1/* The ziplist is a specially encoded dually linked list that is designed
2 * to be very memory efficient. It stores both strings and integer values,
3 * where integers are encoded as actual integers instead of a series of
4 * characters. It allows push and pop operations on either side of the list
5 * in O(1) time. However, because every operation requires a reallocation of
6 * the memory used by the ziplist, the actual complexity is related to the
7 * amount of memory used by the ziplist.
8 *
9 * ----------------------------------------------------------------------------
10 *
11 * ZIPLIST OVERALL LAYOUT
12 * ======================
13 *
14 * The general layout of the ziplist is as follows:
15 *
16 * <zlbytes> <zltail> <zllen> <entry> <entry> ... <entry> <zlend>
17 *
18 * NOTE: all fields are stored in little endian, if not specified otherwise.
19 *
20 * <uint32_t zlbytes> is an unsigned integer to hold the number of bytes that
21 * the ziplist occupies, including the four bytes of the zlbytes field itself.
22 * This value needs to be stored to be able to resize the entire structure
23 * without the need to traverse it first.
24 *
25 * <uint32_t zltail> is the offset to the last entry in the list. This allows
26 * a pop operation on the far side of the list without the need for full
27 * traversal.
28 *
29 * <uint16_t zllen> is the number of entries. When there are more than
30 * 2^16-2 entries, this value is set to 2^16-1 and we need to traverse the
31 * entire list to know how many items it holds.
32 *
33 * <uint8_t zlend> is a special entry representing the end of the ziplist.
34 * Is encoded as a single byte equal to 255. No other normal entry starts
35 * with a byte set to the value of 255.
36 *
37 * ZIPLIST ENTRIES
38 * ===============
39 *
40 * Every entry in the ziplist is prefixed by metadata that contains two pieces
41 * of information. First, the length of the previous entry is stored to be
42 * able to traverse the list from back to front. Second, the entry encoding is
43 * provided. It represents the entry type, integer or string, and in the case
44 * of strings it also represents the length of the string payload.
45 * So a complete entry is stored like this:
46 *
47 * <prevlen> <encoding> <entry-data>
48 *
49 * Sometimes the encoding represents the entry itself, like for small integers
50 * as we'll see later. In such a case the <entry-data> part is missing, and we
51 * could have just:
52 *
53 * <prevlen> <encoding>
54 *
55 * The length of the previous entry, <prevlen>, is encoded in the following way:
56 * If this length is smaller than 254 bytes, it will only consume a single
57 * byte representing the length as an unsigned 8 bit integer. When the length
58 * is greater than or equal to 254, it will consume 5 bytes. The first byte is
59 * set to 254 (FE) to indicate a larger value is following. The remaining 4
60 * bytes take the length of the previous entry as value.
61 *
62 * So practically an entry is encoded in the following way:
63 *
64 * <prevlen from 0 to 253> <encoding> <entry>
65 *
66 * Or alternatively if the previous entry length is greater than 253 bytes
67 * the following encoding is used:
68 *
69 * 0xFE <4 bytes unsigned little endian prevlen> <encoding> <entry>
70 *
71 * The encoding field of the entry depends on the content of the
72 * entry. When the entry is a string, the first 2 bits of the encoding first
73 * byte will hold the type of encoding used to store the length of the string,
74 * followed by the actual length of the string. When the entry is an integer
75 * the first 2 bits are both set to 1. The following 2 bits are used to specify
76 * what kind of integer will be stored after this header. An overview of the
77 * different types and encodings is as follows. The first byte is always enough
78 * to determine the kind of entry.
79 *
80 * |00pppppp| - 1 byte
81 * String value with length less than or equal to 63 bytes (6 bits).
82 * "pppppp" represents the unsigned 6 bit length.
83 * |01pppppp|qqqqqqqq| - 2 bytes
84 * String value with length less than or equal to 16383 bytes (14 bits).
85 * IMPORTANT: The 14 bit number is stored in big endian.
86 * |10000000|qqqqqqqq|rrrrrrrr|ssssssss|tttttttt| - 5 bytes
87 * String value with length greater than or equal to 16384 bytes.
88 * Only the 4 bytes following the first byte represents the length
89 * up to 2^32-1. The 6 lower bits of the first byte are not used and
90 * are set to zero.
91 * IMPORTANT: The 32 bit number is stored in big endian.
92 * |11000000| - 3 bytes
93 * Integer encoded as int16_t (2 bytes).
94 * |11010000| - 5 bytes
95 * Integer encoded as int32_t (4 bytes).
96 * |11100000| - 9 bytes
97 * Integer encoded as int64_t (8 bytes).
98 * |11110000| - 4 bytes
99 * Integer encoded as 24 bit signed (3 bytes).
100 * |11111110| - 2 bytes
101 * Integer encoded as 8 bit signed (1 byte).
102 * |1111xxxx| - (with xxxx between 0001 and 1101) immediate 4 bit integer.
103 * Unsigned integer from 0 to 12. The encoded value is actually from
104 * 1 to 13 because 0000 and 1111 can not be used, so 1 should be
105 * subtracted from the encoded 4 bit value to obtain the right value.
106 * |11111111| - End of ziplist special entry.
107 *
108 * Like for the ziplist header, all the integers are represented in little
109 * endian byte order, even when this code is compiled in big endian systems.
110 *
111 * EXAMPLES OF ACTUAL ZIPLISTS
112 * ===========================
113 *
114 * The following is a ziplist containing the two elements representing
115 * the strings "2" and "5". It is composed of 15 bytes, that we visually
116 * split into sections:
117 *
118 * [0f 00 00 00] [0c 00 00 00] [02 00] [00 f3] [02 f6] [ff]
119 * | | | | | |
120 * zlbytes zltail entries "2" "5" end
121 *
122 * The first 4 bytes represent the number 15, that is the number of bytes
123 * the whole ziplist is composed of. The second 4 bytes are the offset
124 * at which the last ziplist entry is found, that is 12, in fact the
125 * last entry, that is "5", is at offset 12 inside the ziplist.
126 * The next 16 bit integer represents the number of elements inside the
127 * ziplist, its value is 2 since there are just two elements inside.
128 * Finally "00 f3" is the first entry representing the number 2. It is
129 * composed of the previous entry length, which is zero because this is
130 * our first entry, and the byte F3 which corresponds to the encoding
131 * |1111xxxx| with xxxx between 0001 and 1101. We need to remove the "F"
132 * higher order bits 1111, and subtract 1 from the "3", so the entry value
133 * is "2". The next entry has a prevlen of 02, since the first entry is
134 * composed of exactly two bytes. The entry itself, F6, is encoded exactly
135 * like the first entry, and 6-1 = 5, so the value of the entry is 5.
136 * Finally the special entry FF signals the end of the ziplist.
137 *
138 * Adding another element to the above string with the value "Hello World"
139 * allows us to show how the ziplist encodes small strings. We'll just show
140 * the hex dump of the entry itself. Imagine the bytes as following the
141 * entry that stores "5" in the ziplist above:
142 *
143 * [02] [0b] [48 65 6c 6c 6f 20 57 6f 72 6c 64]
144 *
145 * The first byte, 02, is the length of the previous entry. The next
146 * byte represents the encoding in the pattern |00pppppp| that means
147 * that the entry is a string of length <pppppp>, so 0B means that
148 * an 11 bytes string follows. From the third byte (48) to the last (64)
149 * there are just the ASCII characters for "Hello World".
150 *
151 * ----------------------------------------------------------------------------
152 *
153 * Copyright (c) 2009-2012, Pieter Noordhuis <pcnoordhuis at gmail dot com>
154 * Copyright (c) 2009-2017, Salvatore Sanfilippo <antirez at gmail dot com>
155 * Copyright (c) 2020, Redis Labs, Inc
156 * All rights reserved.
157 *
158 * Redistribution and use in source and binary forms, with or without
159 * modification, are permitted provided that the following conditions are met:
160 *
161 * * Redistributions of source code must retain the above copyright notice,
162 * this list of conditions and the following disclaimer.
163 * * Redistributions in binary form must reproduce the above copyright
164 * notice, this list of conditions and the following disclaimer in the
165 * documentation and/or other materials provided with the distribution.
166 * * Neither the name of Redis nor the names of its contributors may be used
167 * to endorse or promote products derived from this software without
168 * specific prior written permission.
169 *
170 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
171 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
172 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
173 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
174 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
175 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
176 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
177 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
178 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
179 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
180 * POSSIBILITY OF SUCH DAMAGE.
181 */
182
183#include <stdio.h>
184#include <stdlib.h>
185#include <string.h>
186#include <stdint.h>
187#include <limits.h>
188#include "zmalloc.h"
189#include "util.h"
190#include "ziplist.h"
191#include "config.h"
192#include "endianconv.h"
193#include "redisassert.h"
194
195#define ZIP_END 255 /* Special "end of ziplist" entry. */
196#define ZIP_BIG_PREVLEN 254 /* ZIP_BIG_PREVLEN - 1 is the max number of bytes of
197 the previous entry, for the "prevlen" field prefixing
198 each entry, to be represented with just a single byte.
199 Otherwise it is represented as FE AA BB CC DD, where
200 AA BB CC DD are a 4 bytes unsigned integer
201 representing the previous entry len. */
202
203/* Different encoding/length possibilities */
204#define ZIP_STR_MASK 0xc0
205#define ZIP_INT_MASK 0x30
206#define ZIP_STR_06B (0 << 6)
207#define ZIP_STR_14B (1 << 6)
208#define ZIP_STR_32B (2 << 6)
209#define ZIP_INT_16B (0xc0 | 0<<4)
210#define ZIP_INT_32B (0xc0 | 1<<4)
211#define ZIP_INT_64B (0xc0 | 2<<4)
212#define ZIP_INT_24B (0xc0 | 3<<4)
213#define ZIP_INT_8B 0xfe
214
215/* 4 bit integer immediate encoding |1111xxxx| with xxxx between
216 * 0001 and 1101. */
217#define ZIP_INT_IMM_MASK 0x0f /* Mask to extract the 4 bits value. To add
218 one is needed to reconstruct the value. */
219#define ZIP_INT_IMM_MIN 0xf1 /* 11110001 */
220#define ZIP_INT_IMM_MAX 0xfd /* 11111101 */
221
222#define INT24_MAX 0x7fffff
223#define INT24_MIN (-INT24_MAX - 1)
224
225/* Macro to determine if the entry is a string. String entries never start
226 * with "11" as most significant bits of the first byte. */
227#define ZIP_IS_STR(enc) (((enc) & ZIP_STR_MASK) < ZIP_STR_MASK)
228
229/* Utility macros.*/
230
231/* Return total bytes a ziplist is composed of. */
232#define ZIPLIST_BYTES(zl) (*((uint32_t*)(zl)))
233
234/* Return the offset of the last item inside the ziplist. */
235#define ZIPLIST_TAIL_OFFSET(zl) (*((uint32_t*)((zl)+sizeof(uint32_t))))
236
237/* Return the length of a ziplist, or UINT16_MAX if the length cannot be
238 * determined without scanning the whole ziplist. */
239#define ZIPLIST_LENGTH(zl) (*((uint16_t*)((zl)+sizeof(uint32_t)*2)))
240
241/* The size of a ziplist header: two 32 bit integers for the total
242 * bytes count and last item offset. One 16 bit integer for the number
243 * of items field. */
244#define ZIPLIST_HEADER_SIZE (sizeof(uint32_t)*2+sizeof(uint16_t))
245
246/* Size of the "end of ziplist" entry. Just one byte. */
247#define ZIPLIST_END_SIZE (sizeof(uint8_t))
248
249/* Return the pointer to the first entry of a ziplist. */
250#define ZIPLIST_ENTRY_HEAD(zl) ((zl)+ZIPLIST_HEADER_SIZE)
251
252/* Return the pointer to the last entry of a ziplist, using the
253 * last entry offset inside the ziplist header. */
254#define ZIPLIST_ENTRY_TAIL(zl) ((zl)+intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl)))
255
256/* Return the pointer to the last byte of a ziplist, which is, the
257 * end of ziplist FF entry. */
258#define ZIPLIST_ENTRY_END(zl) ((zl)+intrev32ifbe(ZIPLIST_BYTES(zl))-ZIPLIST_END_SIZE)
259
260/* Increment the number of items field in the ziplist header. Note that this
261 * macro should never overflow the unsigned 16 bit integer, since entries are
262 * always pushed one at a time. When UINT16_MAX is reached we want the count
263 * to stay there to signal that a full scan is needed to get the number of
264 * items inside the ziplist. */
265#define ZIPLIST_INCR_LENGTH(zl,incr) { \
266 if (intrev16ifbe(ZIPLIST_LENGTH(zl)) < UINT16_MAX) \
267 ZIPLIST_LENGTH(zl) = intrev16ifbe(intrev16ifbe(ZIPLIST_LENGTH(zl))+incr); \
268}
269
270/* Don't let ziplists grow over 1GB in any case, don't wanna risk overflow in
271 * zlbytes */
272#define ZIPLIST_MAX_SAFETY_SIZE (1<<30)
273int ziplistSafeToAdd(unsigned char* zl, size_t add) {
274 size_t len = zl? ziplistBlobLen(zl): 0;
275 if (len + add > ZIPLIST_MAX_SAFETY_SIZE)
276 return 0;
277 return 1;
278}
279
280
281/* We use this function to receive information about a ziplist entry.
282 * Note that this is not how the data is actually encoded, is just what we
283 * get filled by a function in order to operate more easily. */
284typedef struct zlentry {
285 unsigned int prevrawlensize; /* Bytes used to encode the previous entry len*/
286 unsigned int prevrawlen; /* Previous entry len. */
287 unsigned int lensize; /* Bytes used to encode this entry type/len.
288 For example strings have a 1, 2 or 5 bytes
289 header. Integers always use a single byte.*/
290 unsigned int len; /* Bytes used to represent the actual entry.
