| 1 | /* |
| 2 | Copyright (c) 2000, 2010, Oracle and/or its affiliates. |
| 3 | |
| 4 | This program is free software; you can redistribute it and/or modify |
| 5 | it under the terms of the GNU General Public License as published by |
| 6 | the Free Software Foundation; version 2 of the License. |
| 7 | |
| 8 | This program is distributed in the hope that it will be useful, |
| 9 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 10 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 11 | GNU General Public License for more details. |
| 12 | |
| 13 | You should have received a copy of the GNU General Public License |
| 14 | along with this program; if not, write to the Free Software |
| 15 | Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ |
| 16 | |
| 17 | |
| 18 | /* classes to use when handling where clause */ |
| 19 | |
| 20 | #ifndef _opt_range_h |
| 21 | #define _opt_range_h |
| 22 | |
| 23 | #ifdef USE_PRAGMA_INTERFACE |
| 24 | #pragma interface /* gcc class implementation */ |
| 25 | #endif |
| 26 | |
| 27 | #include "records.h" /* READ_RECORD */ |
| 28 | #include "queues.h" /* QUEUE */ |
| 29 | #include "filesort.h" /* SORT_INFO */ |
| 30 | |
| 31 | /* |
| 32 | It is necessary to include set_var.h instead of item.h because there |
| 33 | are dependencies on include order for set_var.h and item.h. This |
| 34 | will be resolved later. |
| 35 | */ |
| 36 | #include "sql_class.h" // set_var.h: THD |
| 37 | #include "set_var.h" /* Item */ |
| 38 | |
| 39 | class JOIN; |
| 40 | class Item_sum; |
| 41 | |
| 42 | struct KEY_PART { |
| 43 | uint16 key,part; |
| 44 | /* See KEY_PART_INFO for meaning of the next two: */ |
| 45 | uint16 store_length, length; |
| 46 | uint8 null_bit; |
| 47 | /* |
| 48 | Keypart flags (0 when this structure is used by partition pruning code |
| 49 | for fake partitioning index description) |
| 50 | */ |
| 51 | uint8 flag; |
| 52 | Field *field; |
| 53 | Field::imagetype image_type; |
| 54 | }; |
| 55 | |
| 56 | |
| 57 | class RANGE_OPT_PARAM; |
| 58 | /* |
| 59 | A construction block of the SEL_ARG-graph. |
| 60 | |
| 61 | The following description only covers graphs of SEL_ARG objects with |
| 62 | sel_arg->type==KEY_RANGE: |
| 63 | |
| 64 | One SEL_ARG object represents an "elementary interval" in form |
| 65 | |
| 66 | min_value <=? table.keypartX <=? max_value |
| 67 | |
| 68 | The interval is a non-empty interval of any kind: with[out] minimum/maximum |
| 69 | bound, [half]open/closed, single-point interval, etc. |
| 70 | |
| 71 | 1. SEL_ARG GRAPH STRUCTURE |
| 72 | |
| 73 | SEL_ARG objects are linked together in a graph. The meaning of the graph |
| 74 | is better demostrated by an example: |
| 75 | |
| 76 | tree->keys[i] |
| 77 | | |
| 78 | | $ $ |
| 79 | | part=1 $ part=2 $ part=3 |
| 80 | | $ $ |
| 81 | | +-------+ $ +-------+ $ +--------+ |
| 82 | | | kp1<1 |--$-->| kp2=5 |--$-->| kp3=10 | |
| 83 | | +-------+ $ +-------+ $ +--------+ |
| 84 | | | $ $ | |
| 85 | | | $ $ +--------+ |
| 86 | | | $ $ | kp3=12 | |
| 87 | | | $ $ +--------+ |
| 88 | | +-------+ $ $ |
| 89 | \->| kp1=2 |--$--------------$-+ |
| 90 | +-------+ $ $ | +--------+ |
| 91 | | $ $ ==>| kp3=11 | |
| 92 | +-------+ $ $ | +--------+ |
| 93 | | kp1=3 |--$--------------$-+ | |
| 94 | +-------+ $ $ +--------+ |
| 95 | | $ $ | kp3=14 | |
| 96 | ... $ $ +--------+ |
| 97 | |
| 98 | The entire graph is partitioned into "interval lists". |
| 99 | |
| 100 | An interval list is a sequence of ordered disjoint intervals over the same |
| 101 | key part. SEL_ARG are linked via "next" and "prev" pointers. Additionally, |
| 102 | all intervals in the list form an RB-tree, linked via left/right/parent |
| 103 | pointers. The RB-tree root SEL_ARG object will be further called "root of the |
| 104 | interval list". |
| 105 | |
| 106 | In the example pic, there are 4 interval lists: |
| 107 | "kp<1 OR kp1=2 OR kp1=3", "kp2=5", "kp3=10 OR kp3=12", "kp3=11 OR kp3=13". |
| 108 | The vertical lines represent SEL_ARG::next/prev pointers. |
| 109 | |
| 110 | In an interval list, each member X may have SEL_ARG::next_key_part pointer |
| 111 | pointing to the root of another interval list Y. The pointed interval list |
| 112 | must cover a key part with greater number (i.e. Y->part > X->part). |
| 113 | |
| 114 | In the example pic, the next_key_part pointers are represented by |
| 115 | horisontal lines. |
| 116 | |
| 117 | 2. SEL_ARG GRAPH SEMANTICS |
| 118 | |
| 119 | It represents a condition in a special form (we don't have a name for it ATM) |
| 120 | The SEL_ARG::next/prev is "OR", and next_key_part is "AND". |
| 121 | |
| 122 | For example, the picture represents the condition in form: |
| 123 | (kp1 < 1 AND kp2=5 AND (kp3=10 OR kp3=12)) OR |
| 124 | (kp1=2 AND (kp3=11 OR kp3=14)) OR |
| 125 | (kp1=3 AND (kp3=11 OR kp3=14)) |
| 126 | |
| 127 | |
| 128 | 3. SEL_ARG GRAPH USE |
| 129 | |
| 130 | Use get_mm_tree() to construct SEL_ARG graph from WHERE condition. |
| 131 | Then walk the SEL_ARG graph and get a list of dijsoint ordered key |
| 132 | intervals (i.e. intervals in form |
| 133 | |
| 134 | (constA1, .., const1_K) < (keypart1,.., keypartK) < (constB1, .., constB_K) |
| 135 | |
| 136 | Those intervals can be used to access the index. The uses are in: |
| 137 | - check_quick_select() - Walk the SEL_ARG graph and find an estimate of |
| 138 | how many table records are contained within all |
| 139 | intervals. |
| 140 | - get_quick_select() - Walk the SEL_ARG, materialize the key intervals, |
| 141 | and create QUICK_RANGE_SELECT object that will |
| 142 | read records within these intervals. |
| 143 | |
| 144 | 4. SPACE COMPLEXITY NOTES |
| 145 | |
| 146 | SEL_ARG graph is a representation of an ordered disjoint sequence of |
| 147 | intervals over the ordered set of index tuple values. |
