]>
git.saurik.com Git - redis.git/blob - src/t_zset.c
5 /*-----------------------------------------------------------------------------
7 *----------------------------------------------------------------------------*/
9 /* ZSETs are ordered sets using two data structures to hold the same elements
10 * in order to get O(log(N)) INSERT and REMOVE operations into a sorted
13 * The elements are added to an hash table mapping Redis objects to scores.
14 * At the same time the elements are added to a skip list mapping scores
15 * to Redis objects (so objects are sorted by scores in this "view"). */
17 /* This skiplist implementation is almost a C translation of the original
18 * algorithm described by William Pugh in "Skip Lists: A Probabilistic
19 * Alternative to Balanced Trees", modified in three ways:
20 * a) this implementation allows for repeated values.
21 * b) the comparison is not just by key (our 'score') but by satellite data.
22 * c) there is a back pointer, so it's a doubly linked list with the back
23 * pointers being only at "level 1". This allows to traverse the list
24 * from tail to head, useful for ZREVRANGE. */
26 zskiplistNode
*zslCreateNode(int level
, double score
, robj
*obj
) {
27 zskiplistNode
*zn
= zmalloc(sizeof(*zn
)+level
*sizeof(struct zskiplistLevel
));
33 zskiplist
*zslCreate(void) {
37 zsl
= zmalloc(sizeof(*zsl
));
40 zsl
->header
= zslCreateNode(ZSKIPLIST_MAXLEVEL
,0,NULL
);
41 for (j
= 0; j
< ZSKIPLIST_MAXLEVEL
; j
++) {
42 zsl
->header
->level
[j
].forward
= NULL
;
43 zsl
->header
->level
[j
].span
= 0;
45 zsl
->header
->backward
= NULL
;
50 void zslFreeNode(zskiplistNode
*node
) {
51 decrRefCount(node
->obj
);
55 void zslFree(zskiplist
*zsl
) {
56 zskiplistNode
*node
= zsl
->header
->level
[0].forward
, *next
;
60 next
= node
->level
[0].forward
;
67 int zslRandomLevel(void) {
69 while ((random()&0xFFFF) < (ZSKIPLIST_P
* 0xFFFF))
71 return (level
<ZSKIPLIST_MAXLEVEL
) ? level
: ZSKIPLIST_MAXLEVEL
;
74 zskiplistNode
*zslInsert(zskiplist
*zsl
, double score
, robj
*obj
) {
75 zskiplistNode
*update
[ZSKIPLIST_MAXLEVEL
], *x
;
76 unsigned int rank
[ZSKIPLIST_MAXLEVEL
];
80 for (i
= zsl
->level
-1; i
>= 0; i
--) {
81 /* store rank that is crossed to reach the insert position */
82 rank
[i
] = i
== (zsl
->level
-1) ? 0 : rank
[i
+1];
83 while (x
->level
[i
].forward
&&
84 (x
->level
[i
].forward
->score
< score
||
85 (x
->level
[i
].forward
->score
== score
&&
86 compareStringObjects(x
->level
[i
].forward
->obj
,obj
) < 0))) {
87 rank
[i
] += x
->level
[i
].span
;
88 x
= x
->level
[i
].forward
;
92 /* we assume the key is not already inside, since we allow duplicated
93 * scores, and the re-insertion of score and redis object should never
94 * happpen since the caller of zslInsert() should test in the hash table
95 * if the element is already inside or not. */
96 level
= zslRandomLevel();
97 if (level
> zsl
->level
) {
98 for (i
= zsl
->level
; i
< level
; i
++) {
100 update
[i
] = zsl
->header
;
101 update
[i
]->level
[i
].span
= zsl
->length
;
105 x
= zslCreateNode(level
,score
,obj
);
106 for (i
= 0; i
< level
; i
++) {
107 x
->level
[i
].forward
= update
[i
]->level
[i
].forward
;
108 update
[i
]->level
[i
].forward
= x
;
110 /* update span covered by update[i] as x is inserted here */
111 x
->level
[i
].span
= update
[i
]->level
[i
].span
- (rank
[0] - rank
[i
]);
112 update
[i
]->level
[i
].span
= (rank
[0] - rank
[i
]) + 1;
115 /* increment span for untouched levels */
116 for (i
= level
; i
< zsl
->level
; i
++) {
117 update
[i
]->level
[i
].span
++;
120 x
->backward
= (update
[0] == zsl
->header
) ? NULL
: update
[0];
121 if (x
->level
[0].forward
)
122 x
->level
[0].forward
->backward
= x
;
129 /* Internal function used by zslDelete, zslDeleteByScore and zslDeleteByRank */
130 void zslDeleteNode(zskiplist
*zsl
, zskiplistNode
*x
, zskiplistNode
**update
) {
132 for (i
= 0; i
< zsl
->level
; i
++) {
133 if (update
[i
]->level
[i
].forward
== x
) {
134 update
[i
]->level
[i
].span
+= x
->level
[i
].span
- 1;
135 update
[i
]->level
[i
].forward
= x
->level
[i
].forward
;
137 update
[i
]->level
[i
].span
-= 1;
140 if (x
->level
[0].forward
) {
141 x
->level
[0].forward
->backward
= x
->backward
;
143 zsl
->tail
= x
->backward
;
145 while(zsl
->level
> 1 && zsl
->header
->level
[zsl
->level
-1].forward
== NULL
)
150 /* Delete an element with matching score/object from the skiplist. */
151 int zslDelete(zskiplist
*zsl
, double score
, robj
*obj
) {
152 zskiplistNode
*update
[ZSKIPLIST_MAXLEVEL
], *x
;
156 for (i
= zsl
->level
-1; i
>= 0; i
--) {
157 while (x
->level
[i
].forward
&&
158 (x
->level
[i
].forward
->score
< score
||
159 (x
->level
[i
].forward
->score
== score
&&
160 compareStringObjects(x
->level
[i
].forward
->obj
,obj
) < 0)))
161 x
= x
->level
[i
].forward
;
164 /* We may have multiple elements with the same score, what we need
165 * is to find the element with both the right score and object. */
166 x
= x
->level
[0].forward
;
167 if (x
&& score
== x
->score
&& equalStringObjects(x
->obj
,obj
)) {
168 zslDeleteNode(zsl
, x
, update
);
172 return 0; /* not found */
174 return 0; /* not found */
177 static int zslValueGteMin(double value
, zrangespec
*spec
) {
178 return spec
->minex
? (value
> spec
->min
) : (value
>= spec
->min
);
181 static int zslValueLteMax(double value
, zrangespec
*spec
) {
182 return spec
->maxex
? (value
< spec
->max
) : (value
<= spec
->max
);
185 /* Returns if there is a part of the zset is in range. */
186 int zslIsInRange(zskiplist
*zsl
, zrangespec
*range
) {
189 /* Test for ranges that will always be empty. */
190 if (range
->min
> range
->max
||
191 (range
->min
== range
->max
&& (range
->minex
|| range
->maxex
)))
194 if (x
== NULL
|| !zslValueGteMin(x
->score
,range
))
196 x
= zsl
->header
->level
[0].forward
;
197 if (x
== NULL
|| !zslValueLteMax(x
->score
,range
))
202 /* Find the first node that is contained in the specified range.
203 * Returns NULL when no element is contained in the range. */
204 zskiplistNode
*zslFirstInRange(zskiplist
*zsl
, zrangespec range
) {
208 /* If everything is out of range, return early. */
209 if (!zslIsInRange(zsl
,&range
)) return NULL
;
212 for (i
= zsl
->level
-1; i
>= 0; i
--) {
213 /* Go forward while *OUT* of range. */
214 while (x
->level
[i
].forward
&&
215 !zslValueGteMin(x
->level
[i
].forward
->score
,&range
))
216 x
= x
->level
[i
].forward
;
219 /* This is an inner range, so the next node cannot be NULL. */
220 x
= x
->level
[0].forward
;
221 redisAssert(x
!= NULL
);
223 /* Check if score <= max. */
224 if (!zslValueLteMax(x
->score
,&range
)) return NULL
;
228 /* Find the last node that is contained in the specified range.
229 * Returns NULL when no element is contained in the range. */
230 zskiplistNode
*zslLastInRange(zskiplist
*zsl
, zrangespec range
) {
234 /* If everything is out of range, return early. */
235 if (!zslIsInRange(zsl
,&range
)) return NULL
;
238 for (i
= zsl
->level
-1; i
>= 0; i
--) {
239 /* Go forward while *IN* range. */
240 while (x
->level
[i
].forward
&&
241 zslValueLteMax(x
->level
[i
].forward
->score
,&range
))
242 x
= x
->level
[i
].forward
;
245 /* This is an inner range, so this node cannot be NULL. */
246 redisAssert(x
!= NULL
);
248 /* Check if score >= min. */
249 if (!zslValueGteMin(x
->score
,&range
)) return NULL
;
253 /* Delete all the elements with score between min and max from the skiplist.