291 For strings this is just the string length
292 while for integers it is 1, 2, 3, 4, 8 or
293 0 (for 4 bit immediate) depending on the
294 number range. */
295 unsigned int headersize; /* prevrawlensize + lensize. */
296 unsigned char encoding; /* Set to ZIP_STR_* or ZIP_INT_* depending on
297 the entry encoding. However for 4 bits
298 immediate integers this can assume a range
299 of values and must be range-checked. */
300 unsigned char *p; /* Pointer to the very start of the entry, that
301 is, this points to prev-entry-len field. */
302} zlentry;
303
304#define ZIPLIST_ENTRY_ZERO(zle) { \
305 (zle)->prevrawlensize = (zle)->prevrawlen = 0; \
306 (zle)->lensize = (zle)->len = (zle)->headersize = 0; \
307 (zle)->encoding = 0; \
308 (zle)->p = NULL; \
309}
310
311/* Extract the encoding from the byte pointed by 'ptr' and set it into
312 * 'encoding' field of the zlentry structure. */
313#define ZIP_ENTRY_ENCODING(ptr, encoding) do { \
314 (encoding) = ((ptr)[0]); \
315 if ((encoding) < ZIP_STR_MASK) (encoding) &= ZIP_STR_MASK; \
316} while(0)
317
318#define ZIP_ENCODING_SIZE_INVALID 0xff
319/* Return the number of bytes required to encode the entry type + length.
320 * On error, return ZIP_ENCODING_SIZE_INVALID */
321static inline unsigned int zipEncodingLenSize(unsigned char encoding) {
322 if (encoding == ZIP_INT_16B || encoding == ZIP_INT_32B ||
323 encoding == ZIP_INT_24B || encoding == ZIP_INT_64B ||
324 encoding == ZIP_INT_8B)
325 return 1;
326 if (encoding >= ZIP_INT_IMM_MIN && encoding <= ZIP_INT_IMM_MAX)
327 return 1;
328 if (encoding == ZIP_STR_06B)
329 return 1;
330 if (encoding == ZIP_STR_14B)
331 return 2;
332 if (encoding == ZIP_STR_32B)
333 return 5;
334 return ZIP_ENCODING_SIZE_INVALID;
335}
336
337#define ZIP_ASSERT_ENCODING(encoding) do { \
338 assert(zipEncodingLenSize(encoding) != ZIP_ENCODING_SIZE_INVALID); \
339} while (0)
340
341/* Return bytes needed to store integer encoded by 'encoding' */
342static inline unsigned int zipIntSize(unsigned char encoding) {
343 switch(encoding) {
344 case ZIP_INT_8B: return 1;
345 case ZIP_INT_16B: return 2;
346 case ZIP_INT_24B: return 3;
347 case ZIP_INT_32B: return 4;
348 case ZIP_INT_64B: return 8;
349 }
350 if (encoding >= ZIP_INT_IMM_MIN && encoding <= ZIP_INT_IMM_MAX)
351 return 0; /* 4 bit immediate */
352 /* bad encoding, covered by a previous call to ZIP_ASSERT_ENCODING */
353 redis_unreachable();
354 return 0;
355}
356
357/* Write the encoding header of the entry in 'p'. If p is NULL it just returns
358 * the amount of bytes required to encode such a length. Arguments:
359 *
360 * 'encoding' is the encoding we are using for the entry. It could be
361 * ZIP_INT_* or ZIP_STR_* or between ZIP_INT_IMM_MIN and ZIP_INT_IMM_MAX
362 * for single-byte small immediate integers.
363 *
364 * 'rawlen' is only used for ZIP_STR_* encodings and is the length of the
365 * string that this entry represents.
366 *
367 * The function returns the number of bytes used by the encoding/length
368 * header stored in 'p'. */
369unsigned int zipStoreEntryEncoding(unsigned char *p, unsigned char encoding, unsigned int rawlen) {
370 unsigned char len = 1, buf[5];
371
372 if (ZIP_IS_STR(encoding)) {
373 /* Although encoding is given it may not be set for strings,
374 * so we determine it here using the raw length. */
375 if (rawlen <= 0x3f) {
376 if (!p) return len;
377 buf[0] = ZIP_STR_06B | rawlen;
378 } else if (rawlen <= 0x3fff) {
379 len += 1;
380 if (!p) return len;
381 buf[0] = ZIP_STR_14B | ((rawlen >> 8) & 0x3f);
382 buf[1] = rawlen & 0xff;
383 } else {
384 len += 4;
385 if (!p) return len;
386 buf[0] = ZIP_STR_32B;
387 buf[1] = (rawlen >> 24) & 0xff;
388 buf[2] = (rawlen >> 16) & 0xff;
389 buf[3] = (rawlen >> 8) & 0xff;
390 buf[4] = rawlen & 0xff;
391 }
392 } else {
393 /* Implies integer encoding, so length is always 1. */
394 if (!p) return len;
395 buf[0] = encoding;
396 }
397
398 /* Store this length at p. */
399 memcpy(p,buf,len);
400 return len;
401}
402
403/* Decode the entry encoding type and data length (string length for strings,
404 * number of bytes used for the integer for integer entries) encoded in 'ptr'.
405 * The 'encoding' variable is input, extracted by the caller, the 'lensize'
406 * variable will hold the number of bytes required to encode the entry
407 * length, and the 'len' variable will hold the entry length.
408 * On invalid encoding error, lensize is set to 0. */
409#define ZIP_DECODE_LENGTH(ptr, encoding, lensize, len) do { \
410 if ((encoding) < ZIP_STR_MASK) { \
411 if ((encoding) == ZIP_STR_06B) { \
412 (lensize) = 1; \
413 (len) = (ptr)[0] & 0x3f; \
414 } else if ((encoding) == ZIP_STR_14B) { \
415 (lensize) = 2; \
416 (len) = (((ptr)[0] & 0x3f) << 8) | (ptr)[1]; \
417 } else if ((encoding) == ZIP_STR_32B) { \
418 (lensize) = 5; \
419 (len) = ((uint32_t)(ptr)[1] << 24) | \
420 ((uint32_t)(ptr)[2] << 16) | \
421 ((uint32_t)(ptr)[3] << 8) | \
422 ((uint32_t)(ptr)[4]); \
423 } else { \
424 (lensize) = 0; /* bad encoding, should be covered by a previous */ \
425 (len) = 0; /* ZIP_ASSERT_ENCODING / zipEncodingLenSize, or */ \
426 /* match the lensize after this macro with 0. */ \
427 } \
428 } else { \
429 (lensize) = 1; \
430 if ((encoding) == ZIP_INT_8B) (len) = 1; \
431 else if ((encoding) == ZIP_INT_16B) (len) = 2; \
432 else if ((encoding) == ZIP_INT_24B) (len) = 3; \
433 else if ((encoding) == ZIP_INT_32B) (len) = 4; \
434 else if ((encoding) == ZIP_INT_64B) (len) = 8; \
435 else if (encoding >= ZIP_INT_IMM_MIN && encoding <= ZIP_INT_IMM_MAX) \
436 (len) = 0; /* 4 bit immediate */ \
437 else \
438 (lensize) = (len) = 0; /* bad encoding */ \
439 } \
440} while(0)
441
442/* Encode the length of the previous entry and write it to "p". This only
443 * uses the larger encoding (required in __ziplistCascadeUpdate). */
444int zipStorePrevEntryLengthLarge(unsigned char *p, unsigned int len) {
445 uint32_t u32;
446 if (p != NULL) {
447 p[0] = ZIP_BIG_PREVLEN;
448 u32 = len;
449 memcpy(p+1,&u32,sizeof(u32));
450 memrev32ifbe(p+1);
451 }
452 return 1 + sizeof(uint32_t);
453}
454
455/* Encode the length of the previous entry and write it to "p". Return the
456 * number of bytes needed to encode this length if "p" is NULL. */
457unsigned int zipStorePrevEntryLength(unsigned char *p, unsigned int len) {
458 if (p == NULL) {
459 return (len < ZIP_BIG_PREVLEN) ? 1 : sizeof(uint32_t) + 1;
460 } else {
461 if (len < ZIP_BIG_PREVLEN) {
462 p[0] = len;
463 return 1;
464 } else {
465 return zipStorePrevEntryLengthLarge(p,len);
466 }
467 }
468}
469
470/* Return the number of bytes used to encode the length of the previous
471 * entry. The length is returned by setting the var 'prevlensize'. */
472#define ZIP_DECODE_PREVLENSIZE(ptr, prevlensize) do { \
473 if ((ptr)[0] < ZIP_BIG_PREVLEN) { \
474 (prevlensize) = 1; \
475 } else { \
476 (prevlensize) = 5; \
477 } \
478} while(0)
479
480/* Return the length of the previous element, and the number of bytes that
481 * are used in order to encode the previous element length.
482 * 'ptr' must point to the prevlen prefix of an entry (that encodes the
483 * length of the previous entry in order to navigate the elements backward).
484 * The length of the previous entry is stored in 'prevlen', the number of
485 * bytes needed to encode the previous entry length are stored in
486 * 'prevlensize'. */
487#define ZIP_DECODE_PREVLEN(ptr, prevlensize, prevlen) do { \
488 ZIP_DECODE_PREVLENSIZE(ptr, prevlensize); \
489 if ((prevlensize) == 1) { \
490 (prevlen) = (ptr)[0]; \
491 } else { /* prevlensize == 5 */ \
492 (prevlen) = ((ptr)[4] << 24) | \
493 ((ptr)[3] << 16) | \
494 ((ptr)[2] << 8) | \
495 ((ptr)[1]); \
496 } \
497} while(0)
498
499/* Given a pointer 'p' to the prevlen info that prefixes an entry, this
500 * function returns the difference in number of bytes needed to encode
501 * the prevlen if the previous entry changes of size.
502 *
503 * So if A is the number of bytes used right now to encode the 'prevlen'
504 * field.
505 *
506 * And B is the number of bytes that are needed in order to encode the
507 * 'prevlen' if the previous element will be updated to one of size 'len'.
508 *
509 * Then the function returns B - A
510 *
511 * So the function returns a positive number if more space is needed,
512 * a negative number if less space is needed, or zero if the same space
513 * is needed. */
514int zipPrevLenByteDiff(unsigned char *p, unsigned int len) {
515 unsigned int prevlensize;
516 ZIP_DECODE_PREVLENSIZE(p, prevlensize);
517 return zipStorePrevEntryLength(NULL, len) - prevlensize;
518}
519
520/* Check if string pointed to by 'entry' can be encoded as an integer.
521 * Stores the integer value in 'v' and its encoding in 'encoding'. */
522int zipTryEncoding(unsigned char *entry, unsigned int entrylen, long long *v, unsigned char *encoding) {
523 long long value;
524
525 if (entrylen >= 32 || entrylen == 0) return 0;
526 if (string2ll((char*)entry,entrylen,&value)) {
527 /* Great, the string can be encoded. Check what's the smallest
528 * of our encoding types that can hold this value. */
529 if (value >= 0 && value <= 12) {
530 *encoding = ZIP_INT_IMM_MIN+value;
531 } else if (value >= INT8_MIN && value <= INT8_MAX) {
532 *encoding = ZIP_INT_8B;
533 } else if (value >= INT16_MIN && value <= INT16_MAX) {
534 *encoding = ZIP_INT_16B;
535 } else if (value >= INT24_MIN && value <= INT24_MAX) {
536 *encoding = ZIP_INT_24B;
537 } else if (value >= INT32_MIN && value <= INT32_MAX) {
538 *encoding = ZIP_INT_32B;
539 } else {
540 *encoding = ZIP_INT_64B;
541 }
542 *v = value;
543 return 1;
544 }
545 return 0;
546}
547
548/* Store integer 'value' at 'p', encoded as 'encoding' */
549void zipSaveInteger(unsigned char *p, int64_t value, unsigned char encoding) {
550 int16_t i16;
551 int32_t i32;
552 int64_t i64;
553 if (encoding == ZIP_INT_8B) {
554 ((int8_t*)p)[0] = (int8_t)value;
555 } else if (encoding == ZIP_INT_16B) {
556 i16 = value;
557 memcpy(p,&i16,sizeof(i16));
558 memrev16ifbe(p);
559 } else if (encoding == ZIP_INT_24B) {
560 i32 = ((uint64_t)value)<<8;
561 memrev32ifbe(&i32);
562 memcpy(p,((uint8_t*)&i32)+1,sizeof(i32)-sizeof(uint8_t));
563 } else if (encoding == ZIP_INT_32B) {
564 i32 = value;
565 memcpy(p,&i32,sizeof(i32));
566 memrev32ifbe(p);
567 } else if (encoding == ZIP_INT_64B) {
568 i64 = value;
569 memcpy(p,&i64,sizeof(i64));
570 memrev64ifbe(p);
571 } else if (encoding >= ZIP_INT_IMM_MIN && encoding <= ZIP_INT_IMM_MAX) {
572 /* Nothing to do, the value is stored in the encoding itself. */
573 } else {
574 assert(NULL);
575 }
576}
577
578/* Read integer encoded as 'encoding' from 'p' */
579int64_t zipLoadInteger(unsigned char *p, unsigned char encoding) {
580 int16_t i16;
581 int32_t i32;
582 int64_t i64, ret = 0;
583 if (encoding == ZIP_INT_8B) {
584 ret = ((int8_t*)p)[0];
585 } else if (encoding == ZIP_INT_16B) {
586 memcpy(&i16,p,sizeof(i16));
587 memrev16ifbe(&i16);
588 ret = i16;
589 } else if (encoding == ZIP_INT_32B) {
590 memcpy(&i32,p,sizeof(i32));
591 memrev32ifbe(&i32);
592 ret = i32;
593 } else if (encoding == ZIP_INT_24B) {
594 i32 = 0;
595 memcpy(((uint8_t*)&i32)+1,p,sizeof(i32)-sizeof(uint8_t));
596 memrev32ifbe(&i32);
597 ret = i32>>8;
598 } else if (encoding == ZIP_INT_64B) {
599 memcpy(&i64,p,sizeof(i64));
600 memrev64ifbe(&i64);
601 ret = i64;
602 } else if (encoding >= ZIP_INT_IMM_MIN && encoding <= ZIP_INT_IMM_MAX) {
603 ret = (encoding & ZIP_INT_IMM_MASK)-1;
604 } else {
605 assert(NULL);
606 }
607 return ret;
608}
609
610/* Fills a struct with all information about an entry.
611 * This function is the "unsafe" alternative to the one below.