| 148 | |
| 149 | For multi-part keys, one can construct a WHERE expression such that its |
| 150 | list of intervals will be of combinatorial size. Here is an example: |
| 151 | |
| 152 | (keypart1 IN (1,2, ..., n1)) AND |
| 153 | (keypart2 IN (1,2, ..., n2)) AND |
| 154 | (keypart3 IN (1,2, ..., n3)) |
| 155 | |
| 156 | For this WHERE clause the list of intervals will have n1*n2*n3 intervals |
| 157 | of form |
| 158 | |
| 159 | (keypart1, keypart2, keypart3) = (k1, k2, k3), where 1 <= k{i} <= n{i} |
| 160 | |
| 161 | SEL_ARG graph structure aims to reduce the amount of required space by |
| 162 | "sharing" the elementary intervals when possible (the pic at the |
| 163 | beginning of this comment has examples of such sharing). The sharing may |
| 164 | prevent combinatorial blowup: |
| 165 | |
| 166 | There are WHERE clauses that have combinatorial-size interval lists but |
| 167 | will be represented by a compact SEL_ARG graph. |
| 168 | Example: |
| 169 | (keypartN IN (1,2, ..., n1)) AND |
| 170 | ... |
| 171 | (keypart2 IN (1,2, ..., n2)) AND |
| 172 | (keypart1 IN (1,2, ..., n3)) |
| 173 | |
| 174 | but not in all cases: |
| 175 | |
| 176 | - There are WHERE clauses that do have a compact SEL_ARG-graph |
| 177 | representation but get_mm_tree() and its callees will construct a |
| 178 | graph of combinatorial size. |
| 179 | Example: |
| 180 | (keypart1 IN (1,2, ..., n1)) AND |
| 181 | (keypart2 IN (1,2, ..., n2)) AND |
| 182 | ... |
| 183 | (keypartN IN (1,2, ..., n3)) |
| 184 | |
| 185 | - There are WHERE clauses for which the minimal possible SEL_ARG graph |
| 186 | representation will have combinatorial size. |
| 187 | Example: |
| 188 | By induction: Let's take any interval on some keypart in the middle: |
| 189 | |
| 190 | kp15=c0 |
| 191 | |
| 192 | Then let's AND it with this interval 'structure' from preceding and |
| 193 | following keyparts: |
| 194 | |
| 195 | (kp14=c1 AND kp16=c3) OR keypart14=c2) (*) |
| 196 | |
| 197 | We will obtain this SEL_ARG graph: |
| 198 | |
| 199 | kp14 $ kp15 $ kp16 |
| 200 | $ $ |
| 201 | +---------+ $ +---------+ $ +---------+ |
| 202 | | kp14=c1 |--$-->| kp15=c0 |--$-->| kp16=c3 | |
| 203 | +---------+ $ +---------+ $ +---------+ |
| 204 | | $ $ |
| 205 | +---------+ $ +---------+ $ |
| 206 | | kp14=c2 |--$-->| kp15=c0 | $ |
| 207 | +---------+ $ +---------+ $ |
| 208 | $ $ |
| 209 | |
| 210 | Note that we had to duplicate "kp15=c0" and there was no way to avoid |
| 211 | that. |
| 212 | The induction step: AND the obtained expression with another "wrapping" |
| 213 | expression like (*). |
| 214 | When the process ends because of the limit on max. number of keyparts |
| 215 | we'll have: |
| 216 | |
| 217 | WHERE clause length is O(3*#max_keyparts) |
| 218 | SEL_ARG graph size is O(2^(#max_keyparts/2)) |
| 219 | |
| 220 | (it is also possible to construct a case where instead of 2 in 2^n we |
| 221 | have a bigger constant, e.g. 4, and get a graph with 4^(31/2)= 2^31 |
| 222 | nodes) |
| 223 | |
| 224 | We avoid consuming too much memory by setting a limit on the number of |
| 225 | SEL_ARG object we can construct during one range analysis invocation. |
| 226 | */ |
| 227 | |
| 228 | class SEL_ARG :public Sql_alloc |
| 229 | { |
| 230 | static int sel_cmp(Field *field, uchar *a, uchar *b, uint8 a_flag, |
| 231 | uint8 b_flag); |
| 232 | public: |
| 233 | uint8 min_flag,max_flag,maybe_flag; |
| 234 | uint8 part; // Which key part |
| 235 | uint8 maybe_null; |
| 236 | /* |
| 237 | The ordinal number the least significant component encountered in |
| 238 | the ranges of the SEL_ARG tree (the first component has number 1) |
| 239 | */ |
| 240 | uint16 max_part_no; |
| 241 | /* |
| 242 | Number of children of this element in the RB-tree, plus 1 for this |
| 243 | element itself. |
| 244 | */ |
| 245 | uint16 elements; |
| 246 | /* |
| 247 | Valid only for elements which are RB-tree roots: Number of times this |
| 248 | RB-tree is referred to (it is referred by SEL_ARG::next_key_part or by |
| 249 | SEL_TREE::keys[i] or by a temporary SEL_ARG* variable) |
| 250 | */ |
| 251 | ulong use_count; |
| 252 | |
| 253 | Field *field; |
| 254 | uchar *min_value,*max_value; // Pointer to range |
| 255 | |
| 256 | /* |
| 257 | eq_tree() requires that left == right == 0 if the type is MAYBE_KEY. |
| 258 | */ |
| 259 | SEL_ARG *left,*right; /* R-B tree children */ |
| 260 | SEL_ARG *next,*prev; /* Links for bi-directional interval list */ |
| 261 | SEL_ARG *parent; /* R-B tree parent */ |
| 262 | SEL_ARG *next_key_part; |
| 263 | enum leaf_color { BLACK,RED } color; |
| 264 | enum Type { IMPOSSIBLE, MAYBE, MAYBE_KEY, KEY_RANGE } type; |
| 265 | |
| 266 | enum { MAX_SEL_ARGS = 16000 }; |
| 267 | |
| 268 | SEL_ARG() {} |
| 269 | SEL_ARG(SEL_ARG &); |
| 270 | SEL_ARG(Field *,const uchar *, const uchar *); |
| 271 | SEL_ARG(Field *field, uint8 part, uchar *min_value, uchar *max_value, |
| 272 | uint8 min_flag, uint8 max_flag, uint8 maybe_flag); |
| 273 | SEL_ARG(enum Type type_arg) |
| 274 | :min_flag(0), max_part_no(0) /* first key part means 1. 0 mean 'no parts'*/, |
| 275 | elements(1),use_count(1),left(0),right(0), |
| 276 | next_key_part(0), color(BLACK), type(type_arg) |
| 277 | {} |
| 278 | /** |
| 279 | returns true if a range predicate is equal. Use all_same() |
| 280 | to check for equality of all the predicates on this keypart. |
| 281 | */ |
| 282 | inline bool is_same(const SEL_ARG *arg) const |
| 283 | { |
| 284 | if (type != arg->type || part != arg->part) |
| 285 | return false; |
| 286 | if (type != KEY_RANGE) |
| 287 | return true; |
| 288 | return cmp_min_to_min(arg) == 0 && cmp_max_to_max(arg) == 0; |
| 289 | } |
| 290 | /** |
| 291 | returns true if all the predicates in the keypart tree are equal |
