]>
Commit | Line | Data |
---|---|---|
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 | * The general layout of the ziplist is as follows: | |
13 | * <zlbytes><zltail><zllen><entry><entry><zlend> | |
14 | * | |
15 | * <zlbytes> is an unsigned integer to hold the number of bytes that the | |
16 | * ziplist occupies. This value needs to be stored to be able to resize the | |
17 | * entire structure without the need to traverse it first. | |
18 | * | |
19 | * <zltail> is the offset to the last entry in the list. This allows a pop | |
20 | * operation on the far side of the list without the need for full traversal. | |
21 | * | |
22 | * <zllen> is the number of entries.When this value is larger than 2**16-2, | |
23 | * we need to traverse the entire list to know how many items it holds. | |
24 | * | |
25 | * <zlend> is a single byte special value, equal to 255, which indicates the | |
26 | * end of the list. | |
27 | * | |
28 | * ZIPLIST ENTRIES: | |
29 | * Every entry in the ziplist is prefixed by a header that contains two pieces | |
30 | * of information. First, the length of the previous entry is stored to be | |
31 | * able to traverse the list from back to front. Second, the encoding with an | |
32 | * optional string length of the entry itself is stored. | |
33 | * | |
34 | * The length of the previous entry is encoded in the following way: | |
35 | * If this length is smaller than 254 bytes, it will only consume a single | |
36 | * byte that takes the length as value. When the length is greater than or | |
37 | * equal to 254, it will consume 5 bytes. The first byte is set to 254 to | |
38 | * indicate a larger value is following. The remaining 4 bytes take the | |
39 | * length of the previous entry as value. | |
40 | * | |
41 | * The other header field of the entry itself depends on the contents of the | |
42 | * entry. When the entry is a string, the first 2 bits of this header will hold | |
43 | * the type of encoding used to store the length of the string, followed by the | |
44 | * actual length of the string. When the entry is an integer the first 2 bits | |
45 | * are both set to 1. The following 2 bits are used to specify what kind of | |
46 | * integer will be stored after this header. An overview of the different | |
47 | * types and encodings is as follows: | |
48 | * | |
49 | * |00pppppp| - 1 byte | |
50 | * String value with length less than or equal to 63 bytes (6 bits). | |
51 | * |01pppppp|qqqqqqqq| - 2 bytes | |
52 | * String value with length less than or equal to 16383 bytes (14 bits). | |
53 | * |10______|qqqqqqqq|rrrrrrrr|ssssssss|tttttttt| - 5 bytes | |
54 | * String value with length greater than or equal to 16384 bytes. | |
55 | * |11000000| - 1 byte | |
56 | * Integer encoded as int16_t (2 bytes). | |
57 | * |11010000| - 1 byte | |
58 | * Integer encoded as int32_t (4 bytes). | |
59 | * |11100000| - 1 byte | |
60 | * Integer encoded as int64_t (8 bytes). | |
61 | * |11110000| - 1 byte | |
62 | * Integer encoded as 24 bit signed (3 bytes). | |
63 | * |11111110| - 1 byte | |
64 | * Integer encoded as 8 bit signed (1 byte). | |
65 | * |1111xxxx| - (with xxxx between 0000 and 1101) immediate 4 bit integer. | |
66 | * Unsigned integer from 0 to 12. The encoded value is actually from | |
67 | * 1 to 13 because 0000 and 1111 can not be used, so 1 should be | |
68 | * subtracted from the encoded 4 bit value to obtain the right value. | |
69 | * |11111111| - End of ziplist. | |
70 | * | |
71 | * All the integers are represented in little endian byte order. | |
72 | */ | |
73 | ||
74 | #include <stdio.h> | |
75 | #include <stdlib.h> | |
76 | #include <string.h> | |
77 | #include <stdint.h> | |
78 | #include <assert.h> | |
79 | #include <limits.h> | |
80 | #include "zmalloc.h" | |
81 | #include "util.h" | |
82 | #include "ziplist.h" | |
83 | #include "endianconv.h" | |
84 | ||
85 | #define ZIP_END 255 | |
86 | #define ZIP_BIGLEN 254 | |
87 | ||
88 | /* Different encoding/length possibilities */ | |
89 | #define ZIP_STR_MASK 0xc0 | |
90 | #define ZIP_INT_MASK 0x30 | |
91 | #define ZIP_STR_06B (0 << 6) | |
92 | #define ZIP_STR_14B (1 << 6) | |
93 | #define ZIP_STR_32B (2 << 6) | |
94 | #define ZIP_INT_16B (0xc0 | 0<<4) | |
95 | #define ZIP_INT_32B (0xc0 | 1<<4) | |
96 | #define ZIP_INT_64B (0xc0 | 2<<4) | |
97 | #define ZIP_INT_24B (0xc0 | 3<<4) | |
98 | #define ZIP_INT_8B 0xfe | |
99 | /* 4 bit integer immediate encoding */ | |
100 | #define ZIP_INT_IMM_MASK 0x0f | |
101 | #define ZIP_INT_IMM_MIN 0xf1 /* 11110001 */ | |
102 | #define ZIP_INT_IMM_MAX 0xfd /* 11111101 */ | |
103 | #define ZIP_INT_IMM_VAL(v) (v & ZIP_INT_IMM_MASK) | |
104 | ||
105 | #define INT24_MAX 0x7fffff | |
106 | #define INT24_MIN (-INT24_MAX - 1) | |
107 | ||
108 | /* Macro to determine type */ | |
109 | #define ZIP_IS_STR(enc) (((enc) & ZIP_STR_MASK) < ZIP_STR_MASK) | |
110 | ||
111 | /* Utility macros */ | |
112 | #define ZIPLIST_BYTES(zl) (*((uint32_t*)(zl))) | |
113 | #define ZIPLIST_TAIL_OFFSET(zl) (*((uint32_t*)((zl)+sizeof(uint32_t)))) | |
114 | #define ZIPLIST_LENGTH(zl) (*((uint16_t*)((zl)+sizeof(uint32_t)*2))) | |
115 | #define ZIPLIST_HEADER_SIZE (sizeof(uint32_t)*2+sizeof(uint16_t)) | |
116 | #define ZIPLIST_ENTRY_HEAD(zl) ((zl)+ZIPLIST_HEADER_SIZE) | |
117 | #define ZIPLIST_ENTRY_TAIL(zl) ((zl)+intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))) | |
118 | #define ZIPLIST_ENTRY_END(zl) ((zl)+intrev32ifbe(ZIPLIST_BYTES(zl))-1) | |
119 | ||
120 | /* We know a positive increment can only be 1 because entries can only be | |
121 | * pushed one at a time. */ | |
122 | #define ZIPLIST_INCR_LENGTH(zl,incr) { \ | |
123 | if (ZIPLIST_LENGTH(zl) < UINT16_MAX) \ | |
124 | ZIPLIST_LENGTH(zl) = intrev16ifbe(intrev16ifbe(ZIPLIST_LENGTH(zl))+incr); \ | |
125 | } | |
126 | ||
127 | typedef struct zlentry { | |
128 | unsigned int prevrawlensize, prevrawlen; | |
129 | unsigned int lensize, len; | |
130 | unsigned int headersize; | |
131 | unsigned char encoding; | |
132 | unsigned char *p; | |
133 | } zlentry; | |
134 | ||
135 | /* Extract the encoding from the byte pointed by 'ptr' and set it into | |
136 | * 'encoding'. */ | |
137 | #define ZIP_ENTRY_ENCODING(ptr, encoding) do { \ | |
138 | (encoding) = (ptr[0]); \ | |
139 | if ((encoding) < ZIP_STR_MASK) (encoding) &= ZIP_STR_MASK; \ | |
140 | } while(0) | |
141 | ||
142 | /* Return bytes needed to store integer encoded by 'encoding' */ | |
143 | static unsigned int zipIntSize(unsigned char encoding) { | |
144 | switch(encoding) { | |
145 | case ZIP_INT_8B: return 1; | |
146 | case ZIP_INT_16B: return 2; | |
147 | case ZIP_INT_24B: return 3; | |
148 | case ZIP_INT_32B: return 4; | |
149 | case ZIP_INT_64B: return 8; | |
150 | default: return 0; /* 4 bit immediate */ | |
151 | } | |
152 | assert(NULL); | |
153 | return 0; | |
154 | } | |
155 | ||
156 | /* Encode the length 'l' writing it in 'p'. If p is NULL it just returns | |