254 * Min and mx are inclusive, so a score >= min || score <= max is deleted.
255 * Note that this function takes the reference to the hash table view of the
256 * sorted set, in order to remove the elements from the hash table too. */
257 unsigned long zslDeleteRangeByScore(zskiplist
*zsl
, zrangespec range
, dict
*dict
) {
258 zskiplistNode
*update
[ZSKIPLIST_MAXLEVEL
], *x
;
259 unsigned long removed
= 0;
263 for (i
= zsl
->level
-1; i
>= 0; i
--) {
264 while (x
->level
[i
].forward
&& (range
.minex
?
265 x
->level
[i
].forward
->score
<= range
.min
:
266 x
->level
[i
].forward
->score
< range
.min
))
267 x
= x
->level
[i
].forward
;
271 /* Current node is the last with score < or <= min. */
272 x
= x
->level
[0].forward
;
274 /* Delete nodes while in range. */
275 while (x
&& (range
.maxex
? x
->score
< range
.max
: x
->score
<= range
.max
)) {
276 zskiplistNode
*next
= x
->level
[0].forward
;
277 zslDeleteNode(zsl
,x
,update
);
278 dictDelete(dict
,x
->obj
);
286 /* Delete all the elements with rank between start and end from the skiplist.
287 * Start and end are inclusive. Note that start and end need to be 1-based */
288 unsigned long zslDeleteRangeByRank(zskiplist
*zsl
, unsigned int start
, unsigned int end
, dict
*dict
) {
289 zskiplistNode
*update
[ZSKIPLIST_MAXLEVEL
], *x
;
290 unsigned long traversed
= 0, removed
= 0;
294 for (i
= zsl
->level
-1; i
>= 0; i
--) {
295 while (x
->level
[i
].forward
&& (traversed
+ x
->level
[i
].span
) < start
) {
296 traversed
+= x
->level
[i
].span
;
297 x
= x
->level
[i
].forward
;
303 x
= x
->level
[0].forward
;
304 while (x
&& traversed
<= end
) {
305 zskiplistNode
*next
= x
->level
[0].forward
;
306 zslDeleteNode(zsl
,x
,update
);
307 dictDelete(dict
,x
->obj
);
316 /* Find the rank for an element by both score and key.
317 * Returns 0 when the element cannot be found, rank otherwise.
318 * Note that the rank is 1-based due to the span of zsl->header to the
320 unsigned long zslGetRank(zskiplist
*zsl
, double score
, robj
*o
) {
322 unsigned long rank
= 0;
326 for (i
= zsl
->level
-1; i
>= 0; i
--) {
327 while (x
->level
[i
].forward
&&
328 (x
->level
[i
].forward
->score
< score
||
329 (x
->level
[i
].forward
->score
== score
&&
330 compareStringObjects(x
->level
[i
].forward
->obj
,o
) <= 0))) {
331 rank
+= x
->level
[i
].span
;
332 x
= x
->level
[i
].forward
;
335 /* x might be equal to zsl->header, so test if obj is non-NULL */
336 if (x
->obj
&& equalStringObjects(x
->obj
,o
)) {
343 /* Finds an element by its rank. The rank argument needs to be 1-based. */
344 zskiplistNode
* zslGetElementByRank(zskiplist
*zsl
, unsigned long rank
) {
346 unsigned long traversed
= 0;
350 for (i
= zsl
->level
-1; i
>= 0; i
--) {
351 while (x
->level
[i
].forward
&& (traversed
+ x
->level
[i
].span
) <= rank
)
353 traversed
+= x
->level
[i
].span
;
354 x
= x
->level
[i
].forward
;
356 if (traversed
== rank
) {
363 /* Populate the rangespec according to the objects min and max. */
364 static int zslParseRange(robj
*min
, robj
*max
, zrangespec
*spec
) {
366 spec
->minex
= spec
->maxex
= 0;
368 /* Parse the min-max interval. If one of the values is prefixed
369 * by the "(" character, it's considered "open". For instance
370 * ZRANGEBYSCORE zset (1.5 (2.5 will match min < x < max
371 * ZRANGEBYSCORE zset 1.5 2.5 will instead match min <= x <= max */
372 if (min
->encoding
== REDIS_ENCODING_INT
) {
373 spec
->min
= (long)min
->ptr
;
375 if (((char*)min
->ptr
)[0] == '(') {
376 spec
->min
= strtod((char*)min
->ptr
+1,&eptr
);
377 if (eptr
[0] != '\0' || isnan(spec
->min
)) return REDIS_ERR
;
380 spec
->min
= strtod((char*)min
->ptr
,&eptr
);
381 if (eptr
[0] != '\0' || isnan(spec
->min
)) return REDIS_ERR
;
384 if (max
->encoding
== REDIS_ENCODING_INT
) {
385 spec
->max
= (long)max
->ptr
;
387 if (((char*)max
->ptr
)[0] == '(') {
388 spec
->max
= strtod((char*)max
->ptr
+1,&eptr
);
389 if (eptr
[0] != '\0' || isnan(spec
->max
)) return REDIS_ERR
;
392 spec
->max
= strtod((char*)max
->ptr
,&eptr
);
393 if (eptr
[0] != '\0' || isnan(spec
->max
)) return REDIS_ERR
;
400 /*-----------------------------------------------------------------------------
401 * Ziplist-backed sorted set API
402 *----------------------------------------------------------------------------*/
404 double zzlGetScore(unsigned char *sptr
) {
411 redisAssert(sptr
!= NULL
);
412 redisAssert(ziplistGet(sptr
,&vstr
,&vlen
,&vlong
));
415 memcpy(buf
,vstr
,vlen
);
417 score
= strtod(buf
,NULL
);
425 /* Compare element in sorted set with given element. */
426 int zzlCompareElements(unsigned char *eptr
, unsigned char *cstr
, unsigned int clen
) {
430 unsigned char vbuf
[32];
433 redisAssert(ziplistGet(eptr
,&vstr
,&vlen
,&vlong
));
435 /* Store string representation of long long in buf. */
436 vlen
= ll2string((char*)vbuf
,sizeof(vbuf
),vlong
);
440 minlen
= (vlen
< clen
) ? vlen
: clen
;
441 cmp
= memcmp(vstr
,cstr
,minlen
);
442 if (cmp
== 0) return vlen
-clen
;
446 unsigned int zzlLength(unsigned char *zl
) {
447 return ziplistLen(zl
)/2;
450 /* Move to next entry based on the values in eptr and sptr. Both are set to
451 * NULL when there is no next entry. */
452 void zzlNext(unsigned char *zl
, unsigned char **eptr
, unsigned char **sptr
) {
453 unsigned char *_eptr
, *_sptr
;
454 redisAssert(*eptr
!= NULL
&& *sptr
!= NULL
);
456 _eptr
= ziplistNext(zl
,*sptr
);
458 _sptr
= ziplistNext(zl
,_eptr
);
459 redisAssert(_sptr
!= NULL
);
469 /* Move to the previous entry based on the values in eptr and sptr. Both are
470 * set to NULL when there is no next entry. */
471 void zzlPrev(unsigned char *zl
, unsigned char **eptr
, unsigned char **sptr
) {
472 unsigned char *_eptr
, *_sptr
;
473 redisAssert(*eptr
!= NULL
&& *sptr
!= NULL
);
475 _sptr
= ziplistPrev(zl
,*eptr
);
477 _eptr
= ziplistPrev(zl
,_sptr
);
478 redisAssert(_eptr
!= NULL
);
480 /* No previous entry. */
488 /* Returns if there is a part of the zset is in range. Should only be used
489 * internally by zzlFirstInRange and zzlLastInRange. */
490 int zzlIsInRange(unsigned char *zl
, zrangespec
*range
) {
494 /* Test for ranges that will always be empty. */
495 if (range
->min
> range
->max
||
496 (range
->min
== range
->max
&& (range
->minex
|| range
->maxex
)))
499 p
= ziplistIndex(zl
,-1); /* Last score. */
500 redisAssert(p
!= NULL
);
501 score
= zzlGetScore(p
);
502 if (!zslValueGteMin(score
,range
))
505 p
= ziplistIndex(zl
,1); /* First score. */
506 redisAssert(p
!= NULL
);
507 score
= zzlGetScore(p
);
508 if (!zslValueLteMax(score
,range
))
514 /* Find pointer to the first element contained in the specified range.