| 292 | */ |
| 293 | bool all_same(const SEL_ARG *arg) const |
| 294 | { |
| 295 | if (type != arg->type || part != arg->part) |
| 296 | return false; |
| 297 | if (type != KEY_RANGE) |
| 298 | return true; |
| 299 | if (arg == this) |
| 300 | return true; |
| 301 | const SEL_ARG *cmp_arg= arg->first(); |
| 302 | const SEL_ARG *cur_arg= first(); |
| 303 | for (; cur_arg && cmp_arg && cur_arg->is_same(cmp_arg); |
| 304 | cur_arg= cur_arg->next, cmp_arg= cmp_arg->next) ; |
| 305 | if (cur_arg || cmp_arg) |
| 306 | return false; |
| 307 | return true; |
| 308 | } |
| 309 | inline void merge_flags(SEL_ARG *arg) { maybe_flag|=arg->maybe_flag; } |
| 310 | inline void maybe_smaller() { maybe_flag=1; } |
| 311 | /* Return true iff it's a single-point null interval */ |
| 312 | inline bool is_null_interval() { return maybe_null && max_value[0] == 1; } |
| 313 | inline int cmp_min_to_min(const SEL_ARG* arg) const |
| 314 | { |
| 315 | return sel_cmp(field,min_value, arg->min_value, min_flag, arg->min_flag); |
| 316 | } |
| 317 | inline int cmp_min_to_max(const SEL_ARG* arg) const |
| 318 | { |
| 319 | return sel_cmp(field,min_value, arg->max_value, min_flag, arg->max_flag); |
| 320 | } |
| 321 | inline int cmp_max_to_max(const SEL_ARG* arg) const |
| 322 | { |
| 323 | return sel_cmp(field,max_value, arg->max_value, max_flag, arg->max_flag); |
| 324 | } |
| 325 | inline int cmp_max_to_min(const SEL_ARG* arg) const |
| 326 | { |
| 327 | return sel_cmp(field,max_value, arg->min_value, max_flag, arg->min_flag); |
| 328 | } |
| 329 | SEL_ARG *clone_and(THD *thd, SEL_ARG* arg) |
| 330 | { // Get overlapping range |
| 331 | uchar *new_min,*new_max; |
| 332 | uint8 flag_min,flag_max; |
| 333 | if (cmp_min_to_min(arg) >= 0) |
| 334 | { |
| 335 | new_min=min_value; flag_min=min_flag; |
| 336 | } |
| 337 | else |
| 338 | { |
| 339 | new_min=arg->min_value; flag_min=arg->min_flag; /* purecov: deadcode */ |
| 340 | } |
| 341 | if (cmp_max_to_max(arg) <= 0) |
| 342 | { |
| 343 | new_max=max_value; flag_max=max_flag; |
| 344 | } |
| 345 | else |
| 346 | { |
| 347 | new_max=arg->max_value; flag_max=arg->max_flag; |
| 348 | } |
| 349 | return new (thd->mem_root) SEL_ARG(field, part, new_min, new_max, flag_min, |
| 350 | flag_max, |
| 351 | MY_TEST(maybe_flag && arg->maybe_flag)); |
| 352 | } |
| 353 | SEL_ARG *clone_first(SEL_ARG *arg) |
| 354 | { // min <= X < arg->min |
| 355 | return new SEL_ARG(field,part, min_value, arg->min_value, |
| 356 | min_flag, arg->min_flag & NEAR_MIN ? 0 : NEAR_MAX, |
| 357 | maybe_flag | arg->maybe_flag); |
| 358 | } |
| 359 | SEL_ARG *clone_last(SEL_ARG *arg) |
| 360 | { // min <= X <= key_max |
| 361 | return new SEL_ARG(field, part, min_value, arg->max_value, |
| 362 | min_flag, arg->max_flag, maybe_flag | arg->maybe_flag); |
| 363 | } |
| 364 | SEL_ARG *clone(RANGE_OPT_PARAM *param, SEL_ARG *new_parent, SEL_ARG **next); |
| 365 | |
| 366 | bool copy_min(SEL_ARG* arg) |
| 367 | { // Get overlapping range |
| 368 | if (cmp_min_to_min(arg) > 0) |
| 369 | { |
| 370 | min_value=arg->min_value; min_flag=arg->min_flag; |
| 371 | if ((max_flag & (NO_MAX_RANGE | NO_MIN_RANGE)) == |
| 372 | (NO_MAX_RANGE | NO_MIN_RANGE)) |
| 373 | return 1; // Full range |
| 374 | } |
| 375 | maybe_flag|=arg->maybe_flag; |
| 376 | return 0; |
| 377 | } |
| 378 | bool copy_max(SEL_ARG* arg) |
| 379 | { // Get overlapping range |
| 380 | if (cmp_max_to_max(arg) <= 0) |
| 381 | { |
| 382 | max_value=arg->max_value; max_flag=arg->max_flag; |
| 383 | if ((max_flag & (NO_MAX_RANGE | NO_MIN_RANGE)) == |
| 384 | (NO_MAX_RANGE | NO_MIN_RANGE)) |
| 385 | return 1; // Full range |
| 386 | } |
| 387 | maybe_flag|=arg->maybe_flag; |
| 388 | return 0; |
| 389 | } |
| 390 | |
| 391 | void copy_min_to_min(SEL_ARG *arg) |
| 392 | { |
| 393 | min_value=arg->min_value; min_flag=arg->min_flag; |
| 394 | } |
| 395 | void copy_min_to_max(SEL_ARG *arg) |
| 396 | { |
| 397 | max_value=arg->min_value; |
| 398 | max_flag=arg->min_flag & NEAR_MIN ? 0 : NEAR_MAX; |
| 399 | } |
| 400 | void copy_max_to_min(SEL_ARG *arg) |
| 401 | { |
| 402 | min_value=arg->max_value; |
| 403 | min_flag=arg->max_flag & NEAR_MAX ? 0 : NEAR_MIN; |
| 404 | } |
| 405 | /* returns a number of keypart values (0 or 1) appended to the key buffer */ |
| 406 | int store_min(uint length, uchar **min_key,uint min_key_flag) |
| 407 | { |
| 408 | /* "(kp1 > c1) AND (kp2 OP c2) AND ..." -> (kp1 > c1) */ |
| 409 | if ((min_flag & GEOM_FLAG) || |
| 410 | (!(min_flag & NO_MIN_RANGE) && |
| 411 | !(min_key_flag & (NO_MIN_RANGE | NEAR_MIN)))) |
| 412 | { |
| 413 | if (maybe_null && *min_value) |
| 414 | { |
| 415 | **min_key=1; |
| 416 | bzero(*min_key+1,length-1); |
| 417 | } |
| 418 | else |
| 419 | memcpy(*min_key,min_value,length); |
| 420 | (*min_key)+= length; |
| 421 | return 1; |
| 422 | } |
| 423 | return 0; |
| 424 | } |
| 425 | /* returns a number of keypart values (0 or 1) appended to the key buffer */ |
| 426 | int store_max(uint length, uchar **max_key, uint max_key_flag) |
| 427 | { |
| 428 | if (!(max_flag & NO_MAX_RANGE) && |
| 429 | !(max_key_flag & (NO_MAX_RANGE | NEAR_MAX))) |
| 430 | { |
| 431 | if (maybe_null && *max_value) |
| 432 | { |
| 433 | **max_key=1; |
| 434 | bzero(*max_key+1,length-1); |
| 435 | } |
| 436 | else |
| 437 | memcpy(*max_key,max_value,length); |
| 438 | (*max_key)+= length; |
| 439 | return 1; |
| 440 | } |
| 441 | return 0; |
| 442 | } |
| 443 | |
| 444 | /* |
| 445 | Returns a number of keypart values appended to the key buffer |
| 446 | for min key and max key. This function is used by both Range |
| 447 | Analysis and Partition pruning. For partition pruning we have |
| 448 | to ensure that we don't store also subpartition fields. Thus |
| 449 | we have to stop at the last partition part and not step into |