612 * Generally, all function that return a pointer to an element in the ziplist
613 * will assert that this element is valid, so it can be freely used.
614 * Generally functions such ziplistGet assume the input pointer is already
615 * validated (since it's the return value of another function). */
616static inline void zipEntry(unsigned char *p, zlentry *e) {
617 ZIP_DECODE_PREVLEN(p, e->prevrawlensize, e->prevrawlen);
618 ZIP_ENTRY_ENCODING(p + e->prevrawlensize, e->encoding);
619 ZIP_DECODE_LENGTH(p + e->prevrawlensize, e->encoding, e->lensize, e->len);
620 assert(e->lensize != 0); /* check that encoding was valid. */
621 e->headersize = e->prevrawlensize + e->lensize;
622 e->p = p;
623}
624
625/* Fills a struct with all information about an entry.
626 * This function is safe to use on untrusted pointers, it'll make sure not to
627 * try to access memory outside the ziplist payload.
628 * Returns 1 if the entry is valid, and 0 otherwise. */
629static inline int zipEntrySafe(unsigned char* zl, size_t zlbytes, unsigned char *p, zlentry *e, int validate_prevlen) {
630 unsigned char *zlfirst = zl + ZIPLIST_HEADER_SIZE;
631 unsigned char *zllast = zl + zlbytes - ZIPLIST_END_SIZE;
632#define OUT_OF_RANGE(p) (unlikely((p) < zlfirst || (p) > zllast))
633
634 /* If there's no possibility for the header to reach outside the ziplist,
635 * take the fast path. (max lensize and prevrawlensize are both 5 bytes) */
636 if (p >= zlfirst && p + 10 < zllast) {
637 ZIP_DECODE_PREVLEN(p, e->prevrawlensize, e->prevrawlen);
638 ZIP_ENTRY_ENCODING(p + e->prevrawlensize, e->encoding);
639 ZIP_DECODE_LENGTH(p + e->prevrawlensize, e->encoding, e->lensize, e->len);
640 e->headersize = e->prevrawlensize + e->lensize;
641 e->p = p;
642 /* We didn't call ZIP_ASSERT_ENCODING, so we check lensize was set to 0. */
643 if (unlikely(e->lensize == 0))
644 return 0;
645 /* Make sure the entry doesn't reach outside the edge of the ziplist */
646 if (OUT_OF_RANGE(p + e->headersize + e->len))
647 return 0;
648 /* Make sure prevlen doesn't reach outside the edge of the ziplist */
649 if (validate_prevlen && OUT_OF_RANGE(p - e->prevrawlen))
650 return 0;
651 return 1;
652 }
653
654 /* Make sure the pointer doesn't reach outside the edge of the ziplist */
655 if (OUT_OF_RANGE(p))
656 return 0;
657
658 /* Make sure the encoded prevlen header doesn't reach outside the allocation */
659 ZIP_DECODE_PREVLENSIZE(p, e->prevrawlensize);
660 if (OUT_OF_RANGE(p + e->prevrawlensize))
661 return 0;
662
663 /* Make sure encoded entry header is valid. */
664 ZIP_ENTRY_ENCODING(p + e->prevrawlensize, e->encoding);
665 e->lensize = zipEncodingLenSize(e->encoding);
666 if (unlikely(e->lensize == ZIP_ENCODING_SIZE_INVALID))
667 return 0;
668
669 /* Make sure the encoded entry header doesn't reach outside the allocation */
670 if (OUT_OF_RANGE(p + e->prevrawlensize + e->lensize))
671 return 0;
672
673 /* Decode the prevlen and entry len headers. */
674 ZIP_DECODE_PREVLEN(p, e->prevrawlensize, e->prevrawlen);
675 ZIP_DECODE_LENGTH(p + e->prevrawlensize, e->encoding, e->lensize, e->len);
676 e->headersize = e->prevrawlensize + e->lensize;
677
678 /* Make sure the entry doesn't reach outside the edge of the ziplist */
679 if (OUT_OF_RANGE(p + e->headersize + e->len))
680 return 0;
681
682 /* Make sure prevlen doesn't reach outside the edge of the ziplist */
683 if (validate_prevlen && OUT_OF_RANGE(p - e->prevrawlen))
684 return 0;
685
686 e->p = p;
687 return 1;
688#undef OUT_OF_RANGE
689}
690
691/* Return the total number of bytes used by the entry pointed to by 'p'. */
692static inline unsigned int zipRawEntryLengthSafe(unsigned char* zl, size_t zlbytes, unsigned char *p) {
693 zlentry e;
694 assert(zipEntrySafe(zl, zlbytes, p, &e, 0));
695 return e.headersize + e.len;
696}
697
698/* Return the total number of bytes used by the entry pointed to by 'p'. */
699static inline unsigned int zipRawEntryLength(unsigned char *p) {
700 zlentry e;
701 zipEntry(p, &e);
702 return e.headersize + e.len;
703}
704
705/* Validate that the entry doesn't reach outside the ziplist allocation. */
706static inline void zipAssertValidEntry(unsigned char* zl, size_t zlbytes, unsigned char *p) {
707 zlentry e;
708 assert(zipEntrySafe(zl, zlbytes, p, &e, 1));
709}
710
711/* Create a new empty ziplist. */
712unsigned char *ziplistNew(void) {
713 unsigned int bytes = ZIPLIST_HEADER_SIZE+ZIPLIST_END_SIZE;
714 unsigned char *zl = zmalloc(bytes);
715 ZIPLIST_BYTES(zl) = intrev32ifbe(bytes);
716 ZIPLIST_TAIL_OFFSET(zl) = intrev32ifbe(ZIPLIST_HEADER_SIZE);
717 ZIPLIST_LENGTH(zl) = 0;
718 zl[bytes-1] = ZIP_END;
719 return zl;
720}
721
722/* Resize the ziplist. */
723unsigned char *ziplistResize(unsigned char *zl, size_t len) {
724 assert(len < UINT32_MAX);
725 zl = zrealloc(zl,len);
726 ZIPLIST_BYTES(zl) = intrev32ifbe(len);
727 zl[len-1] = ZIP_END;
728 return zl;
729}
730
731/* When an entry is inserted, we need to set the prevlen field of the next
732 * entry to equal the length of the inserted entry. It can occur that this
733 * length cannot be encoded in 1 byte and the next entry needs to be grow
734 * a bit larger to hold the 5-byte encoded prevlen. This can be done for free,
735 * because this only happens when an entry is already being inserted (which
736 * causes a realloc and memmove). However, encoding the prevlen may require
737 * that this entry is grown as well. This effect may cascade throughout
738 * the ziplist when there are consecutive entries with a size close to
739 * ZIP_BIG_PREVLEN, so we need to check that the prevlen can be encoded in
740 * every consecutive entry.
741 *
742 * Note that this effect can also happen in reverse, where the bytes required
743 * to encode the prevlen field can shrink. This effect is deliberately ignored,
744 * because it can cause a "flapping" effect where a chain prevlen fields is
745 * first grown and then shrunk again after consecutive inserts. Rather, the
746 * field is allowed to stay larger than necessary, because a large prevlen
747 * field implies the ziplist is holding large entries anyway.
748 *
749 * The pointer "p" points to the first entry that does NOT need to be
750 * updated, i.e. consecutive fields MAY need an update. */
751unsigned char *__ziplistCascadeUpdate(unsigned char *zl, unsigned char *p) {
752 zlentry cur;
753 size_t prevlen, prevlensize, prevoffset; /* Informat of the last changed entry. */
754 size_t firstentrylen; /* Used to handle insert at head. */
755 size_t rawlen, curlen = intrev32ifbe(ZIPLIST_BYTES(zl));
756 size_t extra = 0, cnt = 0, offset;
757 size_t delta = 4; /* Extra bytes needed to update a entry's prevlen (5-1). */
758 unsigned char *tail = zl + intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl));
759
760 /* Empty ziplist */
761 if (p[0] == ZIP_END) return zl;
762
763 zipEntry(p, &cur); /* no need for "safe" variant since the input pointer was validated by the function that returned it. */
764 firstentrylen = prevlen = cur.headersize + cur.len;
765 prevlensize = zipStorePrevEntryLength(NULL, prevlen);
766 prevoffset = p - zl;
767 p += prevlen;
768
769 /* Iterate ziplist to find out how many extra bytes do we need to update it. */
770 while (p[0] != ZIP_END) {
771 assert(zipEntrySafe(zl, curlen, p, &cur, 0));
772
773 /* Abort when "prevlen" has not changed. */
774 if (cur.prevrawlen == prevlen) break;
775
776 /* Abort when entry's "prevlensize" is big enough. */
777 if (cur.prevrawlensize >= prevlensize) {
778 if (cur.prevrawlensize == prevlensize) {
779 zipStorePrevEntryLength(p, prevlen);
780 } else {
781 /* This would result in shrinking, which we want to avoid.
782 * So, set "prevlen" in the available bytes. */
783 zipStorePrevEntryLengthLarge(p, prevlen);
784 }
785 break;
786 }
787
788 /* cur.prevrawlen means cur is the former head entry. */
789 assert(cur.prevrawlen == 0 || cur.prevrawlen + delta == prevlen);
790
791 /* Update prev entry's info and advance the cursor. */
792 rawlen = cur.headersize + cur.len;
793 prevlen = rawlen + delta;
794 prevlensize = zipStorePrevEntryLength(NULL, prevlen);
795 prevoffset = p - zl;
796 p += rawlen;
797 extra += delta;
798 cnt++;
799 }
800
801 /* Extra bytes is zero all update has been done(or no need to update). */
802 if (extra == 0) return zl;
803
804 /* Update tail offset after loop. */
805 if (tail == zl + prevoffset) {
806 /* When the last entry we need to update is also the tail, update tail offset
807 * unless this is the only entry that was updated (so the tail offset didn't change). */
808 if (extra - delta != 0) {
809 ZIPLIST_TAIL_OFFSET(zl) =
810 intrev32ifbe(intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))+extra-delta);
811 }
812 } else {
813 /* Update the tail offset in cases where the last entry we updated is not the tail. */
814 ZIPLIST_TAIL_OFFSET(zl) =
815 intrev32ifbe(intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))+extra);
816 }
817
818 /* Now "p" points at the first unchanged byte in original ziplist,
819 * move data after that to new ziplist. */
820 offset = p - zl;
821 zl = ziplistResize(zl, curlen + extra);
822 p = zl + offset;
823 memmove(p + extra, p, curlen - offset - 1);
824 p += extra;
825
826 /* Iterate all entries that need to be updated tail to head. */
827 while (cnt) {
828 zipEntry(zl + prevoffset, &cur); /* no need for "safe" variant since we already iterated on all these entries above. */
829 rawlen = cur.headersize + cur.len;
830 /* Move entry to tail and reset prevlen. */
831 memmove(p - (rawlen - cur.prevrawlensize),
832 zl + prevoffset + cur.prevrawlensize,
833 rawlen - cur.prevrawlensize);
834 p -= (rawlen + delta);
835 if (cur.prevrawlen == 0) {
836 /* "cur" is the previous head entry, update its prevlen with firstentrylen. */
837 zipStorePrevEntryLength(p, firstentrylen);
838 } else {
839 /* An entry's prevlen can only increment 4 bytes. */
840 zipStorePrevEntryLength(p, cur.prevrawlen+delta);
841 }
842 /* Forward to previous entry. */
843 prevoffset -= cur.prevrawlen;
844 cnt--;
845 }
846 return zl;
847}
848
849/* Delete "num" entries, starting at "p". Returns pointer to the ziplist. */
850unsigned char *__ziplistDelete(unsigned char *zl, unsigned char *p, unsigned int num) {
851 unsigned int i, totlen, deleted = 0;
852 size_t offset;
853 int nextdiff = 0;
854 zlentry first, tail;
855 size_t zlbytes = intrev32ifbe(ZIPLIST_BYTES(zl));
856
857 zipEntry(p, &first); /* no need for "safe" variant since the input pointer was validated by the function that returned it. */
858 for (i = 0; p[0] != ZIP_END && i < num; i++) {
859 p += zipRawEntryLengthSafe(zl, zlbytes, p);
860 deleted++;
861 }
862
863 assert(p >= first.p);
864 totlen = p-first.p; /* Bytes taken by the element(s) to delete. */
865 if (totlen > 0) {
866 uint32_t set_tail;
867 if (p[0] != ZIP_END) {
868 /* Storing `prevrawlen` in this entry may increase or decrease the
869 * number of bytes required compare to the current `prevrawlen`.