515 * Returns NULL when no element is contained in the range. */
516 unsigned char *zzlFirstInRange(unsigned char *zl
, zrangespec range
) {
517 unsigned char *eptr
= ziplistIndex(zl
,0), *sptr
;
520 /* If everything is out of range, return early. */
521 if (!zzlIsInRange(zl
,&range
)) return NULL
;
523 while (eptr
!= NULL
) {
524 sptr
= ziplistNext(zl
,eptr
);
525 redisAssert(sptr
!= NULL
);
527 score
= zzlGetScore(sptr
);
528 if (zslValueGteMin(score
,&range
)) {
529 /* Check if score <= max. */
530 if (zslValueLteMax(score
,&range
))
535 /* Move to next element. */
536 eptr
= ziplistNext(zl
,sptr
);
542 /* Find pointer to the last element contained in the specified range.
543 * Returns NULL when no element is contained in the range. */
544 unsigned char *zzlLastInRange(unsigned char *zl
, zrangespec range
) {
545 unsigned char *eptr
= ziplistIndex(zl
,-2), *sptr
;
548 /* If everything is out of range, return early. */
549 if (!zzlIsInRange(zl
,&range
)) return NULL
;
551 while (eptr
!= NULL
) {
552 sptr
= ziplistNext(zl
,eptr
);
553 redisAssert(sptr
!= NULL
);
555 score
= zzlGetScore(sptr
);
556 if (zslValueLteMax(score
,&range
)) {
557 /* Check if score >= min. */
558 if (zslValueGteMin(score
,&range
))
563 /* Move to previous element by moving to the score of previous element.
564 * When this returns NULL, we know there also is no element. */
565 sptr
= ziplistPrev(zl
,eptr
);
567 redisAssert((eptr
= ziplistPrev(zl
,sptr
)) != NULL
);
575 unsigned char *zzlFind(unsigned char *zl
, robj
*ele
, double *score
) {
576 unsigned char *eptr
= ziplistIndex(zl
,0), *sptr
;
578 ele
= getDecodedObject(ele
);
579 while (eptr
!= NULL
) {
580 sptr
= ziplistNext(zl
,eptr
);
581 redisAssert(sptr
!= NULL
);
583 if (ziplistCompare(eptr
,ele
->ptr
,sdslen(ele
->ptr
))) {
584 /* Matching element, pull out score. */
585 if (score
!= NULL
) *score
= zzlGetScore(sptr
);
590 /* Move to next element. */
591 eptr
= ziplistNext(zl
,sptr
);
598 /* Delete (element,score) pair from ziplist. Use local copy of eptr because we
599 * don't want to modify the one given as argument. */
600 unsigned char *zzlDelete(unsigned char *zl
, unsigned char *eptr
) {
601 unsigned char *p
= eptr
;
603 /* TODO: add function to ziplist API to delete N elements from offset. */
604 zl
= ziplistDelete(zl
,&p
);
605 zl
= ziplistDelete(zl
,&p
);
609 unsigned char *zzlInsertAt(unsigned char *zl
, unsigned char *eptr
, robj
*ele
, double score
) {
615 redisAssert(ele
->encoding
== REDIS_ENCODING_RAW
);
616 scorelen
= d2string(scorebuf
,sizeof(scorebuf
),score
);
618 zl
= ziplistPush(zl
,ele
->ptr
,sdslen(ele
->ptr
),ZIPLIST_TAIL
);
619 zl
= ziplistPush(zl
,(unsigned char*)scorebuf
,scorelen
,ZIPLIST_TAIL
);
621 /* Keep offset relative to zl, as it might be re-allocated. */
623 zl
= ziplistInsert(zl
,eptr
,ele
->ptr
,sdslen(ele
->ptr
));
626 /* Insert score after the element. */
627 redisAssert((sptr
= ziplistNext(zl
,eptr
)) != NULL
);
628 zl
= ziplistInsert(zl
,sptr
,(unsigned char*)scorebuf
,scorelen
);
634 /* Insert (element,score) pair in ziplist. This function assumes the element is
635 * not yet present in the list. */
636 unsigned char *zzlInsert(unsigned char *zl
, robj
*ele
, double score
) {
637 unsigned char *eptr
= ziplistIndex(zl
,0), *sptr
;
640 ele
= getDecodedObject(ele
);
641 while (eptr
!= NULL
) {
642 sptr
= ziplistNext(zl
,eptr
);
643 redisAssert(sptr
!= NULL
);
644 s
= zzlGetScore(sptr
);
647 /* First element with score larger than score for element to be
648 * inserted. This means we should take its spot in the list to
649 * maintain ordering. */
650 zl
= zzlInsertAt(zl
,eptr
,ele
,score
);
652 } else if (s
== score
) {
653 /* Ensure lexicographical ordering for elements. */
654 if (zzlCompareElements(eptr
,ele
->ptr
,sdslen(ele
->ptr
)) > 0) {
655 zl
= zzlInsertAt(zl
,eptr
,ele
,score
);
660 /* Move to next element. */
661 eptr
= ziplistNext(zl
,sptr
);
664 /* Push on tail of list when it was not yet inserted. */
666 zl
= zzlInsertAt(zl
,NULL
,ele
,score
);
672 unsigned char *zzlDeleteRangeByScore(unsigned char *zl
, zrangespec range
, unsigned long *deleted
) {
673 unsigned char *eptr
, *sptr
;
675 unsigned long num
= 0;
677 if (deleted
!= NULL
) *deleted
= 0;
679 eptr
= zzlFirstInRange(zl
,range
);
680 if (eptr
== NULL
) return zl
;
682 /* When the tail of the ziplist is deleted, eptr will point to the sentinel
683 * byte and ziplistNext will return NULL. */
684 while ((sptr
= ziplistNext(zl
,eptr
)) != NULL
) {
685 score
= zzlGetScore(sptr
);
686 if (zslValueLteMax(score
,&range
)) {
687 /* Delete both the element and the score. */
688 zl
= ziplistDelete(zl
,&eptr
);
689 zl
= ziplistDelete(zl
,&eptr
);
692 /* No longer in range. */
697 if (deleted
!= NULL
) *deleted
= num
;
701 /* Delete all the elements with rank between start and end from the skiplist.
702 * Start and end are inclusive. Note that start and end need to be 1-based */
703 unsigned char *zzlDeleteRangeByRank(unsigned char *zl
, unsigned int start
, unsigned int end
, unsigned long *deleted
) {
704 unsigned int num
= (end
-start
)+1;
705 if (deleted
) *deleted
= num
;
706 zl
= ziplistDeleteRange(zl
,2*(start
-1),2*num
);
710 /*-----------------------------------------------------------------------------
711 * Common sorted set API
712 *----------------------------------------------------------------------------*/
714 unsigned int zsetLength(robj
*zobj
) {
716 if (zobj
->encoding
== REDIS_ENCODING_ZIPLIST
) {
717 length
= zzlLength(zobj
->ptr
);
718 } else if (zobj
->encoding
== REDIS_ENCODING_SKIPLIST
) {
719 length
= ((zset
*)zobj
->ptr
)->zsl
->length
;
721 redisPanic("Unknown sorted set encoding");
726 void zsetConvert(robj
*zobj
, int encoding
) {
728 zskiplistNode
*node
, *next
;
732 if (zobj
->encoding
== encoding
) return;
733 if (zobj
->encoding
== REDIS_ENCODING_ZIPLIST
) {
734 unsigned char *zl
= zobj
->ptr
;
735 unsigned char *eptr
, *sptr
;
740 if (encoding
!= REDIS_ENCODING_SKIPLIST
)
741 redisPanic("Unknown target encoding");
743 zs
= zmalloc(sizeof(*zs
));
744 zs
->dict
= dictCreate(&zsetDictType
,NULL
);
745 zs
->zsl
= zslCreate();
747 eptr
= ziplistIndex(zl
,0);
748 redisAssert(eptr
!= NULL
);
749 sptr
= ziplistNext(zl
,eptr
);
750 redisAssert(sptr
!= NULL
);
752 while (eptr
!= NULL
) {
753 score
= zzlGetScore(sptr
);
754 redisAssert(ziplistGet(eptr
,&vstr
,&vlen
,&vlong
));
756 ele
= createStringObjectFromLongLong(vlong
);
758 ele
= createStringObject((char*)vstr
,vlen
);
760 /* Has incremented refcount since it was just created. */
761 node
= zslInsert(zs
->zsl
,score
,ele
);
762 redisAssert(dictAdd(zs
->dict
,ele
,&node
->score
) == DICT_OK
);
763 incrRefCount(ele
); /* Added to dictionary. */
764 zzlNext(zl
,&eptr
,&sptr
);
769 zobj
->encoding
= REDIS_ENCODING_SKIPLIST
;
770 } else if (zobj
->encoding
== REDIS_ENCODING_SKIPLIST
) {
771 unsigned char *zl
= ziplistNew();
