| 1 | #include <assert.h> |
| 2 | #include <stdbool.h> |
| 3 | #include <stdio.h> |
| 4 | #include <stdlib.h> |
| 5 | #include <string.h> |
| 6 | #include <inttypes.h> |
| 7 | |
| 8 | #include <roaring/containers/bitset.h> |
| 9 | #include <roaring/containers/containers.h> |
| 10 | #include <roaring/roaring_array.h> |
| 11 | |
| 12 | // Convention: [0,ra->size) all elements are initialized |
| 13 | // [ra->size, ra->allocation_size) is junk and contains nothing needing freeing |
| 14 | |
| 15 | extern inline int32_t ra_get_size(const roaring_array_t *ra); |
| 16 | extern inline int32_t ra_get_index(const roaring_array_t *ra, uint16_t x); |
| 17 | extern inline void *ra_get_container_at_index(const roaring_array_t *ra, |
| 18 | uint16_t i, uint8_t *typecode); |
| 19 | extern inline void ra_unshare_container_at_index(roaring_array_t *ra, |
| 20 | uint16_t i); |
| 21 | extern inline void ra_replace_key_and_container_at_index(roaring_array_t *ra, |
| 22 | int32_t i, |
| 23 | uint16_t key, void *c, |
| 24 | uint8_t typecode); |
| 25 | extern inline void ra_set_container_at_index(const roaring_array_t *ra, |
| 26 | int32_t i, void *c, |
| 27 | uint8_t typecode); |
| 28 | |
| 29 | static bool realloc_array(roaring_array_t *ra, int32_t new_capacity) { |
| 30 | // because we combine the allocations, it is not possible to use realloc |
| 31 | /*ra->keys = |
| 32 | (uint16_t *)realloc(ra->keys, sizeof(uint16_t) * new_capacity); |
| 33 | ra->containers = |
| 34 | (void **)realloc(ra->containers, sizeof(void *) * new_capacity); |
| 35 | ra->typecodes = |
| 36 | (uint8_t *)realloc(ra->typecodes, sizeof(uint8_t) * new_capacity); |
| 37 | if (!ra->keys || !ra->containers || !ra->typecodes) { |
| 38 | free(ra->keys); |
| 39 | free(ra->containers); |
| 40 | free(ra->typecodes); |
| 41 | return false; |
| 42 | }*/ |
| 43 | |
| 44 | if ( new_capacity == 0 ) { |
| 45 | free(ra->containers); |
| 46 | ra->containers = NULL; |
| 47 | ra->keys = NULL; |
| 48 | ra->typecodes = NULL; |
| 49 | ra->allocation_size = 0; |
| 50 | return true; |
| 51 | } |
| 52 | const size_t memoryneeded = |
| 53 | new_capacity * (sizeof(uint16_t) + sizeof(void *) + sizeof(uint8_t)); |
| 54 | void *bigalloc = malloc(memoryneeded); |
| 55 | if (!bigalloc) return false; |
| 56 | void *oldbigalloc = ra->containers; |
| 57 | void **newcontainers = (void **)bigalloc; |
| 58 | uint16_t *newkeys = (uint16_t *)(newcontainers + new_capacity); |
| 59 | uint8_t *newtypecodes = (uint8_t *)(newkeys + new_capacity); |
| 60 | assert((char *)(newtypecodes + new_capacity) == |
| 61 | (char *)bigalloc + memoryneeded); |
| 62 | if(ra->size > 0) { |
| 63 | memcpy(newcontainers, ra->containers, sizeof(void *) * ra->size); |
| 64 | memcpy(newkeys, ra->keys, sizeof(uint16_t) * ra->size); |
| 65 | memcpy(newtypecodes, ra->typecodes, sizeof(uint8_t) * ra->size); |
| 66 | } |
| 67 | ra->containers = newcontainers; |
| 68 | ra->keys = newkeys; |
| 69 | ra->typecodes = newtypecodes; |
| 70 | ra->allocation_size = new_capacity; |
| 71 | free(oldbigalloc); |
| 72 | return true; |
| 73 | } |
| 74 | |
| 75 | bool ra_init_with_capacity(roaring_array_t *new_ra, uint32_t cap) { |
| 76 | if (!new_ra) return false; |
| 77 | ra_init(new_ra); |
| 78 | |
| 79 | if (cap > INT32_MAX) { return false; } |
| 80 | |
| 81 | if(cap > 0) { |
| 82 | void *bigalloc = |
| 83 | malloc(cap * (sizeof(uint16_t) + sizeof(void *) + sizeof(uint8_t))); |
| 84 | if( bigalloc == NULL ) return false; |
| 85 | new_ra->containers = (void **)bigalloc; |
| 86 | new_ra->keys = (uint16_t *)(new_ra->containers + cap); |
| 87 | new_ra->typecodes = (uint8_t *)(new_ra->keys + cap); |
| 88 | // Narrowing is safe because of above check |
| 89 | new_ra->allocation_size = (int32_t)cap; |
| 90 | } |
| 91 | return true; |
| 92 | } |
| 93 | |
| 94 | int ra_shrink_to_fit(roaring_array_t *ra) { |
| 95 | int savings = (ra->allocation_size - ra->size) * |
| 96 | (sizeof(uint16_t) + sizeof(void *) + sizeof(uint8_t)); |
| 97 | if (!realloc_array(ra, ra->size)) { |
| 98 | return 0; |
| 99 | } |
| 100 | ra->allocation_size = ra->size; |
| 101 | return savings; |
| 102 | } |
| 103 | |
| 104 | void ra_init(roaring_array_t *new_ra) { |
| 105 | if (!new_ra) { return; } |
| 106 | new_ra->keys = NULL; |
| 107 | new_ra->containers = NULL; |
| 108 | new_ra->typecodes = NULL; |
| 109 | |
| 110 | new_ra->allocation_size = 0; |
| 111 | new_ra->size = 0; |
| 112 | new_ra->flags = 0; |
| 113 | } |
| 114 | |
| 115 | bool ra_copy(const roaring_array_t *source, roaring_array_t *dest, |
| 116 | bool copy_on_write) { |