870 * There always is room to store this, because it was previously
871 * stored by an entry that is now being deleted. */
872 nextdiff = zipPrevLenByteDiff(p,first.prevrawlen);
873
874 /* Note that there is always space when p jumps backward: if
875 * the new previous entry is large, one of the deleted elements
876 * had a 5 bytes prevlen header, so there is for sure at least
877 * 5 bytes free and we need just 4. */
878 p -= nextdiff;
879 assert(p >= first.p && p<zl+zlbytes-1);
880 zipStorePrevEntryLength(p,first.prevrawlen);
881
882 /* Update offset for tail */
883 set_tail = intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))-totlen;
884
885 /* When the tail contains more than one entry, we need to take
886 * "nextdiff" in account as well. Otherwise, a change in the
887 * size of prevlen doesn't have an effect on the *tail* offset. */
888 assert(zipEntrySafe(zl, zlbytes, p, &tail, 1));
889 if (p[tail.headersize+tail.len] != ZIP_END) {
890 set_tail = set_tail + nextdiff;
891 }
892
893 /* Move tail to the front of the ziplist */
894 /* since we asserted that p >= first.p. we know totlen >= 0,
895 * so we know that p > first.p and this is guaranteed not to reach
896 * beyond the allocation, even if the entries lens are corrupted. */
897 size_t bytes_to_move = zlbytes-(p-zl)-1;
898 memmove(first.p,p,bytes_to_move);
899 } else {
900 /* The entire tail was deleted. No need to move memory. */
901 set_tail = (first.p-zl)-first.prevrawlen;
902 }
903
904 /* Resize the ziplist */
905 offset = first.p-zl;
906 zlbytes -= totlen - nextdiff;
907 zl = ziplistResize(zl, zlbytes);
908 p = zl+offset;
909
910 /* Update record count */
911 ZIPLIST_INCR_LENGTH(zl,-deleted);
912
913 /* Set the tail offset computed above */
914 assert(set_tail <= zlbytes - ZIPLIST_END_SIZE);
915 ZIPLIST_TAIL_OFFSET(zl) = intrev32ifbe(set_tail);
916
917 /* When nextdiff != 0, the raw length of the next entry has changed, so
918 * we need to cascade the update throughout the ziplist */
919 if (nextdiff != 0)
920 zl = __ziplistCascadeUpdate(zl,p);
921 }
922 return zl;
923}
924
925/* Insert item at "p". */
926unsigned char *__ziplistInsert(unsigned char *zl, unsigned char *p, unsigned char *s, unsigned int slen) {
927 size_t curlen = intrev32ifbe(ZIPLIST_BYTES(zl)), reqlen, newlen;
928 unsigned int prevlensize, prevlen = 0;
929 size_t offset;
930 int nextdiff = 0;
931 unsigned char encoding = 0;
932 long long value = 123456789; /* initialized to avoid warning. Using a value
933 that is easy to see if for some reason
934 we use it uninitialized. */
935 zlentry tail;
936
937 /* Find out prevlen for the entry that is inserted. */
938 if (p[0] != ZIP_END) {
939 ZIP_DECODE_PREVLEN(p, prevlensize, prevlen);
940 } else {
941 unsigned char *ptail = ZIPLIST_ENTRY_TAIL(zl);
942 if (ptail[0] != ZIP_END) {
943 prevlen = zipRawEntryLengthSafe(zl, curlen, ptail);
944 }
945 }
946
947 /* See if the entry can be encoded */
948 if (zipTryEncoding(s,slen,&value,&encoding)) {
949 /* 'encoding' is set to the appropriate integer encoding */
950 reqlen = zipIntSize(encoding);
951 } else {
952 /* 'encoding' is untouched, however zipStoreEntryEncoding will use the
953 * string length to figure out how to encode it. */
954 reqlen = slen;
955 }
956 /* We need space for both the length of the previous entry and
957 * the length of the payload. */
958 reqlen += zipStorePrevEntryLength(NULL,prevlen);
959 reqlen += zipStoreEntryEncoding(NULL,encoding,slen);
960
961 /* When the insert position is not equal to the tail, we need to
962 * make sure that the next entry can hold this entry's length in
963 * its prevlen field. */
964 int forcelarge = 0;
965 nextdiff = (p[0] != ZIP_END) ? zipPrevLenByteDiff(p,reqlen) : 0;
966 if (nextdiff == -4 && reqlen < 4) {
967 nextdiff = 0;
968 forcelarge = 1;
969 }
970
971 /* Store offset because a realloc may change the address of zl. */
972 offset = p-zl;
973 newlen = curlen+reqlen+nextdiff;
974 zl = ziplistResize(zl,newlen);
975 p = zl+offset;
976
977 /* Apply memory move when necessary and update tail offset. */
978 if (p[0] != ZIP_END) {
979 /* Subtract one because of the ZIP_END bytes */
980 memmove(p+reqlen,p-nextdiff,curlen-offset-1+nextdiff);
981
982 /* Encode this entry's raw length in the next entry. */
983 if (forcelarge)
984 zipStorePrevEntryLengthLarge(p+reqlen,reqlen);
985 else
986 zipStorePrevEntryLength(p+reqlen,reqlen);
987
988 /* Update offset for tail */
989 ZIPLIST_TAIL_OFFSET(zl) =
990 intrev32ifbe(intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))+reqlen);
991
992 /* When the tail contains more than one entry, we need to take
993 * "nextdiff" in account as well. Otherwise, a change in the
994 * size of prevlen doesn't have an effect on the *tail* offset. */
995 assert(zipEntrySafe(zl, newlen, p+reqlen, &tail, 1));
996 if (p[reqlen+tail.headersize+tail.len] != ZIP_END) {
997 ZIPLIST_TAIL_OFFSET(zl) =
998 intrev32ifbe(intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))+nextdiff);
999 }
1000 } else {
1001 /* This element will be the new tail. */
1002 ZIPLIST_TAIL_OFFSET(zl) = intrev32ifbe(p-zl);
1003 }
1004
1005 /* When nextdiff != 0, the raw length of the next entry has changed, so
1006 * we need to cascade the update throughout the ziplist */
1007 if (nextdiff != 0) {
1008 offset = p-zl;
1009 zl = __ziplistCascadeUpdate(zl,p+reqlen);
1010 p = zl+offset;
1011 }
1012
1013 /* Write the entry */
1014 p += zipStorePrevEntryLength(p,prevlen);
1015 p += zipStoreEntryEncoding(p,encoding,slen);
1016 if (ZIP_IS_STR(encoding)) {
1017 memcpy(p,s,slen);
1018 } else {
1019 zipSaveInteger(p,value,encoding);
1020 }
1021 ZIPLIST_INCR_LENGTH(zl,1);
1022 return zl;
1023}
1024
1025/* Merge ziplists 'first' and 'second' by appending 'second' to 'first'.
1026 *
1027 * NOTE: The larger ziplist is reallocated to contain the new merged ziplist.
1028 * Either 'first' or 'second' can be used for the result. The parameter not
1029 * used will be free'd and set to NULL.
1030 *
1031 * After calling this function, the input parameters are no longer valid since
1032 * they are changed and free'd in-place.
1033 *
1034 * The result ziplist is the contents of 'first' followed by 'second'.
1035 *
1036 * On failure: returns NULL if the merge is impossible.
1037 * On success: returns the merged ziplist (which is expanded version of either
1038 * 'first' or 'second', also frees the other unused input ziplist, and sets the
1039 * input ziplist argument equal to newly reallocated ziplist return value. */
1040unsigned char *ziplistMerge(unsigned char **first, unsigned char **second) {
1041 /* If any params are null, we can't merge, so NULL. */
1042 if (first == NULL || *first == NULL || second == NULL || *second == NULL)
1043 return NULL;
1044
1045 /* Can't merge same list into itself. */
1046 if (*first == *second)
1047 return NULL;
1048
1049 size_t first_bytes = intrev32ifbe(ZIPLIST_BYTES(*first));
1050 size_t first_len = intrev16ifbe(ZIPLIST_LENGTH(*first));
1051
1052 size_t second_bytes = intrev32ifbe(ZIPLIST_BYTES(*second));
1053 size_t second_len = intrev16ifbe(ZIPLIST_LENGTH(*second));
1054
1055 int append;
1056 unsigned char *source, *target;
1057 size_t target_bytes, source_bytes;
1058 /* Pick the largest ziplist so we can resize easily in-place.
1059 * We must also track if we are now appending or prepending to
1060 * the target ziplist. */
1061 if (first_len >= second_len) {
1062 /* retain first, append second to first. */
1063 target = *first;
1064 target_bytes = first_bytes;
1065 source = *second;
1066 source_bytes = second_bytes;
1067 append = 1;
1068 } else {
1069 /* else, retain second, prepend first to second. */
1070 target = *second;
1071 target_bytes = second_bytes;
1072 source = *first;
1073 source_bytes = first_bytes;
1074 append = 0;
1075 }
1076
1077 /* Calculate final bytes (subtract one pair of metadata) */
1078 size_t zlbytes = first_bytes + second_bytes -
1079 ZIPLIST_HEADER_SIZE - ZIPLIST_END_SIZE;
1080 size_t zllength = first_len + second_len;
1081
1082 /* Combined zl length should be limited within UINT16_MAX */
1083 zllength = zllength < UINT16_MAX ? zllength : UINT16_MAX;
1084
1085 /* larger values can't be stored into ZIPLIST_BYTES */
1086 assert(zlbytes < UINT32_MAX);
1087
1088 /* Save offset positions before we start ripping memory apart. */
1089 size_t first_offset = intrev32ifbe(ZIPLIST_TAIL_OFFSET(*first));
1090 size_t second_offset = intrev32ifbe(ZIPLIST_TAIL_OFFSET(*second));
1091
1092 /* Extend target to new zlbytes then append or prepend source. */
1093 target = zrealloc(target, zlbytes);
1094 if (append) {
1095 /* append == appending to target */
1096 /* Copy source after target (copying over original [END]):
1097 * [TARGET - END, SOURCE - HEADER] */
1098 memcpy(target + target_bytes - ZIPLIST_END_SIZE,
1099 source + ZIPLIST_HEADER_SIZE,
1100 source_bytes - ZIPLIST_HEADER_SIZE);
1101 } else {
1102 /* !append == prepending to target */
1103 /* Move target *contents* exactly size of (source - [END]),
1104 * then copy source into vacated space (source - [END]):
1105 * [SOURCE - END, TARGET - HEADER] */
1106 memmove(target + source_bytes - ZIPLIST_END_SIZE,
1107 target + ZIPLIST_HEADER_SIZE,
1108 target_bytes - ZIPLIST_HEADER_SIZE);
1109 memcpy(target, source, source_bytes - ZIPLIST_END_SIZE);
1110 }
1111
1112 /* Update header metadata. */
1113 ZIPLIST_BYTES(target) = intrev32ifbe(zlbytes);
1114 ZIPLIST_LENGTH(target) = intrev16ifbe(zllength);
1115 /* New tail offset is:
1116 * + N bytes of first ziplist
1117 * - 1 byte for [END] of first ziplist
1118 * + M bytes for the offset of the original tail of the second ziplist
1119 * - J bytes for HEADER because second_offset keeps no header. */
1120 ZIPLIST_TAIL_OFFSET(target) = intrev32ifbe(
1121 (first_bytes - ZIPLIST_END_SIZE) +
1122 (second_offset - ZIPLIST_HEADER_SIZE));
1123
1124 /* __ziplistCascadeUpdate just fixes the prev length values until it finds a
1125 * correct prev length value (then it assumes the rest of the list is okay).
1126 * We tell CascadeUpdate to start at the first ziplist's tail element to fix
1127 * the merge seam. */
1128 target = __ziplistCascadeUpdate(target, target+first_offset);
1129
1130 /* Now free and NULL out what we didn't realloc */
1131 if (append) {
1132 zfree(*second);
1133 *second = NULL;
1134 *first = target;
1135 } else {
1136 zfree(*first);
1137 *first = NULL;
1138 *second = target;
1139 }
1140 return target;
1141}
1142
1143unsigned char *ziplistPush(unsigned char *zl, unsigned char *s, unsigned int slen, int where) {
1144 unsigned char *p;
1145 p = (where == ZIPLIST_HEAD) ? ZIPLIST_ENTRY_HEAD(zl) : ZIPLIST_ENTRY_END(zl);
1146 return __ziplistInsert(zl,p,s,slen);
1147}
1148
1149/* Returns an offset to use for iterating with ziplistNext. When the given
1150 * index is negative, the list is traversed back to front. When the list
1151 * doesn't contain an element at the provided index, NULL is returned. */
1152unsigned char *ziplistIndex(unsigned char *zl, int index) {
1153 unsigned char *p;
1154 unsigned int prevlensize, prevlen = 0;
1155 size_t zlbytes = intrev32ifbe(ZIPLIST_BYTES(zl));
1156 if (index < 0) {
1157 index = (-index)-1;
1158 p = ZIPLIST_ENTRY_TAIL(zl);
1159 if (p[0] != ZIP_END) {
1160 /* No need for "safe" check: when going backwards, we know the header
1161 * we're parsing is in the range, we just need to assert (below) that
1162 * the size we take doesn't cause p to go outside the allocation. */
1163 ZIP_DECODE_PREVLENSIZE(p, prevlensize);
1164 assert(p + prevlensize < zl + zlbytes - ZIPLIST_END_SIZE);
1165 ZIP_DECODE_PREVLEN(p, prevlensize, prevlen);
1166 while (prevlen > 0 && index--) {
1167 p -= prevlen;
1168 assert(p >= zl + ZIPLIST_HEADER_SIZE && p < zl + zlbytes - ZIPLIST_END_SIZE);
1169 ZIP_DECODE_PREVLEN(p, prevlensize, prevlen);
1170 }
1171 }
1172 } else {
1173 p = ZIPLIST_ENTRY_HEAD(zl);
1174 while (index--) {
1175 /* Use the "safe" length: When we go forward, we need to be careful
1176 * not to decode an entry header if it's past the ziplist allocation. */
1177 p += zipRawEntryLengthSafe(zl, zlbytes, p);
1178 if (p[0] == ZIP_END)
1179 break;
1180 }
1181 }
1182 if (p[0] == ZIP_END || index > 0)
1183 return NULL;
1184 zipAssertValidEntry(zl, zlbytes, p);
1185 return p;
1186}
1187
1188/* Return pointer to next entry in ziplist.
1189 *
1190 * zl is the pointer to the ziplist
1191 * p is the pointer to the current element
1192 *
1193 * The element after 'p' is returned, otherwise NULL if we are at the end. */
1194unsigned char *ziplistNext(unsigned char *zl, unsigned char *p) {
1195 ((void) zl);
1196 size_t zlbytes = intrev32ifbe(ZIPLIST_BYTES(zl));
1197
1198 /* "p" could be equal to ZIP_END, caused by ziplistDelete,
1199 * and we should return NULL. Otherwise, we should return NULL
1200 * when the *next* element is ZIP_END (there is no next entry). */
1201 if (p[0] == ZIP_END) {
1202 return NULL;
1203 }
1204
1205 p += zipRawEntryLength(p);
1206 if (p[0] == ZIP_END) {
1207 return NULL;
1208 }
1209
1210 zipAssertValidEntry(zl, zlbytes, p);
1211 return p;
1212}
1213
1214/* Return pointer to previous entry in ziplist. */
1215unsigned char *ziplistPrev(unsigned char *zl, unsigned char *p) {
1216 unsigned int prevlensize, prevlen = 0;
1217
1218 /* Iterating backwards from ZIP_END should return the tail. When "p" is
1219 * equal to the first element of the list, we're already at the head,
1220 * and should return NULL. */
1221 if (p[0] == ZIP_END) {
1222 p = ZIPLIST_ENTRY_TAIL(zl);
1223 return (p[0] == ZIP_END) ? NULL : p;
1224 } else if (p == ZIPLIST_ENTRY_HEAD(zl)) {
1225 return NULL;
1226 } else {
1227 ZIP_DECODE_PREVLEN(p, prevlensize, prevlen);
1228 assert(prevlen > 0);
1229 p-=prevlen;
1230 size_t zlbytes = intrev32ifbe(ZIPLIST_BYTES(zl));
1231 zipAssertValidEntry(zl, zlbytes, p);
1232 return p;
1233 }
1234}
1235
1236/* Get entry pointed to by 'p' and store in either '*sstr' or 'sval' depending
1237 * on the encoding of the entry. '*sstr' is always set to NULL to be able
1238 * to find out whether the string pointer or the integer value was set.