773 if (encoding
!= REDIS_ENCODING_ZIPLIST
)
774 redisPanic("Unknown target encoding");
776 /* Approach similar to zslFree(), since we want to free the skiplist at
777 * the same time as creating the ziplist. */
779 dictRelease(zs
->dict
);
780 node
= zs
->zsl
->header
->level
[0].forward
;
781 zfree(zs
->zsl
->header
);
785 ele
= getDecodedObject(node
->obj
);
786 zl
= zzlInsertAt(zl
,NULL
,ele
,node
->score
);
789 next
= node
->level
[0].forward
;
796 zobj
->encoding
= REDIS_ENCODING_ZIPLIST
;
798 redisPanic("Unknown sorted set encoding");
802 /*-----------------------------------------------------------------------------
803 * Sorted set commands
804 *----------------------------------------------------------------------------*/
806 /* This generic command implements both ZADD and ZINCRBY. */
807 void zaddGenericCommand(redisClient
*c
, int incr
) {
808 static char *nanerr
= "resulting score is not a number (NaN)";
809 robj
*key
= c
->argv
[1];
813 double score
, curscore
= 0.0;
815 if (getDoubleFromObjectOrReply(c
,c
->argv
[2],&score
,NULL
) != REDIS_OK
)
818 zobj
= lookupKeyWrite(c
->db
,key
);
820 if (server
.zset_max_ziplist_entries
== 0 ||
821 server
.zset_max_ziplist_value
< sdslen(c
->argv
[3]->ptr
))
823 zobj
= createZsetObject();
825 zobj
= createZsetZiplistObject();
827 dbAdd(c
->db
,key
,zobj
);
829 if (zobj
->type
!= REDIS_ZSET
) {
830 addReply(c
,shared
.wrongtypeerr
);
835 if (zobj
->encoding
== REDIS_ENCODING_ZIPLIST
) {
838 /* Prefer non-encoded element when dealing with ziplists. */
840 if ((eptr
= zzlFind(zobj
->ptr
,ele
,&curscore
)) != NULL
) {
844 addReplyError(c
,nanerr
);
845 /* Don't need to check if the sorted set is empty, because
846 * we know it has at least one element. */
851 /* Remove and re-insert when score changed. */
852 if (score
!= curscore
) {
853 zobj
->ptr
= zzlDelete(zobj
->ptr
,eptr
);
854 zobj
->ptr
= zzlInsert(zobj
->ptr
,ele
,score
);
856 signalModifiedKey(c
->db
,key
);
860 if (incr
) /* ZINCRBY */
861 addReplyDouble(c
,score
);
863 addReply(c
,shared
.czero
);
865 /* Optimize: check if the element is too large or the list becomes
866 * too long *before* executing zzlInsert. */
867 zobj
->ptr
= zzlInsert(zobj
->ptr
,ele
,score
);
868 if (zzlLength(zobj
->ptr
) > server
.zset_max_ziplist_entries
)
869 zsetConvert(zobj
,REDIS_ENCODING_SKIPLIST
);
870 if (sdslen(ele
->ptr
) > server
.zset_max_ziplist_value
)
871 zsetConvert(zobj
,REDIS_ENCODING_SKIPLIST
);
873 signalModifiedKey(c
->db
,key
);
876 if (incr
) /* ZINCRBY */
877 addReplyDouble(c
,score
);
879 addReply(c
,shared
.cone
);
881 } else if (zobj
->encoding
== REDIS_ENCODING_SKIPLIST
) {
882 zset
*zs
= zobj
->ptr
;
883 zskiplistNode
*znode
;
886 ele
= c
->argv
[3] = tryObjectEncoding(c
->argv
[3]);
887 de
= dictFind(zs
->dict
,ele
);
889 curobj
= dictGetEntryKey(de
);
890 curscore
= *(double*)dictGetEntryVal(de
);
895 addReplyError(c
,nanerr
);
896 /* Don't need to check if the sorted set is empty, because
897 * we know it has at least one element. */
902 /* Remove and re-insert when score changed. We can safely delete
903 * the key object from the skiplist, since the dictionary still has
904 * a reference to it. */
905 if (score
!= curscore
) {
906 redisAssert(zslDelete(zs
->zsl
,curscore
,curobj
));
907 znode
= zslInsert(zs
->zsl
,score
,curobj
);
908 incrRefCount(curobj
); /* Re-inserted in skiplist. */
909 dictGetEntryVal(de
) = &znode
->score
; /* Update score ptr. */
911 signalModifiedKey(c
->db
,key
);
915 if (incr
) /* ZINCRBY */
916 addReplyDouble(c
,score
);
918 addReply(c
,shared
.czero
);
920 znode
= zslInsert(zs
->zsl
,score
,ele
);
921 incrRefCount(ele
); /* Inserted in skiplist. */
922 redisAssert(dictAdd(zs
->dict
,ele
,&znode
->score
) == DICT_OK
);
923 incrRefCount(ele
); /* Added to dictionary. */
925 signalModifiedKey(c
->db
,key
);
928 if (incr
) /* ZINCRBY */
929 addReplyDouble(c
,score
);
931 addReply(c
,shared
.cone
);
934 redisPanic("Unknown sorted set encoding");
938 void zaddCommand(redisClient
*c
) {
939 zaddGenericCommand(c
,0);
942 void zincrbyCommand(redisClient
*c
) {
943 zaddGenericCommand(c
,1);
946 void zremCommand(redisClient
*c
) {
947 robj
*key
= c
->argv
[1];
948 robj
*ele
= c
->argv
[2];
951 if ((zobj
= lookupKeyWriteOrReply(c
,key
,shared
.czero
)) == NULL
||
952 checkType(c
,zobj
,REDIS_ZSET
)) return;
954 if (zobj
->encoding
== REDIS_ENCODING_ZIPLIST
) {
957 if ((eptr
= zzlFind(zobj
->ptr
,ele
,NULL
)) != NULL
) {
958 zobj
->ptr
= zzlDelete(zobj
->ptr
,eptr
);
959 if (zzlLength(zobj
->ptr
) == 0) dbDelete(c
->db
,key
);
961 addReply(c
,shared
.czero
);
964 } else if (zobj
->encoding
== REDIS_ENCODING_SKIPLIST
) {
965 zset
*zs
= zobj
->ptr
;
969 de
= dictFind(zs
->dict
,ele
);
971 /* Delete from the skiplist */
972 score
= *(double*)dictGetEntryVal(de
);
973 redisAssert(zslDelete(zs
->zsl
,score
,ele
));
975 /* Delete from the hash table */
976 dictDelete(zs
->dict
,ele
);
977 if (htNeedsResize(zs
->dict
)) dictResize(zs
->dict
);
978 if (dictSize(zs
->dict
) == 0) dbDelete(c
->db
,key
);
980 addReply(c
,shared
.czero
);
984 redisPanic("Unknown sorted set encoding");
987 signalModifiedKey(c
->db
,key
);
989 addReply(c
,shared
.cone
);
992 void zremrangebyscoreCommand(redisClient
*c
) {
993 robj
*key
= c
->argv
[1];
996 unsigned long deleted
;
998 /* Parse the range arguments. */
999 if (zslParseRange(c
->argv
[2],c
->argv
[3],&range
) != REDIS_OK
) {
1000 addReplyError(c
,"min or max is not a double");
1004 if ((zobj
= lookupKeyWriteOrReply(c
,key
,shared
.czero
)) == NULL
||
1005 checkType(c
,zobj
,REDIS_ZSET
)) return;
1007 if (zobj
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1008 zobj
->ptr
= zzlDeleteRangeByScore(zobj
->ptr
,range
,&deleted
);
1009 if (zzlLength(zobj
->ptr
) == 0) dbDelete(c
->db
,key
);
1010 } else if (zobj
->encoding
== REDIS_ENCODING_SKIPLIST
) {
1011 zset
*zs
= zobj
->ptr
;
1012 deleted
= zslDeleteRangeByScore(zs
->zsl
,range
,zs
->dict
);
1013 if (htNeedsResize(zs
->dict
)) dictResize(zs
->dict
);
1014 if (dictSize(zs
->dict
) == 0) dbDelete(c
->db
,key
);
1016 redisPanic("Unknown sorted set encoding");
1019 if (deleted
) signalModifiedKey(c
->db
,key
);
1020 server
.dirty
+= deleted
;
1021 addReplyLongLong(c
,deleted
);
1024 void zremrangebyrankCommand(redisClient
*c
) {
1025 robj
*key
= c
->argv
[1];
1030 unsigned long deleted
;
1032 if ((getLongFromObjectOrReply(c
, c
->argv
[2], &start
, NULL
) != REDIS_OK
) ||
1033 (getLongFromObjectOrReply(c
, c
->argv
[3], &end
, NULL
) != REDIS_OK
)) return;
1035 if ((zobj
= lookupKeyWriteOrReply(c
,key
,shared
.czero
)) == NULL
||
1036 checkType(c
,zobj
,REDIS_ZSET
)) return;
1038 /* Sanitize indexes. */
1039 llen
= zsetLength(zobj
);
1040 if (start
< 0) start
= llen
+start
;
1041 if (end
< 0) end
= llen
+end
;
1042 if (start
< 0) start
= 0;
1044 /* Invariant: start >= 0, so this test will be true when end < 0.
