]>
git.saurik.com Git - redis.git/blob - ziplist.c
8ac84975643ae814cdba2c09c0ed0eb9ccac6865
1 /* Memory layout of a ziplist, containing "foo", "bar", "quux":
2 * <zlbytes><zllen><len>"foo"<len>"bar"<len>"quux"
4 * <zlbytes> is an unsigned integer to hold the number of bytes that
5 * the ziplist occupies. This is stored to not have to traverse the ziplist
6 * to know the new length when pushing.
8 * <zllen> is the number of items in the ziplist. When this value is
9 * greater than 254, we need to traverse the entire list to know
10 * how many items it holds.
12 * <len> is the number of bytes occupied by a single entry. When this
13 * number is greater than 253, the length will occupy 5 bytes, where
14 * the extra bytes contain an unsigned integer to hold the length.
26 /* Important note: the ZIP_END value is used to depict the end of the
27 * ziplist structure. When a pointer contains an entry, the first couple
28 * of bytes contain the encoded length of the previous entry. This length
29 * is encoded as ZIP_ENC_RAW length, so the first two bits will contain 00
30 * and the byte will therefore never have a value of 255. */
32 #define ZIP_BIGLEN 254
36 #define ZIP_ENC_SHORT 1
38 #define ZIP_ENC_LLONG 3
39 #define ZIP_ENCODING(p) ((p)[0] >> 6)
41 /* Length encoding for raw entries */
42 #define ZIP_LEN_INLINE 0
43 #define ZIP_LEN_UINT16 1
44 #define ZIP_LEN_UINT32 2
47 #define ZIPLIST_BYTES(zl) (*((unsigned int*)(zl)))
48 #define ZIPLIST_TAIL_OFFSET(zl) (*((zl)+sizeof(unsigned int)))
49 #define ZIPLIST_LENGTH(zl) (*((zl)+2*sizeof(unsigned int)))
50 #define ZIPLIST_HEADER_SIZE (2*sizeof(unsigned int)+1)
51 #define ZIPLIST_ENTRY_HEAD(zl) ((zl)+ZIPLIST_HEADER_SIZE)
52 #define ZIPLIST_ENTRY_TAIL(zl) ((zl)+ZIPLIST_TAIL_OFFSET(zl))
53 #define ZIPLIST_ENTRY_END(zl) ((zl)+ZIPLIST_BYTES(zl)-1)
54 #define ZIPLIST_INCR_LENGTH(zl,incr) { \
55 if (ZIPLIST_LENGTH(zl) < ZIP_BIGLEN) ZIPLIST_LENGTH(zl)+=incr; }
57 typedef struct zlentry
{
58 unsigned int prevrawlensize
, prevrawlen
;
59 unsigned int lensize
, len
;
60 unsigned int headersize
;
61 unsigned char encoding
;
65 /* Return bytes needed to store integer encoded by 'encoding' */
66 static unsigned int zipEncodingSize(char encoding
) {
67 if (encoding
== ZIP_ENC_SHORT
) {
68 return sizeof(short int);
69 } else if (encoding
== ZIP_ENC_INT
) {
71 } else if (encoding
== ZIP_ENC_LLONG
) {
72 return sizeof(long long);
77 /* Decode the encoded length pointed by 'p'. If a pointer to 'lensize' is
78 * provided, it is set to the number of bytes required to encode the length. */
79 static unsigned int zipDecodeLength(unsigned char *p
, unsigned int *lensize
) {
80 unsigned char encoding
= ZIP_ENCODING(p
), lenenc
;
83 if (encoding
== ZIP_ENC_RAW
) {
84 lenenc
= (p
[0] >> 4) & 0x3;
85 if (lenenc
== ZIP_LEN_INLINE
) {
87 if (lensize
) *lensize
= 1;
88 } else if (lenenc
== ZIP_LEN_UINT16
) {
89 len
= p
[1] | (p
[2] << 8);
90 if (lensize
) *lensize
= 3;
92 len
= p
[1] | (p
[2] << 8) | (p
[3] << 16) | (p
[4] << 24);