| 117 | if (!ra_init_with_capacity(dest, source->size)) return false; |
| 118 | dest->size = source->size; |
| 119 | dest->allocation_size = source->size; |
| 120 | if(dest->size > 0) { |
| 121 | memcpy(dest->keys, source->keys, dest->size * sizeof(uint16_t)); |
| 122 | } |
| 123 | // we go through the containers, turning them into shared containers... |
| 124 | if (copy_on_write) { |
| 125 | for (int32_t i = 0; i < dest->size; ++i) { |
| 126 | source->containers[i] = get_copy_of_container( |
| 127 | source->containers[i], &source->typecodes[i], copy_on_write); |
| 128 | } |
| 129 | // we do a shallow copy to the other bitmap |
| 130 | if(dest->size > 0) { |
| 131 | memcpy(dest->containers, source->containers, |
| 132 | dest->size * sizeof(void *)); |
| 133 | memcpy(dest->typecodes, source->typecodes, |
| 134 | dest->size * sizeof(uint8_t)); |
| 135 | } |
| 136 | } else { |
| 137 | if(dest->size > 0) { |
| 138 | memcpy(dest->typecodes, source->typecodes, |
| 139 | dest->size * sizeof(uint8_t)); |
| 140 | } |
| 141 | for (int32_t i = 0; i < dest->size; i++) { |
| 142 | dest->containers[i] = |
| 143 | container_clone(source->containers[i], source->typecodes[i]); |
| 144 | if (dest->containers[i] == NULL) { |
| 145 | for (int32_t j = 0; j < i; j++) { |
| 146 | container_free(dest->containers[j], dest->typecodes[j]); |
| 147 | } |
| 148 | ra_clear_without_containers(dest); |
| 149 | return false; |
| 150 | } |
| 151 | } |
| 152 | } |
| 153 | return true; |
| 154 | } |
| 155 | |
| 156 | bool ra_overwrite(const roaring_array_t *source, roaring_array_t *dest, |
| 157 | bool copy_on_write) { |
| 158 | ra_clear_containers(dest); // we are going to overwrite them |
| 159 | if (dest->allocation_size < source->size) { |
| 160 | if (!realloc_array(dest, source->size)) { |
| 161 | return false; |
| 162 | } |
| 163 | } |
| 164 | dest->size = source->size; |
| 165 | memcpy(dest->keys, source->keys, dest->size * sizeof(uint16_t)); |
| 166 | // we go through the containers, turning them into shared containers... |
| 167 | if (copy_on_write) { |
| 168 | for (int32_t i = 0; i < dest->size; ++i) { |
| 169 | source->containers[i] = get_copy_of_container( |
| 170 | source->containers[i], &source->typecodes[i], copy_on_write); |
| 171 | } |
| 172 | // we do a shallow copy to the other bitmap |
| 173 | memcpy(dest->containers, source->containers, |
| 174 | dest->size * sizeof(void *)); |
| 175 | memcpy(dest->typecodes, source->typecodes, |
| 176 | dest->size * sizeof(uint8_t)); |
| 177 | } else { |
| 178 | memcpy(dest->typecodes, source->typecodes, |
| 179 | dest->size * sizeof(uint8_t)); |
| 180 | for (int32_t i = 0; i < dest->size; i++) { |
| 181 | dest->containers[i] = |
| 182 | container_clone(source->containers[i], source->typecodes[i]); |
| 183 | if (dest->containers[i] == NULL) { |
| 184 | for (int32_t j = 0; j < i; j++) { |
| 185 | container_free(dest->containers[j], dest->typecodes[j]); |
| 186 | } |
| 187 | ra_clear_without_containers(dest); |
| 188 | return false; |
| 189 | } |
| 190 | } |
| 191 | } |
| 192 | return true; |
| 193 | } |
| 194 | |
| 195 | void ra_clear_containers(roaring_array_t *ra) { |
| 196 | for (int32_t i = 0; i < ra->size; ++i) { |
| 197 | container_free(ra->containers[i], ra->typecodes[i]); |
| 198 | } |
| 199 | } |
| 200 | |
| 201 | void ra_reset(roaring_array_t *ra) { |
| 202 | ra_clear_containers(ra); |
| 203 | ra->size = 0; |
| 204 | ra_shrink_to_fit(ra); |
| 205 | } |
| 206 | |
| 207 | void ra_clear_without_containers(roaring_array_t *ra) { |
| 208 | free(ra->containers); // keys and typecodes are allocated with containers |
| 209 | ra->size = 0; |
| 210 | ra->allocation_size = 0; |
| 211 | ra->containers = NULL; |
| 212 | ra->keys = NULL; |
| 213 | ra->typecodes = NULL; |
| 214 | } |
| 215 | |
| 216 | void ra_clear(roaring_array_t *ra) { |
| 217 | ra_clear_containers(ra); |
| 218 | ra_clear_without_containers(ra); |
| 219 | } |
| 220 | |
| 221 | bool extend_array(roaring_array_t *ra, int32_t k) { |
| 222 | int32_t desired_size = ra->size + k; |
| 223 | assert(desired_size <= MAX_CONTAINERS); |
| 224 | if (desired_size > ra->allocation_size) { |
| 225 | int32_t new_capacity = |
| 226 | (ra->size < 1024) ? 2 * desired_size : 5 * desired_size / 4; |
| 227 | if (new_capacity > MAX_CONTAINERS) { |
| 228 | new_capacity = MAX_CONTAINERS; |
| 229 | } |
| 230 | |
| 231 | return realloc_array(ra, new_capacity); |
| 232 | } |
| 233 | return true; |
| 234 | } |
| 235 | |
| 236 | void ra_append(roaring_array_t *ra, uint16_t key, void *container, |
| 237 | uint8_t typecode) { |