1045 * The range is empty when start > end or start >= length. */
1046 if (start
> end
|| start
>= llen
) {
1047 addReply(c
,shared
.czero
);
1050 if (end
>= llen
) end
= llen
-1;
1052 if (zobj
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1053 /* Correct for 1-based rank. */
1054 zobj
->ptr
= zzlDeleteRangeByRank(zobj
->ptr
,start
+1,end
+1,&deleted
);
1055 if (zzlLength(zobj
->ptr
) == 0) dbDelete(c
->db
,key
);
1056 } else if (zobj
->encoding
== REDIS_ENCODING_SKIPLIST
) {
1057 zset
*zs
= zobj
->ptr
;
1059 /* Correct for 1-based rank. */
1060 deleted
= zslDeleteRangeByRank(zs
->zsl
,start
+1,end
+1,zs
->dict
);
1061 if (htNeedsResize(zs
->dict
)) dictResize(zs
->dict
);
1062 if (dictSize(zs
->dict
) == 0) dbDelete(c
->db
,key
);
1064 redisPanic("Unknown sorted set encoding");
1067 if (deleted
) signalModifiedKey(c
->db
,key
);
1068 server
.dirty
+= deleted
;
1069 addReplyLongLong(c
,deleted
);
1074 int type
; /* Set, sorted set */
1079 /* Set iterators. */
1092 /* Sorted set iterators. */
1096 unsigned char *eptr
, *sptr
;
1100 zskiplistNode
*node
;
1107 /* Use dirty flags for pointers that need to be cleaned up in the next
1108 * iteration over the zsetopval. The dirty flag for the long long value is
1109 * special, since long long values don't need cleanup. Instead, it means that
1110 * we already checked that "ell" holds a long long, or tried to convert another
1111 * representation into a long long value. When this was successful,
1112 * OPVAL_VALID_LL is set as well. */
1113 #define OPVAL_DIRTY_ROBJ 1
1114 #define OPVAL_DIRTY_LL 2
1115 #define OPVAL_VALID_LL 4
1117 /* Store value retrieved from the iterator. */
1120 unsigned char _buf
[32]; /* Private buffer. */
1122 unsigned char *estr
;
1128 typedef union _iterset iterset
;
1129 typedef union _iterzset iterzset
;
1131 void zuiInitIterator(zsetopsrc
*op
) {
1132 if (op
->subject
== NULL
)
1135 if (op
->type
== REDIS_SET
) {
1136 iterset
*it
= &op
->iter
.set
;
1137 if (op
->encoding
== REDIS_ENCODING_INTSET
) {
1138 it
->is
.is
= op
->subject
->ptr
;
1140 } else if (op
->encoding
== REDIS_ENCODING_HT
) {
1141 it
->ht
.dict
= op
->subject
->ptr
;
1142 it
->ht
.di
= dictGetIterator(op
->subject
->ptr
);
1143 it
->ht
.de
= dictNext(it
->ht
.di
);
1145 redisPanic("Unknown set encoding");
1147 } else if (op
->type
== REDIS_ZSET
) {
1148 iterzset
*it
= &op
->iter
.zset
;
1149 if (op
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1150 it
->zl
.zl
= op
->subject
->ptr
;
1151 it
->zl
.eptr
= ziplistIndex(it
->zl
.zl
,0);
1152 if (it
->zl
.eptr
!= NULL
) {
1153 it
->zl
.sptr
= ziplistNext(it
->zl
.zl
,it
->zl
.eptr
);
1154 redisAssert(it
->zl
.sptr
!= NULL
);
1156 } else if (op
->encoding
== REDIS_ENCODING_SKIPLIST
) {
1157 it
->sl
.zs
= op
->subject
->ptr
;
1158 it
->sl
.node
= it
->sl
.zs
->zsl
->header
->level
[0].forward
;
1160 redisPanic("Unknown sorted set encoding");
1163 redisPanic("Unsupported type");
1167 void zuiClearIterator(zsetopsrc
*op
) {
1168 if (op
->subject
== NULL
)
1171 if (op
->type
== REDIS_SET
) {
1172 iterset
*it
= &op
->iter
.set
;
1173 if (op
->encoding
== REDIS_ENCODING_INTSET
) {
1174 REDIS_NOTUSED(it
); /* skip */
1175 } else if (op
->encoding
== REDIS_ENCODING_HT
) {
1176 dictReleaseIterator(it
->ht
.di
);
1178 redisPanic("Unknown set encoding");
1180 } else if (op
->type
== REDIS_ZSET
) {
1181 iterzset
*it
= &op
->iter
.zset
;
1182 if (op
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1183 REDIS_NOTUSED(it
); /* skip */
1184 } else if (op
->encoding
== REDIS_ENCODING_SKIPLIST
) {
1185 REDIS_NOTUSED(it
); /* skip */
1187 redisPanic("Unknown sorted set encoding");
1190 redisPanic("Unsupported type");
1194 int zuiLength(zsetopsrc
*op
) {
1195 if (op
->subject
== NULL
)
1198 if (op
->type
== REDIS_SET
) {
1199 iterset
*it
= &op
->iter
.set
;
1200 if (op
->encoding
== REDIS_ENCODING_INTSET
) {
1201 return intsetLen(it
->is
.is
);
1202 } else if (op
->encoding
== REDIS_ENCODING_HT
) {
1203 return dictSize(it
->ht
.dict
);
1205 redisPanic("Unknown set encoding");
1207 } else if (op
->type
== REDIS_ZSET
) {
1208 iterzset
*it
= &op
->iter
.zset
;
1209 if (op
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1210 return zzlLength(it
->zl
.zl
);
1211 } else if (op
->encoding
== REDIS_ENCODING_SKIPLIST
) {
1212 return it
->sl
.zs
->zsl
->length
;
1214 redisPanic("Unknown sorted set encoding");
1217 redisPanic("Unsupported type");
1221 /* Check if the current value is valid. If so, store it in the passed structure
1222 * and move to the next element. If not valid, this means we have reached the
1223 * end of the structure and can abort. */
1224 int zuiNext(zsetopsrc
*op
, zsetopval
*val
) {
1225 if (op
->subject
== NULL
)
1228 if (val
->flags
& OPVAL_DIRTY_ROBJ
)
1229 decrRefCount(val
->ele
);
1231 bzero(val
,sizeof(zsetopval
));
1233 if (op
->type
== REDIS_SET
) {
1234 iterset
*it
= &op
->iter
.set
;
1235 if (op
->encoding
== REDIS_ENCODING_INTSET
) {
1236 if (!intsetGet(it
->is
.is
,it
->is
.ii
,(int64_t*)&val
->ell
))
1240 /* Move to next element. */
1242 } else if (op
->encoding
== REDIS_ENCODING_HT
) {
1243 if (it
->ht
.de
== NULL
)
1245 val
->ele
= dictGetEntryKey(it
->ht
.de
);
1248 /* Move to next element. */
1249 it
->ht
.de
= dictNext(it
->ht
.di
);
1251 redisPanic("Unknown set encoding");
1253 } else if (op
->type
== REDIS_ZSET
) {
1254 iterzset
*it
= &op
->iter
.zset
;
1255 if (op
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1256 /* No need to check both, but better be explicit. */
1257 if (it
->zl
.eptr
== NULL
|| it
->zl
.sptr
== NULL
)
1259 redisAssert(ziplistGet(it
->zl
.eptr
,&val
->estr
,&val
->elen
,&val
->ell
));
1260 val
->score
= zzlGetScore(it
->zl
.sptr
);
1262 /* Move to next element. */
1263 zzlNext(it
->zl
.zl
,&it
->zl
.eptr
,&it
->zl
.sptr
);
1264 } else if (op
->encoding
== REDIS_ENCODING_SKIPLIST
) {
1265 if (it
->sl
.node
== NULL
)
1267 val
->ele
= it
->sl
.node
->obj
;
1268 val
->score
= it
->sl
.node
->score
;
1270 /* Move to next element. */
1271 it
->sl
.node
= it