157 | * the amount of bytes required to encode such a length. */ | |
158 | static unsigned int zipEncodeLength(unsigned char *p, unsigned char encoding, unsigned int rawlen) { | |
159 | unsigned char len = 1, buf[5]; | |
160 | ||
161 | if (ZIP_IS_STR(encoding)) { | |
162 | /* Although encoding is given it may not be set for strings, | |
163 | * so we determine it here using the raw length. */ | |
164 | if (rawlen <= 0x3f) { | |
165 | if (!p) return len; | |
166 | buf[0] = ZIP_STR_06B | rawlen; | |
167 | } else if (rawlen <= 0x3fff) { | |
168 | len += 1; | |
169 | if (!p) return len; | |
170 | buf[0] = ZIP_STR_14B | ((rawlen >> 8) & 0x3f); | |
171 | buf[1] = rawlen & 0xff; | |
172 | } else { | |
173 | len += 4; | |
174 | if (!p) return len; | |
175 | buf[0] = ZIP_STR_32B; | |
176 | buf[1] = (rawlen >> 24) & 0xff; | |
177 | buf[2] = (rawlen >> 16) & 0xff; | |
178 | buf[3] = (rawlen >> 8) & 0xff; | |
179 | buf[4] = rawlen & 0xff; | |
180 | } | |
181 | } else { | |
182 | /* Implies integer encoding, so length is always 1. */ | |
183 | if (!p) return len; | |
184 | buf[0] = encoding; | |
185 | } | |
186 | ||
187 | /* Store this length at p */ | |
188 | memcpy(p,buf,len); | |
189 | return len; | |
190 | } | |
191 | ||
192 | /* Decode the length encoded in 'ptr'. The 'encoding' variable will hold the | |
193 | * entries encoding, the 'lensize' variable will hold the number of bytes | |
194 | * required to encode the entries length, and the 'len' variable will hold the | |
195 | * entries length. */ | |
196 | #define ZIP_DECODE_LENGTH(ptr, encoding, lensize, len) do { \ | |
197 | ZIP_ENTRY_ENCODING((ptr), (encoding)); \ | |
198 | if ((encoding) < ZIP_STR_MASK) { \ | |
199 | if ((encoding) == ZIP_STR_06B) { \ | |
200 | (lensize) = 1; \ | |
201 | (len) = (ptr)[0] & 0x3f; \ | |
202 | } else if ((encoding) == ZIP_STR_14B) { \ | |
203 | (lensize) = 2; \ | |
204 | (len) = (((ptr)[0] & 0x3f) << 8) | (ptr)[1]; \ | |
205 | } else if (encoding == ZIP_STR_32B) { \ | |
206 | (lensize) = 5; \ | |
207 | (len) = ((ptr)[1] << 24) | \ | |
208 | ((ptr)[2] << 16) | \ | |
209 | ((ptr)[3] << 8) | \ | |
210 | ((ptr)[4]); \ | |
211 | } else { \ | |
212 | assert(NULL); \ | |
213 | } \ | |
214 | } else { \ | |
215 | (lensize) = 1; \ | |
216 | (len) = zipIntSize(encoding); \ | |
217 | } \ | |
218 | } while(0); | |
219 | ||
220 | /* Encode the length of the previous entry and write it to "p". Return the | |
221 | * number of bytes needed to encode this length if "p" is NULL. */ | |
222 | static unsigned int zipPrevEncodeLength(unsigned char *p, unsigned int len) { | |
223 | if (p == NULL) { | |
224 | return (len < ZIP_BIGLEN) ? 1 : sizeof(len)+1; | |
225 | } else { | |
226 | if (len < ZIP_BIGLEN) { | |
227 | p[0] = len; | |
228 | return 1; | |
229 | } else { | |
230 | p[0] = ZIP_BIGLEN; | |
231 | memcpy(p+1,&len,sizeof(len)); | |
232 | memrev32ifbe(p+1); | |
233 | return 1+sizeof(len); | |
234 | } | |
235 | } | |
236 | } | |
237 | ||
238 | /* Encode the length of the previous entry and write it to "p". This only | |
239 | * uses the larger encoding (required in __ziplistCascadeUpdate). */ | |
240 | static void zipPrevEncodeLengthForceLarge(unsigned char *p, unsigned int len) { | |
241 | if (p == NULL) return; | |
242 | p[0] = ZIP_BIGLEN; | |
243 | memcpy(p+1,&len,sizeof(len)); | |
244 | memrev32ifbe(p+1); | |
245 | } | |
246 | ||
247 | /* Decode the number of bytes required to store the length of the previous | |
248 | * element, from the perspective of the entry pointed to by 'ptr'. */ | |
249 | #define ZIP_DECODE_PREVLENSIZE(ptr, prevlensize) do { \ | |
250 | if ((ptr)[0] < ZIP_BIGLEN) { \ | |
251 | (prevlensize) = 1; \ | |
252 | } else { \ | |
253 | (prevlensize) = 5; \ | |
254 | } \ | |
255 | } while(0); | |
256 | ||
257 | /* Decode the length of the previous element, from the perspective of the entry | |
258 | * pointed to by 'ptr'. */ | |
259 | #define ZIP_DECODE_PREVLEN(ptr, prevlensize, prevlen) do { \ | |
260 | ZIP_DECODE_PREVLENSIZE(ptr, prevlensize); \ | |
261 | if ((prevlensize) == 1) { \ | |
262 | (prevlen) = (ptr)[0]; \ | |
263 | } else if ((prevlensize) == 5) { \ | |
264 | assert(sizeof((prevlensize)) == 4); \ | |
265 | memcpy(&(prevlen), ((char*)(ptr)) + 1, 4); \ | |
266 | memrev32ifbe(&prevlen); \ | |
267 | } \ | |
268 | } while(0); | |
269 | ||
270 | /* Return the difference in number of bytes needed to store the length of the | |
271 | * previous element 'len', in the entry pointed to by 'p'. */ | |
272 | static int zipPrevLenByteDiff(unsigned char *p, unsigned int len) { | |
273 | unsigned int prevlensize; | |
274 | ZIP_DECODE_PREVLENSIZE(p, prevlensize); | |
275 | return zipPrevEncodeLength(NULL, len) - prevlensize; | |
276 | } | |
277 | ||
278 | /* Return the total number of bytes used by the entry pointed to by 'p'. */ | |
279 | static unsigned int zipRawEntryLength(unsigned char *p) { | |
280 | unsigned int prevlensize, encoding, lensize, len; | |
281 | ZIP_DECODE_PREVLENSIZE(p, prevlensize); | |
282 | ZIP_DECODE_LENGTH(p + prevlensize, encoding, lensize, len); | |
283 | return prevlensize + lensize + len; | |
284 | } | |
285 | ||
286 | /* Check if string pointed to by 'entry' can be encoded as an integer. | |
287 | * Stores the integer value in 'v' and its encoding in 'encoding'. */ | |
288 | static int zipTryEncoding(unsigned char *entry, unsigned int entrylen, long long *v, unsigned char *encoding) { | |
289 | long long value; | |
290 | ||
291 | if (entrylen >= 32 || entrylen == 0) return 0; | |
292 | if (string2ll((char*)entry,entrylen,&value)) { | |
293 | /* Great, the string can be encoded. Check what's the smallest | |
294 | * of our encoding types that can hold this value. */ | |
295 | if (value >= 0 && value <= 12) { | |
296 | *encoding = ZIP_INT_IMM_MIN+value; | |
297 | } else if (value >= INT8_MIN && value <= INT8_MAX) { | |
298 | *encoding = ZIP_INT_8B; | |
299 | } else if (value >= INT16_MIN && value <= INT16_MAX) { | |
300 | *encoding = ZIP_INT_16B; | |
301 | } else if (value >= INT24_MIN && value <= INT24_MAX) { | |
302 | *encoding = ZIP_INT_24B; | |
303 | } else if (value >= INT32_MIN && value <= INT32_MAX) { | |
304 | *encoding = ZIP_INT_32B; | |
305 | } else { | |
306 | *encoding = ZIP_INT_64B; | |
307 | } | |
308 | *v = value; | |
309 | return 1; | |
310 | } | |
311 | return 0; | |
312 | } | |
313 | ||
314 | /* Store integer 'value' at 'p', encoded as 'encoding' */ | |
315 | static void zipSaveInteger(unsigned char *p, int64_t value, unsigned char encoding) { | |
316 | int16_t i16; | |
317 | int32_t i32; | |
318 | int64_t i64; | |
319 | if (encoding == ZIP_INT_8B) { | |
320 | ((int8_t*)p)[0] = (int8_t)value; | |
321 | } else if (encoding == ZIP_INT_16B) { | |
322 | i16 = value; | |
323 | memcpy(p,&i16,sizeof(i16)); | |
324 | memrev16ifbe(p); | |
325 | } else if (encoding == ZIP_INT_24B) { | |