| 450 | the subpartition fields. For Range Analysis we set last_part |
| 451 | to MAX_KEY which we should never reach. |
| 452 | */ |
| 453 | int store_min_key(KEY_PART *key, |
| 454 | uchar **range_key, |
| 455 | uint *range_key_flag, |
| 456 | uint last_part) |
| 457 | { |
| 458 | SEL_ARG *key_tree= first(); |
| 459 | uint res= key_tree->store_min(key[key_tree->part].store_length, |
| 460 | range_key, *range_key_flag); |
| 461 | *range_key_flag|= key_tree->min_flag; |
| 462 | if (key_tree->next_key_part && |
| 463 | key_tree->next_key_part->type == SEL_ARG::KEY_RANGE && |
| 464 | key_tree->part != last_part && |
| 465 | key_tree->next_key_part->part == key_tree->part+1 && |
| 466 | !(*range_key_flag & (NO_MIN_RANGE | NEAR_MIN))) |
| 467 | res+= key_tree->next_key_part->store_min_key(key, |
| 468 | range_key, |
| 469 | range_key_flag, |
| 470 | last_part); |
| 471 | return res; |
| 472 | } |
| 473 | |
| 474 | /* returns a number of keypart values appended to the key buffer */ |
| 475 | int store_max_key(KEY_PART *key, |
| 476 | uchar **range_key, |
| 477 | uint *range_key_flag, |
| 478 | uint last_part) |
| 479 | { |
| 480 | SEL_ARG *key_tree= last(); |
| 481 | uint res=key_tree->store_max(key[key_tree->part].store_length, |
| 482 | range_key, *range_key_flag); |
| 483 | (*range_key_flag)|= key_tree->max_flag; |
| 484 | if (key_tree->next_key_part && |
| 485 | key_tree->next_key_part->type == SEL_ARG::KEY_RANGE && |
| 486 | key_tree->part != last_part && |
| 487 | key_tree->next_key_part->part == key_tree->part+1 && |
| 488 | !(*range_key_flag & (NO_MAX_RANGE | NEAR_MAX))) |
| 489 | res+= key_tree->next_key_part->store_max_key(key, |
| 490 | range_key, |
| 491 | range_key_flag, |
| 492 | last_part); |
| 493 | return res; |
| 494 | } |
| 495 | |
| 496 | SEL_ARG *insert(SEL_ARG *key); |
| 497 | SEL_ARG *tree_delete(SEL_ARG *key); |
| 498 | SEL_ARG *find_range(SEL_ARG *key); |
| 499 | SEL_ARG *rb_insert(SEL_ARG *leaf); |
| 500 | friend SEL_ARG *rb_delete_fixup(SEL_ARG *root,SEL_ARG *key, SEL_ARG *par); |
| 501 | #ifdef EXTRA_DEBUG |
| 502 | friend int test_rb_tree(SEL_ARG *element,SEL_ARG *parent); |
| 503 | void test_use_count(SEL_ARG *root); |
| 504 | #endif |
| 505 | SEL_ARG *first(); |
| 506 | const SEL_ARG *first() const; |
| 507 | SEL_ARG *last(); |
| 508 | void make_root(); |
| 509 | inline bool simple_key() |
| 510 | { |
| 511 | return !next_key_part && elements == 1; |
| 512 | } |
| 513 | void increment_use_count(long count) |
| 514 | { |
| 515 | if (next_key_part) |
| 516 | { |
| 517 | next_key_part->use_count+=count; |
| 518 | count*= (next_key_part->use_count-count); |
| 519 | for (SEL_ARG *pos=next_key_part->first(); pos ; pos=pos->next) |
| 520 | if (pos->next_key_part) |
| 521 | pos->increment_use_count(count); |
| 522 | } |
| 523 | } |
| 524 | void incr_refs() |
| 525 | { |
| 526 | increment_use_count(1); |
| 527 | use_count++; |
| 528 | } |
| 529 | void incr_refs_all() |
| 530 | { |
| 531 | for (SEL_ARG *pos=first(); pos ; pos=pos->next) |
| 532 | { |
| 533 | pos->increment_use_count(1); |
| 534 | } |
| 535 | use_count++; |
| 536 | } |
| 537 | void free_tree() |
| 538 | { |
| 539 | for (SEL_ARG *pos=first(); pos ; pos=pos->next) |
| 540 | if (pos->next_key_part) |
| 541 | { |
| 542 | pos->next_key_part->use_count--; |
| 543 | pos->next_key_part->free_tree(); |
| 544 | } |
| 545 | } |
| 546 | |
| 547 | inline SEL_ARG **parent_ptr() |
| 548 | { |
| 549 | return parent->left == this ? &parent->left : &parent->right; |
| 550 | } |
| 551 | |
| 552 | |
| 553 | /* |
| 554 | Check if this SEL_ARG object represents a single-point interval |
| 555 | |
| 556 | SYNOPSIS |
| 557 | is_singlepoint() |
| 558 | |
| 559 | DESCRIPTION |
| 560 | Check if this SEL_ARG object (not tree) represents a single-point |
| 561 | interval, i.e. if it represents a "keypart = const" or |
| 562 | "keypart IS NULL". |
| 563 | |
| 564 | RETURN |
| 565 | TRUE This SEL_ARG object represents a singlepoint interval |
| 566 | FALSE Otherwise |
| 567 | */ |
| 568 | |
| 569 | bool is_singlepoint() |
| 570 | { |
| 571 | /* |
| 572 | Check for NEAR_MIN ("strictly less") and NO_MIN_RANGE (-inf < field) |
| 573 | flags, and the same for right edge. |
| 574 | */ |
| 575 | if (min_flag || max_flag) |
| 576 | return FALSE; |
| 577 | uchar *min_val= min_value; |
| 578 | uchar *max_val= max_value; |
| 579 | |
| 580 | if (maybe_null) |
| 581 | { |
| 582 | /* First byte is a NULL value indicator */ |
| 583 | if (*min_val != *max_val) |
| 584 | return FALSE; |
| 585 | |
| 586 | if (*min_val) |
| 587 | return TRUE; /* This "x IS NULL" */ |
| 588 | min_val++; |
| 589 | max_val++; |
| 590 | } |
| 591 | return !field->key_cmp(min_val, max_val); |
| 592 | } |
| 593 | SEL_ARG *clone_tree(RANGE_OPT_PARAM *param); |
| 594 | }; |
| 595 | |
| 596 | |
| 597 | class SEL_ARG_IMPOSSIBLE: public SEL_ARG |
| 598 | { |
| 599 | public: |
| 600 | SEL_ARG_IMPOSSIBLE(Field *field) |
| 601 | :SEL_ARG(field, 0, 0) |
| 602 | { |
| 603 | type= SEL_ARG::IMPOSSIBLE; |
| 604 | } |
| 605 | }; |
| 606 | |
| 607 | |
| 608 | class RANGE_OPT_PARAM |
| 609 | { |
| 610 | public: |
| 611 | THD *thd; /* Current thread handle */ |
| 612 | TABLE *table; /* Table being analyzed */ |
| 613 | table_map prev_tables; |
| 614 | table_map read_tables; |
| 615 | table_map current_table; /* Bit of the table being analyzed */ |
| 616 | |
| 617 | /* Array of parts of all keys for which range analysis is performed */ |
| 618 | KEY_PART *key_parts; |
| 619 | KEY_PART *key_parts_end; |
| 620 | MEM_ROOT *mem_root; /* Memory that will be freed when range analysis completes */ |
| 621 | MEM_ROOT *old_root; /* Memory that will last until the query end */ |
| 622 | /* |
| 623 | Number of indexes used in range analysis (In SEL_TREE::keys only first |