93 if (lensize
) *lensize
= 5;
96 len
= zipEncodingSize(encoding
);
97 if (lensize
) *lensize
= 1;
102 /* Encode the length 'l' writing it in 'p'. If p is NULL it just returns
103 * the amount of bytes required to encode such a length. */
104 static unsigned int zipEncodeLength(unsigned char *p
, char encoding
, unsigned int rawlen
) {
105 unsigned char len
= 1, lenenc
, buf
[5];
106 if (encoding
== ZIP_ENC_RAW
) {
109 lenenc
= ZIP_LEN_INLINE
;
111 } else if (rawlen
<= 0xffff) {
114 lenenc
= ZIP_LEN_UINT16
;
115 buf
[1] = (rawlen
) & 0xff;
116 buf
[2] = (rawlen
>> 8) & 0xff;
120 lenenc
= ZIP_LEN_UINT32
;
121 buf
[1] = (rawlen
) & 0xff;
122 buf
[2] = (rawlen
>> 8) & 0xff;
123 buf
[3] = (rawlen
>> 16) & 0xff;
124 buf
[4] = (rawlen
>> 24) & 0xff;
126 buf
[0] = (lenenc
<< 4) | (buf
[0] & 0xf);
130 /* Apparently we need to store the length in 'p' */
131 buf
[0] = (encoding
<< 6) | (buf
[0] & 0x3f);
136 /* Return the difference in number of bytes needed to store the new length
137 * "len" on the entry pointed to by "p". */
138 static int zipPrevLenByteDiff(unsigned char *p
, unsigned int len
) {
139 unsigned int prevlensize
;
140 zipDecodeLength(p
,&prevlensize
);
141 return zipEncodeLength(NULL
,ZIP_ENC_RAW
,len
)-prevlensize
;
144 /* Check if string pointed to by 'entry' can be encoded as an integer.
145 * Stores the integer value in 'v' and its encoding in 'encoding'.
146 * Warning: this function requires a NULL-terminated string! */
147 static int zipTryEncoding(unsigned char *entry
, long long *v
, char *encoding
) {
151 if (entry
[0] == '-' || (entry
[0] >= '0' && entry
[0] <= '9')) {
152 value
= strtoll((char*)entry
,&eptr
,10);
153 if (eptr
[0] != '\0') return 0;
154 if (value
>= SHRT_MIN
&& value
<= SHRT_MAX
) {
155 *encoding
= ZIP_ENC_SHORT
;
156 } else if (value
>= INT_MIN
&& value
<= INT_MAX
) {
157 *encoding
= ZIP_ENC_INT
;
159 *encoding
= ZIP_ENC_LLONG
;
167 /* Store integer 'value' at 'p', encoded as 'encoding' */
168 static void zipSaveInteger(unsigned char *p
, long long value
, char encoding
) {
172 if (encoding
== ZIP_ENC_SHORT
) {
174 memcpy(p
,&s
,sizeof(s
));
175 } else if (encoding
== ZIP_ENC_INT
) {
177 memcpy(p
,&i
,sizeof(i
));
178 } else if (encoding
== ZIP_ENC_LLONG
) {
180 memcpy(p
,&l
,sizeof(l
));
186 /* Read integer encoded as 'encoding' from 'p' */
187 static long long zipLoadInteger(unsigned char *p
, char encoding
) {
191 if (encoding
== ZIP_ENC_SHORT
) {
192 memcpy(&s
,p
,sizeof(s
));
194 } else if (encoding
== ZIP_ENC_INT
) {
195 memcpy(&i
,p
,sizeof(i
));
197 } else if (encoding
== ZIP_ENC_LLONG
) {
198 memcpy(&l
,p
,sizeof(l
));
206 /* Return a struct with all information about an entry. */
207 static zlentry
zipEntry(unsigned char *p
) {
209 e
.prevrawlen
= zipDecodeLength(p
,&e
.prevrawlensize
);
210 e
.len
= zipDecodeLength(p
+e
.prevrawlensize
,&e
.lensize
);
211 e
.headersize
= e
.prevrawlensize
+e
.lensize
;
212 e
.encoding
= ZIP_ENCODING(p
+e
.prevrawlensize
);
217 /* Return the total number of bytes used by the entry at "p". */