326 | i32 = value<<8; | |
327 | memrev32ifbe(&i32); | |
328 | memcpy(p,((uint8_t*)&i32)+1,sizeof(i32)-sizeof(uint8_t)); | |
329 | } else if (encoding == ZIP_INT_32B) { | |
330 | i32 = value; | |
331 | memcpy(p,&i32,sizeof(i32)); | |
332 | memrev32ifbe(p); | |
333 | } else if (encoding == ZIP_INT_64B) { | |
334 | i64 = value; | |
335 | memcpy(p,&i64,sizeof(i64)); | |
336 | memrev64ifbe(p); | |
337 | } else if (encoding >= ZIP_INT_IMM_MIN && encoding <= ZIP_INT_IMM_MAX) { | |
338 | /* Nothing to do, the value is stored in the encoding itself. */ | |
339 | } else { | |
340 | assert(NULL); | |
341 | } | |
342 | } | |
343 | ||
344 | /* Read integer encoded as 'encoding' from 'p' */ | |
345 | static int64_t zipLoadInteger(unsigned char *p, unsigned char encoding) { | |
346 | int16_t i16; | |
347 | int32_t i32; | |
348 | int64_t i64, ret = 0; | |
349 | if (encoding == ZIP_INT_8B) { | |
350 | ret = ((int8_t*)p)[0]; | |
351 | } else if (encoding == ZIP_INT_16B) { | |
352 | memcpy(&i16,p,sizeof(i16)); | |
353 | memrev16ifbe(&i16); | |
354 | ret = i16; | |
355 | } else if (encoding == ZIP_INT_32B) { | |
356 | memcpy(&i32,p,sizeof(i32)); | |
357 | memrev32ifbe(&i32); | |
358 | ret = i32; | |
359 | } else if (encoding == ZIP_INT_24B) { | |
360 | i32 = 0; | |
361 | memcpy(((uint8_t*)&i32)+1,p,sizeof(i32)-sizeof(uint8_t)); | |
362 | memrev32ifbe(&i32); | |
363 | ret = i32>>8; | |
364 | } else if (encoding == ZIP_INT_64B) { | |
365 | memcpy(&i64,p,sizeof(i64)); | |
366 | memrev64ifbe(&i64); | |
367 | ret = i64; | |
368 | } else if (encoding >= ZIP_INT_IMM_MIN && encoding <= ZIP_INT_IMM_MAX) { | |
369 | ret = (encoding & ZIP_INT_IMM_MASK)-1; | |
370 | } else { | |
371 | assert(NULL); | |
372 | } | |
373 | return ret; | |
374 | } | |
375 | ||
376 | /* Return a struct with all information about an entry. */ | |
377 | static zlentry zipEntry(unsigned char *p) { | |
378 | zlentry e; | |
379 | ||
380 | ZIP_DECODE_PREVLEN(p, e.prevrawlensize, e.prevrawlen); | |
381 | ZIP_DECODE_LENGTH(p + e.prevrawlensize, e.encoding, e.lensize, e.len); | |
382 | e.headersize = e.prevrawlensize + e.lensize; | |
383 | e.p = p; | |
384 | return e; | |
385 | } | |
386 | ||
387 | /* Create a new empty ziplist. */ | |
388 | unsigned char *ziplistNew(void) { | |
389 | unsigned int bytes = ZIPLIST_HEADER_SIZE+1; | |
390 | unsigned char *zl = zmalloc(bytes); | |
391 | ZIPLIST_BYTES(zl) = intrev32ifbe(bytes); | |
392 | ZIPLIST_TAIL_OFFSET(zl) = intrev32ifbe(ZIPLIST_HEADER_SIZE); | |
393 | ZIPLIST_LENGTH(zl) = 0; | |
394 | zl[bytes-1] = ZIP_END; | |
395 | return zl; | |
396 | } | |
397 | ||
398 | /* Resize the ziplist. */ | |
399 | static unsigned char *ziplistResize(unsigned char *zl, unsigned int len) { | |
400 | zl = zrealloc(zl,len); | |
401 | ZIPLIST_BYTES(zl) = intrev32ifbe(len); | |
402 | zl[len-1] = ZIP_END; | |
403 | return zl; | |
404 | } | |
405 | ||
406 | /* When an entry is inserted, we need to set the prevlen field of the next | |
407 | * entry to equal the length of the inserted entry. It can occur that this | |
408 | * length cannot be encoded in 1 byte and the next entry needs to be grow | |
409 | * a bit larger to hold the 5-byte encoded prevlen. This can be done for free, | |
410 | * because this only happens when an entry is already being inserted (which | |
411 | * causes a realloc and memmove). However, encoding the prevlen may require | |
412 | * that this entry is grown as well. This effect may cascade throughout | |
413 | * the ziplist when there are consecutive entries with a size close to | |
414 | * ZIP_BIGLEN, so we need to check that the prevlen can be encoded in every | |
415 | * consecutive entry. | |
416 | * | |
417 | * Note that this effect can also happen in reverse, where the bytes required | |
418 | * to encode the prevlen field can shrink. This effect is deliberately ignored, | |
419 | * because it can cause a "flapping" effect where a chain prevlen fields is | |
420 | * first grown and then shrunk again after consecutive inserts. Rather, the | |
421 | * field is allowed to stay larger than necessary, because a large prevlen | |
422 | * field implies the ziplist is holding large entries anyway. | |
423 | * | |
424 | * The pointer "p" points to the first entry that does NOT need to be | |
425 | * updated, i.e. consecutive fields MAY need an update. */ | |
426 | static unsigned char *__ziplistCascadeUpdate(unsigned char *zl, unsigned char *p) { | |
427 | size_t curlen = intrev32ifbe(ZIPLIST_BYTES(zl)), rawlen, rawlensize; | |
428 | size_t offset, noffset, extra; | |
429 | unsigned char *np; | |
430 | zlentry cur, next; | |
431 | ||
432 | while (p[0] != ZIP_END) { | |
433 | cur = zipEntry(p); | |
434 | rawlen = cur.headersize + cur.len; | |
435 | rawlensize = zipPrevEncodeLength(NULL,rawlen); | |
436 | ||
437 | /* Abort if there is no next entry. */ | |
438 | if (p[rawlen] == ZIP_END) break; | |
439 | next = zipEntry(p+rawlen); | |
440 | ||
441 | /* Abort when "prevlen" has not changed. */ | |
442 | if (next.prevrawlen == rawlen) break; | |
443 | ||
444 | if (next.prevrawlensize < rawlensize) { | |
445 | /* The "prevlen" field of "next" needs more bytes to hold | |
446 | * the raw length of "cur". */ | |
447 | offset = p-zl; | |
448 | extra = rawlensize-next.prevrawlensize; | |
449 | zl = ziplistResize(zl,curlen+extra); | |
450 | p = zl+offset; | |
451 | ||
452 | /* Current pointer and offset for next element. */ | |
453 | np = p+rawlen; | |
454 | noffset = np-zl; | |
455 | ||
456 | /* Update tail offset when next element is not the tail element. */ | |
457 | if ((zl+intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))) != np) { | |
458 | ZIPLIST_TAIL_OFFSET(zl) = | |
459 | intrev32ifbe(intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))+extra); | |
460 | } | |
461 | ||
462 | /* Move the tail to the back. */ | |
463 | memmove(np+rawlensize, | |
464 | np+next.prevrawlensize, | |
465 | curlen-noffset-next.prevrawlensize-1); | |
466 | zipPrevEncodeLength(np,rawlen); | |
467 | ||
468 | /* Advance the cursor */ | |
469 | p += rawlen; | |
470 | curlen += extra; | |
471 | } else { | |
472 | if (next.prevrawlensize > rawlensize) { | |
473 | /* This would result in shrinking, which we want to avoid. | |
474 | * So, set "rawlen" in the available bytes. */ | |
475 | zipPrevEncodeLengthForceLarge(p+rawlen,rawlen); | |
476 | } else { | |
477 | zipPrevEncodeLength(p+rawlen,rawlen); | |
478 | } | |
479 | ||
480 | /* Stop here, as the raw length of "next" has not changed. */ | |
481 | break; | |
482 | } | |
483 | } | |
484 | return zl; | |
485 | } | |
486 | ||
487 | /* Delete "num" entries, starting at "p". Returns pointer to the ziplist. */ | |
488 | static unsigned char *__ziplistDelete(unsigned char *zl, unsigned char *p, unsigned int num) { | |
489 | unsigned int i, totlen, deleted = 0; | |
490 | size_t offset; | |
491 | int nextdiff = 0; | |
492 | zlentry first, tail; | |