| 238 | extend_array(ra, 1); |
| 239 | const int32_t pos = ra->size; |
| 240 | |
| 241 | ra->keys[pos] = key; |
| 242 | ra->containers[pos] = container; |
| 243 | ra->typecodes[pos] = typecode; |
| 244 | ra->size++; |
| 245 | } |
| 246 | |
| 247 | void ra_append_copy(roaring_array_t *ra, const roaring_array_t *sa, |
| 248 | uint16_t index, bool copy_on_write) { |
| 249 | extend_array(ra, 1); |
| 250 | const int32_t pos = ra->size; |
| 251 | |
| 252 | // old contents is junk not needing freeing |
| 253 | ra->keys[pos] = sa->keys[index]; |
| 254 | // the shared container will be in two bitmaps |
| 255 | if (copy_on_write) { |
| 256 | sa->containers[index] = get_copy_of_container( |
| 257 | sa->containers[index], &sa->typecodes[index], copy_on_write); |
| 258 | ra->containers[pos] = sa->containers[index]; |
| 259 | ra->typecodes[pos] = sa->typecodes[index]; |
| 260 | } else { |
| 261 | ra->containers[pos] = |
| 262 | container_clone(sa->containers[index], sa->typecodes[index]); |
| 263 | ra->typecodes[pos] = sa->typecodes[index]; |
| 264 | } |
| 265 | ra->size++; |
| 266 | } |
| 267 | |
| 268 | void ra_append_copies_until(roaring_array_t *ra, const roaring_array_t *sa, |
| 269 | uint16_t stopping_key, bool copy_on_write) { |
| 270 | for (int32_t i = 0; i < sa->size; ++i) { |
| 271 | if (sa->keys[i] >= stopping_key) break; |
| 272 | ra_append_copy(ra, sa, i, copy_on_write); |
| 273 | } |
| 274 | } |
| 275 | |
| 276 | void ra_append_copy_range(roaring_array_t *ra, const roaring_array_t *sa, |
| 277 | int32_t start_index, int32_t end_index, |
| 278 | bool copy_on_write) { |
| 279 | extend_array(ra, end_index - start_index); |
| 280 | for (int32_t i = start_index; i < end_index; ++i) { |
| 281 | const int32_t pos = ra->size; |
| 282 | ra->keys[pos] = sa->keys[i]; |
| 283 | if (copy_on_write) { |
| 284 | sa->containers[i] = get_copy_of_container( |
| 285 | sa->containers[i], &sa->typecodes[i], copy_on_write); |
| 286 | ra->containers[pos] = sa->containers[i]; |
| 287 | ra->typecodes[pos] = sa->typecodes[i]; |
| 288 | } else { |
| 289 | ra->containers[pos] = |
| 290 | container_clone(sa->containers[i], sa->typecodes[i]); |
| 291 | ra->typecodes[pos] = sa->typecodes[i]; |
| 292 | } |
| 293 | ra->size++; |
| 294 | } |
| 295 | } |
| 296 | |
| 297 | void ra_append_copies_after(roaring_array_t *ra, const roaring_array_t *sa, |
| 298 | uint16_t before_start, bool copy_on_write) { |
| 299 | int start_location = ra_get_index(sa, before_start); |
| 300 | if (start_location >= 0) |
| 301 | ++start_location; |
| 302 | else |
| 303 | start_location = -start_location - 1; |
| 304 | ra_append_copy_range(ra, sa, start_location, sa->size, copy_on_write); |
| 305 | } |
| 306 | |
| 307 | void ra_append_move_range(roaring_array_t *ra, roaring_array_t *sa, |
| 308 | int32_t start_index, int32_t end_index) { |
| 309 | extend_array(ra, end_index - start_index); |
| 310 | |
| 311 | for (int32_t i = start_index; i < end_index; ++i) { |
| 312 | const int32_t pos = ra->size; |
| 313 | |
| 314 | ra->keys[pos] = sa->keys[i]; |
| 315 | ra->containers[pos] = sa->containers[i]; |
| 316 | ra->typecodes[pos] = sa->typecodes[i]; |
| 317 | ra->size++; |
| 318 | } |
| 319 | } |
| 320 | |
| 321 | void ra_append_range(roaring_array_t *ra, roaring_array_t *sa, |
| 322 | int32_t start_index, int32_t end_index, |
| 323 | bool copy_on_write) { |
| 324 | extend_array(ra, end_index - start_index); |
| 325 | |
| 326 | for (int32_t i = start_index; i < end_index; ++i) { |
| 327 | const int32_t pos = ra->size; |
| 328 | ra->keys[pos] = sa->keys[i]; |
| 329 | if (copy_on_write) { |
| 330 | sa->containers[i] = get_copy_of_container( |
| 331 | sa->containers[i], &sa->typecodes[i], copy_on_write); |
| 332 | ra->containers[pos] = sa->containers[i]; |
| 333 | ra->typecodes[pos] = sa->typecodes[i]; |
| 334 | } else { |
| 335 | ra->containers[pos] = |
| 336 | container_clone(sa->containers[i], sa->typecodes[i]); |
| 337 | ra->typecodes[pos] = sa->typecodes[i]; |
| 338 | } |
| 339 | ra->size++; |
| 340 | } |
| 341 | } |
| 342 | |
| 343 | void *ra_get_container(roaring_array_t *ra, uint16_t x, uint8_t *typecode) { |
| 344 | int i = binarySearch(ra->keys, (int32_t)ra->size, x); |
| 345 | if (i < 0) return NULL; |
| 346 | *typecode = ra->typecodes[i]; |
| 347 | return ra->containers[i]; |
| 348 | } |
| 349 | |
| 350 | extern inline void *ra_get_container_at_index(const roaring_array_t *ra, uint16_t i, |
| 351 | uint8_t *typecode); |
| 352 | |
| 353 | void *ra_get_writable_container(roaring_array_t *ra, uint16_t x, |
| 354 | uint8_t *typecode) { |