1239 * Return 0 if 'p' points to the end of the ziplist, 1 otherwise. */
1240unsigned int ziplistGet(unsigned char *p, unsigned char **sstr, unsigned int *slen, long long *sval) {
1241 zlentry entry;
1242 if (p == NULL || p[0] == ZIP_END) return 0;
1243 if (sstr) *sstr = NULL;
1244
1245 zipEntry(p, &entry); /* no need for "safe" variant since the input pointer was validated by the function that returned it. */
1246 if (ZIP_IS_STR(entry.encoding)) {
1247 if (sstr) {
1248 *slen = entry.len;
1249 *sstr = p+entry.headersize;
1250 }
1251 } else {
1252 if (sval) {
1253 *sval = zipLoadInteger(p+entry.headersize,entry.encoding);
1254 }
1255 }
1256 return 1;
1257}
1258
1259/* Insert an entry at "p". */
1260unsigned char *ziplistInsert(unsigned char *zl, unsigned char *p, unsigned char *s, unsigned int slen) {
1261 return __ziplistInsert(zl,p,s,slen);
1262}
1263
1264/* Delete a single entry from the ziplist, pointed to by *p.
1265 * Also update *p in place, to be able to iterate over the
1266 * ziplist, while deleting entries. */
1267unsigned char *ziplistDelete(unsigned char *zl, unsigned char **p) {
1268 size_t offset = *p-zl;
1269 zl = __ziplistDelete(zl,*p,1);
1270
1271 /* Store pointer to current element in p, because ziplistDelete will
1272 * do a realloc which might result in a different "zl"-pointer.
1273 * When the delete direction is back to front, we might delete the last
1274 * entry and end up with "p" pointing to ZIP_END, so check this. */
1275 *p = zl+offset;
1276 return zl;
1277}
1278
1279/* Delete a range of entries from the ziplist. */
1280unsigned char *ziplistDeleteRange(unsigned char *zl, int index, unsigned int num) {
1281 unsigned char *p = ziplistIndex(zl,index);
1282 return (p == NULL) ? zl : __ziplistDelete(zl,p,num);
1283}
1284
1285/* Replaces the entry at p. This is equivalent to a delete and an insert,
1286 * but avoids some overhead when replacing a value of the same size. */
1287unsigned char *ziplistReplace(unsigned char *zl, unsigned char *p, unsigned char *s, unsigned int slen) {
1288
1289 /* get metadata of the current entry */
1290 zlentry entry;
1291 zipEntry(p, &entry);
1292
1293 /* compute length of entry to store, excluding prevlen */
1294 unsigned int reqlen;
1295 unsigned char encoding = 0;
1296 long long value = 123456789; /* initialized to avoid warning. */
1297 if (zipTryEncoding(s,slen,&value,&encoding)) {
1298 reqlen = zipIntSize(encoding); /* encoding is set */
1299 } else {
1300 reqlen = slen; /* encoding == 0 */
1301 }
1302 reqlen += zipStoreEntryEncoding(NULL,encoding,slen);
1303
1304 if (reqlen == entry.lensize + entry.len) {
1305 /* Simply overwrite the element. */
1306 p += entry.prevrawlensize;
1307 p += zipStoreEntryEncoding(p,encoding,slen);
1308 if (ZIP_IS_STR(encoding)) {
1309 memcpy(p,s,slen);
1310 } else {
1311 zipSaveInteger(p,value,encoding);
1312 }
1313 } else {
1314 /* Fallback. */
1315 zl = ziplistDelete(zl,&p);
1316 zl = ziplistInsert(zl,p,s,slen);
1317 }
1318 return zl;
1319}
1320
1321/* Compare entry pointer to by 'p' with 'sstr' of length 'slen'. */
1322/* Return 1 if equal. */
1323unsigned int ziplistCompare(unsigned char *p, unsigned char *sstr, unsigned int slen) {
1324 zlentry entry;
1325 unsigned char sencoding;
1326 long long zval, sval;
1327 if (p[0] == ZIP_END) return 0;
1328
1329 zipEntry(p, &entry); /* no need for "safe" variant since the input pointer was validated by the function that returned it. */
1330 if (ZIP_IS_STR(entry.encoding)) {
1331 /* Raw compare */
1332 if (entry.len == slen) {
1333 return memcmp(p+entry.headersize,sstr,slen) == 0;
1334 } else {
1335 return 0;
1336 }
1337 } else {
1338 /* Try to compare encoded values. Don't compare encoding because
1339 * different implementations may encoded integers differently. */
1340 if (zipTryEncoding(sstr,slen,&sval,&sencoding)) {
1341 zval = zipLoadInteger(p+entry.headersize,entry.encoding);
1342 return zval == sval;
1343 }
1344 }
1345 return 0;
1346}
1347
1348/* Find pointer to the entry equal to the specified entry. Skip 'skip' entries
1349 * between every comparison. Returns NULL when the field could not be found. */
1350unsigned char *ziplistFind(unsigned char *zl, unsigned char *p, unsigned char *vstr, unsigned int vlen, unsigned int skip) {
1351 int skipcnt = 0;
1352 unsigned char vencoding = 0;
1353 long long vll = 0;
1354 size_t zlbytes = ziplistBlobLen(zl);
1355
1356 while (p[0] != ZIP_END) {
1357 struct zlentry e;
1358 unsigned char *q;
1359
1360 assert(zipEntrySafe(zl, zlbytes, p, &e, 1));
1361 q = p + e.prevrawlensize + e.lensize;
1362
1363 if (skipcnt == 0) {
1364 /* Compare current entry with specified entry */
1365 if (ZIP_IS_STR(e.encoding)) {
1366 if (e.len == vlen && memcmp(q, vstr, vlen) == 0) {
1367 return p;
1368 }
1369 } else {
1370 /* Find out if the searched field can be encoded. Note that
1371 * we do it only the first time, once done vencoding is set
1372 * to non-zero and vll is set to the integer value. */
1373 if (vencoding == 0) {
1374 if (!zipTryEncoding(vstr, vlen, &vll, &vencoding)) {
1375 /* If the entry can't be encoded we set it to
1376 * UCHAR_MAX so that we don't retry again the next
1377 * time. */
1378 vencoding = UCHAR_MAX;
1379 }
1380 /* Must be non-zero by now */
1381 assert(vencoding);
1382 }
1383
1384 /* Compare current entry with specified entry, do it only
1385 * if vencoding != UCHAR_MAX because if there is no encoding
1386 * possible for the field it can't be a valid integer. */
1387 if (vencoding != UCHAR_MAX) {
1388 long long ll = zipLoadInteger(q, e.encoding);
1389 if (ll == vll) {
1390 return p;
1391 }
1392 }
1393 }
1394
1395 /* Reset skip count */
1396 skipcnt = skip;
1397 } else {
1398 /* Skip entry */
1399 skipcnt--;
1400 }
1401
1402 /* Move to next entry */
1403 p = q + e.len;
1404 }
1405
1406 return NULL;
1407}
1408
1409/* Return length of ziplist. */
1410unsigned int ziplistLen(unsigned char *zl) {
1411 unsigned int len = 0;
1412 if (intrev16ifbe(ZIPLIST_LENGTH(zl)) < UINT16_MAX) {
1413 len = intrev16ifbe(ZIPLIST_LENGTH(zl));
1414 } else {
1415 unsigned char *p = zl+ZIPLIST_HEADER_SIZE;
1416 size_t zlbytes = intrev32ifbe(ZIPLIST_BYTES(zl));
1417 while (*p != ZIP_END) {
1418 p += zipRawEntryLengthSafe(zl, zlbytes, p);
1419 len++;
1420 }
1421
1422 /* Re-store length if small enough */
1423 if (len < UINT16_MAX) ZIPLIST_LENGTH(zl) = intrev16ifbe(len);
1424 }
1425 return len;
1426}
1427
1428/* Return ziplist blob size in bytes. */
1429size_t ziplistBlobLen(unsigned char *zl) {
1430 return intrev32ifbe(ZIPLIST_BYTES(zl));
1431}
1432
1433void ziplistRepr(unsigned char *zl) {
1434 unsigned char *p;
1435 int index = 0;
1436 zlentry entry;
1437 size_t zlbytes = ziplistBlobLen(zl);
1438
1439 printf(
1440 "{total bytes %u} "
1441 "{num entries %u}\n"
1442 "{tail offset %u}\n",
1443 intrev32ifbe(ZIPLIST_BYTES(zl)),
1444 intrev16ifbe(ZIPLIST_LENGTH(zl)),
1445 intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl)));
1446 p = ZIPLIST_ENTRY_HEAD(zl);
1447 while(*p != ZIP_END) {
1448 assert(zipEntrySafe(zl, zlbytes, p, &entry, 1));
1449 printf(
1450 "{\n"
1451 "\taddr 0x%08lx,\n"
1452 "\tindex %2d,\n"
1453 "\toffset %5lu,\n"
1454 "\thdr+entry len: %5u,\n"
1455 "\thdr len%2u,\n"
1456 "\tprevrawlen: %5u,\n"
1457 "\tprevrawlensize: %2u,\n"
1458 "\tpayload %5u\n",
1459 (long unsigned)p,
1460 index,
1461 (unsigned long) (p-zl),
1462 entry.headersize+entry.len,
1463 entry.headersize,
1464 entry.prevrawlen,
1465 entry.prevrawlensize,
1466 entry.len);
1467 printf("\tbytes: ");
1468 for (unsigned int i = 0; i < entry.headersize+entry.len; i++) {
1469 printf("%02x|",p[i]);
1470 }
1471 printf("\n");
1472 p += entry.headersize;
1473 if (ZIP_IS_STR(entry.encoding)) {
1474 printf("\t[str]");
1475 if (entry.len > 40) {
1476 if (fwrite(p,40,1,stdout) == 0) perror("fwrite");
1477 printf("...");
1478 } else {
1479 if (entry.len &&
1480 fwrite(p,entry.len,1,stdout) == 0) perror("fwrite");
1481 }
1482 } else {
1483 printf("\t[int]%lld", (long long) zipLoadInteger(p,entry.encoding));
1484 }
1485 printf("\n}\n");
1486 p += entry.len;
1487 index++;
1488 }
1489 printf("{end}\n\n");
1490}
1491
1492/* Validate the integrity of the data structure.
1493 * when `deep` is 0, only the integrity of the header is validated.
1494 * when `deep` is 1, we scan all the entries one by one. */
1495int ziplistValidateIntegrity(unsigned char *zl, size_t size, int deep,
1496 ziplistValidateEntryCB entry_cb, void *cb_userdata) {
1497 /* check that we can actually read the header. (and ZIP_END) */
1498 if (size < ZIPLIST_HEADER_SIZE + ZIPLIST_END_SIZE)
1499 return 0;
1500
1501 /* check that the encoded size in the header must match the allocated size. */
1502 size_t bytes = intrev32ifbe(ZIPLIST_BYTES(zl));
1503 if (bytes != size)
1504 return 0;
1505
1506 /* the last byte must be the terminator. */
1507 if (zl[size - ZIPLIST_END_SIZE] != ZIP_END)
1508 return 0;
1509
1510 /* make sure the tail offset isn't reaching outside the allocation. */
1511 if (intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl)) > size - ZIPLIST_END_SIZE)
1512 return 0;
1513
1514 if (!deep)
1515 return 1;
1516
1517 unsigned int count = 0;
1518 unsigned int header_count = intrev16ifbe(ZIPLIST_LENGTH(zl));
1519 unsigned char *p = ZIPLIST_ENTRY_HEAD(zl);
1520 unsigned char *prev = NULL;
1521 size_t prev_raw_size = 0;
1522 while(*p != ZIP_END) {
1523 struct zlentry e;
1524 /* Decode the entry headers and fail if invalid or reaches outside the allocation */
1525 if (!zipEntrySafe(zl, size, p, &e, 1))
1526 return 0;
1527
1528 /* Make sure the record stating the prev entry size is correct. */
1529 if (e.prevrawlen != prev_raw_size)
1530 return 0;
1531
1532 /* Optionally let the caller validate the entry too. */
1533 if (entry_cb && !entry_cb(p, header_count, cb_userdata))
1534 return 0;
1535
1536 /* Move to the next entry */
1537 prev_raw_size = e.headersize + e.len;
1538 prev = p;
1539 p += e.headersize + e.len;
1540 count++;
1541 }
1542
1543 /* Make sure 'p' really does point to the end of the ziplist. */
1544 if (p != zl + bytes - ZIPLIST_END_SIZE)
1545 return 0;
1546
1547 /* Make sure the <zltail> entry really do point to the start of the last entry. */
1548 if (prev != NULL && prev != ZIPLIST_ENTRY_TAIL(zl))
1549 return 0;
1550
1551 /* Check that the count in the header is correct */
1552 if (header_count != UINT16_MAX && count != header_count)
1553 return 0;
1554
1555 return 1;
1556}
1557
1558/* Randomly select a pair of key and value.
1559 * total_count is a pre-computed length/2 of the ziplist (to avoid calls to ziplistLen)
1560 * 'key' and 'val' are used to store the result key value pair.