218 static unsigned int zipRawEntryLength(unsigned char *p
) {
219 zlentry e
= zipEntry(p
);
220 return e
.headersize
+ e
.len
;
223 /* Create a new empty ziplist. */
224 unsigned char *ziplistNew(void) {
225 unsigned int bytes
= ZIPLIST_HEADER_SIZE
+1;
226 unsigned char *zl
= zmalloc(bytes
);
227 ZIPLIST_BYTES(zl
) = bytes
;
228 ZIPLIST_TAIL_OFFSET(zl
) = ZIPLIST_HEADER_SIZE
;
229 ZIPLIST_LENGTH(zl
) = 0;
230 zl
[bytes
-1] = ZIP_END
;
234 /* Resize the ziplist. */
235 static unsigned char *ziplistResize(unsigned char *zl
, unsigned int len
) {
236 zl
= zrealloc(zl
,len
);
237 ZIPLIST_BYTES(zl
) = len
;
242 /* Delete "num" entries, starting at "p". Returns pointer to the ziplist. */
243 static unsigned char *__ziplistDelete(unsigned char *zl
, unsigned char *p
, int num
) {
244 unsigned int i
, totlen
, deleted
= 0;
246 zlentry first
= zipEntry(p
);
247 for (i
= 0; p
[0] != ZIP_END
&& i
< num
; i
++) {
248 p
+= zipRawEntryLength(p
);
254 if (p
[0] != ZIP_END
) {
255 /* Tricky: storing the prevlen in this entry might reduce or
256 * increase the number of bytes needed, compared to the current
257 * prevlen. Note that we can always store this length because
258 * it was previously stored by an entry that is being deleted. */
259 nextdiff
= zipPrevLenByteDiff(p
,first
.prevrawlen
);
260 zipEncodeLength(p
-nextdiff
,ZIP_ENC_RAW
,first
.prevrawlen
);
262 /* Update offset for tail */
263 ZIPLIST_TAIL_OFFSET(zl
) -= totlen
+nextdiff
;
265 /* Move tail to the front of the ziplist */
266 memmove(first
.p
,p
-nextdiff
,ZIPLIST_BYTES(zl
)-(p
-zl
)-1+nextdiff
);
268 /* The entire tail was deleted. No need to move memory. */
269 ZIPLIST_TAIL_OFFSET(zl
) = (first
.p
-zl
)-first
.prevrawlen
;
272 /* Resize and update length */
273 zl
= ziplistResize(zl
, ZIPLIST_BYTES(zl
)-totlen
+nextdiff
);
274 ZIPLIST_INCR_LENGTH(zl
,-deleted
);
279 /* Insert item at "p". */
280 static unsigned char *__ziplistInsert(unsigned char *zl
, unsigned char *p
, unsigned char *s
, unsigned int slen
) {
281 unsigned int curlen
= ZIPLIST_BYTES(zl
), reqlen
, prevlen
= 0;
282 unsigned int offset
, nextdiff
= 0;
284 char encoding
= ZIP_ENC_RAW
;
288 /* Find out prevlen for the entry that is inserted. */
289 if (p
[0] != ZIP_END
) {
291 prevlen
= entry
.prevrawlen
;
293 tail
= ZIPLIST_ENTRY_TAIL(zl
);
294 if (tail
[0] != ZIP_END
) {
295 prevlen
= zipRawEntryLength(tail
);
299 /* See if the entry can be encoded */
300 if (zipTryEncoding(s
,&value
,&encoding
)) {
301 reqlen
= zipEncodingSize(encoding
);
306 /* We need space for both the length of the previous entry and
307 * the length of the payload. */
308 reqlen
+= zipEncodeLength(NULL
,ZIP_ENC_RAW
,prevlen
);
309 reqlen
+= zipEncodeLength(NULL
,encoding
,slen
);
311 /* When the insert position is not equal to the tail, we need to
312 * make sure that the next entry can hold this entry's length in
313 * its prevlen field. */
314 nextdiff
= p
[0] != ZIP_END
? zipPrevLenByteDiff(p
,reqlen
) : 0;