493 | ||
494 | first = zipEntry(p); | |
495 | for (i = 0; p[0] != ZIP_END && i < num; i++) { | |
496 | p += zipRawEntryLength(p); | |
497 | deleted++; | |
498 | } | |
499 | ||
500 | totlen = p-first.p; | |
501 | if (totlen > 0) { | |
502 | if (p[0] != ZIP_END) { | |
503 | /* Storing `prevrawlen` in this entry may increase or decrease the | |
504 | * number of bytes required compare to the current `prevrawlen`. | |
505 | * There always is room to store this, because it was previously | |
506 | * stored by an entry that is now being deleted. */ | |
507 | nextdiff = zipPrevLenByteDiff(p,first.prevrawlen); | |
508 | p -= nextdiff; | |
509 | zipPrevEncodeLength(p,first.prevrawlen); | |
510 | ||
511 | /* Update offset for tail */ | |
512 | ZIPLIST_TAIL_OFFSET(zl) = | |
513 | intrev32ifbe(intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))-totlen); | |
514 | ||
515 | /* When the tail contains more than one entry, we need to take | |
516 | * "nextdiff" in account as well. Otherwise, a change in the | |
517 | * size of prevlen doesn't have an effect on the *tail* offset. */ | |
518 | tail = zipEntry(p); | |
519 | if (p[tail.headersize+tail.len] != ZIP_END) { | |
520 | ZIPLIST_TAIL_OFFSET(zl) = | |
521 | intrev32ifbe(intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))+nextdiff); | |
522 | } | |
523 | ||
524 | /* Move tail to the front of the ziplist */ | |
525 | memmove(first.p,p, | |
526 | intrev32ifbe(ZIPLIST_BYTES(zl))-(p-zl)-1); | |
527 | } else { | |
528 | /* The entire tail was deleted. No need to move memory. */ | |
529 | ZIPLIST_TAIL_OFFSET(zl) = | |
530 | intrev32ifbe((first.p-zl)-first.prevrawlen); | |
531 | } | |
532 | ||
533 | /* Resize and update length */ | |
534 | offset = first.p-zl; | |
535 | zl = ziplistResize(zl, intrev32ifbe(ZIPLIST_BYTES(zl))-totlen+nextdiff); | |
536 | ZIPLIST_INCR_LENGTH(zl,-deleted); | |
537 | p = zl+offset; | |
538 | ||
539 | /* When nextdiff != 0, the raw length of the next entry has changed, so | |
540 | * we need to cascade the update throughout the ziplist */ | |
541 | if (nextdiff != 0) | |
542 | zl = __ziplistCascadeUpdate(zl,p); | |
543 | } | |
544 | return zl; | |
545 | } | |
546 | ||
547 | /* Insert item at "p". */ | |
548 | static unsigned char *__ziplistInsert(unsigned char *zl, unsigned char *p, unsigned char *s, unsigned int slen) { | |
549 | size_t curlen = intrev32ifbe(ZIPLIST_BYTES(zl)), reqlen, prevlen = 0; | |
550 | size_t offset; | |
551 | int nextdiff = 0; | |
552 | unsigned char encoding = 0; | |
553 | long long value = 123456789; /* initialized to avoid warning. Using a value | |
554 | that is easy to see if for some reason | |
555 | we use it uninitialized. */ | |
556 | zlentry entry, tail; | |
557 | ||
558 | /* Find out prevlen for the entry that is inserted. */ | |
559 | if (p[0] != ZIP_END) { | |
560 | entry = zipEntry(p); | |
561 | prevlen = entry.prevrawlen; | |
562 | } else { | |
563 | unsigned char *ptail = ZIPLIST_ENTRY_TAIL(zl); | |
564 | if (ptail[0] != ZIP_END) { | |
565 | prevlen = zipRawEntryLength(ptail); | |
566 | } | |
567 | } | |
568 | ||
569 | /* See if the entry can be encoded */ | |
570 | if (zipTryEncoding(s,slen,&value,&encoding)) { | |
571 | /* 'encoding' is set to the appropriate integer encoding */ | |
572 | reqlen = zipIntSize(encoding); | |
573 | } else { | |
574 | /* 'encoding' is untouched, however zipEncodeLength will use the | |
575 | * string length to figure out how to encode it. */ | |
576 | reqlen = slen; | |
577 | } | |
578 | /* We need space for both the length of the previous entry and | |
579 | * the length of the payload. */ | |
580 | reqlen += zipPrevEncodeLength(NULL,prevlen); | |
581 | reqlen += zipEncodeLength(NULL,encoding,slen); | |
582 | ||
583 | /* When the insert position is not equal to the tail, we need to | |
584 | * make sure that the next entry can hold this entry's length in | |
585 | * its prevlen field. */ | |
586 | nextdiff = (p[0] != ZIP_END) ? zipPrevLenByteDiff(p,reqlen) : 0; | |
587 | ||
588 | /* Store offset because a realloc may change the address of zl. */ | |
589 | offset = p-zl; | |
590 | zl = ziplistResize(zl,curlen+reqlen+nextdiff); | |
591 | p = zl+offset; | |
592 | ||
593 | /* Apply memory move when necessary and update tail offset. */ | |
594 | if (p[0] != ZIP_END) { | |
595 | /* Subtract one because of the ZIP_END bytes */ | |
596 | memmove(p+reqlen,p-nextdiff,curlen-offset-1+nextdiff); | |
597 | ||
598 | /* Encode this entry's raw length in the next entry. */ | |
599 | zipPrevEncodeLength(p+reqlen,reqlen); | |
600 | ||
601 | /* Update offset for tail */ | |
602 | ZIPLIST_TAIL_OFFSET(zl) = | |
603 | intrev32ifbe(intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))+reqlen); | |
604 | ||
605 | /* When the tail contains more than one entry, we need to take | |
606 | * "nextdiff" in account as well. Otherwise, a change in the | |
607 | * size of prevlen doesn't have an effect on the *tail* offset. */ | |
608 | tail = zipEntry(p+reqlen); | |
609 | if (p[reqlen+tail.headersize+tail.len] != ZIP_END) { | |
610 | ZIPLIST_TAIL_OFFSET(zl) = | |
611 | intrev32ifbe(intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))+nextdiff); | |
612 | } | |
613 | } else { | |
614 | /* This element will be the new tail. */ | |
615 | ZIPLIST_TAIL_OFFSET(zl) = intrev32ifbe(p-zl); | |
616 | } | |
617 | ||
618 | /* When nextdiff != 0, the raw length of the next entry has changed, so | |
619 | * we need to cascade the update throughout the ziplist */ | |
620 | if (nextdiff != 0) { | |
621 | offset = p-zl; | |
622 | zl = __ziplistCascadeUpdate(zl,p+reqlen); | |
623 | p = zl+offset; | |
624 | } | |
625 | ||
626 | /* Write the entry */ | |
627 | p += zipPrevEncodeLength(p,prevlen); | |
628 | p += zipEncodeLength(p,encoding,slen); | |
629 | if (ZIP_IS_STR(encoding)) { | |
630 | memcpy(p,s,slen); | |
631 | } else { | |
632 | zipSaveInteger(p,value,encoding); | |
633 | } | |
634 | ZIPLIST_INCR_LENGTH(zl,1); | |
635 | return zl; | |
636 | } | |
637 | ||
638 | unsigned char *ziplistPush(unsigned char *zl, unsigned char *s, unsigned int slen, int where) { | |
639 | unsigned char *p; | |
640 | p = (where == ZIPLIST_HEAD) ? ZIPLIST_ENTRY_HEAD(zl) : ZIPLIST_ENTRY_END(zl); | |
641 | return __ziplistInsert(zl,p,s,slen); | |
642 | } | |
643 | ||
644 | /* Returns an offset to use for iterating with ziplistNext. When the given | |
645 | * index is negative, the list is traversed back to front. When the list | |
646 | * doesn't contain an element at the provided index, NULL is returned. */ | |
647 | unsigned char *ziplistIndex(unsigned char *zl, int index) { | |
648 | unsigned char *p; | |
649 | zlentry entry; | |
650 | if (index < 0) { | |
651 | index = (-index)-1; | |
652 | p = ZIPLIST_ENTRY_TAIL(zl); | |
653 | if (p[0] != ZIP_END) { | |
654 | entry = zipEntry(p); | |
655 | while (entry.prevrawlen > 0 && index--) { | |