| 355 | int i = binarySearch(ra->keys, (int32_t)ra->size, x); |
| 356 | if (i < 0) return NULL; |
| 357 | *typecode = ra->typecodes[i]; |
| 358 | return get_writable_copy_if_shared(ra->containers[i], typecode); |
| 359 | } |
| 360 | |
| 361 | void *ra_get_writable_container_at_index(roaring_array_t *ra, uint16_t i, |
| 362 | uint8_t *typecode) { |
| 363 | assert(i < ra->size); |
| 364 | *typecode = ra->typecodes[i]; |
| 365 | return get_writable_copy_if_shared(ra->containers[i], typecode); |
| 366 | } |
| 367 | |
| 368 | uint16_t ra_get_key_at_index(const roaring_array_t *ra, uint16_t i) { |
| 369 | return ra->keys[i]; |
| 370 | } |
| 371 | |
| 372 | extern inline int32_t ra_get_index(const roaring_array_t *ra, uint16_t x); |
| 373 | |
| 374 | extern inline int32_t ra_advance_until(const roaring_array_t *ra, uint16_t x, |
| 375 | int32_t pos); |
| 376 | |
| 377 | // everything skipped over is freed |
| 378 | int32_t ra_advance_until_freeing(roaring_array_t *ra, uint16_t x, int32_t pos) { |
| 379 | while (pos < ra->size && ra->keys[pos] < x) { |
| 380 | container_free(ra->containers[pos], ra->typecodes[pos]); |
| 381 | ++pos; |
| 382 | } |
| 383 | return pos; |
| 384 | } |
| 385 | |
| 386 | void ra_insert_new_key_value_at(roaring_array_t *ra, int32_t i, uint16_t key, |
| 387 | void *container, uint8_t typecode) { |
| 388 | extend_array(ra, 1); |
| 389 | // May be an optimization opportunity with DIY memmove |
| 390 | memmove(&(ra->keys[i + 1]), &(ra->keys[i]), |
| 391 | sizeof(uint16_t) * (ra->size - i)); |
| 392 | memmove(&(ra->containers[i + 1]), &(ra->containers[i]), |
| 393 | sizeof(void *) * (ra->size - i)); |
| 394 | memmove(&(ra->typecodes[i + 1]), &(ra->typecodes[i]), |
| 395 | sizeof(uint8_t) * (ra->size - i)); |
| 396 | ra->keys[i] = key; |
| 397 | ra->containers[i] = container; |
| 398 | ra->typecodes[i] = typecode; |
| 399 | ra->size++; |
| 400 | } |
| 401 | |
| 402 | // note: Java routine set things to 0, enabling GC. |
| 403 | // Java called it "resize" but it was always used to downsize. |
| 404 | // Allowing upsize would break the conventions about |
| 405 | // valid containers below ra->size. |
| 406 | |
| 407 | void ra_downsize(roaring_array_t *ra, int32_t new_length) { |
| 408 | assert(new_length <= ra->size); |
| 409 | ra->size = new_length; |
| 410 | } |
| 411 | |
| 412 | void ra_remove_at_index(roaring_array_t *ra, int32_t i) { |
| 413 | memmove(&(ra->containers[i]), &(ra->containers[i + 1]), |
| 414 | sizeof(void *) * (ra->size - i - 1)); |
| 415 | memmove(&(ra->keys[i]), &(ra->keys[i + 1]), |
| 416 | sizeof(uint16_t) * (ra->size - i - 1)); |
| 417 | memmove(&(ra->typecodes[i]), &(ra->typecodes[i + 1]), |
| 418 | sizeof(uint8_t) * (ra->size - i - 1)); |
| 419 | ra->size--; |
| 420 | } |
| 421 | |
| 422 | void ra_remove_at_index_and_free(roaring_array_t *ra, int32_t i) { |
| 423 | container_free(ra->containers[i], ra->typecodes[i]); |
| 424 | ra_remove_at_index(ra, i); |
| 425 | } |
| 426 | |
| 427 | // used in inplace andNot only, to slide left the containers from |
| 428 | // the mutated RoaringBitmap that are after the largest container of |
| 429 | // the argument RoaringBitmap. In use it should be followed by a call to |
| 430 | // downsize. |
| 431 | // |
| 432 | void ra_copy_range(roaring_array_t *ra, uint32_t begin, uint32_t end, |
| 433 | uint32_t new_begin) { |
| 434 | assert(begin <= end); |
| 435 | assert(new_begin < begin); |
| 436 | |
| 437 | const int range = end - begin; |
| 438 | |
| 439 | // We ensure to previously have freed overwritten containers |
| 440 | // that are not copied elsewhere |
| 441 | |
| 442 | memmove(&(ra->containers[new_begin]), &(ra->containers[begin]), |
| 443 | sizeof(void *) * range); |
| 444 | memmove(&(ra->keys[new_begin]), &(ra->keys[begin]), |
| 445 | sizeof(uint16_t) * range); |
| 446 | memmove(&(ra->typecodes[new_begin]), &(ra->typecodes[begin]), |
| 447 | sizeof(uint8_t) * range); |
| 448 | } |
| 449 | |
| 450 | void ra_shift_tail(roaring_array_t *ra, int32_t count, int32_t distance) { |
| 451 | if (distance > 0) { |
| 452 | extend_array(ra, distance); |
| 453 | } |
| 454 | int32_t srcpos = ra->size - count; |
| 455 | int32_t dstpos = srcpos + distance; |
| 456 | memmove(&(ra->keys[dstpos]), &(ra->keys[srcpos]), |
| 457 | sizeof(uint16_t) * count); |
| 458 | memmove(&(ra->containers[dstpos]), &(ra->containers[srcpos]), |
| 459 | sizeof(void *) * count); |
| 460 | memmove(&(ra->typecodes[dstpos]), &(ra->typecodes[srcpos]), |
| 461 | sizeof(uint8_t) * count); |
| 462 | ra->size += distance; |
| 463 | } |
| 464 | |
| 465 | |