316 /* Store offset because a realloc may change the address of zl. */
318 zl
= ziplistResize(zl
,curlen
+reqlen
+nextdiff
);
321 /* Apply memory move when necessary and update tail offset. */
322 if (p
[0] != ZIP_END
) {
323 /* Subtract one because of the ZIP_END bytes */
324 memmove(p
+reqlen
,p
-nextdiff
,curlen
-offset
-1+nextdiff
);
325 /* Encode this entry's raw length in the next entry. */
326 zipEncodeLength(p
+reqlen
,ZIP_ENC_RAW
,reqlen
);
327 /* Update offset for tail */
328 ZIPLIST_TAIL_OFFSET(zl
) += reqlen
+nextdiff
;
330 /* This element will be the new tail. */
331 ZIPLIST_TAIL_OFFSET(zl
) = p
-zl
;
334 /* Write the entry */
335 p
+= zipEncodeLength(p
,ZIP_ENC_RAW
,prevlen
);
336 p
+= zipEncodeLength(p
,encoding
,slen
);
337 if (encoding
!= ZIP_ENC_RAW
) {
338 zipSaveInteger(p
,value
,encoding
);
342 ZIPLIST_INCR_LENGTH(zl
,1);
346 unsigned char *ziplistPush(unsigned char *zl
, unsigned char *s
, unsigned int slen
, int where
) {
348 p
= (where
== ZIPLIST_HEAD
) ? ZIPLIST_ENTRY_HEAD(zl
) : ZIPLIST_ENTRY_END(zl
);
349 return __ziplistInsert(zl
,p
,s
,slen
);
352 unsigned char *ziplistPop(unsigned char *zl
, sds
*target
, int where
) {
356 if (target
) *target
= NULL
;
358 /* Get pointer to element to remove */
359 p
= (where
== ZIPLIST_HEAD
) ? ZIPLIST_ENTRY_HEAD(zl
) : ZIPLIST_ENTRY_TAIL(zl
);
360 if (*p
== ZIP_END
) return zl
;
364 if (entry
.encoding
== ZIP_ENC_RAW
) {
365 *target
= sdsnewlen(p
+entry
.headersize
,entry
.len
);
367 value
= zipLoadInteger(p
+entry
.headersize
,entry
.encoding
);
368 *target
= sdscatprintf(sdsempty(), "%lld", value
);
372 zl
= __ziplistDelete(zl
,p
,1);
376 /* Returns an offset to use for iterating with ziplistNext. */
377 unsigned char *ziplistIndex(unsigned char *zl
, unsigned int index
) {
378 unsigned char *p
= zl
+ZIPLIST_HEADER_SIZE
;
380 for (; i
< index
; i
++) {
381 if (*p
== ZIP_END
) break;
382 p
+= zipRawEntryLength(p
);
387 /* Return pointer to next entry in ziplist. */
388 unsigned char *ziplistNext(unsigned char *p
) {
389 return *p
== ZIP_END
? p
: p
+zipRawEntryLength(p
);
392 /* Get entry pointer to by 'p' and store in either 'e' or 'v' depending
393 * on the encoding of the entry. 'e' is always set to NULL to be able
394 * to find out whether the string pointer or the integer value was set.
395 * Return 0 if 'p' points to the end of the zipmap, 1 otherwise. */
396 unsigned int ziplistGet(unsigned char *p
, unsigned char **sstr
, unsigned int *slen
, long long *sval
) {
398 if (*p
== ZIP_END
) return 0;
399 if (sstr
) *sstr
= NULL
;
402 if (entry
.encoding
== ZIP_ENC_RAW
) {
405 *sstr
= p
+entry
.headersize
;
409 *sval
= zipLoadInteger(p
+entry
.headersize
,entry
.encoding
);
415 /* Delete a range of entries from the ziplist. */
416 unsigned char *ziplistDeleteRange(unsigned char *zl
, unsigned int index
, unsigned int num
) {
417 unsigned char *p
= ziplistIndex(zl
,index
);
418 return __ziplistDelete(zl
,p
,num
);
421 /* Delete a single entry from the ziplist, pointed to by *p.
