdd(dstzset->dict,o,score); - incrRefCount(o); /* added to dictionary */ - zslInsert(dstzset->zsl,*score,o); - incrRefCount(o); /* added to skiplist */ + /* Only continue when present in every input. */ + if (j == setnum) { + tmp = zuiObjectFromValue(&zval); + znode = zslInsert(dstzset->zsl,score,tmp); + incrRefCount(tmp); /* added to skiplist */ + dictAdd(dstzset->dict,tmp,&znode->score); + incrRefCount(tmp); /* added to dictionary */ + + if (tmp->encoding == REDIS_ENCODING_RAW) + if (sdslen(tmp->ptr) > maxelelen) + maxelelen = sdslen(tmp->ptr); } } - dictReleaseIterator(di); } } else if (op == REDIS_OP_UNION) { for (i = 0; i < setnum; i++) { - if (!src[i].dict) continue; + if (zuiLength(&src[i]) == 0) + continue; + + while (zuiNext(&src[i],&zval)) { + double score, value; - di = dictGetIterator(src[i].dict); - while((de = dictNext(di)) != NULL) { - /* skip key when already processed */ - if (dictFind(dstzset->dict,dictGetEntryKey(de)) != NULL) continue; + /* Skip key when already processed */ + if (dictFind(dstzset->dict,zuiObjectFromValue(&zval)) != NULL) + continue; - double *score = zmalloc(sizeof(double)), value; - *score = src[i].weight * zunionInterDictValue(de); + /* Initialize score */ + score = src[i].weight * zval.score; + if (isnan(score)) score = 0; - /* because the zsets are sorted by size, its only possible - * for sets at larger indices to hold this entry */ + /* Because the inputs are sorted by size, it's only possible + * for sets at larger indices to hold this element. */ for (j = (i+1); j < setnum; j++) { - dictEntry *other = dictFind(src[j].dict,dictGetEntryKey(de)); - if (other) { - value = src[j].weight * zunionInterDictValue(other); - zunionInterAggregate(score, value, aggregate); + /* It is not safe to access the zset we are + * iterating, so explicitly check for equal object. */ + if(src[j].subject == src[i].subject) { + value = zval.score*src[j].weight; + zunionInterAggregate(&score,value,aggregate); + } else if (zuiFind(&src[j],&zval,&value)) { + value *= src[j].weight; + zunionInterAggregate(&score,value,aggregate); } } - robj *o = dictGetEntryKey(de); - dictAdd(dstzset->dict,o,score); - incrRefCount(o); /* added to dictionary */ - zslInsert(dstzset->zsl,*score,o); - incrRefCount(o); /* added to skiplist */ + tmp = zuiObjectFromValue(&zval); + znode = zslInsert(dstzset->zsl,score,tmp); + incrRefCount(zval.ele); /* added to skiplist */ + dictAdd(dstzset->dict,tmp,&znode->score); + incrRefCount(zval.ele); /* added to dictionary */ + + if (tmp->encoding == REDIS_ENCODING_RAW) + if (sdslen(tmp->ptr) > maxelelen) + maxelelen = sdslen(tmp->ptr); } - dictReleaseIterator(di); } } else { - /* unknown operator */ - redisAssert(op == REDIS_OP_INTER || op == REDIS_OP_UNION); + redisPanic("Unknown operator"); } - dbDelete(c->db,dstkey); + for (i = 0; i < setnum; i++) + zuiClearIterator(&src[i]); + + if (dbDelete(c->db,dstkey)) { + signalModifiedKey(c->db,dstkey); + touched = 1; + server.dirty++; + } if (dstzset->zsl->length) { + /* Convert to ziplist when in limits. */ + if (dstzset->zsl->length <= server.zset_max_ziplist_entries && + maxelelen <= server.zset_max_ziplist_value) + zsetConvert(dstobj,REDIS_ENCODING_ZIPLIST); + dbAdd(c->db,dstkey,dstobj); - addReplyLongLong(c, dstzset->zsl->length); - touchWatchedKey(c->db,dstkey); + addReplyLongLong(c,zsetLength(dstobj)); + if (!touched) signalModifiedKey(c->db,dstkey); server.dirty++; } else { decrRefCount(dstobj); - addReply(c, shared.czero); + addReply(c,shared.czero); } zfree(src); } @@ -725,16 +1661,13 @@ void zinterstoreCommand(redisClient *c) { } void zrangeGenericCommand(redisClient *c, int reverse) { - robj *o; + robj *key = c->argv[1]; + robj *zobj; + int withscores = 0; long start; long end; - int withscores = 0; int llen; - int rangelen, j; - zset *zsetobj; - zskiplist *zsl; - zskiplistNode *ln; - robj *ele; + int rangelen; if ((getLongFromObjectOrReply(c, c->argv[2], &start, NULL) != REDIS_OK) || (getLongFromObjectOrReply(c, c->argv[3], &end, NULL) != REDIS_OK)) return; @@ -746,13 +1679,11 @@ void zrangeGenericCommand(redisClient *c, int reverse) { 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; + if ((zobj = lookupKeyReadOrReply(c,key,shared.emptymultibulk)) == NULL + || checkType(c,zobj,REDIS_ZSET)) return; - /* convert negative indexes */ + /* Sanitize indexes. */ + llen = zsetLength(zobj); if (start < 0) start = llen+start; if (end < 0) end = llen+end; if (start < 0) start = 0; @@ -766,24 +1697,68 @@ void zrangeGenericCommand(redisClient *c, int reverse) { if (end >= llen) end = llen-1; rangelen = (end-start)+1; - /* check if starting point is trivial, before searching - * the element in log(N) time */ - if (reverse) { - ln = start == 0 ? zsl->tail : zslistTypeGetElementByRank(zsl, llen-start); - } else { - ln = start == 0 ? - zsl->header->forward[0] : zslistTypeGetElementByRank(zsl, start+1); - } - /* Return the result in form of a multi-bulk reply */ - addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n", - withscores ? (rangelen*2) : rangelen)); - for (j = 0; j < rangelen; j++) { - ele = ln->obj; - addReplyBulk(c,ele); - if (withscores) - addReplyDouble(c,ln->score); - ln = reverse ? ln->backward : ln->forward[0]; + addReplyMultiBulkLen(c, withscores ? (rangelen*2) : rangelen); + + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + unsigned char *zl = zobj->ptr; + unsigned char *eptr, *sptr; + unsigned char *vstr; + unsigned int vlen; + long long vlong; + + if (reverse) + eptr = ziplistIndex(zl,-2-(2*start)); + else + eptr = ziplistIndex(zl,2*start); + + redisAssertWithInfo(c,zobj,eptr != NULL); + sptr = ziplistNext(zl,eptr); + + while (rangelen--) { + redisAssertWithInfo(c,zobj,eptr != NULL && sptr != NULL); + redisAssertWithInfo(c,zobj,ziplistGet(eptr,&vstr,&vlen,&vlong)); + if (vstr == NULL) + addReplyBulkLongLong(c,vlong); + else + addReplyBulkCBuffer(c,vstr,vlen); + + if (withscores) + addReplyDouble(c,zzlGetScore(sptr)); + + if (reverse) + zzlPrev(zl,&eptr,&sptr); + else + zzlNext(zl,&eptr,&sptr); + } + + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + zset *zs = zobj->ptr; + zskiplist *zsl = zs->zsl; + zskiplistNode *ln; + robj *ele; + + /* Check if starting point is trivial, before doing log(N) lookup. */ + if (reverse) { + ln = zsl->tail; + if (start > 0) + ln = zslGetElementByRank(zsl,llen-start); + } else { + ln = zsl->header->level[0].forward; + if (start > 0) + ln = zslGetElementByRank(zsl,start+1); + } + + while(rangelen--) { + redisAssertWithInfo(c,zobj,ln != NULL); + ele = ln->obj; + addReplyBulk(c,ele); + if (withscores) + addReplyDouble(c,ln->score); + ln = reverse ? ln->backward : ln->level[0].forward; + } + } else { + redisPanic("Unknown sorted set encoding"); } } @@ -795,189 +1770,376 @@ void zrevrangeCommand(redisClient *c) { zrangeGenericCommand(c,1); } -/* This command implements both ZRANGEBYSCORE and ZCOUNT. - * If justcount is non-zero, just the count is returned. */ -void genericZrangebyscoreCommand(redisClient *c, int justcount) { - robj *o; - double min, max; - int minex = 0, maxex = 0; /* are min or max exclusive? */ - int offset = 0, limit = -1; +/* This command implements ZRANGEBYSCORE, ZREVRANGEBYSCORE. */ +void genericZrangebyscoreCommand(redisClient *c, int reverse) { + zrangespec range; + robj *key = c->argv[1]; + robj *zobj; + long offset = 0, limit = -1; int withscores = 0; - int badsyntax = 0; + unsigned long rangelen = 0; + void *replylen = NULL; + int minidx, maxidx; - /* Parse the min-max interval. If one of the values is prefixed - * by the "(" character, it's considered "open". For instance - * ZRANGEBYSCORE zset (1.5 (2.5 will match min < x < max - * ZRANGEBYSCORE zset 1.5 2.5 will instead match min <= x <= max */ - if (((char*)c->argv[2]->ptr)[0] == '(') { - min = strtod((char*)c->argv[2]->ptr+1,NULL); - minex = 1; - } else { - min = strtod(c->argv[2]->ptr,NULL); - } - if (((char*)c->argv[3]->ptr)[0] == '(') { - max = strtod((char*)c->argv[3]->ptr+1,NULL); - maxex = 1; + /* Parse the range arguments. */ + if (reverse) { + /* Range is given as [max,min] */ + maxidx = 2; minidx = 3; } else { - max = strtod(c->argv[3]->ptr,NULL); + /* Range is given as [min,max] */ + minidx = 2; maxidx = 3; } - /* Parse "WITHSCORES": note that if the command was called with - * the name ZCOUNT then we are sure that c->argc == 4, so we'll never - * enter the following paths to parse WITHSCORES and LIMIT. */ - if (c->argc == 5 || c->argc == 8) { - if (strcasecmp(c->argv[c->argc-1]->ptr,"withscores") == 0) - withscores = 1; - else - badsyntax = 1; - } - if (c->argc != (4 + withscores) && c->argc != (7 + withscores)) - badsyntax = 1; - if (badsyntax) { - addReplySds(c, - sdsnew("-ERR wrong number of arguments for ZRANGEBYSCORE\r\n")); + if (zslParseRange(c->argv[minidx],c->argv[maxidx],&range) != REDIS_OK) { + addReplyError(c,"min or max is not a float"); return; } - /* Parse "LIMIT" */ - if (c->argc == (7 + withscores) && strcasecmp(c->argv[4]->ptr,"limit")) { - addReply(c,shared.syntaxerr); - return; - } else if (c->argc == (7 + withscores)) { - offset = atoi(c->argv[5]->ptr); - limit = atoi(c->argv[6]->ptr); - if (offset < 0) offset = 0; + /* Parse optional extra arguments. Note that ZCOUNT will exactly have + * 4 arguments, so we'll never enter the following code path. */ + if (c->argc > 4) { + int remaining = c->argc - 4; + int pos = 4; + + while (remaining) { + if (remaining >= 1 && !strcasecmp(c->argv[pos]->ptr,"withscores")) { + pos++; remaining--; + withscores = 1; + } else if (remaining >= 3 && !strcasecmp(c->argv[pos]->ptr,"limit")) { + if ((getLongFromObjectOrReply(c, c->argv[pos+1], &offset, NULL) != REDIS_OK) || + (getLongFromObjectOrReply(c, c->argv[pos+2], &limit, NULL) != REDIS_OK)) return; + pos += 3; remaining -= 3; + } else { + addReply(c,shared.syntaxerr); + return; + } + } } /* Ok, lookup the key and get the range */ - o = lookupKeyRead(c->db,c->argv[1]); - if (o == NULL) { - addReply(c,justcount ? shared.czero : shared.emptymultibulk); - } else { - if (o->type != REDIS_ZSET) { - addReply(c,shared.wrongtypeerr); + if ((zobj = lookupKeyReadOrReply(c,key,shared.emptymultibulk)) == NULL || + checkType(c,zobj,REDIS_ZSET)) return; + + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + unsigned char *zl = zobj->ptr; + unsigned char *eptr, *sptr; + unsigned char *vstr; + unsigned int vlen; + long long vlong; + double score; + + /* If reversed, get the last node in range as starting point. */ + if (reverse) { + eptr = zzlLastInRange(zl,range); } else { - zset *zsetobj = o->ptr; - zskiplist *zsl = zsetobj->zsl; - zskiplistNode *ln; - robj *ele, *lenobj = NULL; - unsigned long rangelen = 0; - - /* Get the first node with the score >= min, or with - * score > min if 'minex' is true. */ - ln = zslFirstWithScore(zsl,min); - while (minex && ln && ln->score == min) ln = ln->forward[0]; - - if (ln == NULL) { - /* No element matching the speciifed interval */ - addReply(c,justcount ? shared.czero : shared.emptymultibulk); - return; + eptr = zzlFirstInRange(zl,range); + } + + /* No "first" element in the specified interval. */ + if (eptr == NULL) { + addReply(c, shared.emptymultibulk); + return; + } + + /* Get score pointer for the first element. */ + redisAssertWithInfo(c,zobj,eptr != NULL); + sptr = ziplistNext(zl,eptr); + + /* We don't know in advance how many matching elements there are in the + * list, so we push this object that will represent the multi-bulk + * length in the output buffer, and will "fix" it later */ + replylen = addDeferredMultiBulkLength(c); + + /* If there is an offset, just traverse the number of elements without + * checking the score because that is done in the next loop. */ + while (eptr && offset--) { + if (reverse) { + zzlPrev(zl,&eptr,&sptr); + } else { + zzlNext(zl,&eptr,&sptr); } + } + + while (eptr && limit--) { + score = zzlGetScore(sptr); - /* We don't know in advance how many matching elements there - * are in the list, so we push this object that will represent - * the multi-bulk length in the output buffer, and will "fix" - * it later */ - if (!justcount) { - lenobj = createObject(REDIS_STRING,NULL); - addReply(c,lenobj); - decrRefCount(lenobj); + /* Abort when the node is no longer in range. */ + if (reverse) { + if (!zslValueGteMin(score,&range)) break; + } else { + if (!zslValueLteMax(score,&range)) break; } - while(ln && (maxex ? (ln->score < max) : (ln->score <= max))) { - if (offset) { - offset--; - ln = ln->forward[0]; - continue; - } - if (limit == 0) break; - if (!justcount) { - ele = ln->obj; - addReplyBulk(c,ele); - if (withscores) - addReplyDouble(c,ln->score); - } - ln = ln->forward[0]; - rangelen++; - if (limit > 0) limit--; + /* We know the element exists, so ziplistGet should always succeed */ + redisAssertWithInfo(c,zobj,ziplistGet(eptr,&vstr,&vlen,&vlong)); + + rangelen++; + if (vstr == NULL) { + addReplyBulkLongLong(c,vlong); + } else { + addReplyBulkCBuffer(c,vstr,vlen); + } + + if (withscores) { + addReplyDouble(c,score); + } + + /* Move to next node */ + if (reverse) { + zzlPrev(zl,&eptr,&sptr); + } else { + zzlNext(zl,&eptr,&sptr); + } + } + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + zset *zs = zobj->ptr; + zskiplist *zsl = zs->zsl; + zskiplistNode *ln; + + /* If reversed, get the last node in range as starting point. */ + if (reverse) { + ln = zslLastInRange(zsl,range); + } else { + ln = zslFirstInRange(zsl,range); + } + + /* No "first" element in the specified interval. */ + if (ln == NULL) { + addReply(c, shared.emptymultibulk); + return; + } + + /* We don't know in