656 | p -= entry.prevrawlen; | |
657 | entry = zipEntry(p); | |
658 | } | |
659 | } | |
660 | } else { | |
661 | p = ZIPLIST_ENTRY_HEAD(zl); | |
662 | while (p[0] != ZIP_END && index--) { | |
663 | p += zipRawEntryLength(p); | |
664 | } | |
665 | } | |
666 | return (p[0] == ZIP_END || index > 0) ? NULL : p; | |
667 | } | |
668 | ||
669 | /* Return pointer to next entry in ziplist. | |
670 | * | |
671 | * zl is the pointer to the ziplist | |
672 | * p is the pointer to the current element | |
673 | * | |
674 | * The element after 'p' is returned, otherwise NULL if we are at the end. */ | |
675 | unsigned char *ziplistNext(unsigned char *zl, unsigned char *p) { | |
676 | ((void) zl); | |
677 | ||
678 | /* "p" could be equal to ZIP_END, caused by ziplistDelete, | |
679 | * and we should return NULL. Otherwise, we should return NULL | |
680 | * when the *next* element is ZIP_END (there is no next entry). */ | |
681 | if (p[0] == ZIP_END) { | |
682 | return NULL; | |
683 | } | |
684 | ||
685 | p += zipRawEntryLength(p); | |
686 | if (p[0] == ZIP_END) { | |
687 | return NULL; | |
688 | } | |
689 | ||
690 | return p; | |
691 | } | |
692 | ||
693 | /* Return pointer to previous entry in ziplist. */ | |
694 | unsigned char *ziplistPrev(unsigned char *zl, unsigned char *p) { | |
695 | zlentry entry; | |
696 | ||
697 | /* Iterating backwards from ZIP_END should return the tail. When "p" is | |
698 | * equal to the first element of the list, we're already at the head, | |
699 | * and should return NULL. */ | |
700 | if (p[0] == ZIP_END) { | |
701 | p = ZIPLIST_ENTRY_TAIL(zl); | |
702 | return (p[0] == ZIP_END) ? NULL : p; | |
703 | } else if (p == ZIPLIST_ENTRY_HEAD(zl)) { | |
704 | return NULL; | |
705 | } else { | |
706 | entry = zipEntry(p); | |
707 | assert(entry.prevrawlen > 0); | |
708 | return p-entry.prevrawlen; | |
709 | } | |
710 | } | |
711 | ||
712 | /* Get entry pointer to by 'p' and store in either 'e' or 'v' depending | |
713 | * on the encoding of the entry. 'e' is always set to NULL to be able | |
714 | * to find out whether the string pointer or the integer value was set. | |
715 | * Return 0 if 'p' points to the end of the zipmap, 1 otherwise. */ | |
716 | unsigned int ziplistGet(unsigned char *p, unsigned char **sstr, unsigned int *slen, long long *sval) { | |
717 | zlentry entry; | |
718 | if (p == NULL || p[0] == ZIP_END) return 0; | |
719 | if (sstr) *sstr = NULL; | |
720 | ||
721 | entry = zipEntry(p); | |
722 | if (ZIP_IS_STR(entry.encoding)) { | |
723 | if (sstr) { | |
724 | *slen = entry.len; | |
725 | *sstr = p+entry.headersize; | |
726 | } | |
727 | } else { | |
728 | if (sval) { | |
729 | *sval = zipLoadInteger(p+entry.headersize,entry.encoding); | |
730 | } | |
731 | } | |
732 | return 1; | |
733 | } | |
734 | ||
735 | /* Insert an entry at "p". */ | |
736 | unsigned char *ziplistInsert(unsigned char *zl, unsigned char *p, unsigned char *s, unsigned int slen) { | |
737 | return __ziplistInsert(zl,p,s,slen); | |
738 | } | |
739 | ||
740 | /* Delete a single entry from the ziplist, pointed to by *p. | |
741 | * Also update *p in place, to be able to iterate over the | |
742 | * ziplist, while deleting entries. */ | |
743 | unsigned char *ziplistDelete(unsigned char *zl, unsigned char **p) { | |
744 | size_t offset = *p-zl; | |
745 | zl = __ziplistDelete(zl,*p,1); | |
746 | ||
747 | /* Store pointer to current element in p, because ziplistDelete will | |
748 | * do a realloc which might result in a different "zl"-pointer. | |
749 | * When the delete direction is back to front, we might delete the last | |
750 | * entry and end up with "p" pointing to ZIP_END, so check this. */ | |
751 | *p = zl+offset; | |
752 | return zl; | |
753 | } | |
754 | ||
755 | /* Delete a range of entries from the ziplist. */ | |
756 | unsigned char *ziplistDeleteRange(unsigned char *zl, unsigned int index, unsigned int num) { | |
757 | unsigned char *p = ziplistIndex(zl,index); | |
758 | return (p == NULL) ? zl : __ziplistDelete(zl,p,num); | |
759 | } | |
760 | ||
761 | /* Compare entry pointer to by 'p' with 'entry'. Return 1 if equal. */ | |
762 | unsigned int ziplistCompare(unsigned char *p, unsigned char *sstr, unsigned int slen) { | |
763 | zlentry entry; | |
764 | unsigned char sencoding; | |
765 | long long zval, sval; | |
766 | if (p[0] == ZIP_END) return 0; | |
767 | ||
768 | entry = zipEntry(p); | |
769 | if (ZIP_IS_STR(entry.encoding)) { | |
770 | /* Raw compare */ | |
771 | if (entry.len == slen) { | |
772 | return memcmp(p+entry.headersize,sstr,slen) == 0; | |
773 | } else { | |
774 | return 0; | |
775 | } | |
776 | } else { | |
777 | /* Try to compare encoded values. Don't compare encoding because | |
778 | * different implementations may encoded integers differently. */ | |
779 | if (zipTryEncoding(sstr,slen,&sval,&sencoding)) { | |
780 | zval = zipLoadInteger(p+entry.headersize,entry.encoding); | |
781 | return zval == sval; | |
782 | } | |
783 | } | |
784 | return 0; | |
785 | } | |
786 | ||
787 | /* Find pointer to the entry equal to the specified entry. Skip 'skip' entries | |
788 | * between every comparison. Returns NULL when the field could not be found. */ | |
789 | unsigned char *ziplistFind(unsigned char *p, unsigned char *vstr, unsigned int vlen, unsigned int skip) { | |
790 | int skipcnt = 0; | |
791 | unsigned char vencoding = 0; | |
792 | long long vll = 0; | |
793 | ||
794 | while (p[0] != ZIP_END) { | |
795 | unsigned int prevlensize, encoding, lensize, len; | |
796 | unsigned char *q; | |
797 | ||
798 | ZIP_DECODE_PREVLENSIZE(p, prevlensize); | |
799 | ZIP_DECODE_LENGTH(p + prevlensize, encoding, lensize, len); | |
800 | q = p + prevlensize + lensize; | |
801 | ||
802 | if (skipcnt == 0) { | |
803 | /* Compare current entry with specified entry */ | |
804 | if (ZIP_IS_STR(encoding)) { | |
805 | if (len == vlen && memcmp(q, vstr, vlen) == 0) { | |
806 | return p; | |
807 | } | |
808 | } else { | |
809 | /* Find out if the searched field can be encoded. Note that | |
810 | * we do it only the first time, once done vencoding is set | |
811 | * to non-zero and vll is set to the integer value. */ | |
812 | if (vencoding == 0) { | |
813 | if (!zipTryEncoding(vstr, vlen, &vll, &vencoding)) { | |
814 | /* If the entry can't be encoded we set it to | |
815 | * UCHAR_MAX so that we don't retry again the next | |
816 | * time. */ | |
817 | vencoding = UCHAR_MAX; | |
818 | } | |
819 | /* Must be non-zero by now */ | |
820 | assert(vencoding); | |
821 | } | |
822 | ||
823 | /* Compare current entry with specified entry, do it only | |
824 | * if vencoding != UCHAR_MAX because if there is no encoding | |
825 | * possible for the field it can't be a valid integer. */ | |
826 | if (vencoding != UCHAR_MAX) { | |
827 | long long ll = zipLoadInteger(q, encoding); | |
828 | if (ll == vll) { | |
829 | return p; | |
830 | } | |
831 | } | |
832 | } | |