->sl
.node
->level
[0].forward
;
1273 redisPanic("Unknown sorted set encoding");
1276 redisPanic("Unsupported type");
1281 int zuiLongLongFromValue(zsetopval
*val
) {
1282 if (!(val
->flags
& OPVAL_DIRTY_LL
)) {
1283 val
->flags
|= OPVAL_DIRTY_LL
;
1285 if (val
->ele
!= NULL
) {
1286 if (val
->ele
->encoding
== REDIS_ENCODING_INT
) {
1287 val
->ell
= (long)val
->ele
->ptr
;
1288 val
->flags
|= OPVAL_VALID_LL
;
1289 } else if (val
->ele
->encoding
== REDIS_ENCODING_RAW
) {
1290 if (string2ll(val
->ele
->ptr
,sdslen(val
->ele
->ptr
),&val
->ell
))
1291 val
->flags
|= OPVAL_VALID_LL
;
1293 redisPanic("Unsupported element encoding");
1295 } else if (val
->estr
!= NULL
) {
1296 if (string2ll((char*)val
->estr
,val
->elen
,&val
->ell
))
1297 val
->flags
|= OPVAL_VALID_LL
;
1299 /* The long long was already set, flag as valid. */
1300 val
->flags
|= OPVAL_VALID_LL
;
1303 return val
->flags
& OPVAL_VALID_LL
;
1306 robj
*zuiObjectFromValue(zsetopval
*val
) {
1307 if (val
->ele
== NULL
) {
1308 if (val
->estr
!= NULL
) {
1309 val
->ele
= createStringObject((char*)val
->estr
,val
->elen
);
1311 val
->ele
= createStringObjectFromLongLong(val
->ell
);
1313 val
->flags
|= OPVAL_DIRTY_ROBJ
;
1318 int zuiBufferFromValue(zsetopval
*val
) {
1319 if (val
->estr
== NULL
) {
1320 if (val
->ele
!= NULL
) {
1321 if (val
->ele
->encoding
== REDIS_ENCODING_INT
) {
1322 val
->elen
= ll2string((char*)val
->_buf
,sizeof(val
->_buf
),(long)val
->ele
->ptr
);
1323 val
->estr
= val
->_buf
;
1324 } else if (val
->ele
->encoding
== REDIS_ENCODING_RAW
) {
1325 val
->elen
= sdslen(val
->ele
->ptr
);
1326 val
->estr
= val
->ele
->ptr
;
1328 redisPanic("Unsupported element encoding");
1331 val
->elen
= ll2string((char*)val
->_buf
,sizeof(val
->_buf
),val
->ell
);
1332 val
->estr
= val
->_buf
;
1338 /* Find value pointed to by val in the source pointer to by op. When found,
1339 * return 1 and store its score in target. Return 0 otherwise. */
1340 int zuiFind(zsetopsrc
*op
, zsetopval
*val
, double *score
) {
1341 if (op
->subject
== NULL
)
1344 if (op
->type
== REDIS_SET
) {
1345 iterset
*it
= &op
->iter
.set
;
1347 if (op
->encoding
== REDIS_ENCODING_INTSET
) {
1348 if (zuiLongLongFromValue(val
) && intsetFind(it
->is
.is
,val
->ell
)) {
1354 } else if (op
->encoding
== REDIS_ENCODING_HT
) {
1355 zuiObjectFromValue(val
);
1356 if (dictFind(it
->ht
.dict
,val
->ele
) != NULL
) {
1363 redisPanic("Unknown set encoding");
1365 } else if (op
->type
== REDIS_ZSET
) {
1366 iterzset
*it
= &op
->iter
.zset
;
1367 zuiObjectFromValue(val
);
1369 if (op
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1370 if (zzlFind(it
->zl
.zl
,val
->ele
,score
) != NULL
) {
1371 /* Score is already set by zzlFind. */
1376 } else if (op
->encoding
== REDIS_ENCODING_SKIPLIST
) {
1378 if ((de
= dictFind(it
->sl
.zs
->dict
,val
->ele
)) != NULL
) {
1379 *score
= *(double*)dictGetEntryVal(de
);
1385 redisPanic("Unknown sorted set encoding");
1388 redisPanic("Unsupported type");
1392 int zuiCompareByCardinality(const void *s1
, const void *s2
) {
1393 return zuiLength((zsetopsrc
*)s1
) - zuiLength((zsetopsrc
*)s2
);
1396 #define REDIS_AGGR_SUM 1
1397 #define REDIS_AGGR_MIN 2
1398 #define REDIS_AGGR_MAX 3
1399 #define zunionInterDictValue(_e) (dictGetEntryVal(_e) == NULL ? 1.0 : *(double*)dictGetEntryVal(_e))
1401 inline static void zunionInterAggregate(double *target
, double val
, int aggregate
) {
1402 if (aggregate
== REDIS_AGGR_SUM
) {
1403 *target
= *target
+ val
;
1404 /* The result of adding two doubles is NaN when one variable
1405 * is +inf and the other is -inf. When these numbers are added,
1406 * we maintain the convention of the result being 0.0. */
1407 if (isnan(*target
)) *target
= 0.0;
1408 } else if (aggregate
== REDIS_AGGR_MIN
) {
1409 *target
= val
< *target
? val
: *target
;
1410 } else if (aggregate
== REDIS_AGGR_MAX
) {
1411 *target
= val
> *target
? val
: *target
;
1414 redisPanic("Unknown ZUNION/INTER aggregate type");
1418 void zunionInterGenericCommand(redisClient
*c
, robj
*dstkey
, int op
) {
1420 int aggregate
= REDIS_AGGR_SUM
;
1424 unsigned int maxelelen
= 0;
1427 zskiplistNode
*znode
;
1430 /* expect setnum input keys to be given */
1431 setnum
= atoi(c
->argv
[2]->ptr
);
1434 "at least 1 input key is needed for ZUNIONSTORE/ZINTERSTORE");
1438 /* test if the expected number of keys would overflow */
1439 if (3+setnum
> c
->argc
) {
1440 addReply(c
,shared
.syntaxerr
);
1444 /* read keys to be used for input */
1445 src
= zcalloc(sizeof(zsetopsrc
) * setnum
);
1446 for (i
= 0, j
= 3; i
< setnum
; i
++, j
++) {
1447 robj
*obj
= lookupKeyWrite(c
->db
,c
->argv
[j
]);
1449 if (obj
->type
!= REDIS_ZSET
&& obj
->type
!= REDIS_SET
) {
1451 addReply(c
,shared
.wrongtypeerr
);
1455 src
[i
].subject
= obj
;
1456 src
[i
].type
= obj
->type
;
1457 src
[i
].encoding
= obj
->encoding
;
1459 src
[i
].subject
= NULL
;
1462 /* Default all weights to 1. */
1463 src
[i
].weight
= 1.0;
1466 /* parse optional extra arguments */
1468 int remaining
= c
->argc
- j
;
1471 if (remaining
>= (setnum
+ 1) && !strcasecmp(c
->argv
[j
]->ptr
,"weights")) {
1473 for (i
= 0; i
< setnum
; i
++, j
++, remaining
--) {
1474 if (getDoubleFromObjectOrReply(c
,c
->argv
[j
],&src
[i
].weight
,
1475 "weight value is not a double") != REDIS_OK
)
1481 } else if (remaining
>= 2 && !strcasecmp(c
->argv
[j
]->ptr
,"aggregate")) {
1483 if (!strcasecmp(c
->argv
[j
]->ptr
,"sum")) {
1484 aggregate
= REDIS_AGGR_SUM
;
1485 } else if (!strcasecmp(c
->argv
[j
]->ptr
,"min")) {
1486 aggregate
= REDIS_AGGR_MIN
;
1487 } else if (!strcasecmp(c
->argv
[j
]->ptr
,"max")) {
1488 aggregate
= REDIS_AGGR_MAX
;
1491 addReply(c
,shared
.syntaxerr
);
1497 addReply(c
,shared
.syntaxerr
);
1503 for (i
= 0; i
< setnum
; i
++)
1504 zuiInitIterator(&src
[i
]);
1506 /* sort sets from the smallest to largest, this will improve our
1507 * algorithm's performance */
1508 qsort(src
,setnum
,sizeof(zsetopsrc
),zuiCompareByCardinality
);
1510 dstobj
= createZsetObject();
1511 dstzset
= dstobj
->ptr
;