advance how many matching elements there are in the + * list, so we push this object that will represent the multi-bulk + * length in the output buffer, and will "fix" it later */ + replylen = addDeferredMultiBulkLength(c); + + /* If there is an offset, just traverse the number of elements without + * checking the score because that is done in the next loop. */ + while (ln && offset--) { + if (reverse) { + ln = ln->backward; + } else { + ln = ln->level[0].forward; } - if (justcount) { - addReplyLongLong(c,(long)rangelen); + } + + while (ln && limit--) { + /* Abort when the node is no longer in range. */ + if (reverse) { + if (!zslValueGteMin(ln->score,&range)) break; } else { - lenobj->ptr = sdscatprintf(sdsempty(),"*%lu\r\n", - withscores ? (rangelen*2) : rangelen); + if (!zslValueLteMax(ln->score,&range)) break; + } + + rangelen++; + addReplyBulk(c,ln->obj); + + if (withscores) { + addReplyDouble(c,ln->score); + } + + /* Move to next node */ + if (reverse) { + ln = ln->backward; + } else { + ln = ln->level[0].forward; } } + } else { + redisPanic("Unknown sorted set encoding"); + } + + if (withscores) { + rangelen *= 2; } + + setDeferredMultiBulkLength(c, replylen, rangelen); } void zrangebyscoreCommand(redisClient *c) { genericZrangebyscoreCommand(c,0); } -void zcountCommand(redisClient *c) { +void zrevrangebyscoreCommand(redisClient *c) { genericZrangebyscoreCommand(c,1); } +void zcountCommand(redisClient *c) { + robj *key = c->argv[1]; + robj *zobj; + zrangespec range; + int count = 0; + + /* Parse the range arguments */ + if (zslParseRange(c->argv[2],c->argv[3],&range) != REDIS_OK) { + addReplyError(c,"min or max is not a float"); + return; + } + + /* Lookup the sorted set */ + if ((zobj = lookupKeyReadOrReply(c, key, shared.czero)) == NULL || + checkType(c, zobj, REDIS_ZSET)) return; + + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + unsigned char *zl = zobj->ptr; + unsigned char *eptr, *sptr; + double score; + + /* Use the first element in range as the starting point */ + eptr = zzlFirstInRange(zl,range); + + /* No "first" element */ + if (eptr == NULL) { + addReply(c, shared.czero); + return; + } + + /* First element is in range */ + sptr = ziplistNext(zl,eptr); + score = zzlGetScore(sptr); + redisAssertWithInfo(c,zobj,zslValueLteMax(score,&range)); + + /* Iterate over elements in range */ + while (eptr) { + score = zzlGetScore(sptr); + + /* Abort when the node is no longer in range. */ + if (!zslValueLteMax(score,&range)) { + break; + } else { + count++; + zzlNext(zl,&eptr,&sptr); + } + } + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + zset *zs = zobj->ptr; + zskiplist *zsl = zs->zsl; + zskiplistNode *zn; + unsigned long rank; + + /* Find first element in range */ + zn = zslFirstInRange(zsl, range); + + /* Use rank of first element, if any, to determine preliminary count */ + if (zn != NULL) { + rank = zslGetRank(zsl, zn->score, zn->obj); + count = (zsl->length - (rank - 1)); + + /* Find last element in range */ + zn = zslLastInRange(zsl, range); + + /* Use rank of last element, if any, to determine the actual count */ + if (zn != NULL) { + rank = zslGetRank(zsl, zn->score, zn->obj); + count -= (zsl->length - rank); + } + } + } else { + redisPanic("Unknown sorted set encoding"); + } + + addReplyLongLong(c, count); +} + void zcardCommand(redisClient *c) { - robj *o; - zset *zs; + robj *key = c->argv[1]; + robj *zobj; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || - checkType(c,o,REDIS_ZSET)) return; + if ((zobj = lookupKeyReadOrReply(c,key,shared.czero)) == NULL || + checkType(c,zobj,REDIS_ZSET)) return; - zs = o->ptr; - addReplyUlong(c,zs->zsl->length); + addReplyLongLong(c,zsetLength(zobj)); } void zscoreCommand(redisClient *c) { - robj *o; - zset *zs; - dictEntry *de; + robj *key = c->argv[1]; + robj *zobj; + double score; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL || - checkType(c,o,REDIS_ZSET)) return; + if ((zobj = lookupKeyReadOrReply(c,key,shared.nullbulk)) == NULL || + checkType(c,zobj,REDIS_ZSET)) return; - zs = o->ptr; - de = dictFind(zs->dict,c->argv[2]); - if (!de) { - addReply(c,shared.nullbulk); - } else { - double *score = dictGetEntryVal(de); + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + if (zzlFind(zobj->ptr,c->argv[2],&score) != NULL) + addReplyDouble(c,score); + else + addReply(c,shared.nullbulk); + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + zset *zs = zobj->ptr; + dictEntry *de; - addReplyDouble(c,*score); + c->argv[2] = tryObjectEncoding(c->argv[2]); + de = dictFind(zs->dict,c->argv[2]); + if (de != NULL) { + score = *(double*)dictGetVal(de); + addReplyDouble(c,score); + } else { + addReply(c,shared.nullbulk); + } + } else { + redisPanic("Unknown sorted set encoding"); } } void zrankGenericCommand(redisClient *c, int reverse) { - robj *o; - zset *zs; - zskiplist *zsl; - dictEntry *de; + robj *key = c->argv[1]; + robj *ele = c->argv[2]; + robj *zobj; + unsigned long llen; unsigned long rank; - double *score; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL || - checkType(c,o,REDIS_ZSET)) return; + if ((zobj = lookupKeyReadOrReply(c,key,shared.nullbulk)) == NULL || + checkType(c,zobj,REDIS_ZSET)) return; + llen = zsetLength(zobj); + + redisAssertWithInfo(c,ele,ele->encoding == REDIS_ENCODING_RAW); + if (zobj->encoding == REDIS_ENCODING_ZIPLIST) { + unsigned char *zl = zobj->ptr; + unsigned char *eptr, *sptr; + + eptr = ziplistIndex(zl,0); + redisAssertWithInfo(c,zobj,eptr != NULL); + sptr = ziplistNext(zl,eptr); + redisAssertWithInfo(c,zobj,sptr != NULL); + + rank = 1; + while(eptr != NULL) { + if (ziplistCompare(eptr,ele->ptr,sdslen(ele->ptr))) + break; + rank++; + zzlNext(zl,&eptr,&sptr); + } - zs = o->ptr; - zsl = zs->zsl; - de = dictFind(zs->dict,c->argv[2]); - if (!de) { - addReply(c,shared.nullbulk); - return; - } + if (eptr != NULL) { + if (reverse) + addReplyLongLong(c,llen-rank); + else + addReplyLongLong(c,rank-1); + } else { + addReply(c,shared.nullbulk); + } + } else if (zobj->encoding == REDIS_ENCODING_SKIPLIST) { + zset *zs = zobj->ptr; + zskiplist *zsl = zs->zsl; + dictEntry *de; + double score; - score = dictGetEntryVal(de); - rank = zslistTypeGetRank(zsl, *score, c->argv[2]); - if (rank) { - if (reverse) { - addReplyLongLong(c, zsl->length - rank); + ele = c->argv[2] = tryObjectEncoding(c->argv[2]); + de = dictFind(zs->dict,ele); + if (de != NULL) { + score = *(double*)dictGetVal(de); + rank = zslGetRank(zsl,score,ele); + redisAssertWithInfo(c,ele,rank); /* Existing elements always have a rank. */ + if (reverse) + addReplyLongLong(c,llen-rank); + else + addReplyLongLong(c,rank-1); } else { - addReplyLongLong(c, rank-1); + addReply(c,shared.nullbulk); } } else { - addReply(c,shared.nullbulk); + redisPanic("Unknown sorted set encoding"); } }