| 466 | size_t ra_size_in_bytes(roaring_array_t *ra) { |
| 467 | size_t cardinality = 0; |
| 468 | size_t tot_len = |
| 469 | 1 /* initial byte type */ + 4 /* tot_len */ + sizeof(roaring_array_t) + |
| 470 | ra->size * (sizeof(uint16_t) + sizeof(void *) + sizeof(uint8_t)); |
| 471 | for (int32_t i = 0; i < ra->size; i++) { |
| 472 | tot_len += |
| 473 | (container_serialization_len(ra->containers[i], ra->typecodes[i]) + |
| 474 | sizeof(uint16_t)); |
| 475 | cardinality += |
| 476 | container_get_cardinality(ra->containers[i], ra->typecodes[i]); |
| 477 | } |
| 478 | |
| 479 | if ((cardinality * sizeof(uint32_t) + sizeof(uint32_t)) < tot_len) { |
| 480 | return cardinality * sizeof(uint32_t) + 1 + sizeof(uint32_t); |
| 481 | } |
| 482 | return tot_len; |
| 483 | } |
| 484 | |
| 485 | void ra_to_uint32_array(const roaring_array_t *ra, uint32_t *ans) { |
| 486 | size_t ctr = 0; |
| 487 | for (int32_t i = 0; i < ra->size; ++i) { |
| 488 | int num_added = container_to_uint32_array( |
| 489 | ans + ctr, ra->containers[i], ra->typecodes[i], |
| 490 | ((uint32_t)ra->keys[i]) << 16); |
| 491 | ctr += num_added; |
| 492 | } |
| 493 | } |
| 494 | |
| 495 | bool ra_range_uint32_array(const roaring_array_t *ra, size_t offset, size_t limit, uint32_t *ans) { |
| 496 | size_t ctr = 0; |
| 497 | size_t dtr = 0; |
| 498 | |
| 499 | size_t t_limit = 0; |
| 500 | |
| 501 | bool first = false; |
| 502 | size_t first_skip = 0; |
| 503 | |
| 504 | uint32_t *t_ans = NULL; |
| 505 | size_t cur_len = 0; |
| 506 | |
| 507 | for (int i = 0; i < ra->size; ++i) { |
| 508 | |
| 509 | const void *container = container_unwrap_shared(ra->containers[i], &ra->typecodes[i]); |
| 510 | switch (ra->typecodes[i]) { |
| 511 | case BITSET_CONTAINER_TYPE_CODE: |
| 512 | t_limit = ((const bitset_container_t *)container)->cardinality; |
| 513 | break; |
| 514 | case ARRAY_CONTAINER_TYPE_CODE: |
| 515 | t_limit = ((const array_container_t *)container)->cardinality; |
| 516 | break; |
| 517 | case RUN_CONTAINER_TYPE_CODE: |
| 518 | t_limit = run_container_cardinality((const run_container_t *)container); |
| 519 | break; |
| 520 | } |
| 521 | if (ctr + t_limit - 1 >= offset && ctr < offset + limit){ |
| 522 | if (!first){ |
| 523 | //first_skip = t_limit - (ctr + t_limit - offset); |
| 524 | first_skip = offset - ctr; |
| 525 | first = true; |
| 526 | t_ans = (uint32_t *)malloc(sizeof(*t_ans) * (first_skip + limit)); |
| 527 | if(t_ans == NULL) { |
| 528 | return false; |
| 529 | } |
| 530 | memset(t_ans, 0, sizeof(*t_ans) * (first_skip + limit)) ; |
| 531 | cur_len = first_skip + limit; |
| 532 | } |
| 533 | if (dtr + t_limit > cur_len){ |
| 534 | uint32_t * append_ans = (uint32_t *)malloc(sizeof(*append_ans) * (cur_len + t_limit)); |
| 535 | if(append_ans == NULL) { |
| 536 | if(t_ans != NULL) free(t_ans); |
| 537 | return false; |
| 538 | } |
| 539 | memset(append_ans, 0, sizeof(*append_ans) * (cur_len + t_limit)); |
| 540 | cur_len = cur_len + t_limit; |
| 541 | memcpy(append_ans, t_ans, dtr * sizeof(uint32_t)); |
| 542 | free(t_ans); |
| 543 | t_ans = append_ans; |
| 544 | } |
| 545 | switch (ra->typecodes[i]) { |
| 546 | case BITSET_CONTAINER_TYPE_CODE: |
| 547 | container_to_uint32_array( |
| 548 | t_ans + dtr, (const bitset_container_t *)container, ra->typecodes[i], |
| 549 | ((uint32_t)ra->keys[i]) << 16); |
| 550 | break; |
| 551 | case ARRAY_CONTAINER_TYPE_CODE: |
| 552 | container_to_uint32_array( |
| 553 | t_ans + dtr, (const array_container_t *)container, ra->typecodes[i], |
| 554 | ((uint32_t)ra->keys[i]) << 16); |
| 555 | break; |
| 556 | case RUN_CONTAINER_TYPE_CODE: |
| 557 | container_to_uint32_array( |
| 558 | t_ans + dtr, (const run_container_t *)container, ra->typecodes[i], |
| 559 | ((uint32_t)ra->keys[i]) << 16); |
| 560 | break; |
| 561 | } |
| 562 | dtr += t_limit; |
| 563 | } |
| 564 | ctr += t_limit; |
| 565 | if (dtr-first_skip >= limit) break; |
| 566 | } |
| 567 | if(t_ans != NULL) { |
| 568 | memcpy(ans, t_ans+first_skip, limit * sizeof(uint32_t)); |
| 569 | free(t_ans); |
| 570 | } |
| 571 | return true; |
| 572 | } |
| 573 | |
| 574 | bool ra_has_run_container(const roaring_array_t *ra) { |
| 575 | for (int32_t k = 0; k < ra->size; ++k) { |
| 576 | if (get_container_type(ra->containers[k], ra->typecodes[k]) == |
| 577 | RUN_CONTAINER_TYPE_CODE) |
| 578 | return true; |
| 579 | } |
| 580 | return false; |
| 581 | } |
| 582 | |
| 583 | uint32_t (const roaring_array_t *ra) { |
| 584 | if (ra_has_run_container(ra)) { |
| 585 | if (ra->size < |
| 586 | NO_OFFSET_THRESHOLD) { // for small bitmaps, we omit the offsets |