1512 memset(&zval
, 0, sizeof(zval
));
1514 if (op
== REDIS_OP_INTER
) {
1515 /* Skip everything if the smallest input is empty. */
1516 if (zuiLength(&src
[0]) > 0) {
1517 /* Precondition: as src[0] is non-empty and the inputs are ordered
1518 * by size, all src[i > 0] are non-empty too. */
1519 while (zuiNext(&src
[0],&zval
)) {
1520 double score
, value
;
1522 score
= src
[0].weight
* zval
.score
;
1523 for (j
= 1; j
< setnum
; j
++) {
1524 /* It is not safe to access the hash we zset we are
1525 * iterating, so explicitly check for equal object. */
1526 if (src
[j
].subject
== src
[0].subject
) {
1527 value
= zval
.score
*src
[j
].weight
;
1528 zunionInterAggregate(&score
,value
,aggregate
);
1529 } else if (zuiFind(&src
[j
],&zval
,&value
)) {
1530 value
*= src
[j
].weight
;
1531 zunionInterAggregate(&score
,value
,aggregate
);
1537 /* Only continue when present in every input. */
1539 tmp
= zuiObjectFromValue(&zval
);
1540 znode
= zslInsert(dstzset
->zsl
,score
,tmp
);
1541 incrRefCount(tmp
); /* added to skiplist */
1542 dictAdd(dstzset
->dict
,tmp
,&znode
->score
);
1543 incrRefCount(tmp
); /* added to dictionary */
1545 if (tmp
->encoding
== REDIS_ENCODING_RAW
)
1546 if (sdslen(tmp
->ptr
) > maxelelen
)
1547 maxelelen
= sdslen(tmp
->ptr
);
1551 } else if (op
== REDIS_OP_UNION
) {
1552 for (i
= 0; i
< setnum
; i
++) {
1553 if (zuiLength(&src
[i
]) == 0)
1556 while (zuiNext(&src
[i
],&zval
)) {
1557 double score
, value
;
1559 /* Skip key when already processed */
1560 if (dictFind(dstzset
->dict
,zuiObjectFromValue(&zval
)) != NULL
)
1563 /* Initialize score */
1564 score
= src
[i
].weight
* zval
.score
;
1566 /* Because the inputs are sorted by size, it's only possible
1567 * for sets at larger indices to hold this element. */
1568 for (j
= (i
+1); j
< setnum
; j
++) {
1569 /* It is not safe to access the hash we zset we are
1570 * iterating, so explicitly check for equal object. */
1571 if(src
[j
].subject
== src
[i
].subject
) {
1572 value
= zval
.score
*src
[j
].weight
;
1573 zunionInterAggregate(&score
,value
,aggregate
);
1574 } else if (zuiFind(&src
[j
],&zval
,&value
)) {
1575 value
*= src
[j
].weight
;
1576 zunionInterAggregate(&score
,value
,aggregate
);
1580 tmp
= zuiObjectFromValue(&zval
);
1581 znode
= zslInsert(dstzset
->zsl
,score
,tmp
);
1582 incrRefCount(zval
.ele
); /* added to skiplist */
1583 dictAdd(dstzset
->dict
,tmp
,&znode
->score
);
1584 incrRefCount(zval
.ele
); /* added to dictionary */
1586 if (tmp
->encoding
== REDIS_ENCODING_RAW
)
1587 if (sdslen(tmp
->ptr
) > maxelelen
)
1588 maxelelen
= sdslen(tmp
->ptr
);
1592 redisPanic("Unknown operator");
1595 for (i
= 0; i
< setnum
; i
++)
1596 zuiClearIterator(&src
[i
]);
1598 if (dbDelete(c
->db
,dstkey
)) {
1599 signalModifiedKey(c
->db
,dstkey
);
1603 if (dstzset
->zsl
->length
) {
1604 /* Convert to ziplist when in limits. */
1605 if (dstzset
->zsl
->length
<= server
.zset_max_ziplist_entries
&&
1606 maxelelen
<= server
.zset_max_ziplist_value
)
1607 zsetConvert(dstobj
,REDIS_ENCODING_ZIPLIST
);
1609 dbAdd(c
->db
,dstkey
,dstobj
);
1610 addReplyLongLong(c
,zsetLength(dstobj
));
1611 if (!touched
) signalModifiedKey(c
->db
,dstkey
);
1614 decrRefCount(dstobj
);
1615 addReply(c
,shared
.czero
);
1620 void zunionstoreCommand(redisClient
*c
) {
1621 zunionInterGenericCommand(c
,c
->argv
[1], REDIS_OP_UNION
);
1624 void zinterstoreCommand(redisClient
*c
) {
1625 zunionInterGenericCommand(c
,c
->argv
[1], REDIS_OP_INTER
);
1628 void zrangeGenericCommand(redisClient
*c
, int reverse
) {
1629 robj
*key
= c
->argv
[1];
1637 if ((getLongFromObjectOrReply(c
, c
->argv
[2], &start
, NULL
) != REDIS_OK
) ||
1638 (getLongFromObjectOrReply(c
, c
->argv
[3], &end
, NULL
) != REDIS_OK
)) return;
1640 if (c
->argc
== 5 && !strcasecmp(c
->argv
[4]->ptr
,"withscores")) {
1642 } else if (c
->argc
>= 5) {
1643 addReply(c
,shared
.syntaxerr
);
1647 if ((zobj
= lookupKeyReadOrReply(c
,key
,shared
.emptymultibulk
)) == NULL
1648 || checkType(c
,zobj
,REDIS_ZSET
)) return;
1650 /* Sanitize indexes. */
1651 llen
= zsetLength(zobj
);
1652 if (start
< 0) start
= llen
+start
;
1653 if (end
< 0) end
= llen
+end
;
1654 if (start
< 0) start
= 0;
1656 /* Invariant: start >= 0, so this test will be true when end < 0.
1657 * The range is empty when start > end or start >= length. */
1658 if (start
> end
|| start
>= llen
) {
1659 addReply(c
,shared
.emptymultibulk
);
1662 if (end
>= llen
) end
= llen
-1;
1663 rangelen
= (end
-start
)+1;
1665 /* Return the result in form of a multi-bulk reply */
1666 addReplyMultiBulkLen(c
, withscores
? (rangelen
*2) : rangelen
);
1668 if (zobj
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1669 unsigned char *zl
= zobj
->ptr
;
1670 unsigned char *eptr
, *sptr
;
1671 unsigned char *vstr
;
1676 eptr
= ziplistIndex(zl
,-2-(2*start
));
1678 eptr
= ziplistIndex(zl
,2*start
);
1680 redisAssert(eptr
!= NULL
);
1681 sptr
= ziplistNext(zl
,eptr
);
1683 while (rangelen
--) {
1684 redisAssert(eptr
!= NULL
&& sptr
!= NULL
);
1685 redisAssert(ziplistGet(eptr
,&vstr
,&vlen
,&vlong
));
1687 addReplyBulkLongLong(c
,vlong
);
1689 addReplyBulkCBuffer(c
,vstr
,vlen
);
1692 addReplyDouble(c
,zzlGetScore(sptr
));
1695 zzlPrev(zl
,&eptr
,&sptr
);
1697 zzlNext(zl
,&eptr
,&sptr
);
1700 } else if (zobj
->encoding
== REDIS_ENCODING_SKIPLIST
) {
1701 zset
*zs
= zobj
->ptr
;
1702 zskiplist
*zsl
= zs
->zsl
;
1706 /* Check if starting point is trivial, before doing log(N) lookup. */
1710 ln
= zslGetElementByRank(zsl
,llen
-start
);
1712 ln
= zsl
->header
->level
[0].forward
;
1714 ln
= zslGetElementByRank(zsl
,start
+1);
1718 redisAssert(ln
!= NULL
);
1720 addReplyBulk(c
,ele
);
1722 addReplyDouble(c
,ln
->score
);
1723 ln
= reverse
? ln
->backward
: ln
->level
[0].forward
;
1726 redisPanic("Unknown sorted set encoding");
1730 void zrangeCommand(redisClient
*c
) {
1731 zrangeGenericCommand(c
,0);
1734 void zrevrangeCommand(redisClient
*c
) {
1735 zrangeGenericCommand(c
,1);
1738 /* This command implements ZRANGEBYSCORE, ZREVRANGEBYSCORE and ZCOUNT.