1561 * 'val' can be NULL if the value is not needed. */
1562void ziplistRandomPair(unsigned char *zl, unsigned long total_count, ziplistEntry *key, ziplistEntry *val) {
1563 int ret;
1564 unsigned char *p;
1565
1566 /* Avoid div by zero on corrupt ziplist */
1567 assert(total_count);
1568
1569 /* Generate even numbers, because ziplist saved K-V pair */
1570 int r = (rand() % total_count) * 2;
1571 p = ziplistIndex(zl, r);
1572 ret = ziplistGet(p, &key->sval, &key->slen, &key->lval);
1573 assert(ret != 0);
1574
1575 if (!val)
1576 return;
1577 p = ziplistNext(zl, p);
1578 ret = ziplistGet(p, &val->sval, &val->slen, &val->lval);
1579 assert(ret != 0);
1580}
1581
1582/* int compare for qsort */
1583int uintCompare(const void *a, const void *b) {
1584 return (*(unsigned int *) a - *(unsigned int *) b);
1585}
1586
1587/* Helper method to store a string into from val or lval into dest */
1588static inline void ziplistSaveValue(unsigned char *val, unsigned int len, long long lval, ziplistEntry *dest) {
1589 dest->sval = val;
1590 dest->slen = len;
1591 dest->lval = lval;
1592}
1593
1594/* Randomly select count of key value pairs and store into 'keys' and
1595 * 'vals' args. The order of the picked entries is random, and the selections
1596 * are non-unique (repetitions are possible).
1597 * The 'vals' arg can be NULL in which case we skip these. */
1598void ziplistRandomPairs(unsigned char *zl, unsigned int count, ziplistEntry *keys, ziplistEntry *vals) {
1599 unsigned char *p, *key, *value;
1600 unsigned int klen = 0, vlen = 0;
1601 long long klval = 0, vlval = 0;
1602
1603 /* Notice: the index member must be first due to the use in uintCompare */
1604 typedef struct {
1605 unsigned int index;
1606 unsigned int order;
1607 } rand_pick;
1608 rand_pick *picks = zmalloc(sizeof(rand_pick)*count);
1609 unsigned int total_size = ziplistLen(zl)/2;
1610
1611 /* Avoid div by zero on corrupt ziplist */
1612 assert(total_size);
1613
1614 /* create a pool of random indexes (some may be duplicate). */
1615 for (unsigned int i = 0; i < count; i++) {
1616 picks[i].index = (rand() % total_size) * 2; /* Generate even indexes */
1617 /* keep track of the order we picked them */
1618 picks[i].order = i;
1619 }
1620
1621 /* sort by indexes. */
1622 qsort(picks, count, sizeof(rand_pick), uintCompare);
1623
1624 /* fetch the elements form the ziplist into a output array respecting the original order. */
1625 unsigned int zipindex = picks[0].index, pickindex = 0;
1626 p = ziplistIndex(zl, zipindex);
1627 while (ziplistGet(p, &key, &klen, &klval) && pickindex < count) {
1628 p = ziplistNext(zl, p);
1629 assert(ziplistGet(p, &value, &vlen, &vlval));
1630 while (pickindex < count && zipindex == picks[pickindex].index) {
1631 int storeorder = picks[pickindex].order;
1632 ziplistSaveValue(key, klen, klval, &keys[storeorder]);
1633 if (vals)
1634 ziplistSaveValue(value, vlen, vlval, &vals[storeorder]);
1635 pickindex++;
1636 }
1637 zipindex += 2;
1638 p = ziplistNext(zl, p);
1639 }
1640
1641 zfree(picks);
1642}
1643
1644/* Randomly select count of key value pairs and store into 'keys' and
1645 * 'vals' args. The selections are unique (no repetitions), and the order of
1646 * the picked entries is NOT-random.
1647 * The 'vals' arg can be NULL in which case we skip these.
1648 * The return value is the number of items picked which can be lower than the
1649 * requested count if the ziplist doesn't hold enough pairs. */
1650unsigned int ziplistRandomPairsUnique(unsigned char *zl, unsigned int count, ziplistEntry *keys, ziplistEntry *vals) {
1651 unsigned char *p, *key;
1652 unsigned int klen = 0;
1653 long long klval = 0;
1654 unsigned int total_size = ziplistLen(zl)/2;
1655 unsigned int index = 0;
1656 if (count > total_size)
1657 count = total_size;
1658
1659 /* To only iterate once, every time we try to pick a member, the probability
1660 * we pick it is the quotient of the count left we want to pick and the
1661 * count still we haven't visited in the dict, this way, we could make every
1662 * member be equally picked.*/
1663 p = ziplistIndex(zl, 0);
1664 unsigned int picked = 0, remaining = count;
1665 while (picked < count && p) {
1666 double randomDouble = ((double)rand()) / RAND_MAX;
1667 double threshold = ((double)remaining) / (total_size - index);
1668 if (randomDouble <= threshold) {
1669 assert(ziplistGet(p, &key, &klen, &klval));
1670 ziplistSaveValue(key, klen, klval, &keys[picked]);
1671 p = ziplistNext(zl, p);
1672 assert(p);
1673 if (vals) {
1674 assert(ziplistGet(p, &key, &klen, &klval));
1675 ziplistSaveValue(key, klen, klval, &vals[picked]);
1676 }
1677 remaining--;
1678 picked++;
1679 } else {
1680 p = ziplistNext(zl, p);
1681 assert(p);
1682 }
1683 p = ziplistNext(zl, p);
1684 index++;
1685 }
1686 return picked;
1687}
1688
1689#ifdef REDIS_TEST
1690#include <sys/time.h>
1691#include "adlist.h"
1692#include "sds.h"
1693#include "testhelp.h"
1694
1695#define debug(f, ...) { if (DEBUG) printf(f, __VA_ARGS__); }
1696
1697static unsigned char *createList() {
1698 unsigned char *zl = ziplistNew();
1699 zl = ziplistPush(zl, (unsigned char*)"foo", 3, ZIPLIST_TAIL);
1700 zl = ziplistPush(zl, (unsigned char*)"quux", 4, ZIPLIST_TAIL);
1701 zl = ziplistPush(zl, (unsigned char*)"hello", 5, ZIPLIST_HEAD);
1702 zl = ziplistPush(zl, (unsigned char*)"1024", 4, ZIPLIST_TAIL);
1703 return zl;
1704}
1705
1706static unsigned char *createIntList() {
1707 unsigned char *zl = ziplistNew();
1708 char buf[32];
1709
1710 sprintf(buf, "100");
1711 zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL);
1712 sprintf(buf, "128000");
1713 zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL);
1714 sprintf(buf, "-100");
1715 zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_HEAD);
1716 sprintf(buf, "4294967296");
1717 zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_HEAD);
1718 sprintf(buf, "non integer");
1719 zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL);
1720 sprintf(buf, "much much longer non integer");
1721 zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL);
1722 return zl;
1723}
1724
1725static long long usec(void) {
1726 struct timeval tv;
1727 gettimeofday(&tv,NULL);
1728 return (((long long)tv.tv_sec)*1000000)+tv.tv_usec;
1729}
1730
1731static void stress(int pos, int num, int maxsize, int dnum) {
1732 int i,j,k;
1733 unsigned char *zl;
1734 char posstr[2][5] = { "HEAD", "TAIL" };
1735 long long start;
1736 for (i = 0; i < maxsize; i+=dnum) {
1737 zl = ziplistNew();
1738 for (j = 0; j < i; j++) {
1739 zl = ziplistPush(zl,(unsigned char*)"quux",4,ZIPLIST_TAIL);
1740 }
1741
1742 /* Do num times a push+pop from pos */
1743 start = usec();
1744 for (k = 0; k < num; k++) {
1745 zl = ziplistPush(zl,(unsigned char*)"quux",4,pos);
1746 zl = ziplistDeleteRange(zl,0,1);
1747 }
1748 printf("List size: %8d, bytes: %8d, %dx push+pop (%s): %6lld usec\n",
1749 i,intrev32ifbe(ZIPLIST_BYTES(zl)),num,posstr[pos],usec()-start);
1750 zfree(zl);
1751 }
1752}
1753
1754static unsigned char *pop(unsigned char *zl, int where) {
1755 unsigned char *p, *vstr;
1756 unsigned int vlen;
1757 long long vlong;
1758
1759 p = ziplistIndex(zl,where == ZIPLIST_HEAD ? 0 : -1);
1760 if (ziplistGet(p,&vstr,&vlen,&vlong)) {
1761 if (where == ZIPLIST_HEAD)
1762 printf("Pop head: ");
1763 else
1764 printf("Pop tail: ");
1765
1766 if (vstr) {
1767 if (vlen && fwrite(vstr,vlen,1,stdout) == 0) perror("fwrite");
1768 }
1769 else {
1770 printf("%lld", vlong);
1771 }
1772
1773 printf("\n");
1774 return ziplistDelete(zl,&p);
1775 } else {
1776 printf("ERROR: Could not pop\n");
1777 exit(1);
1778 }
1779}
1780
1781static int randstring(char *target, unsigned int min, unsigned int max) {
1782 int p = 0;
1783 int len = min+rand()%(max-min+1);
1784 int minval, maxval;
1785 switch(rand() % 3) {
1786 case 0:
1787 minval = 0;
1788 maxval = 255;
1789 break;
1790 case 1:
1791 minval = 48;
1792 maxval = 122;
1793 break;
1794 case 2:
1795 minval = 48;
1796 maxval = 52;
1797 break;
1798 default:
1799 assert(NULL);
1800 }
1801
1802 while(p < len)
1803 target[p++] = minval+rand()%(maxval-minval+1);
1804 return len;
1805}
1806
1807static void verify(unsigned char *zl, zlentry *e) {
1808 int len = ziplistLen(zl);
1809 zlentry _e;
1810
1811 ZIPLIST_ENTRY_ZERO(&_e);
1812
1813 for (int i = 0; i < len; i++) {
1814 memset(&e[i], 0, sizeof(zlentry));
1815 zipEntry(ziplistIndex(zl, i), &e[i]);
1816
1817 memset(&_e, 0, sizeof(zlentry));
1818 zipEntry(ziplistIndex(zl, -len+i), &_e);
1819
1820 assert(memcmp(&e[i], &_e, sizeof(zlentry)) == 0);
1821 }
1822}
1823
1824static unsigned char *insertHelper(unsigned char *zl, char ch, size_t len, unsigned char *pos) {
1825 assert(len <= ZIP_BIG_PREVLEN);
1826 unsigned char data[ZIP_BIG_PREVLEN] = {0};
1827 memset(data, ch, len);
1828 return ziplistInsert(zl, pos, data, len);
1829}
1830
1831static int compareHelper(unsigned char *zl, char ch, size_t len, int index) {
1832 assert(len <= ZIP_BIG_PREVLEN);
1833 unsigned char data[ZIP_BIG_PREVLEN] = {0};
1834 memset(data, ch, len);
1835 unsigned char *p = ziplistIndex(zl, index);
1836 assert(p != NULL);
1837 return ziplistCompare(p, data, len);
1838}
1839
1840static size_t strEntryBytesSmall(size_t slen) {
1841 return slen + zipStorePrevEntryLength(NULL, 0) + zipStoreEntryEncoding(NULL, 0, slen);
1842}
1843
1844static size_t strEntryBytesLarge(size_t slen) {
1845 return slen + zipStorePrevEntryLength(NULL, ZIP_BIG_PREVLEN) + zipStoreEntryEncoding(NULL, 0, slen);
1846}
1847
1848/* ./redis-server test ziplist <randomseed> */
1849int ziplistTest(int argc, char **argv, int flags) {
1850 int accurate = (flags & REDIS_TEST_ACCURATE);
1851 unsigned char *zl, *p;
1852 unsigned char *entry;
1853 unsigned int elen;
1854 long long value;
1855 int iteration;
1856
1857 /* If an argument is given, use it as the random seed. */
1858 if (argc >= 4)
1859 srand(atoi(argv[3]));
1860
1861 zl = createIntList();
1862 ziplistRepr(zl);
1863
1864 zfree(zl);
1865
1866 zl = createList();
1867 ziplistRepr(zl);
1868
1869 zl = pop(zl,ZIPLIST_TAIL);
1870 ziplistRepr(zl);
1871
1872 zl = pop(zl,ZIPLIST_HEAD);
1873 ziplistRepr(zl);
1874
1875 zl = pop(zl,ZIPLIST_TAIL);
1876 ziplistRepr(zl);
1877
1878 zl = pop(zl,ZIPLIST_TAIL);
1879 ziplistRepr(zl);
1880
1881 zfree(zl);
1882
1883 printf("Get element at index 3:\n");
1884 {
1885 zl = createList();
1886 p = ziplistIndex(zl, 3);
1887 if (!ziplistGet(p, &entry, &elen, &value)) {
1888 printf("ERROR: Could not access index 3\n");
1889 return 1;
1890 }
1891 if (entry) {
1892 if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite");
1893 printf("\n");
1894 } else {
1895 printf("%lld\n", value);
1896 }
1897 printf("\n");
1898 zfree(zl);
1899 }
1900
1901 printf("Get element at index 4 (out of range):\n");
1902 {
1903 zl = createList();
1904 p = ziplistIndex(zl, 4);
1905 if (p == NULL) {
1906 printf("No entry\n");