| 624 | #keys elements are not empty) |
| 625 | */ |
| 626 | uint keys; |
| 627 | |
| 628 | /* |
| 629 | If true, the index descriptions describe real indexes (and it is ok to |
| 630 | call field->optimize_range(real_keynr[...], ...). |
| 631 | Otherwise index description describes fake indexes. |
| 632 | */ |
| 633 | bool using_real_indexes; |
| 634 | |
| 635 | /* |
| 636 | Aggressively remove "scans" that do not have conditions on first |
| 637 | keyparts. Such scans are usable when doing partition pruning but not |
| 638 | regular range optimization. |
| 639 | */ |
| 640 | bool remove_jump_scans; |
| 641 | |
| 642 | /* |
| 643 | TRUE <=> Range analyzer should remove parts of condition that are found |
| 644 | to be always FALSE. |
| 645 | */ |
| 646 | bool remove_false_where_parts; |
| 647 | |
| 648 | /* |
| 649 | used_key_no -> table_key_no translation table. Only makes sense if |
| 650 | using_real_indexes==TRUE |
| 651 | */ |
| 652 | uint real_keynr[MAX_KEY]; |
| 653 | |
| 654 | /* |
| 655 | Used to store 'current key tuples', in both range analysis and |
| 656 | partitioning (list) analysis |
| 657 | */ |
| 658 | uchar *min_key; |
| 659 | uchar *max_key; |
| 660 | |
| 661 | /* Number of SEL_ARG objects allocated by SEL_ARG::clone_tree operations */ |
| 662 | uint alloced_sel_args; |
| 663 | |
| 664 | bool force_default_mrr; |
| 665 | KEY_PART *key[MAX_KEY]; /* First key parts of keys used in the query */ |
| 666 | |
| 667 | bool statement_should_be_aborted() const |
| 668 | { |
| 669 | return |
| 670 | thd->is_fatal_error || |
| 671 | thd->is_error() || |
| 672 | alloced_sel_args > SEL_ARG::MAX_SEL_ARGS; |
| 673 | } |
| 674 | }; |
| 675 | |
| 676 | |
| 677 | class Explain_quick_select; |
| 678 | /* |
| 679 | A "MIN_TUPLE < tbl.key_tuple < MAX_TUPLE" interval. |
| 680 | |
| 681 | One of endpoints may be absent. 'flags' member has flags which tell whether |
| 682 | the endpoints are '<' or '<='. |
| 683 | */ |
| 684 | class QUICK_RANGE :public Sql_alloc { |
| 685 | public: |
| 686 | uchar *min_key,*max_key; |
| 687 | uint16 min_length,max_length,flag; |
| 688 | key_part_map min_keypart_map, // bitmap of used keyparts in min_key |
| 689 | max_keypart_map; // bitmap of used keyparts in max_key |
| 690 | #ifdef HAVE_valgrind |
| 691 | uint16 dummy; /* Avoid warnings on 'flag' */ |
| 692 | #endif |
| 693 | QUICK_RANGE(); /* Full range */ |
| 694 | QUICK_RANGE(THD *thd, const uchar *min_key_arg, uint min_length_arg, |
| 695 | key_part_map min_keypart_map_arg, |
| 696 | const uchar *max_key_arg, uint max_length_arg, |
| 697 | key_part_map max_keypart_map_arg, |
| 698 | uint flag_arg) |
| 699 | : min_key((uchar*) thd->memdup(min_key_arg, min_length_arg + 1)), |
| 700 | max_key((uchar*) thd->memdup(max_key_arg, max_length_arg + 1)), |
| 701 | min_length((uint16) min_length_arg), |
| 702 | max_length((uint16) max_length_arg), |
| 703 | flag((uint16) flag_arg), |
| 704 | min_keypart_map(min_keypart_map_arg), |
| 705 | max_keypart_map(max_keypart_map_arg) |
| 706 | { |
| 707 | #ifdef HAVE_valgrind |
| 708 | dummy=0; |
| 709 | #endif |
| 710 | } |
| 711 | |
| 712 | /** |
| 713 | Initalizes a key_range object for communication with storage engine. |
| 714 | |
| 715 | This function facilitates communication with the Storage Engine API by |
| 716 | translating the minimum endpoint of the interval represented by this |
| 717 | QUICK_RANGE into an index range endpoint specifier for the engine. |
| 718 | |
| 719 | @param Pointer to an uninitialized key_range C struct. |
| 720 | |
| 721 | @param prefix_length The length of the search key prefix to be used for |
| 722 | lookup. |
| 723 | |
| 724 | @param keypart_map A set (bitmap) of keyparts to be used. |
| 725 | */ |
| 726 | void make_min_endpoint(key_range *kr, uint prefix_length, |
| 727 | key_part_map keypart_map) { |
| 728 | make_min_endpoint(kr); |
| 729 | kr->length= MY_MIN(kr->length, prefix_length); |
| 730 | kr->keypart_map&= keypart_map; |
| 731 | } |
| 732 | |
| 733 | /** |
| 734 | Initalizes a key_range object for communication with storage engine. |
| 735 | |
| 736 | This function facilitates communication with the Storage Engine API by |
| 737 | translating the minimum endpoint of the interval represented by this |
| 738 | QUICK_RANGE into an index range endpoint specifier for the engine. |
| 739 | |
| 740 | @param Pointer to an uninitialized key_range C struct. |
| 741 | */ |
| 742 | void make_min_endpoint(key_range *kr) { |
| 743 | kr->key= (const uchar*)min_key; |
| 744 | kr->length= min_length; |
| 745 | kr->keypart_map= min_keypart_map; |
| 746 | kr->flag= ((flag & NEAR_MIN) ? HA_READ_AFTER_KEY : |
| 747 | (flag & EQ_RANGE) ? HA_READ_KEY_EXACT : HA_READ_KEY_OR_NEXT); |
| 748 | } |
| 749 | |
| 750 | /** |
| 751 | Initalizes a key_range object for communication with storage engine. |
| 752 | |
| 753 | This function facilitates communication with the Storage Engine API by |
| 754 | translating the maximum endpoint of the interval represented by this |
| 755 | QUICK_RANGE into an index range endpoint specifier for the engine. |
| 756 | |
| 757 | @param Pointer to an uninitialized key_range C struct. |
| 758 | |
| 759 | @param prefix_length The length of the search key prefix to be used for |
| 760 | lookup. |
| 761 | |
| 762 | @param keypart_map A set (bitmap) of keyparts to be used. |
| 763 | */ |
| 764 | void make_max_endpoint(key_range *kr, uint prefix_length, |
| 765 | key_part_map keypart_map) { |
| 766 | make_max_endpoint(kr); |
| 767 | kr->length= MY_MIN(kr->length, prefix_length); |
| 768 | kr->keypart_map&= keypart_map; |
| 769 | } |
| 770 | |
| 771 | /** |
| 772 | Initalizes a key_range object for communication with storage engine. |
| 773 | |
| 774 | This function facilitates communication with the Storage Engine API by |