422 * Also update *p in place, to be able to iterate over the
423 * ziplist, while deleting entries. */
424 unsigned char *ziplistDelete(unsigned char *zl
, unsigned char **p
) {
425 unsigned int offset
= *p
-zl
;
426 zl
= __ziplistDelete(zl
,*p
,1);
428 /* Store pointer to current element in p, because ziplistDelete will
429 * do a realloc which might result in a different "zl"-pointer. */
434 /* Compare entry pointer to by 'p' with 'entry'. Return 1 if equal. */
435 unsigned int ziplistCompare(unsigned char *p
, unsigned char *sstr
, unsigned int slen
) {
437 unsigned char sencoding
;
439 if (*p
== ZIP_END
) return 0;
442 if (entry
.encoding
== ZIP_ENC_RAW
) {
444 if (entry
.len
== slen
) {
445 return memcmp(p
+entry
.headersize
,sstr
,slen
) == 0;
450 /* Try to compare encoded values */
451 if (zipTryEncoding(sstr
,&sval
,&sencoding
)) {
452 if (entry
.encoding
== sencoding
) {
453 val
= zipLoadInteger(p
+entry
.headersize
,entry
.encoding
);
461 /* Return length of ziplist. */
462 unsigned int ziplistLen(unsigned char *zl
) {
463 unsigned int len
= 0;
464 if (ZIPLIST_LENGTH(zl
) < ZIP_BIGLEN
) {
465 len
= ZIPLIST_LENGTH(zl
);
467 unsigned char *p
= zl
+ZIPLIST_HEADER_SIZE
;
468 while (*p
!= ZIP_END
) {
469 p
+= zipRawEntryLength(p
);
473 /* Re-store length if small enough */
474 if (len
< ZIP_BIGLEN
) ZIPLIST_LENGTH(zl
) = len
;
479 /* Return size in bytes of ziplist. */
480 unsigned int ziplistSize(unsigned char *zl
) {
481 return ZIPLIST_BYTES(zl
);
484 void ziplistRepr(unsigned char *zl
) {
488 printf("{total bytes %d} {length %u}\n",ZIPLIST_BYTES(zl
), ZIPLIST_LENGTH(zl
));
489 p
= ZIPLIST_ENTRY_HEAD(zl
);
490 while(*p
!= ZIP_END
) {
492 printf("{offset %ld, header %u, payload %u} ",p
-zl
,entry
.headersize
,entry
.len
);
493 p
+= entry
.headersize
;
494 if (entry
.encoding
== ZIP_ENC_RAW
) {
495 fwrite(p
,entry
.len
,1,stdout
);
497 printf("%lld", zipLoadInteger(p
,entry
.encoding
));
505 #ifdef ZIPLIST_TEST_MAIN
507 unsigned char *createList() {
508 unsigned char *zl
= ziplistNew();
509 zl
= ziplistPush(zl
, (unsigned char*)"foo", 3, ZIPLIST_TAIL
);
510 zl
= ziplistPush(zl
, (unsigned char*)"quux", 4, ZIPLIST_TAIL
);
511 zl
= ziplistPush(zl
, (unsigned char*)"hello", 5, ZIPLIST_HEAD
);
512 zl
= ziplistPush(zl
, (unsigned char*)"1024", 4, ZIPLIST_TAIL
);
516 unsigned char *createIntList() {
517 unsigned char *zl
= ziplistNew();
521 zl
= ziplistPush(zl
, buf
, strlen(buf
), ZIPLIST_TAIL
);
522 sprintf(buf
, "128000");
523 zl
= ziplistPush(zl
, buf
, strlen(buf
), ZIPLIST_TAIL
);
524 sprintf(buf
, "-100");
525 zl
= ziplistPush(zl
, buf
, strlen(buf
), ZIPLIST_HEAD
);
526 sprintf(buf
, "4294967296");
527 zl
= ziplistPush(zl