1907 } else {
1908 printf("ERROR: Out of range index should return NULL, returned offset: %ld\n", (long)(p-zl));
1909 return 1;
1910 }
1911 printf("\n");
1912 zfree(zl);
1913 }
1914
1915 printf("Get element at index -1 (last element):\n");
1916 {
1917 zl = createList();
1918 p = ziplistIndex(zl, -1);
1919 if (!ziplistGet(p, &entry, &elen, &value)) {
1920 printf("ERROR: Could not access index -1\n");
1921 return 1;
1922 }
1923 if (entry) {
1924 if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite");
1925 printf("\n");
1926 } else {
1927 printf("%lld\n", value);
1928 }
1929 printf("\n");
1930 zfree(zl);
1931 }
1932
1933 printf("Get element at index -4 (first element):\n");
1934 {
1935 zl = createList();
1936 p = ziplistIndex(zl, -4);
1937 if (!ziplistGet(p, &entry, &elen, &value)) {
1938 printf("ERROR: Could not access index -4\n");
1939 return 1;
1940 }
1941 if (entry) {
1942 if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite");
1943 printf("\n");
1944 } else {
1945 printf("%lld\n", value);
1946 }
1947 printf("\n");
1948 zfree(zl);
1949 }
1950
1951 printf("Get element at index -5 (reverse out of range):\n");
1952 {
1953 zl = createList();
1954 p = ziplistIndex(zl, -5);
1955 if (p == NULL) {
1956 printf("No entry\n");
1957 } else {
1958 printf("ERROR: Out of range index should return NULL, returned offset: %ld\n", (long)(p-zl));
1959 return 1;
1960 }
1961 printf("\n");
1962 zfree(zl);
1963 }
1964
1965 printf("Iterate list from 0 to end:\n");
1966 {
1967 zl = createList();
1968 p = ziplistIndex(zl, 0);
1969 while (ziplistGet(p, &entry, &elen, &value)) {
1970 printf("Entry: ");
1971 if (entry) {
1972 if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite");
1973 } else {
1974 printf("%lld", value);
1975 }
1976 p = ziplistNext(zl,p);
1977 printf("\n");
1978 }
1979 printf("\n");
1980 zfree(zl);
1981 }
1982
1983 printf("Iterate list from 1 to end:\n");
1984 {
1985 zl = createList();
1986 p = ziplistIndex(zl, 1);
1987 while (ziplistGet(p, &entry, &elen, &value)) {
1988 printf("Entry: ");
1989 if (entry) {
1990 if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite");
1991 } else {
1992 printf("%lld", value);
1993 }
1994 p = ziplistNext(zl,p);
1995 printf("\n");
1996 }
1997 printf("\n");
1998 zfree(zl);
1999 }
2000
2001 printf("Iterate list from 2 to end:\n");
2002 {
2003 zl = createList();
2004 p = ziplistIndex(zl, 2);
2005 while (ziplistGet(p, &entry, &elen, &value)) {
2006 printf("Entry: ");
2007 if (entry) {
2008 if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite");
2009 } else {
2010 printf("%lld", value);
2011 }
2012 p = ziplistNext(zl,p);
2013 printf("\n");
2014 }
2015 printf("\n");
2016 zfree(zl);
2017 }
2018
2019 printf("Iterate starting out of range:\n");
2020 {
2021 zl = createList();
2022 p = ziplistIndex(zl, 4);
2023 if (!ziplistGet(p, &entry, &elen, &value)) {
2024 printf("No entry\n");
2025 } else {
2026 printf("ERROR\n");
2027 }
2028 printf("\n");
2029 zfree(zl);
2030 }
2031
2032 printf("Iterate from back to front:\n");
2033 {
2034 zl = createList();
2035 p = ziplistIndex(zl, -1);
2036 while (ziplistGet(p, &entry, &elen, &value)) {
2037 printf("Entry: ");
2038 if (entry) {
2039 if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite");
2040 } else {
2041 printf("%lld", value);
2042 }
2043 p = ziplistPrev(zl,p);
2044 printf("\n");
2045 }
2046 printf("\n");
2047 zfree(zl);
2048 }
2049
2050 printf("Iterate from back to front, deleting all items:\n");
2051 {
2052 zl = createList();
2053 p = ziplistIndex(zl, -1);
2054 while (ziplistGet(p, &entry, &elen, &value)) {
2055 printf("Entry: ");
2056 if (entry) {
2057 if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite");
2058 } else {
2059 printf("%lld", value);
2060 }
2061 zl = ziplistDelete(zl,&p);
2062 p = ziplistPrev(zl,p);
2063 printf("\n");
2064 }
2065 printf("\n");
2066 zfree(zl);
2067 }
2068
2069 printf("Delete inclusive range 0,0:\n");
2070 {
2071 zl = createList();
2072 zl = ziplistDeleteRange(zl, 0, 1);
2073 ziplistRepr(zl);
2074 zfree(zl);
2075 }
2076
2077 printf("Delete inclusive range 0,1:\n");
2078 {
2079 zl = createList();
2080 zl = ziplistDeleteRange(zl, 0, 2);
2081 ziplistRepr(zl);
2082 zfree(zl);
2083 }
2084
2085 printf("Delete inclusive range 1,2:\n");
2086 {
2087 zl = createList();
2088 zl = ziplistDeleteRange(zl, 1, 2);
2089 ziplistRepr(zl);
2090 zfree(zl);
2091 }
2092
2093 printf("Delete with start index out of range:\n");
2094 {
2095 zl = createList();
2096 zl = ziplistDeleteRange(zl, 5, 1);
2097 ziplistRepr(zl);
2098 zfree(zl);
2099 }
2100
2101 printf("Delete with num overflow:\n");
2102 {
2103 zl = createList();
2104 zl = ziplistDeleteRange(zl, 1, 5);
2105 ziplistRepr(zl);
2106 zfree(zl);
2107 }
2108
2109 printf("Delete foo while iterating:\n");
2110 {
2111 zl = createList();
2112 p = ziplistIndex(zl,0);
2113 while (ziplistGet(p,&entry,&elen,&value)) {
2114 if (entry && strncmp("foo",(char*)entry,elen) == 0) {
2115 printf("Delete foo\n");
2116 zl = ziplistDelete(zl,&p);
2117 } else {
2118 printf("Entry: ");
2119 if (entry) {
2120 if (elen && fwrite(entry,elen,1,stdout) == 0)
2121 perror("fwrite");
2122 } else {
2123 printf("%lld",value);
2124 }
2125 p = ziplistNext(zl,p);
2126 printf("\n");
2127 }
2128 }
2129 printf("\n");
2130 ziplistRepr(zl);
2131 zfree(zl);
2132 }
2133
2134 printf("Replace with same size:\n");
2135 {
2136 zl = createList(); /* "hello", "foo", "quux", "1024" */
2137 unsigned char *orig_zl = zl;
2138 p = ziplistIndex(zl, 0);
2139 zl = ziplistReplace(zl, p, (unsigned char*)"zoink", 5);
2140 p = ziplistIndex(zl, 3);
2141 zl = ziplistReplace(zl, p, (unsigned char*)"yy", 2);
2142 p = ziplistIndex(zl, 1);
2143 zl = ziplistReplace(zl, p, (unsigned char*)"65536", 5);
2144 p = ziplistIndex(zl, 0);
2145 assert(!memcmp((char*)p,
2146 "\x00\x05zoink"
2147 "\x07\xf0\x00\x00\x01" /* 65536 as int24 */
2148 "\x05\x04quux" "\x06\x02yy" "\xff",
2149 23));
2150 assert(zl == orig_zl); /* no reallocations have happened */
2151 zfree(zl);
2152 printf("SUCCESS\n\n");
2153 }
2154
2155 printf("Replace with different size:\n");
2156 {
2157 zl = createList(); /* "hello", "foo", "quux", "1024" */
2158 p = ziplistIndex(zl, 1);
2159 zl = ziplistReplace(zl, p, (unsigned char*)"squirrel", 8);
2160 p = ziplistIndex(zl, 0);
2161 assert(!strncmp((char*)p,
2162 "\x00\x05hello" "\x07\x08squirrel" "\x0a\x04quux"
2163 "\x06\xc0\x00\x04" "\xff",
2164 28));
2165 zfree(zl);
2166 printf("SUCCESS\n\n");
2167 }
2168
2169 printf("Regression test for >255 byte strings:\n");
2170 {
2171 char v1[257] = {0}, v2[257] = {0};
2172 memset(v1,'x',256);
2173 memset(v2,'y',256);
2174 zl = ziplistNew();
2175 zl = ziplistPush(zl,(unsigned char*)v1,strlen(v1),ZIPLIST_TAIL);
2176 zl = ziplistPush(zl,(unsigned char*)v2,strlen(v2),ZIPLIST_TAIL);
2177
2178 /* Pop values again and compare their value. */
2179 p = ziplistIndex(zl,0);
2180 assert(ziplistGet(p,&entry,&elen,&value));
2181 assert(strncmp(v1,(char*)entry,elen) == 0);
2182 p = ziplistIndex(zl,1);
2183 assert(ziplistGet(p,&entry,&elen,&value));
2184 assert(strncmp(v2,(char*)entry,elen) == 0);
2185 printf("SUCCESS\n\n");
2186 zfree(zl);
2187 }
2188
2189 printf("Regression test deleting next to last entries:\n");
2190 {
2191 char v[3][257] = {{0}};
2192 zlentry e[3] = {{.prevrawlensize = 0, .prevrawlen = 0, .lensize = 0,
2193 .len = 0, .headersize = 0, .encoding = 0, .p = NULL}};
2194 size_t i;
2195
2196 for (i = 0; i < (sizeof(v)/sizeof(v[0])); i++) {
2197 memset(v[i], 'a' + i, sizeof(v[0]));
2198 }
2199
2200 v[0][256] = '\0';
2201 v[1][ 1] = '\0';
2202 v[2][256] = '\0';
2203
2204 zl = ziplistNew();
2205 for (i = 0; i < (sizeof(v)/sizeof(v[0])); i++) {
2206 zl = ziplistPush(zl, (unsigned char *) v[i], strlen(v[i]), ZIPLIST_TAIL);
2207 }
2208
2209 verify(zl, e);
2210
2211 assert(e[0].prevrawlensize == 1);
2212 assert(e[1].prevrawlensize == 5);
2213 assert(e[2].prevrawlensize == 1);
2214
2215 /* Deleting entry 1 will increase `prevrawlensize` for entry 2 */
2216 unsigned char *p = e[1].p;
2217 zl = ziplistDelete(zl, &p);
2218
2219 verify(zl, e);
2220
2221 assert(e[0].prevrawlensize == 1);
2222 assert(e[1].prevrawlensize == 5);
2223
2224 printf("SUCCESS\n\n");
2225 zfree(zl);
2226 }
2227
2228 printf("Create long list and check indices:\n");
2229 {
2230 unsigned long long start = usec();
2231 zl = ziplistNew();
2232 char buf[32];
2233 int i,len;
2234 for (i = 0; i < 1000; i++) {
2235 len = sprintf(buf,"%d",i);
2236 zl = ziplistPush(zl,(unsigned char*)buf,len,ZIPLIST_TAIL);
2237 }
2238 for (i = 0; i < 1000; i++) {
2239 p = ziplistIndex(zl,i);
2240 assert(ziplistGet(p,NULL,NULL,&value));
2241 assert(i == value);
2242
2243 p = ziplistIndex(zl,-i-1);
2244 assert(ziplistGet(p,NULL,NULL,&value));
2245 assert(999-i == value);
2246 }
2247 printf("SUCCESS. usec=%lld\n\n", usec()-start);
2248 zfree(zl);
2249 }
2250
2251 printf("Compare strings with ziplist entries:\n");
2252 {
2253 zl = createList();
2254 p = ziplistIndex(zl,0);
2255 if (!ziplistCompare(p,(unsigned char*)"hello",5)) {
2256 printf("ERROR: not \"hello\"\n");
2257 return 1;
2258 }
2259 if (ziplistCompare(p,(unsigned char*)"hella",5)) {
2260 printf("ERROR: \"hella\"\n");
2261 return 1;
2262 }
2263
2264 p = ziplistIndex(zl,3);
2265 if (!ziplistCompare(p,(unsigned char*)"1024",4)) {
2266 printf("ERROR: not \"1024\"\n");
2267 return 1;
2268 }
2269 if (ziplistCompare(p,(unsigned char*)"1025",4)) {
2270 printf("ERROR: \"1025\"\n");
2271 return 1;
2272 }
2273 printf("SUCCESS\n\n");
2274 zfree(zl);
2275 }
2276
2277 printf("Merge test:\n");
2278 {
2279 /* create list gives us: [hello, foo, quux, 1024] */
2280 zl = createList();
2281 unsigned char *zl2 = createList();
2282
2283 unsigned char *zl3 = ziplistNew();
2284 unsigned char *zl4 = ziplistNew();
2285
2286 if (ziplistMerge(&zl4, &zl4)) {
2287 printf("ERROR: Allowed merging of one ziplist into itself.\n");
2288 return 1;
2289 }
2290
2291 /* Merge two empty ziplists, get empty result back. */
2292 zl4 = ziplistMerge(&zl3, &zl4);
2293 ziplistRepr(zl4);
2294 if (ziplistLen(zl4)) {
2295 printf("ERROR: Merging two empty ziplists created entries.\n");
2296 return 1;
2297 }
2298 zfree(zl4);
2299
2300 zl2 = ziplistMerge(&zl, &zl2);
2301 /* merge gives us: [hello, foo, quux, 1024, hello, foo, quux, 1024] */
2302 ziplistRepr(zl2);
2303
2304 if (ziplistLen(zl2) != 8) {
2305 printf("ERROR: Merged length not 8, but: %u\n", ziplistLen(zl2));
2306 return 1;
2307 }
2308
2309 p = ziplistIndex(zl2,0);
2310 if (!ziplistCompare(p,(unsigned char*)"hello",5)) {
2311 printf("ERROR: not \"hello\"\n");
2312 return 1;
2313 }
2314 if (ziplistCompare(p,(unsigned char*)"hella",5)) {
2315 printf("ERROR: \"hella\"\n");
2316 return 1;
2317 }
2318
2319 p = ziplistIndex(zl2,3);
2320 if (!ziplistCompare(p,(unsigned char*)"1024",4)) {
2321 printf("ERROR: not \"1024\"\n");
2322 return 1;
2323 }
2324 if (ziplistCompare(p,(unsigned char*)"1025",4)) {