| 775 | translating the maximum endpoint of the interval represented by this |
| 776 | QUICK_RANGE into an index range endpoint specifier for the engine. |
| 777 | |
| 778 | @param Pointer to an uninitialized key_range C struct. |
| 779 | */ |
| 780 | void make_max_endpoint(key_range *kr) { |
| 781 | kr->key= (const uchar*)max_key; |
| 782 | kr->length= max_length; |
| 783 | kr->keypart_map= max_keypart_map; |
| 784 | /* |
| 785 | We use READ_AFTER_KEY here because if we are reading on a key |
| 786 | prefix we want to find all keys with this prefix |
| 787 | */ |
| 788 | kr->flag= (flag & NEAR_MAX ? HA_READ_BEFORE_KEY : HA_READ_AFTER_KEY); |
| 789 | } |
| 790 | }; |
| 791 | |
| 792 | |
| 793 | /* |
| 794 | Quick select interface. |
| 795 | This class is a parent for all QUICK_*_SELECT and FT_SELECT classes. |
| 796 | |
| 797 | The usage scenario is as follows: |
| 798 | 1. Create quick select |
| 799 | quick= new QUICK_XXX_SELECT(...); |
| 800 | |
| 801 | 2. Perform lightweight initialization. This can be done in 2 ways: |
| 802 | 2.a: Regular initialization |
| 803 | if (quick->init()) |
| 804 | { |
| 805 | //the only valid action after failed init() call is delete |
| 806 | delete quick; |
| 807 | } |
| 808 | 2.b: Special initialization for quick selects merged by QUICK_ROR_*_SELECT |
| 809 | if (quick->init_ror_merged_scan()) |
| 810 | delete quick; |
| 811 | |
| 812 | 3. Perform zero, one, or more scans. |
| 813 | while (...) |
| 814 | { |
| 815 | // initialize quick select for scan. This may allocate |
| 816 | // buffers and/or prefetch rows. |
| 817 | if (quick->reset()) |
| 818 | { |
| 819 | //the only valid action after failed reset() call is delete |
| 820 | delete quick; |
| 821 | //abort query |
| 822 | } |
| 823 | |
| 824 | // perform the scan |
| 825 | do |
| 826 | { |
| 827 | res= quick->get_next(); |
| 828 | } while (res && ...) |
| 829 | } |
| 830 | |
| 831 | 4. Delete the select: |
| 832 | delete quick; |
| 833 | |
| 834 | NOTE |
| 835 | quick select doesn't use Sql_alloc/MEM_ROOT allocation because "range |
| 836 | checked for each record" functionality may create/destroy |
| 837 | O(#records_in_some_table) quick selects during query execution. |
| 838 | */ |
| 839 | |
| 840 | class QUICK_SELECT_I |
| 841 | { |
| 842 | public: |
| 843 | ha_rows records; /* estimate of # of records to be retrieved */ |
| 844 | double read_time; /* time to perform this retrieval */ |
| 845 | TABLE *head; |
| 846 | /* |
| 847 | Index this quick select uses, or MAX_KEY for quick selects |
| 848 | that use several indexes |
| 849 | */ |
| 850 | uint index; |
| 851 | |
| 852 | /* |
| 853 | Total length of first used_key_parts parts of the key. |
| 854 | Applicable if index!= MAX_KEY. |
| 855 | */ |
| 856 | uint max_used_key_length; |
| 857 | |
| 858 | /* |
| 859 | Max. number of (first) key parts this quick select uses for retrieval. |
| 860 | eg. for "(key1p1=c1 AND key1p2=c2) OR key1p1=c2" used_key_parts == 2. |
| 861 | Applicable if index!= MAX_KEY. |
| 862 | |
| 863 | For QUICK_GROUP_MIN_MAX_SELECT it includes MIN/MAX argument keyparts. |
| 864 | */ |
| 865 | uint used_key_parts; |
| 866 | |
| 867 | QUICK_SELECT_I(); |
| 868 | virtual ~QUICK_SELECT_I(){}; |
| 869 | |
| 870 | /* |
| 871 | Do post-constructor initialization. |
| 872 | SYNOPSIS |
| 873 | init() |
| 874 | |
| 875 | init() performs initializations that should have been in constructor if |
| 876 | it was possible to return errors from constructors. The join optimizer may |
| 877 | create and then delete quick selects without retrieving any rows so init() |
| 878 | must not contain any IO or CPU intensive code. |
| 879 | |
| 880 | If init() call fails the only valid action is to delete this quick select, |
| 881 | reset() and get_next() must not be called. |
| 882 | |
| 883 | RETURN |
| 884 | 0 OK |
| 885 | other Error code |
| 886 | */ |
| 887 | virtual int init() = 0; |
| 888 | |
| 889 | /* |
| 890 | Initialize quick select for row retrieval. |
| 891 | SYNOPSIS |
| 892 | reset() |
| 893 | |
| 894 | reset() should be called when it is certain that row retrieval will be |
| 895 | necessary. This call may do heavyweight initialization like buffering first |
| 896 | N records etc. If reset() call fails get_next() must not be called. |
| 897 | Note that reset() may be called several times if |
| 898 | * the quick select is executed in a subselect |
| 899 | * a JOIN buffer is used |
| 900 | |
| 901 | RETURN |
| 902 | 0 OK |
| 903 | other Error code |
| 904 | */ |
| 905 | virtual int reset(void) = 0; |
| 906 | |
| 907 | virtual int get_next() = 0; /* get next record to retrieve */ |
| 908 | |
| 909 | /* Range end should be called when we have looped over the whole index */ |
| 910 | virtual void range_end() {} |
| 911 | |
| 912 | virtual bool reverse_sorted() = 0; |
| 913 | virtual bool unique_key_range() { return false; } |
| 914 | |
| 915 | /* |
| 916 | Request that this quick select produces sorted output. Not all quick |
| 917 | selects can do it, the caller is responsible for calling this function |
| 918 | only for those quick selects that can. |
| 919 | */ |
| 920 | virtual void need_sorted_output() = 0; |
| 921 | enum { |
| 922 | QS_TYPE_RANGE = 0, |
| 923 | QS_TYPE_INDEX_INTERSECT = 1, |
| 924 | QS_TYPE_INDEX_MERGE = 2, |
| 925 | QS_TYPE_RANGE_DESC = 3, |
| 926 | QS_TYPE_FULLTEXT = 4, |
| 927 | QS_TYPE_ROR_INTERSECT = 5, |
| 928 | QS_TYPE_ROR_UNION = 6, |
| 929 | QS_TYPE_GROUP_MIN_MAX = 7 |
| 930 | }; |
| 931 | |
| 932 | /* Get type of this quick select - one of the QS_TYPE_* values */ |
| 933 | virtual int get_type() = 0; |
| 934 | |
| 935 | /* |
| 936 | Initialize this quick select as a merged scan inside a ROR-union or a ROR- |
| 937 | intersection scan. The caller must not additionally call init() if this |