, buf
, strlen(buf
), ZIPLIST_HEAD
);
528 sprintf(buf
, "non integer");
529 zl
= ziplistPush(zl
, buf
, strlen(buf
), ZIPLIST_TAIL
);
530 sprintf(buf
, "much much longer non integer");
531 zl
= ziplistPush(zl
, buf
, strlen(buf
), ZIPLIST_TAIL
);
535 int main(int argc
, char **argv
) {
536 unsigned char *zl
, *p
, *q
, *entry
;
541 zl
= createIntList();
547 zl
= ziplistPop(zl
, &s
, ZIPLIST_TAIL
);
548 printf("Pop tail: %s (length %ld)\n", s
, sdslen(s
));
551 zl
= ziplistPop(zl
, &s
, ZIPLIST_HEAD
);
552 printf("Pop head: %s (length %ld)\n", s
, sdslen(s
));
555 zl
= ziplistPop(zl
, &s
, ZIPLIST_TAIL
);
556 printf("Pop tail: %s (length %ld)\n", s
, sdslen(s
));
559 zl
= ziplistPop(zl
, &s
, ZIPLIST_TAIL
);
560 printf("Pop tail: %s (length %ld)\n", s
, sdslen(s
));
563 printf("Iterate list from 0 to end:\n");
566 p
= ziplistIndex(zl
, 0);
567 while (ziplistGet(p
, &entry
, &elen
, &value
)) {
570 fwrite(entry
,elen
,1,stdout
);
572 printf("%lld", value
);
580 printf("Iterate list from 1 to end:\n");
583 p
= ziplistIndex(zl
, 1);
584 while (ziplistGet(p
, &entry
, &elen
, &value
)) {
587 fwrite(entry
,elen
,1,stdout
);
589 printf("%lld", value
);
597 printf("Iterate list from 2 to end:\n");
600 p
= ziplistIndex(zl
, 2);
601 while (ziplistGet(p
, &entry
, &elen
, &value
)) {
604 fwrite(entry
,elen
,1,stdout
);
606 printf("%lld", value
);
614 printf("Iterate starting out of range:\n");
617 p
= ziplistIndex(zl
, 4);
618 if (!ziplistGet(p
, &entry
, &elen
, &value
)) {
619 printf("No entry\n");
626 printf("Delete inclusive range 0,0:\n");
629 zl
= ziplistDeleteRange(zl
, 0, 1);
633 printf("Delete inclusive range 0,1:\n");
636 zl
= ziplistDeleteRange(zl
, 0, 2);
640 printf("Delete inclusive range 1,2:\n");
643 zl
= ziplistDeleteRange(zl
, 1, 2);
647 printf("Delete with start index out of range:\n");
650 zl
= ziplistDeleteRange(zl
, 5, 1);
654 printf("Delete with num overflow:\n");
657 zl
= ziplistDeleteRange(zl
, 1, 5);
661 printf("Delete foo while iterating:\n");
664 p
= ziplistIndex(zl
, 0);
665 while (ziplistGet(p
, &entry
, &elen
, &value
)) {
666 if (entry
&& strncmp("foo", entry
, elen
) == 0) {
667 printf("Delete foo\n");
668 zl
= ziplistDelete(zl
, &p
);
672 fwrite(entry
,elen
,1,stdout
);
674 printf("%lld", value
);
684 printf("Compare strings with ziplist entries:\n");
687 p
= ziplistIndex(zl
, 0);
688 if (!ziplistCompare(p
,"hello",5)) {
689 printf("ERROR: not \"hello\"\n");
692 if (ziplistCompare(p
,"hella",5)) {
693 printf("ERROR: \"hella\"\n");
697 p
= ziplistIndex(zl
, 3);
698 if (!ziplistCompare(p
,"1024",4)) {
699 printf("ERROR: not \"1024\"\n");
702 if (ziplistCompare(p
,"1025",4)) {
703 printf("ERROR: \"1025\"\n");