| 587 | return 4 + (ra->size + 7) / 8 + 4 * ra->size; |
| 588 | } |
| 589 | return 4 + (ra->size + 7) / 8 + |
| 590 | 8 * ra->size; // - 4 because we pack the size with the cookie |
| 591 | } else { |
| 592 | return 4 + 4 + 8 * ra->size; |
| 593 | } |
| 594 | } |
| 595 | |
| 596 | size_t ra_portable_size_in_bytes(const roaring_array_t *ra) { |
| 597 | size_t count = ra_portable_header_size(ra); |
| 598 | |
| 599 | for (int32_t k = 0; k < ra->size; ++k) { |
| 600 | count += container_size_in_bytes(ra->containers[k], ra->typecodes[k]); |
| 601 | } |
| 602 | return count; |
| 603 | } |
| 604 | |
| 605 | size_t ra_portable_serialize(const roaring_array_t *ra, char *buf) { |
| 606 | char *initbuf = buf; |
| 607 | uint32_t startOffset = 0; |
| 608 | bool hasrun = ra_has_run_container(ra); |
| 609 | if (hasrun) { |
| 610 | uint32_t cookie = SERIAL_COOKIE | ((ra->size - 1) << 16); |
| 611 | memcpy(buf, &cookie, sizeof(cookie)); |
| 612 | buf += sizeof(cookie); |
| 613 | uint32_t s = (ra->size + 7) / 8; |
| 614 | uint8_t *bitmapOfRunContainers = (uint8_t *)calloc(s, 1); |
| 615 | assert(bitmapOfRunContainers != NULL); // todo: handle |
| 616 | for (int32_t i = 0; i < ra->size; ++i) { |
| 617 | if (get_container_type(ra->containers[i], ra->typecodes[i]) == |
| 618 | RUN_CONTAINER_TYPE_CODE) { |
| 619 | bitmapOfRunContainers[i / 8] |= (1 << (i % 8)); |
| 620 | } |
| 621 | } |
| 622 | memcpy(buf, bitmapOfRunContainers, s); |
| 623 | buf += s; |
| 624 | free(bitmapOfRunContainers); |
| 625 | if (ra->size < NO_OFFSET_THRESHOLD) { |
| 626 | startOffset = 4 + 4 * ra->size + s; |
| 627 | } else { |
| 628 | startOffset = 4 + 8 * ra->size + s; |
| 629 | } |
| 630 | } else { // backwards compatibility |
| 631 | uint32_t cookie = SERIAL_COOKIE_NO_RUNCONTAINER; |
| 632 | |
| 633 | memcpy(buf, &cookie, sizeof(cookie)); |
| 634 | buf += sizeof(cookie); |
| 635 | memcpy(buf, &ra->size, sizeof(ra->size)); |
| 636 | buf += sizeof(ra->size); |
| 637 | |
| 638 | startOffset = 4 + 4 + 4 * ra->size + 4 * ra->size; |
| 639 | } |
| 640 | for (int32_t k = 0; k < ra->size; ++k) { |
| 641 | memcpy(buf, &ra->keys[k], sizeof(ra->keys[k])); |
| 642 | buf += sizeof(ra->keys[k]); |
| 643 | // get_cardinality returns a value in [1,1<<16], subtracting one |
| 644 | // we get [0,1<<16 - 1] which fits in 16 bits |
| 645 | uint16_t card = (uint16_t)( |
| 646 | container_get_cardinality(ra->containers[k], ra->typecodes[k]) - 1); |
| 647 | memcpy(buf, &card, sizeof(card)); |
| 648 | buf += sizeof(card); |
| 649 | } |
| 650 | if ((!hasrun) || (ra->size >= NO_OFFSET_THRESHOLD)) { |
| 651 | // writing the containers offsets |
| 652 | for (int32_t k = 0; k < ra->size; k++) { |
| 653 | memcpy(buf, &startOffset, sizeof(startOffset)); |
| 654 | buf += sizeof(startOffset); |
| 655 | startOffset = |
| 656 | startOffset + |
| 657 | container_size_in_bytes(ra->containers[k], ra->typecodes[k]); |
| 658 | } |
| 659 | } |
| 660 | for (int32_t k = 0; k < ra->size; ++k) { |
| 661 | buf += container_write(ra->containers[k], ra->typecodes[k], buf); |
| 662 | } |
| 663 | return buf - initbuf; |
| 664 | } |
| 665 | |
| 666 | // Quickly checks whether there is a serialized bitmap at the pointer, |
| 667 | // not exceeding size "maxbytes" in bytes. This function does not allocate |
| 668 | // memory dynamically. |
| 669 | // |
| 670 | // This function returns 0 if and only if no valid bitmap is found. |
| 671 | // Otherwise, it returns how many bytes are occupied. |
| 672 | // |
| 673 | size_t ra_portable_deserialize_size(const char *buf, const size_t maxbytes) { |
| 674 | size_t bytestotal = sizeof(int32_t);// for cookie |
| 675 | if(bytestotal > maxbytes) return 0; |
| 676 | uint32_t cookie; |
| 677 | memcpy(&cookie, buf, sizeof(int32_t)); |
| 678 | buf += sizeof(uint32_t); |
| 679 | if ((cookie & 0xFFFF) != SERIAL_COOKIE && |
| 680 | cookie != SERIAL_COOKIE_NO_RUNCONTAINER) { |
| 681 | return 0; |
| 682 | } |
| 683 | int32_t size; |
| 684 | |
| 685 | if ((cookie & 0xFFFF) == SERIAL_COOKIE) |
| 686 | size = (cookie >> 16) + 1; |
| 687 | else { |
| 688 | bytestotal += sizeof(int32_t); |
| 689 | if(bytestotal > maxbytes) return 0; |
| 690 | memcpy(&size, buf, sizeof(int32_t)); |
| 691 | buf += sizeof(uint32_t); |
| 692 | } |
| 693 | if (size > (1<<16)) { |
| 694 | return 0; // logically impossible |
| 695 | } |
| 696 | char *bitmapOfRunContainers = NULL; |
| 697 | bool hasrun = (cookie & 0xFFFF) == SERIAL_COOKIE; |
| 698 | if (hasrun) { |
| 699 | int32_t s = (size + 7) / 8; |
| 700 | bytestotal += s; |
| 701 | if(bytestotal > maxbytes) return 0; |