1739 * If "justcount", only the number of elements in the range is returned. */
1740 void genericZrangebyscoreCommand(redisClient
*c
, int reverse
, int justcount
) {
1742 robj
*key
= c
->argv
[1];
1743 robj
*emptyreply
, *zobj
;
1744 int offset
= 0, limit
= -1;
1746 unsigned long rangelen
= 0;
1747 void *replylen
= NULL
;
1750 /* Parse the range arguments. */
1752 /* Range is given as [max,min] */
1753 maxidx
= 2; minidx
= 3;
1755 /* Range is given as [min,max] */
1756 minidx
= 2; maxidx
= 3;
1759 if (zslParseRange(c
->argv
[minidx
],c
->argv
[maxidx
],&range
) != REDIS_OK
) {
1760 addReplyError(c
,"min or max is not a double");
1764 /* Parse optional extra arguments. Note that ZCOUNT will exactly have
1765 * 4 arguments, so we'll never enter the following code path. */
1767 int remaining
= c
->argc
- 4;
1771 if (remaining
>= 1 && !strcasecmp(c
->argv
[pos
]->ptr
,"withscores")) {
1774 } else if (remaining
>= 3 && !strcasecmp(c
->argv
[pos
]->ptr
,"limit")) {
1775 offset
= atoi(c
->argv
[pos
+1]->ptr
);
1776 limit
= atoi(c
->argv
[pos
+2]->ptr
);
1777 pos
+= 3; remaining
-= 3;
1779 addReply(c
,shared
.syntaxerr
);
1785 /* Ok, lookup the key and get the range */
1786 emptyreply
= justcount
? shared
.czero
: shared
.emptymultibulk
;
1787 if ((zobj
= lookupKeyReadOrReply(c
,key
,emptyreply
)) == NULL
||
1788 checkType(c
,zobj
,REDIS_ZSET
)) return;
1790 if (zobj
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1791 unsigned char *zl
= zobj
->ptr
;
1792 unsigned char *eptr
, *sptr
;
1793 unsigned char *vstr
;
1798 /* If reversed, get the last node in range as starting point. */
1800 eptr
= zzlLastInRange(zl
,range
);
1802 eptr
= zzlFirstInRange(zl
,range
);
1804 /* No "first" element in the specified interval. */
1806 addReply(c
,emptyreply
);
1810 /* Get score pointer for the first element. */
1811 redisAssert(eptr
!= NULL
);
1812 sptr
= ziplistNext(zl
,eptr
);
1814 /* We don't know in advance how many matching elements there are in the
1815 * list, so we push this object that will represent the multi-bulk
1816 * length in the output buffer, and will "fix" it later */
1818 replylen
= addDeferredMultiBulkLength(c
);
1820 /* If there is an offset, just traverse the number of elements without
1821 * checking the score because that is done in the next loop. */
1822 while (eptr
&& offset
--)
1824 zzlPrev(zl
,&eptr
,&sptr
);
1826 zzlNext(zl
,&eptr
,&sptr
);
1828 while (eptr
&& limit
--) {
1829 score
= zzlGetScore(sptr
);
1831 /* Abort when the node is no longer in range. */
1833 if (!zslValueGteMin(score
,&range
)) break;
1835 if (!zslValueLteMax(score
,&range
)) break;
1841 redisAssert(ziplistGet(eptr
,&vstr
,&vlen
,&vlong
));
1843 addReplyBulkLongLong(c
,vlong
);
1845 addReplyBulkCBuffer(c
,vstr
,vlen
);
1848 addReplyDouble(c
,score
);
1851 /* Move to next node */
1853 zzlPrev(zl
,&eptr
,&sptr
);
1855 zzlNext(zl
,&eptr
,&sptr
);
1857 } else if (zobj
->encoding
== REDIS_ENCODING_SKIPLIST
) {
1858 zset
*zs
= zobj
->ptr
;
1859 zskiplist
*zsl
= zs
->zsl
;
1862 /* If reversed, get the last node in range as starting point. */
1864 ln
= zslLastInRange(zsl
,range
);
1866 ln
= zslFirstInRange(zsl
,range
);
1868 /* No "first" element in the specified interval. */
1870 addReply(c
,emptyreply
);
1874 /* We don't know in advance how many matching elements there are in the
1875 * list, so we push this object that will represent the multi-bulk
1876 * length in the output buffer, and will "fix" it later */
1878 replylen
= addDeferredMultiBulkLength(c
);
1880 /* If there is an offset, just traverse the number of elements without
1881 * checking the score because that is done in the next loop. */
1882 while (ln
&& offset
--)
1883 ln
= reverse
? ln
->backward
: ln
->level
[0].forward
;
1885 while (ln
&& limit
--) {
1886 /* Abort when the node is no longer in range. */
1888 if (!zslValueGteMin(ln
->score
,&range
)) break;
1890 if (!zslValueLteMax(ln
->score
,&range
)) break;
1896 addReplyBulk(c
,ln
->obj
);
1898 addReplyDouble(c
,ln
->score
);
1901 /* Move to next node */
1902 ln
= reverse
? ln
->backward
: ln
->level
[0].forward
;
1905 redisPanic("Unknown sorted set encoding");
1909 addReplyLongLong(c
,(long)rangelen
);
1911 if (withscores
) rangelen
*= 2;
1912 setDeferredMultiBulkLength(c
,replylen
,rangelen
);
1916 void zrangebyscoreCommand(redisClient
*c
) {
1917 genericZrangebyscoreCommand(c
,0,0);
1920 void zrevrangebyscoreCommand(redisClient
*c
) {
1921 genericZrangebyscoreCommand(c
,1,0);
1924 void zcountCommand(redisClient
*c
) {
1925 genericZrangebyscoreCommand(c
,0,1);
1928 void zcardCommand(redisClient
*c
) {
1929 robj
*key
= c
->argv
[1];
1932 if ((zobj
= lookupKeyReadOrReply(c
,key
,shared
.czero
)) == NULL
||
1933 checkType(c
,zobj
,REDIS_ZSET
)) return;
1935 addReplyLongLong(c
,zsetLength(zobj
));
1938 void zscoreCommand(redisClient
*c
) {
1939 robj
*key
= c
->argv
[1];
1943 if ((zobj
= lookupKeyReadOrReply(c
,key
,shared
.nullbulk
)) == NULL
||
1944 checkType(c
,zobj
,REDIS_ZSET
)) return;
1946 if (zobj
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1947 if (zzlFind(zobj
->ptr
,c
->argv
[2],&score
) != NULL
)
1948 addReplyDouble(c
,score
);
1950 addReply(c
,shared
.nullbulk
);
1951 } else if (zobj
->encoding
== REDIS_ENCODING_SKIPLIST
) {
1952 zset
*zs
= zobj
->ptr
;
1955 c
->argv
[2] = tryObjectEncoding(c
->argv
[2]);
1956 de
= dictFind(zs
->dict
,c
->argv
[2]);
1958 score
= *(double*)dictGetEntryVal(de
);
1959 addReplyDouble(c
,score
);
1961 addReply(c
,shared
.nullbulk
);
1964 redisPanic("Unknown sorted set encoding");
1968 void zrankGenericCommand(redisClient
*c
, int reverse
) {
1969 robj
*key
= c
->argv
[1];
1970 robj
*ele
= c
->argv
[2];
1975 if ((zobj
= lookupKeyReadOrReply(c
,key
,shared
.nullbulk
)) == NULL
||
1976 checkType(c
,zobj
,REDIS_ZSET
)) return;
1977 llen
= zsetLength(zobj
);
1979 redisAssert(ele
->encoding
== REDIS_ENCODING_RAW
);
1980 if (zobj
->encoding
== REDIS_ENCODING_ZIPLIST
) {
1981 unsigned char *zl
= zobj
->ptr
;
1982 unsigned char *eptr
, *sptr
;
1984 eptr
= ziplistIndex(zl
,0);
1985 redisAssert(eptr
!= NULL
);
1986 sptr
= ziplistNext(zl
,eptr
);
1987 redisAssert(sptr
!= NULL
);
1990 while(eptr
!= NULL
) {
1991 if (ziplistCompare(eptr
,ele
->ptr
,sdslen(ele
->ptr
)))
1994 zzlNext(zl
,&eptr
,&sptr
);
1999 addReplyLongLong(c
,llen
-rank
);
2001 addReplyLongLong(c
,rank
-1);
2003 addReply(c
,shared
.nullbulk
);
2005 } else if (zobj
->encoding
== REDIS_ENCODING_SKIPLIST
) {
2006 zset
*zs
= zobj
->ptr
;
2007 zskiplist
*zsl
= zs
->zsl
;
2011 ele
= c
->argv
[2] = tryObjectEncoding(c
->argv
[2]);
2012 de
= dictFind(zs
->dict
,ele
);
2014 score
= *(double*)dictGetEntryVal(de
);
2015 rank
= zslGetRank(zsl
,score
,ele
);
2016 redisAssert(rank
); /* Existing elements always have a rank. */
2018 addReplyLongLong(c
,llen
-rank
);
2020 addReplyLongLong(c
,rank
-1);
2022 addReply(c
,shared
.nullbulk
);
2025 redisPanic("Unknown sorted set encoding");
2029 void zrankCommand(redisClient
*c
) {
2030 zrankGenericCommand(c
, 0);
2033 void zrevrankCommand(redisClient
*c
) {
2034 zrankGenericCommand(c
, 1);