833 | ||
834 | /* Reset skip count */ | |
835 | skipcnt = skip; | |
836 | } else { | |
837 | /* Skip entry */ | |
838 | skipcnt--; | |
839 | } | |
840 | ||
841 | /* Move to next entry */ | |
842 | p = q + len; | |
843 | } | |
844 | ||
845 | return NULL; | |
846 | } | |
847 | ||
848 | /* Return length of ziplist. */ | |
849 | unsigned int ziplistLen(unsigned char *zl) { | |
850 | unsigned int len = 0; | |
851 | if (intrev16ifbe(ZIPLIST_LENGTH(zl)) < UINT16_MAX) { | |
852 | len = intrev16ifbe(ZIPLIST_LENGTH(zl)); | |
853 | } else { | |
854 | unsigned char *p = zl+ZIPLIST_HEADER_SIZE; | |
855 | while (*p != ZIP_END) { | |
856 | p += zipRawEntryLength(p); | |
857 | len++; | |
858 | } | |
859 | ||
860 | /* Re-store length if small enough */ | |
861 | if (len < UINT16_MAX) ZIPLIST_LENGTH(zl) = intrev16ifbe(len); | |
862 | } | |
863 | return len; | |
864 | } | |
865 | ||
866 | /* Return ziplist blob size in bytes. */ | |
867 | size_t ziplistBlobLen(unsigned char *zl) { | |
868 | return intrev32ifbe(ZIPLIST_BYTES(zl)); | |
869 | } | |
870 | ||
871 | void ziplistRepr(unsigned char *zl) { | |
872 | unsigned char *p; | |
873 | int index = 0; | |
874 | zlentry entry; | |
875 | ||
876 | printf( | |
877 | "{total bytes %d} " | |
878 | "{length %u}\n" | |
879 | "{tail offset %u}\n", | |
880 | intrev32ifbe(ZIPLIST_BYTES(zl)), | |
881 | intrev16ifbe(ZIPLIST_LENGTH(zl)), | |
882 | intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl))); | |
883 | p = ZIPLIST_ENTRY_HEAD(zl); | |
884 | while(*p != ZIP_END) { | |
885 | entry = zipEntry(p); | |
886 | printf( | |
887 | "{" | |
888 | "addr 0x%08lx, " | |
889 | "index %2d, " | |
890 | "offset %5ld, " | |
891 | "rl: %5u, " | |
892 | "hs %2u, " | |
893 | "pl: %5u, " | |
894 | "pls: %2u, " | |
895 | "payload %5u" | |
896 | "} ", | |
897 | (long unsigned)p, | |
898 | index, | |
899 | (unsigned long) (p-zl), | |
900 | entry.headersize+entry.len, | |
901 | entry.headersize, | |
902 | entry.prevrawlen, | |
903 | entry.prevrawlensize, | |
904 | entry.len); | |
905 | p += entry.headersize; | |
906 | if (ZIP_IS_STR(entry.encoding)) { | |
907 | if (entry.len > 40) { | |
908 | if (fwrite(p,40,1,stdout) == 0) perror("fwrite"); | |
909 | printf("..."); | |
910 | } else { | |
911 | if (entry.len && | |
912 | fwrite(p,entry.len,1,stdout) == 0) perror("fwrite"); | |
913 | } | |
914 | } else { | |
915 | printf("%lld", (long long) zipLoadInteger(p,entry.encoding)); | |
916 | } | |
917 | printf("\n"); | |
918 | p += entry.len; | |
919 | index++; | |
920 | } | |
921 | printf("{end}\n\n"); | |
922 | } | |
923 | ||
924 | #ifdef ZIPLIST_TEST_MAIN | |
925 | #include <sys/time.h> | |
926 | #include "adlist.h" | |
927 | #include "sds.h" | |
928 | ||
929 | #define debug(f, ...) { if (DEBUG) printf(f, __VA_ARGS__); } | |
930 | ||
931 | unsigned char *createList() { | |
932 | unsigned char *zl = ziplistNew(); | |
933 | zl = ziplistPush(zl, (unsigned char*)"foo", 3, ZIPLIST_TAIL); | |
934 | zl = ziplistPush(zl, (unsigned char*)"quux", 4, ZIPLIST_TAIL); | |
935 | zl = ziplistPush(zl, (unsigned char*)"hello", 5, ZIPLIST_HEAD); | |
936 | zl = ziplistPush(zl, (unsigned char*)"1024", 4, ZIPLIST_TAIL); | |
937 | return zl; | |
938 | } | |
939 | ||
940 | unsigned char *createIntList() { | |
941 | unsigned char *zl = ziplistNew(); | |
942 | char buf[32]; | |
943 | ||
944 | sprintf(buf, "100"); | |
945 | zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL); | |
946 | sprintf(buf, "128000"); | |
947 | zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL); | |
948 | sprintf(buf, "-100"); | |
949 | zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_HEAD); | |
950 | sprintf(buf, "4294967296"); | |
951 | zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_HEAD); | |
952 | sprintf(buf, "non integer"); | |
953 | zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL); | |
954 | sprintf(buf, "much much longer non integer"); | |
955 | zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL); | |
956 | return zl; | |
957 | } | |
958 | ||
959 | long long usec(void) { | |
960 | struct timeval tv; | |
961 | gettimeofday(&tv,NULL); | |
962 | return (((long long)tv.tv_sec)*1000000)+tv.tv_usec; | |
963 | } | |
964 | ||
965 | void stress(int pos, int num, int maxsize, int dnum) { | |
966 | int i,j,k; | |
967 | unsigned char *zl; | |
968 | char posstr[2][5] = { "HEAD", "TAIL" }; | |
969 | long long start; | |
970 | for (i = 0; i < maxsize; i+=dnum) { | |
971 | zl = ziplistNew(); | |
972 | for (j = 0; j < i; j++) { | |
973 | zl = ziplistPush(zl,(unsigned char*)"quux",4,ZIPLIST_TAIL); | |
974 | } | |
975 | ||
976 | /* Do num times a push+pop from pos */ | |
977 | start = usec(); | |
978 | for (k = 0; k < num; k++) { | |
979 | zl = ziplistPush(zl,(unsigned char*)"quux",4,pos); | |
980 | zl = ziplistDeleteRange(zl,0,1); | |
981 | } | |
982 | printf("List size: %8d, bytes: %8d, %dx push+pop (%s): %6lld usec\n", | |
983 | i,intrev32ifbe(ZIPLIST_BYTES(zl)),num,posstr[pos],usec()-start); | |
984 | zfree(zl); | |
985 | } | |
986 | } | |
987 | ||
988 | void pop(unsigned char *zl, int where) { | |
989 | unsigned char *p, *vstr; | |
990 | unsigned int vlen; | |
991 | long long vlong; | |
992 | ||
993 | p = ziplistIndex(zl,where == ZIPLIST_HEAD ? 0 : -1); | |
994 | if (ziplistGet(p,&vstr,&vlen,&vlong)) { | |
995 | if (where == ZIPLIST_HEAD) | |
996 | printf("Pop head: "); | |
997 | else | |
998 | printf("Pop tail: "); | |
999 | ||
1000 | if (vstr) | |
1001 | if (vlen && fwrite(vstr,vlen,1,stdout) == 0) perror("fwrite"); | |
1002 | else | |
1003 | printf("%lld", vlong); | |
1004 | ||
1005 | printf("\n"); | |
1006 | ziplistDeleteRange(zl,-1,1); | |
1007 | } else { | |
1008 | printf("ERROR: Could not pop\n"); | |
1009 | exit(1); | |
1010 | } | |
1011 | } | |
1012 | ||
1013 | int randstring(char *target, unsigned int min, unsigned int max) { | |
1014 | int p, len = min+rand()%(max-min+1); | |
1015 | int minval, maxval; | |
1016 | switch(rand() % 3) { | |
1017 | case 0: | |
1018 | minval = 0; | |
1019 | maxval = 255; | |
1020 | break; | |
1021 | case 1: | |
1022 | minval = 48; | |
1023 | maxval = 122; | |
1024 | break; | |
1025 | case 2: | |
1026 | minval = 48; | |
1027 | maxval = 52; | |
1028 | break; | |
1029 | default: | |
1030 | assert(NULL); | |
1031 | } | |
1032 | ||
1033 | while(p < len) | |
1034 | target[p++] = minval+rand()%(maxval-minval+1); | |
1035 | return len; | |
1036 | } | |
1037 | ||
1038 | void verify(unsigned char *zl, zlentry *e) { | |
1039 | int i; | |
1040 | int len = ziplistLen(zl); | |
1041 | zlentry _e; | |
1042 | ||
1043 | for (i = 0; i < len; i++) { | |
1044 | memset(&e[i], 0, sizeof(zlentry)); | |
1045 | e[i] = zipEntry(ziplistIndex(zl, i)); | |
1046 | ||
1047 | memset(&_e, 0, sizeof(zlentry)); | |
1048 | _e = zipEntry(ziplistIndex(zl, -len+i)); | |
1049 | ||
1050 | assert(memcmp(&e[i], &_e, sizeof(zlentry)) == 0); | |
1051 | } | |
1052 | } | |