| 702 | bitmapOfRunContainers = (char *)buf; |
| 703 | buf += s; |
| 704 | } |
| 705 | bytestotal += size * 2 * sizeof(uint16_t); |
| 706 | if(bytestotal > maxbytes) return 0; |
| 707 | uint16_t *keyscards = (uint16_t *)buf; |
| 708 | buf += size * 2 * sizeof(uint16_t); |
| 709 | if ((!hasrun) || (size >= NO_OFFSET_THRESHOLD)) { |
| 710 | // skipping the offsets |
| 711 | bytestotal += size * 4; |
| 712 | if(bytestotal > maxbytes) return 0; |
| 713 | buf += size * 4; |
| 714 | } |
| 715 | // Reading the containers |
| 716 | for (int32_t k = 0; k < size; ++k) { |
| 717 | uint16_t tmp; |
| 718 | memcpy(&tmp, keyscards + 2*k+1, sizeof(tmp)); |
| 719 | uint32_t thiscard = tmp + 1; |
| 720 | bool isbitmap = (thiscard > DEFAULT_MAX_SIZE); |
| 721 | bool isrun = false; |
| 722 | if(hasrun) { |
| 723 | if((bitmapOfRunContainers[k / 8] & (1 << (k % 8))) != 0) { |
| 724 | isbitmap = false; |
| 725 | isrun = true; |
| 726 | } |
| 727 | } |
| 728 | if (isbitmap) { |
| 729 | size_t containersize = BITSET_CONTAINER_SIZE_IN_WORDS * sizeof(uint64_t); |
| 730 | bytestotal += containersize; |
| 731 | if(bytestotal > maxbytes) return 0; |
| 732 | buf += containersize; |
| 733 | } else if (isrun) { |
| 734 | bytestotal += sizeof(uint16_t); |
| 735 | if(bytestotal > maxbytes) return 0; |
| 736 | uint16_t n_runs; |
| 737 | memcpy(&n_runs, buf, sizeof(uint16_t)); |
| 738 | buf += sizeof(uint16_t); |
| 739 | size_t containersize = n_runs * sizeof(rle16_t); |
| 740 | bytestotal += containersize; |
| 741 | if(bytestotal > maxbytes) return 0; |
| 742 | buf += containersize; |
| 743 | } else { |
| 744 | size_t containersize = thiscard * sizeof(uint16_t); |
| 745 | bytestotal += containersize; |
| 746 | if(bytestotal > maxbytes) return 0; |
| 747 | buf += containersize; |
| 748 | } |
| 749 | } |
| 750 | return bytestotal; |
| 751 | } |
| 752 | |
| 753 | |
| 754 | // this function populates answer from the content of buf (reading up to maxbytes bytes). |
| 755 | // The function returns false if a properly serialized bitmap cannot be found. |
| 756 | // if it returns true, readbytes is populated by how many bytes were read, we have that *readbytes <= maxbytes. |
| 757 | bool ra_portable_deserialize(roaring_array_t *answer, const char *buf, const size_t maxbytes, size_t * readbytes) { |
| 758 | *readbytes = sizeof(int32_t);// for cookie |
| 759 | if(*readbytes > maxbytes) { |
| 760 | fprintf(stderr, "Ran out of bytes while reading first 4 bytes.\n" ); |
| 761 | return false; |
| 762 | } |
| 763 | uint32_t cookie; |
| 764 | memcpy(&cookie, buf, sizeof(int32_t)); |
| 765 | buf += sizeof(uint32_t); |
| 766 | if ((cookie & 0xFFFF) != SERIAL_COOKIE && |
| 767 | cookie != SERIAL_COOKIE_NO_RUNCONTAINER) { |
| 768 | fprintf(stderr, "I failed to find one of the right cookies. Found %" PRIu32 "\n" , |
| 769 | cookie); |
| 770 | return false; |
| 771 | } |
| 772 | int32_t size; |
| 773 | |
| 774 | if ((cookie & 0xFFFF) == SERIAL_COOKIE) |
| 775 | size = (cookie >> 16) + 1; |
| 776 | else { |
| 777 | *readbytes += sizeof(int32_t); |
| 778 | if(*readbytes > maxbytes) { |
| 779 | fprintf(stderr, "Ran out of bytes while reading second part of the cookie.\n" ); |
| 780 | return false; |
| 781 | } |
| 782 | memcpy(&size, buf, sizeof(int32_t)); |
| 783 | buf += sizeof(uint32_t); |
| 784 | } |
| 785 | if (size > (1<<16)) { |
| 786 | fprintf(stderr, "You cannot have so many containers, the data must be corrupted: %" PRId32 "\n" , |
| 787 | size); |
| 788 | return false; // logically impossible |
| 789 | } |
| 790 | const char *bitmapOfRunContainers = NULL; |
| 791 | bool hasrun = (cookie & 0xFFFF) == SERIAL_COOKIE; |
| 792 | if (hasrun) { |
| 793 | int32_t s = (size + 7) / 8; |
| 794 | *readbytes += s; |
| 795 | if(*readbytes > maxbytes) {// data is corrupted? |
| 796 | fprintf(stderr, "Ran out of bytes while reading run bitmap.\n" ); |
| 797 | return false; |
| 798 | } |
| 799 | bitmapOfRunContainers = buf; |
| 800 | buf += s; |
| 801 | } |
| 802 | uint16_t *keyscards = (uint16_t *)buf; |
| 803 | |
| 804 | *readbytes += size * 2 * sizeof(uint16_t); |
| 805 | if(*readbytes > maxbytes) { |
| 806 | fprintf(stderr, "Ran out of bytes while reading key-cardinality array.\n" ); |
| 807 | return false; |
| 808 | } |
| 809 | buf += size * 2 * sizeof(uint16_t); |
| 810 | |
| 811 | bool is_ok = ra_init_with_capacity(answer, size); |
| 812 | if (!is_ok) { |
| 813 | fprintf(stderr, "Failed to allocate memory for roaring array. Bailing out.\n" ); |
| 814 | return false; |
| 815 | } |
| 816 | |
| 817 | for (int32_t k = 0; k < size; ++k) { |