| 938 | function is called. |
| 939 | SYNOPSIS |
| 940 | init_ror_merged_scan() |
| 941 | reuse_handler If true, the quick select may use table->handler, |
| 942 | otherwise it must create and use a separate handler |
| 943 | object. |
| 944 | RETURN |
| 945 | 0 Ok |
| 946 | other Error |
| 947 | */ |
| 948 | virtual int init_ror_merged_scan(bool reuse_handler, MEM_ROOT *alloc) |
| 949 | { DBUG_ASSERT(0); return 1; } |
| 950 | |
| 951 | /* |
| 952 | Save ROWID of last retrieved row in file->ref. This used in ROR-merging. |
| 953 | */ |
| 954 | virtual void save_last_pos(){}; |
| 955 | |
| 956 | void add_key_and_length(String *key_names, |
| 957 | String *used_lengths, |
| 958 | bool *first); |
| 959 | |
| 960 | /* |
| 961 | Append comma-separated list of keys this quick select uses to key_names; |
| 962 | append comma-separated list of corresponding used lengths to used_lengths. |
| 963 | This is used by select_describe. |
| 964 | */ |
| 965 | virtual void add_keys_and_lengths(String *key_names, |
| 966 | String *used_lengths)=0; |
| 967 | |
| 968 | void add_key_name(String *str, bool *first); |
| 969 | |
| 970 | /* Save information about quick select's query plan */ |
| 971 | virtual Explain_quick_select* get_explain(MEM_ROOT *alloc)= 0; |
| 972 | |
| 973 | /* |
| 974 | Return 1 if any index used by this quick select |
| 975 | uses field which is marked in passed bitmap. |
| 976 | */ |
| 977 | virtual bool is_keys_used(const MY_BITMAP *fields); |
| 978 | |
| 979 | /** |
| 980 | Simple sanity check that the quick select has been set up |
| 981 | correctly. Function is overridden by quick selects that merge |
| 982 | indices. |
| 983 | */ |
| 984 | virtual bool is_valid() { return index != MAX_KEY; }; |
| 985 | |
| 986 | /* |
| 987 | rowid of last row retrieved by this quick select. This is used only when |
| 988 | doing ROR-index_merge selects |
| 989 | */ |
| 990 | uchar *last_rowid; |
| 991 | |
| 992 | /* |
| 993 | Table record buffer used by this quick select. |
| 994 | */ |
| 995 | uchar *record; |
| 996 | |
| 997 | virtual void replace_handler(handler *new_file) |
| 998 | { |
| 999 | DBUG_ASSERT(0); /* Only supported in QUICK_RANGE_SELECT */ |
| 1000 | } |
| 1001 | |
| 1002 | #ifndef DBUG_OFF |
| 1003 | /* |
| 1004 | Print quick select information to DBUG_FILE. Caller is responsible |
| 1005 | for locking DBUG_FILE before this call and unlocking it afterwards. |
| 1006 | */ |
| 1007 | virtual void dbug_dump(int indent, bool verbose)= 0; |
| 1008 | #endif |
| 1009 | |
| 1010 | /* |
| 1011 | Returns a QUICK_SELECT with reverse order of to the index. |
| 1012 | */ |
| 1013 | virtual QUICK_SELECT_I *make_reverse(uint used_key_parts_arg) { return NULL; } |
| 1014 | |
| 1015 | /* |
| 1016 | Add the key columns used by the quick select into table's read set. |
| 1017 | |
| 1018 | This is used by an optimization in filesort. |
| 1019 | */ |
| 1020 | virtual void add_used_key_part_to_set()=0; |
| 1021 | }; |
| 1022 | |
| 1023 | |
| 1024 | struct st_qsel_param; |
| 1025 | class PARAM; |
| 1026 | |
| 1027 | |
| 1028 | /* |
| 1029 | MRR range sequence, array<QUICK_RANGE> implementation: sequence traversal |
| 1030 | context. |
| 1031 | */ |
| 1032 | typedef struct st_quick_range_seq_ctx |
| 1033 | { |
| 1034 | QUICK_RANGE **first; |
| 1035 | QUICK_RANGE **cur; |
| 1036 | QUICK_RANGE **last; |
| 1037 | } QUICK_RANGE_SEQ_CTX; |
| 1038 | |
| 1039 | range_seq_t quick_range_seq_init(void *init_param, uint n_ranges, uint flags); |
| 1040 | bool quick_range_seq_next(range_seq_t rseq, KEY_MULTI_RANGE *range); |
| 1041 | |
| 1042 | |
| 1043 | /* |
| 1044 | Quick select that does a range scan on a single key. The records are |
| 1045 | returned in key order. |
| 1046 | */ |
| 1047 | class QUICK_RANGE_SELECT : public QUICK_SELECT_I |
| 1048 | { |
| 1049 | protected: |
| 1050 | THD *thd; |
| 1051 | bool no_alloc; |
| 1052 | MEM_ROOT *parent_alloc; |
| 1053 | |
| 1054 | /* true if we enabled key only reads */ |
| 1055 | handler *file; |
| 1056 | |
| 1057 | /* Members to deal with case when this quick select is a ROR-merged scan */ |
| 1058 | bool in_ror_merged_scan; |
| 1059 | MY_BITMAP column_bitmap; |
| 1060 | bool free_file; /* TRUE <=> this->file is "owned" by this quick select */ |
| 1061 | |
| 1062 | /* Range pointers to be used when not using MRR interface */ |
| 1063 | /* Members needed to use the MRR interface */ |
| 1064 | QUICK_RANGE_SEQ_CTX qr_traversal_ctx; |
| 1065 | public: |
| 1066 | uint mrr_flags; /* Flags to be used with MRR interface */ |
| 1067 | protected: |
| 1068 | uint mrr_buf_size; /* copy from thd->variables.mrr_buff_size */ |
| 1069 | HANDLER_BUFFER *mrr_buf_desc; /* the handler buffer */ |
| 1070 | |
| 1071 | /* Info about index we're scanning */ |
| 1072 | |
| 1073 | DYNAMIC_ARRAY ranges; /* ordered array of range ptrs */ |
| 1074 | QUICK_RANGE **cur_range; /* current element in ranges */ |
| 1075 | |
| 1076 | QUICK_RANGE *last_range; |
| 1077 | |
| 1078 | KEY_PART *key_parts; |
| 1079 | KEY_PART_INFO *key_part_info; |
| 1080 | |
| 1081 | bool dont_free; /* Used by QUICK_SELECT_DESC */ |
| 1082 | |
| 1083 | int cmp_next(QUICK_RANGE *range); |
| 1084 | int cmp_prev(QUICK_RANGE *range); |
| 1085 | bool row_in_ranges(); |
| 1086 | public: |
| 1087 | MEM_ROOT alloc; |
| 1088 | |
| 1089 | QUICK_RANGE_SELECT(THD *thd, TABLE *table,uint index_arg,bool no_alloc, |
| 1090 | MEM_ROOT *parent_alloc, bool *create_err); |
| 1091 | ~QUICK_RANGE_SELECT(); |
| 1092 | virtual QUICK_RANGE_SELECT *clone(bool *create_error) |
| 1093 | { return new QUICK_RANGE_SELECT(thd, head, index, no_alloc, parent_alloc, |
| 1094 | create_error); } |
| 1095 | |
| 1096 | void need_sorted_output(); |
| 1097 | int init(); |
| 1098 | int reset(void); |
| 1099 | int get_next(); |
| 1100 | void range_end(); |