2325 printf("ERROR: \"1025\"\n");
2326 return 1;
2327 }
2328
2329 p = ziplistIndex(zl2,4);
2330 if (!ziplistCompare(p,(unsigned char*)"hello",5)) {
2331 printf("ERROR: not \"hello\"\n");
2332 return 1;
2333 }
2334 if (ziplistCompare(p,(unsigned char*)"hella",5)) {
2335 printf("ERROR: \"hella\"\n");
2336 return 1;
2337 }
2338
2339 p = ziplistIndex(zl2,7);
2340 if (!ziplistCompare(p,(unsigned char*)"1024",4)) {
2341 printf("ERROR: not \"1024\"\n");
2342 return 1;
2343 }
2344 if (ziplistCompare(p,(unsigned char*)"1025",4)) {
2345 printf("ERROR: \"1025\"\n");
2346 return 1;
2347 }
2348 printf("SUCCESS\n\n");
2349 zfree(zl);
2350 }
2351
2352 printf("Stress with random payloads of different encoding:\n");
2353 {
2354 unsigned long long start = usec();
2355 int i,j,len,where;
2356 unsigned char *p;
2357 char buf[1024];
2358 int buflen;
2359 list *ref;
2360 listNode *refnode;
2361
2362 /* Hold temp vars from ziplist */
2363 unsigned char *sstr;
2364 unsigned int slen;
2365 long long sval;
2366
2367 iteration = accurate ? 20000 : 20;
2368 for (i = 0; i < iteration; i++) {
2369 zl = ziplistNew();
2370 ref = listCreate();
2371 listSetFreeMethod(ref,(void (*)(void*))sdsfree);
2372 len = rand() % 256;
2373
2374 /* Create lists */
2375 for (j = 0; j < len; j++) {
2376 where = (rand() & 1) ? ZIPLIST_HEAD : ZIPLIST_TAIL;
2377 if (rand() % 2) {
2378 buflen = randstring(buf,1,sizeof(buf)-1);
2379 } else {
2380 switch(rand() % 3) {
2381 case 0:
2382 buflen = sprintf(buf,"%lld",(0LL + rand()) >> 20);
2383 break;
2384 case 1:
2385 buflen = sprintf(buf,"%lld",(0LL + rand()));
2386 break;
2387 case 2:
2388 buflen = sprintf(buf,"%lld",(0LL + rand()) << 20);
2389 break;
2390 default:
2391 assert(NULL);
2392 }
2393 }
2394
2395 /* Add to ziplist */
2396 zl = ziplistPush(zl, (unsigned char*)buf, buflen, where);
2397
2398 /* Add to reference list */
2399 if (where == ZIPLIST_HEAD) {
2400 listAddNodeHead(ref,sdsnewlen(buf, buflen));
2401 } else if (where == ZIPLIST_TAIL) {
2402 listAddNodeTail(ref,sdsnewlen(buf, buflen));
2403 } else {
2404 assert(NULL);
2405 }
2406 }
2407
2408 assert(listLength(ref) == ziplistLen(zl));
2409 for (j = 0; j < len; j++) {
2410 /* Naive way to get elements, but similar to the stresser
2411 * executed from the Tcl test suite. */
2412 p = ziplistIndex(zl,j);
2413 refnode = listIndex(ref,j);
2414
2415 assert(ziplistGet(p,&sstr,&slen,&sval));
2416 if (sstr == NULL) {
2417 buflen = sprintf(buf,"%lld",sval);
2418 } else {
2419 buflen = slen;
2420 memcpy(buf,sstr,buflen);
2421 buf[buflen] = '\0';
2422 }
2423 assert(memcmp(buf,listNodeValue(refnode),buflen) == 0);
2424 }
2425 zfree(zl);
2426 listRelease(ref);
2427 }
2428 printf("Done. usec=%lld\n\n", usec()-start);
2429 }
2430
2431 printf("Stress with variable ziplist size:\n");
2432 {
2433 unsigned long long start = usec();
2434 int maxsize = accurate ? 16384 : 16;
2435 stress(ZIPLIST_HEAD,100000,maxsize,256);
2436 stress(ZIPLIST_TAIL,100000,maxsize,256);
2437 printf("Done. usec=%lld\n\n", usec()-start);
2438 }
2439
2440 /* Benchmarks */
2441 {
2442 zl = ziplistNew();
2443 iteration = accurate ? 100000 : 100;
2444 for (int i=0; i<iteration; i++) {
2445 char buf[4096] = "asdf";
2446 zl = ziplistPush(zl, (unsigned char*)buf, 4, ZIPLIST_TAIL);
2447 zl = ziplistPush(zl, (unsigned char*)buf, 40, ZIPLIST_TAIL);
2448 zl = ziplistPush(zl, (unsigned char*)buf, 400, ZIPLIST_TAIL);
2449 zl = ziplistPush(zl, (unsigned char*)buf, 4000, ZIPLIST_TAIL);
2450 zl = ziplistPush(zl, (unsigned char*)"1", 1, ZIPLIST_TAIL);
2451 zl = ziplistPush(zl, (unsigned char*)"10", 2, ZIPLIST_TAIL);
2452 zl = ziplistPush(zl, (unsigned char*)"100", 3, ZIPLIST_TAIL);
2453 zl = ziplistPush(zl, (unsigned char*)"1000", 4, ZIPLIST_TAIL);
2454 zl = ziplistPush(zl, (unsigned char*)"10000", 5, ZIPLIST_TAIL);
2455 zl = ziplistPush(zl, (unsigned char*)"100000", 6, ZIPLIST_TAIL);
2456 }
2457
2458 printf("Benchmark ziplistFind:\n");
2459 {
2460 unsigned long long start = usec();
2461 for (int i = 0; i < 2000; i++) {
2462 unsigned char *fptr = ziplistIndex(zl, ZIPLIST_HEAD);
2463 fptr = ziplistFind(zl, fptr, (unsigned char*)"nothing", 7, 1);
2464 }
2465 printf("%lld\n", usec()-start);
2466 }
2467
2468 printf("Benchmark ziplistIndex:\n");
2469 {
2470 unsigned long long start = usec();
2471 for (int i = 0; i < 2000; i++) {
2472 ziplistIndex(zl, 99999);
2473 }
2474 printf("%lld\n", usec()-start);
2475 }
2476
2477 printf("Benchmark ziplistValidateIntegrity:\n");
2478 {
2479 unsigned long long start = usec();
2480 for (int i = 0; i < 2000; i++) {
2481 ziplistValidateIntegrity(zl, ziplistBlobLen(zl), 1, NULL, NULL);
2482 }
2483 printf("%lld\n", usec()-start);
2484 }
2485
2486 printf("Benchmark ziplistCompare with string\n");
2487 {
2488 unsigned long long start = usec();
2489 for (int i = 0; i < 2000; i++) {
2490 unsigned char *eptr = ziplistIndex(zl,0);
2491 while (eptr != NULL) {
2492 ziplistCompare(eptr,(unsigned char*)"nothing",7);
2493 eptr = ziplistNext(zl,eptr);
2494 }
2495 }
2496 printf("Done. usec=%lld\n", usec()-start);
2497 }
2498
2499 printf("Benchmark ziplistCompare with number\n");
2500 {
2501 unsigned long long start = usec();
2502 for (int i = 0; i < 2000; i++) {
2503 unsigned char *eptr = ziplistIndex(zl,0);
2504 while (eptr != NULL) {
2505 ziplistCompare(eptr,(unsigned char*)"99999",5);
2506 eptr = ziplistNext(zl,eptr);
2507 }
2508 }
2509 printf("Done. usec=%lld\n", usec()-start);
2510 }
2511
2512 zfree(zl);
2513 }
2514
2515 printf("Stress __ziplistCascadeUpdate:\n");
2516 {
2517 char data[ZIP_BIG_PREVLEN];
2518 zl = ziplistNew();
2519 iteration = accurate ? 100000 : 100;
2520 for (int i = 0; i < iteration; i++) {
2521 zl = ziplistPush(zl, (unsigned char*)data, ZIP_BIG_PREVLEN-4, ZIPLIST_TAIL);
2522 }
2523 unsigned long long start = usec();
2524 zl = ziplistPush(zl, (unsigned char*)data, ZIP_BIG_PREVLEN-3, ZIPLIST_HEAD);
2525 printf("Done. usec=%lld\n\n", usec()-start);
2526 zfree(zl);
2527 }
2528
2529 printf("Edge cases of __ziplistCascadeUpdate:\n");
2530 {
2531 /* Inserting a entry with data length greater than ZIP_BIG_PREVLEN-4
2532 * will leads to cascade update. */
2533 size_t s1 = ZIP_BIG_PREVLEN-4, s2 = ZIP_BIG_PREVLEN-3;
2534 zl = ziplistNew();
2535
2536 zlentry e[4] = {{.prevrawlensize = 0, .prevrawlen = 0, .lensize = 0,
2537 .len = 0, .headersize = 0, .encoding = 0, .p = NULL}};
2538
2539 zl = insertHelper(zl, 'a', s1, ZIPLIST_ENTRY_HEAD(zl));
2540 verify(zl, e);
2541
2542 assert(e[0].prevrawlensize == 1 && e[0].prevrawlen == 0);
2543 assert(compareHelper(zl, 'a', s1, 0));
2544 ziplistRepr(zl);
2545
2546 /* No expand. */
2547 zl = insertHelper(zl, 'b', s1, ZIPLIST_ENTRY_HEAD(zl));
2548 verify(zl, e);
2549
2550 assert(e[0].prevrawlensize == 1 && e[0].prevrawlen == 0);
2551 assert(compareHelper(zl, 'b', s1, 0));
2552
2553 assert(e[1].prevrawlensize == 1 && e[1].prevrawlen == strEntryBytesSmall(s1));
2554 assert(compareHelper(zl, 'a', s1, 1));
2555
2556 ziplistRepr(zl);
2557
2558 /* Expand(tail included). */
2559 zl = insertHelper(zl, 'c', s2, ZIPLIST_ENTRY_HEAD(zl));
2560 verify(zl, e);
2561
2562 assert(e[0].prevrawlensize == 1 && e[0].prevrawlen == 0);
2563 assert(compareHelper(zl, 'c', s2, 0));
2564
2565 assert(e[1].prevrawlensize == 5 && e[1].prevrawlen == strEntryBytesSmall(s2));
2566 assert(compareHelper(zl, 'b', s1, 1));
2567
2568 assert(e[2].prevrawlensize == 5 && e[2].prevrawlen == strEntryBytesLarge(s1));
2569 assert(compareHelper(zl, 'a', s1, 2));
2570
2571 ziplistRepr(zl);
2572
2573 /* Expand(only previous head entry). */
2574 zl = insertHelper(zl, 'd', s2, ZIPLIST_ENTRY_HEAD(zl));
2575 verify(zl, e);
2576
2577 assert(e[0].prevrawlensize == 1 && e[0].prevrawlen == 0);
2578 assert(compareHelper(zl, 'd', s2, 0));
2579
2580 assert(e[1].prevrawlensize == 5 && e[1].prevrawlen == strEntryBytesSmall(s2));
2581 assert(compareHelper(zl, 'c', s2, 1));
2582
2583 assert(e[2].prevrawlensize == 5 && e[2].prevrawlen == strEntryBytesLarge(s2));
2584 assert(compareHelper(zl, 'b', s1, 2));
2585
2586 assert(e[3].prevrawlensize == 5 && e[3].prevrawlen == strEntryBytesLarge(s1));
2587 assert(compareHelper(zl, 'a', s1, 3));
2588
2589 ziplistRepr(zl);
2590
2591 /* Delete from mid. */
2592 unsigned char *p = ziplistIndex(zl, 2);
2593 zl = ziplistDelete(zl, &p);
2594 verify(zl, e);
2595
2596 assert(e[0].prevrawlensize == 1 && e[0].prevrawlen == 0);
2597 assert(compareHelper(zl, 'd', s2, 0));
2598
2599 assert(e[1].prevrawlensize == 5 && e[1].prevrawlen == strEntryBytesSmall(s2));
2600 assert(compareHelper(zl, 'c', s2, 1));
2601
2602 assert(e[2].prevrawlensize == 5 && e[2].prevrawlen == strEntryBytesLarge(s2));
2603 assert(compareHelper(zl, 'a', s1, 2));
2604
2605 ziplistRepr(zl);
2606
2607 zfree(zl);
2608 }
2609
2610 printf("__ziplistInsert nextdiff == -4 && reqlen < 4 (issue #7170):\n");
2611 {
2612 zl = ziplistNew();
2613
2614 /* We set some values to almost reach the critical point - 254 */
2615 char A_252[253] = {0}, A_250[251] = {0};
2616 memset(A_252, 'A', 252);
2617 memset(A_250, 'A', 250);
2618
2619 /* After the rpush, the list look like: [one two A_252 A_250 three 10] */
2620 zl = ziplistPush(zl, (unsigned char*)"one", 3, ZIPLIST_TAIL);
2621 zl = ziplistPush(zl, (unsigned char*)"two", 3, ZIPLIST_TAIL);
2622 zl = ziplistPush(zl, (unsigned char*)A_252, strlen(A_252), ZIPLIST_TAIL);
2623 zl = ziplistPush(zl, (unsigned char*)A_250, strlen(A_250), ZIPLIST_TAIL);
2624 zl = ziplistPush(zl, (unsigned char*)"three", 5, ZIPLIST_TAIL);
2625 zl = ziplistPush(zl, (unsigned char*)"10", 2, ZIPLIST_TAIL);
2626 ziplistRepr(zl);
2627
2628 p = ziplistIndex(zl, 2);
2629 if (!ziplistCompare(p, (unsigned char*)A_252, strlen(A_252))) {
2630 printf("ERROR: not \"A_252\"\n");
2631 return 1;
2632 }
2633
2634 /* When we remove A_252, the list became: [one two A_250 three 10]
2635 * A_250's prev node became node two, because node two quite small
2636 * So A_250's prevlenSize shrink to 1, A_250's total size became 253(1+2+250)
2637 * The prev node of node three is still node A_250.
2638 * We will not shrink the node three's prevlenSize, keep it at 5 bytes */
2639 zl = ziplistDelete(zl, &p);
2640 ziplistRepr(zl);
2641
2642 p = ziplistIndex(zl, 3);
2643 if (!ziplistCompare(p, (unsigned char*)"three", 5)) {
2644 printf("ERROR: not \"three\"\n");
2645 return 1;
2646 }
2647
2648 /* We want to insert a node after A_250, the list became: [one two A_250 10 three 10]
2649 * Because the new node is quite small, node three prevlenSize will shrink to 1 */
2650 zl = ziplistInsert(zl, p, (unsigned char*)"10", 2);
2651 ziplistRepr(zl);
2652
2653 /* Last element should equal 10 */
2654 p = ziplistIndex(zl, -1);
2655 if (!ziplistCompare(p, (unsigned char*)"10", 2)) {
2656 printf("ERROR: not \"10\"\n");
2657 return 1;
2658 }
2659
2660 zfree(zl);
2661 }
2662
2663 printf("ALL TESTS PASSED!\n");
2664 return 0;
2665}
2666#endif
2667