| 818 | uint16_t tmp; |
| 819 | memcpy(&tmp, keyscards + 2*k, sizeof(tmp)); |
| 820 | answer->keys[k] = tmp; |
| 821 | } |
| 822 | if ((!hasrun) || (size >= NO_OFFSET_THRESHOLD)) { |
| 823 | *readbytes += size * 4; |
| 824 | if(*readbytes > maxbytes) {// data is corrupted? |
| 825 | fprintf(stderr, "Ran out of bytes while reading offsets.\n" ); |
| 826 | ra_clear(answer);// we need to clear the containers already allocated, and the roaring array |
| 827 | return false; |
| 828 | } |
| 829 | |
| 830 | // skipping the offsets |
| 831 | buf += size * 4; |
| 832 | } |
| 833 | // Reading the containers |
| 834 | for (int32_t k = 0; k < size; ++k) { |
| 835 | uint16_t tmp; |
| 836 | memcpy(&tmp, keyscards + 2*k+1, sizeof(tmp)); |
| 837 | uint32_t thiscard = tmp + 1; |
| 838 | bool isbitmap = (thiscard > DEFAULT_MAX_SIZE); |
| 839 | bool isrun = false; |
| 840 | if(hasrun) { |
| 841 | if((bitmapOfRunContainers[k / 8] & (1 << (k % 8))) != 0) { |
| 842 | isbitmap = false; |
| 843 | isrun = true; |
| 844 | } |
| 845 | } |
| 846 | if (isbitmap) { |
| 847 | // we check that the read is allowed |
| 848 | size_t containersize = BITSET_CONTAINER_SIZE_IN_WORDS * sizeof(uint64_t); |
| 849 | *readbytes += containersize; |
| 850 | if(*readbytes > maxbytes) { |
| 851 | fprintf(stderr, "Running out of bytes while reading a bitset container.\n" ); |
| 852 | ra_clear(answer);// we need to clear the containers already allocated, and the roaring array |
| 853 | return false; |
| 854 | } |
| 855 | // it is now safe to read |
| 856 | bitset_container_t *c = bitset_container_create(); |
| 857 | if(c == NULL) {// memory allocation failure |
| 858 | fprintf(stderr, "Failed to allocate memory for a bitset container.\n" ); |
| 859 | ra_clear(answer);// we need to clear the containers already allocated, and the roaring array |
| 860 | return false; |
| 861 | } |
| 862 | answer->size++; |
| 863 | buf += bitset_container_read(thiscard, c, buf); |
| 864 | answer->containers[k] = c; |
| 865 | answer->typecodes[k] = BITSET_CONTAINER_TYPE_CODE; |
| 866 | } else if (isrun) { |
| 867 | // we check that the read is allowed |
| 868 | *readbytes += sizeof(uint16_t); |
| 869 | if(*readbytes > maxbytes) { |
| 870 | fprintf(stderr, "Running out of bytes while reading a run container (header).\n" ); |
| 871 | ra_clear(answer);// we need to clear the containers already allocated, and the roaring array |
| 872 | return false; |
| 873 | } |
| 874 | uint16_t n_runs; |
| 875 | memcpy(&n_runs, buf, sizeof(uint16_t)); |
| 876 | size_t containersize = n_runs * sizeof(rle16_t); |
| 877 | *readbytes += containersize; |
| 878 | if(*readbytes > maxbytes) {// data is corrupted? |
| 879 | fprintf(stderr, "Running out of bytes while reading a run container.\n" ); |
| 880 | ra_clear(answer);// we need to clear the containers already allocated, and the roaring array |
| 881 | return false; |
| 882 | } |
| 883 | // it is now safe to read |
| 884 | |
| 885 | run_container_t *c = run_container_create(); |
| 886 | if(c == NULL) {// memory allocation failure |
| 887 | fprintf(stderr, "Failed to allocate memory for a run container.\n" ); |
| 888 | ra_clear(answer);// we need to clear the containers already allocated, and the roaring array |
| 889 | return false; |
| 890 | } |
| 891 | answer->size++; |
| 892 | buf += run_container_read(thiscard, c, buf); |
| 893 | answer->containers[k] = c; |
| 894 | answer->typecodes[k] = RUN_CONTAINER_TYPE_CODE; |
| 895 | } else { |
| 896 | // we check that the read is allowed |
| 897 | size_t containersize = thiscard * sizeof(uint16_t); |
| 898 | *readbytes += containersize; |
| 899 | if(*readbytes > maxbytes) {// data is corrupted? |
| 900 | fprintf(stderr, "Running out of bytes while reading an array container.\n" ); |
| 901 | ra_clear(answer);// we need to clear the containers already allocated, and the roaring array |
| 902 | return false; |
| 903 | } |
| 904 | // it is now safe to read |
| 905 | array_container_t *c = |
| 906 | array_container_create_given_capacity(thiscard); |
| 907 | if(c == NULL) {// memory allocation failure |
| 908 | fprintf(stderr, "Failed to allocate memory for an array container.\n" ); |
| 909 | ra_clear(answer);// we need to clear the containers already allocated, and the roaring array |
| 910 | return false; |
| 911 | } |
| 912 | answer->size++; |
| 913 | buf += array_container_read(thiscard, c, buf); |
| 914 | answer->containers[k] = c; |
| 915 | answer->typecodes[k] = ARRAY_CONTAINER_TYPE_CODE; |
| 916 | } |
| 917 | } |
| 918 | return true; |
| 919 | } |
| 920 | |