1053 | ||
1054 | int main(int argc, char **argv) { | |
1055 | unsigned char *zl, *p; | |
1056 | unsigned char *entry; | |
1057 | unsigned int elen; | |
1058 | long long value; | |
1059 | ||
1060 | /* If an argument is given, use it as the random seed. */ | |
1061 | if (argc == 2) | |
1062 | srand(atoi(argv[1])); | |
1063 | ||
1064 | zl = createIntList(); | |
1065 | ziplistRepr(zl); | |
1066 | ||
1067 | zl = createList(); | |
1068 | ziplistRepr(zl); | |
1069 | ||
1070 | pop(zl,ZIPLIST_TAIL); | |
1071 | ziplistRepr(zl); | |
1072 | ||
1073 | pop(zl,ZIPLIST_HEAD); | |
1074 | ziplistRepr(zl); | |
1075 | ||
1076 | pop(zl,ZIPLIST_TAIL); | |
1077 | ziplistRepr(zl); | |
1078 | ||
1079 | pop(zl,ZIPLIST_TAIL); | |
1080 | ziplistRepr(zl); | |
1081 | ||
1082 | printf("Get element at index 3:\n"); | |
1083 | { | |
1084 | zl = createList(); | |
1085 | p = ziplistIndex(zl, 3); | |
1086 | if (!ziplistGet(p, &entry, &elen, &value)) { | |
1087 | printf("ERROR: Could not access index 3\n"); | |
1088 | return 1; | |
1089 | } | |
1090 | if (entry) { | |
1091 | if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite"); | |
1092 | printf("\n"); | |
1093 | } else { | |
1094 | printf("%lld\n", value); | |
1095 | } | |
1096 | printf("\n"); | |
1097 | } | |
1098 | ||
1099 | printf("Get element at index 4 (out of range):\n"); | |
1100 | { | |
1101 | zl = createList(); | |
1102 | p = ziplistIndex(zl, 4); | |
1103 | if (p == NULL) { | |
1104 | printf("No entry\n"); | |
1105 | } else { | |
1106 | printf("ERROR: Out of range index should return NULL, returned offset: %ld\n", p-zl); | |
1107 | return 1; | |
1108 | } | |
1109 | printf("\n"); | |
1110 | } | |
1111 | ||
1112 | printf("Get element at index -1 (last element):\n"); | |
1113 | { | |
1114 | zl = createList(); | |
1115 | p = ziplistIndex(zl, -1); | |
1116 | if (!ziplistGet(p, &entry, &elen, &value)) { | |
1117 | printf("ERROR: Could not access index -1\n"); | |
1118 | return 1; | |
1119 | } | |
1120 | if (entry) { | |
1121 | if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite"); | |
1122 | printf("\n"); | |
1123 | } else { | |
1124 | printf("%lld\n", value); | |
1125 | } | |
1126 | printf("\n"); | |
1127 | } | |
1128 | ||
1129 | printf("Get element at index -4 (first element):\n"); | |
1130 | { | |
1131 | zl = createList(); | |
1132 | p = ziplistIndex(zl, -4); | |
1133 | if (!ziplistGet(p, &entry, &elen, &value)) { | |
1134 | printf("ERROR: Could not access index -4\n"); | |
1135 | return 1; | |
1136 | } | |
1137 | if (entry) { | |
1138 | if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite"); | |
1139 | printf("\n"); | |
1140 | } else { | |
1141 | printf("%lld\n", value); | |
1142 | } | |
1143 | printf("\n"); | |
1144 | } | |
1145 | ||
1146 | printf("Get element at index -5 (reverse out of range):\n"); | |
1147 | { | |
1148 | zl = createList(); | |
1149 | p = ziplistIndex(zl, -5); | |
1150 | if (p == NULL) { | |
1151 | printf("No entry\n"); | |
1152 | } else { | |
1153 | printf("ERROR: Out of range index should return NULL, returned offset: %ld\n", p-zl); | |
1154 | return 1; | |
1155 | } | |
1156 | printf("\n"); | |
1157 | } | |
1158 | ||
1159 | printf("Iterate list from 0 to end:\n"); | |
1160 | { | |
1161 | zl = createList(); | |
1162 | p = ziplistIndex(zl, 0); | |
1163 | while (ziplistGet(p, &entry, &elen, &value)) { | |
1164 | printf("Entry: "); | |
1165 | if (entry) { | |
1166 | if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite"); | |
1167 | } else { | |
1168 | printf("%lld", value); | |
1169 | } | |
1170 | p = ziplistNext(zl,p); | |
1171 | printf("\n"); | |
1172 | } | |
1173 | printf("\n"); | |
1174 | } | |
1175 | ||
1176 | printf("Iterate list from 1 to end:\n"); | |
1177 | { | |
1178 | zl = createList(); | |
1179 | p = ziplistIndex(zl, 1); | |
1180 | while (ziplistGet(p, &entry, &elen, &value)) { | |
1181 | printf("Entry: "); | |
1182 | if (entry) { | |
1183 | if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite"); | |
1184 | } else { | |
1185 | printf("%lld", value); | |
1186 | } | |
1187 | p = ziplistNext(zl,p); | |
1188 | printf("\n"); | |
1189 | } | |
1190 | printf("\n"); | |
1191 | } | |
1192 | ||
1193 | printf("Iterate list from 2 to end:\n"); | |
1194 | { | |
1195 | zl = createList(); | |
1196 | p = ziplistIndex(zl, 2); | |
1197 | while (ziplistGet(p, &entry, &elen, &value)) { | |
1198 | printf("Entry: "); | |
1199 | if (entry) { | |
1200 | if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite"); | |
1201 | } else { | |
1202 | printf("%lld", value); | |
1203 | } | |
1204 | p = ziplistNext(zl,p); | |
1205 | printf("\n"); | |
1206 | } | |
1207 | printf("\n"); | |
1208 | } | |
1209 | ||
1210 | printf("Iterate starting out of range:\n"); | |
1211 | { | |
1212 | zl = createList(); | |
1213 | p = ziplistIndex(zl, 4); | |
1214 | if (!ziplistGet(p, &entry, &elen, &value)) { | |
1215 | printf("No entry\n"); | |
1216 | } else { | |
1217 | printf("ERROR\n"); | |
1218 | } | |
1219 | printf("\n"); | |
1220 | } | |
1221 | ||
1222 | printf("Iterate from back to front:\n"); | |
1223 | { | |
1224 | zl = createList(); | |
1225 | p = ziplistIndex(zl, -1); | |
1226 | while (ziplistGet(p, &entry, &elen, &value)) { | |
1227 | printf("Entry: "); | |
1228 | if (entry) { | |
1229 | if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite"); | |
1230 | } else { | |
1231 | printf("%lld", value); | |
1232 | } | |
1233 | p = ziplistPrev(zl,p); | |
1234 | printf("\n"); | |
1235 | } | |
1236 | printf("\n"); | |
1237 | } | |
1238 | ||
1239 | printf("Iterate from back to front, deleting all items:\n"); | |
1240 | { | |
1241 | zl = createList(); | |
1242 | p = ziplistIndex(zl, -1); | |
1243 | while (ziplistGet(p, &entry, &elen, &value)) { | |
1244 | printf("Entry: "); | |
1245 | if (entry) { | |
1246 | if (elen && fwrite(entry,elen,1,stdout) == 0) perror("fwrite"); | |
1247 | } else { | |
1248 | printf("%lld", value); | |
1249 | } | |
1250 | zl = ziplistDelete(zl,&p); | |
1251 | p = ziplistPrev(zl,p); | |
1252 | printf("\n"); | |
1253 | } | |
1254 | printf("\n"); | |
1255 | } | |
1256 | ||
1257 | printf("Delete inclusive range 0,0:\n"); | |
1258 | { | |
1259 | zl = createList(); | |
1260 | zl = ziplistDeleteRange(zl, 0, 1); | |
1261 | ziplistRepr(zl); | |
1262 | } | |
1263 | ||
1264 | printf("Delete inclusive range 0,1:\n"); | |
1265 | { | |
1266 | zl = createList(); | |
1267 | zl = ziplistDeleteRange(zl, 0, 2); | |
1268 | ziplistRepr(zl); | |
1269 | } | |
1270 | ||
1271 | printf("Delete inclusive range 1,2:\n"); | |
1272 | { | |
1273 | zl = createList(); | |
1274 | zl = ziplistDeleteRange(zl, 1, 2); | |
1275 | ziplistRepr(zl); | |
1276 | } | |
1277 | ||
1278 | printf("Delete with start index out of range:\n"); | |
1279 | { | |
1280 | zl = createList(); | |
1281 | zl = ziplistDeleteRange(zl, 5, 1); | |
1282 | ziplistRepr(zl); | |
1283 | } | |
1284 | ||
1285 | printf("Delete with num overflow:\n"); | |
1286 | { | |
1287 | zl = createList(); | |
1288 | zl = ziplistDeleteRange(zl, 1, 5); | |
1289 | ziplistRepr(zl); | |
1290 | } | |
1291 | ||
1292 | printf("Delete foo while iterating:\n"); | |
1293 | { | |
1294 | zl = createList(); | |
1295 | p = ziplistIndex(zl,0); | |
1296 | while (ziplistGet(p,&entry,&elen,&value)) { | |
1297 | if (entry && strncmp("foo",(char*)entry,elen) == 0) { | |
1298 | printf("Delete foo\n"); | |
1299 | zl = ziplistDelete(zl,&p); | |
1300 | } else { | |
1301 | printf("Entry: "); | |
1302 | if (entry) { | |
1303 | if (elen && fwrite(entry,elen,1,stdout) == 0) | |
1304 | perror("fwrite"); | |
1305 | } else { | |
1306 | printf("%lld",value); | |
1307 | } | |
1308 | p = ziplistNext(zl,p); | |
1309 | printf("\n"); | |
1310 | } | |
1311 | } | |
1312 | printf("\n"); | |
1313 | ziplistRepr(zl); | |
1314 | } | |
1315 | ||
1316 | printf("Regression test for >255 byte strings:\n"); | |
1317 | { | |
1318 | char v1[257],v2[257]; | |
1319 | memset(v1,'x',256); | |
1320 | memset(v2,'y',256); | |
1321 | zl = ziplistNew(); | |
1322 | zl = ziplistPush(zl,(unsigned char*)v1,strlen(v1),ZIPLIST_TAIL); | |
1323 | zl = ziplistPush(zl,(unsigned char*)v2,strlen(v2),ZIPLIST_TAIL); | |
1324 | ||
1325 | /* Pop values again and compare their value. */ | |
1326 | p = ziplistIndex(zl,0); | |
1327 | assert(ziplistGet(p,&entry,&elen,&value)); | |
1328 | assert(strncmp(v1,(char*)entry,elen) == 0); | |
1329 | p = ziplistIndex(zl,1); | |
1330 | assert(ziplistGet(p,&entry,&elen,&value)); | |
1331 | assert(strncmp(v2,(char*)entry,elen) == 0); | |
1332 | printf("SUCCESS\n\n"); | |
1333 | } | |
1334 | ||
1335 | printf("Regression test deleting next to last entries:\n"); | |
1336 | { | |
1337 | char v[3][257]; | |
1338 | zlentry e[3]; | |
1339 | int i; | |
1340 | ||
1341 | for (i = 0; i < (sizeof(v)/sizeof(v[0])); i++) { | |
1342 | memset(v[i], 'a' + i, sizeof(v[0])); | |
1343 | } | |
1344 | ||
1345 | v[0][256] = '\0'; | |
1346 | v[1][ 1] = '\0'; | |
1347 | v[2][256] = '\0'; | |
1348 | ||
1349 | zl = ziplistNew(); | |
1350 | for (i = 0; i < (sizeof(v)/sizeof(v[0])); i++) { | |
1351 | zl = ziplistPush(zl, (unsigned char *) v[i], strlen(v[i]), ZIPLIST_TAIL); | |
1352 | } | |
1353 | ||
1354 | verify(zl, e); | |
1355 | ||
1356 | assert(e[0].prevrawlensize == 1); | |
1357 | assert(e[1].prevrawlensize == 5); | |
1358 | assert(e[2].prevrawlensize == 1); | |
1359 | ||
1360 | /* Deleting entry 1 will increase `prevrawlensize` for entry 2 */ | |
1361 | unsigned char *p = e[1].p; | |
1362 | zl = ziplistDelete(zl, &p); | |
1363 | ||
1364 | verify(zl, e); | |
1365 | ||
1366 | assert(e[0].prevrawlensize == 1); | |
1367 | assert(e[1].prevrawlensize == 5); | |
1368 | ||
1369 | printf("SUCCESS\n\n"); | |
1370 | } | |
1371 | ||
1372 | printf("Create long list and check indices:\n"); | |
1373 | { | |
1374 | zl = ziplistNew(); | |
1375 | char buf[32]; | |
1376 | int i,len; | |
1377 | for (i = 0; i < 1000; i++) { | |
1378 | len = sprintf(buf,"%d",i); | |
1379 | zl = ziplistPush(zl,(unsigned char*)buf,len,ZIPLIST_TAIL); | |
1380 | } | |
1381 | for (i = 0; i < 1000; i++) { | |
1382 | p = ziplistIndex(zl,i); | |
1383 | assert(ziplistGet(p,NULL,NULL,&value)); | |
1384 | assert(i == value); | |
1385 | ||
1386 | p = ziplistIndex(zl,-i-1); | |
1387 | assert(ziplistGet(p,NULL,NULL,&value)); | |
1388 | assert(999-i == value); | |
1389 | } | |
1390 | printf("SUCCESS\n\n"); | |
1391 | } | |
1392 | ||
1393 | printf("Compare strings with ziplist entries:\n"); | |
1394 | { | |
1395 | zl = createList(); | |
1396 | p = ziplistIndex(zl,0); | |
1397 | if (!ziplistCompare(p,(unsigned char*)"hello",5)) { | |
1398 | printf("ERROR: not \"hello\"\n"); | |
1399 | return 1; | |
1400 | } | |
1401 | if (ziplistCompare(p,(unsigned char*)"hella",5)) { | |
1402 | printf("ERROR: \"hella\"\n"); | |
1403 | return 1; | |
1404 | } | |
1405 | ||
1406 | p = ziplistIndex(zl,3); | |
1407 | if (!ziplistCompare(p,(unsigned char*)"1024",4)) { | |
1408 | printf("ERROR: not \"1024\"\n"); | |
1409 | return 1; | |
1410 | } | |
1411 | if (ziplistCompare(p,(unsigned char*)"1025",4)) { | |
1412 | printf("ERROR: \"1025\"\n"); | |
1413 | return 1; | |
1414 | } | |
1415 | printf("SUCCESS\n\n"); | |
1416 | } | |
1417 | ||
1418 | printf("Stress with random payloads of different encoding:\n"); | |
1419 | { | |
1420 | int i,j,len,where; | |
1421 | unsigned char *p; | |
1422 | char buf[1024]; | |
1423 | int buflen; | |
1424 | list *ref; | |
1425 | listNode *refnode; | |
1426 | ||
1427 | /* Hold temp vars from ziplist */ | |
1428 | unsigned char *sstr; | |
1429 | unsigned int slen; | |
1430 | long long sval; | |
1431 | ||
1432 | for (i = 0; i < 20000; i++) { | |
1433 | zl = ziplistNew(); | |
1434 | ref = listCreate(); | |
1435 | listSetFreeMethod(ref,sdsfree); | |
1436 | len = rand() % 256; | |
1437 | ||
1438 | /* Create lists */ | |
1439 | for (j = 0; j < len; j++) { | |
1440 | where = (rand() & 1) ? ZIPLIST_HEAD : ZIPLIST_TAIL; | |
1441 | if (rand() % 2) { | |
1442 | buflen = randstring(buf,1,sizeof(buf)-1); | |
1443 | } else { | |
1444 | switch(rand() % 3) { | |
1445 | case 0: | |
1446 | buflen = sprintf(buf,"%lld",(0LL + rand()) >> 20); | |
1447 | break; | |
1448 | case 1: | |
1449 | buflen = sprintf(buf,"%lld",(0LL + rand())); | |
1450 | break; | |
1451 | case 2: | |
1452 | buflen = sprintf(buf,"%lld",(0LL + rand()) << 20); | |
1453 | break; | |
1454 | default: | |
1455 | assert(NULL); | |
1456 | } | |
1457 | } | |
1458 | ||
1459 | /* Add to ziplist */ | |
1460 | zl = ziplistPush(zl, (unsigned char*)buf, buflen, where); | |
1461 | ||
1462 | /* Add to reference list */ | |
1463 | if (where == ZIPLIST_HEAD) { | |
1464 | listAddNodeHead(ref,sdsnewlen(buf, buflen)); | |
1465 | } else if (where == ZIPLIST_TAIL) { | |
1466 | listAddNodeTail(ref,sdsnewlen(buf, buflen)); | |
1467 | } else { | |
1468 | assert(NULL); | |
1469 | } | |
1470 | } | |
1471 | ||
1472 | assert(listLength(ref) == ziplistLen(zl)); | |
1473 | for (j = 0; j < len; j++) { | |
1474 | /* Naive way to get elements, but similar to the stresser | |
1475 | * executed from the Tcl test suite. */ | |
1476 | p = ziplistIndex(zl,j); | |
1477 | refnode = listIndex(ref,j); | |
1478 | ||
1479 | assert(ziplistGet(p,&sstr,&slen,&sval)); | |
1480 | if (sstr == NULL) { | |
1481 | buflen = sprintf(buf,"%lld",sval); | |
1482 | } else { | |
1483 | buflen = slen; | |
1484 | memcpy(buf,sstr,buflen); | |
1485 | buf[buflen] = '\0'; | |
1486 | } | |
1487 | assert(memcmp(buf,listNodeValue(refnode),buflen) == 0); | |
1488 | } | |
1489 | zfree(zl); | |
1490 | listRelease(ref); | |
1491 | } | |
1492 | printf("SUCCESS\n\n"); | |
1493 | } | |
1494 | ||
1495 | printf("Stress with variable ziplist size:\n"); | |
1496 | { | |
1497 | stress(ZIPLIST_HEAD,100000,16384,256); | |
1498 | stress(ZIPLIST_TAIL,100000,16384,256); | |
1499 | } | |
1500 | ||
1501 | return 0; | |
1502 | } | |
1503 | ||
1504 | #endif |