| 1101 | int get_next_prefix(uint prefix_length, uint group_key_parts, |
| 1102 | uchar *cur_prefix); |
| 1103 | bool reverse_sorted() { return 0; } |
| 1104 | bool unique_key_range(); |
| 1105 | int init_ror_merged_scan(bool reuse_handler, MEM_ROOT *alloc); |
| 1106 | void save_last_pos() |
| 1107 | { file->position(record); } |
| 1108 | int get_type() { return QS_TYPE_RANGE; } |
| 1109 | void add_keys_and_lengths(String *key_names, String *used_lengths); |
| 1110 | Explain_quick_select *get_explain(MEM_ROOT *alloc); |
| 1111 | #ifndef DBUG_OFF |
| 1112 | void dbug_dump(int indent, bool verbose); |
| 1113 | #endif |
| 1114 | virtual void replace_handler(handler *new_file) { file= new_file; } |
| 1115 | QUICK_SELECT_I *make_reverse(uint used_key_parts_arg); |
| 1116 | |
| 1117 | virtual void add_used_key_part_to_set(); |
| 1118 | |
| 1119 | private: |
| 1120 | /* Default copy ctor used by QUICK_SELECT_DESC */ |
| 1121 | friend class TRP_ROR_INTERSECT; |
| 1122 | friend |
| 1123 | QUICK_RANGE_SELECT *get_quick_select_for_ref(THD *thd, TABLE *table, |
| 1124 | struct st_table_ref *ref, |
| 1125 | ha_rows records); |
| 1126 | friend bool get_quick_keys(PARAM *param, QUICK_RANGE_SELECT *quick, |
| 1127 | KEY_PART *key, SEL_ARG *key_tree, |
| 1128 | uchar *min_key, uint min_key_flag, |
| 1129 | uchar *max_key, uint max_key_flag); |
| 1130 | friend QUICK_RANGE_SELECT *get_quick_select(PARAM*,uint idx, |
| 1131 | SEL_ARG *key_tree, |
| 1132 | uint mrr_flags, |
| 1133 | uint mrr_buf_size, |
| 1134 | MEM_ROOT *alloc); |
| 1135 | friend class QUICK_SELECT_DESC; |
| 1136 | friend class QUICK_INDEX_SORT_SELECT; |
| 1137 | friend class QUICK_INDEX_MERGE_SELECT; |
| 1138 | friend class QUICK_ROR_INTERSECT_SELECT; |
| 1139 | friend class QUICK_INDEX_INTERSECT_SELECT; |
| 1140 | friend class QUICK_GROUP_MIN_MAX_SELECT; |
| 1141 | friend bool quick_range_seq_next(range_seq_t rseq, KEY_MULTI_RANGE *range); |
| 1142 | friend range_seq_t quick_range_seq_init(void *init_param, |
| 1143 | uint n_ranges, uint flags); |
| 1144 | friend |
| 1145 | int read_keys_and_merge_scans(THD *thd, TABLE *head, |
| 1146 | List<QUICK_RANGE_SELECT> quick_selects, |
| 1147 | QUICK_RANGE_SELECT *pk_quick_select, |
| 1148 | READ_RECORD *read_record, |
| 1149 | bool intersection, |
| 1150 | key_map *filtered_scans, |
| 1151 | Unique **unique_ptr); |
| 1152 | |
| 1153 | }; |
| 1154 | |
| 1155 | |
| 1156 | class QUICK_RANGE_SELECT_GEOM: public QUICK_RANGE_SELECT |
| 1157 | { |
| 1158 | public: |
| 1159 | QUICK_RANGE_SELECT_GEOM(THD *thd, TABLE *table, uint index_arg, |
| 1160 | bool no_alloc, MEM_ROOT *parent_alloc, |
| 1161 | bool *create_err) |
| 1162 | :QUICK_RANGE_SELECT(thd, table, index_arg, no_alloc, parent_alloc, |
| 1163 | create_err) |
| 1164 | {}; |
| 1165 | virtual QUICK_RANGE_SELECT *clone(bool *create_error) |
| 1166 | { |
| 1167 | DBUG_ASSERT(0); |
| 1168 | return new QUICK_RANGE_SELECT_GEOM(thd, head, index, no_alloc, |
| 1169 | parent_alloc, create_error); |
| 1170 | } |
| 1171 | virtual int get_next(); |
| 1172 | }; |
| 1173 | |
| 1174 | |
| 1175 | /* |
| 1176 | QUICK_INDEX_SORT_SELECT is the base class for the common functionality of: |
| 1177 | - QUICK_INDEX_MERGE_SELECT, access based on multi-index merge/union |
| 1178 | - QUICK_INDEX_INTERSECT_SELECT, access based on multi-index intersection |
| 1179 | |
| 1180 | |
| 1181 | QUICK_INDEX_SORT_SELECT uses |
| 1182 | * QUICK_RANGE_SELECTs to get rows |
| 1183 | * Unique class |
| 1184 | - to remove duplicate rows for QUICK_INDEX_MERGE_SELECT |
| 1185 | - to intersect rows for QUICK_INDEX_INTERSECT_SELECT |
| 1186 | |
| 1187 | INDEX MERGE OPTIMIZER |
| 1188 | Current implementation doesn't detect all cases where index merge could |
| 1189 | be used, in particular: |
| 1190 | |
| 1191 | * index_merge+'using index' is not supported |
| 1192 | |
| 1193 | * If WHERE part contains complex nested AND and OR conditions, some ways |
| 1194 | to retrieve rows using index merge will not be considered. The choice |
| 1195 | of read plan may depend on the order of conjuncts/disjuncts in WHERE |
| 1196 | part of the query, see comments near imerge_list_or_list and |
| 1197 | SEL_IMERGE::or_sel_tree_with_checks functions for details. |
| 1198 | |
| 1199 | * There is no "index_merge_ref" method (but index merge on non-first |
| 1200 | table in join is possible with 'range checked for each record'). |
| 1201 | |
| 1202 | |
| 1203 | ROW RETRIEVAL ALGORITHM |
| 1204 | |
| 1205 | index merge/intersection uses Unique class for duplicates removal. |
| 1206 | index merge/intersection takes advantage of Clustered Primary Key (CPK) |
| 1207 | if the table has one. |
| 1208 | The index merge/intersection algorithm consists of two phases: |
| 1209 | |
| 1210 | Phase 1 |
| 1211 | (implemented by a QUICK_INDEX_MERGE_SELECT::read_keys_and_merge call): |
| 1212 | |
| 1213 | prepare() |
| 1214 | { |
| 1215 | activate 'index only'; |
| 1216 | while(retrieve next row for non-CPK scan) |
| 1217 | { |
| 1218 | if (there is a CPK scan and row will be retrieved by it) |
| 1219 | skip this row; |
| 1220 | else |
| 1221 | put its rowid into Unique; |
| 1222 | } |
| 1223 | deactivate 'index only'; |
| 1224 | } |
| 1225 | |
| 1226 | Phase 2 |
| 1227 | (implemented as sequence of QUICK_INDEX_MERGE_SELECT::get_next calls): |
| 1228 | |
| 1229 | fetch() |
| 1230 | { |
| 1231 | retrieve all rows from row pointers stored in Unique |
| 1232 | (merging/intersecting them); |
| 1233 | free Unique; |
| 1234 | if (! intersection) |
| 1235 | retrieve all rows for CPK scan; |
| 1236 | } |
| 1237 | */ |
| 1238 | |
| 1239 | class QUICK_INDEX_SORT_SELECT : public QUICK_SELECT_I |
| 1240 | { |
| 1241 | protected: |
| 1242 | Unique *unique; |
| 1243 | public: |
| 1244 | QUICK_INDEX_SORT_SELECT(THD *thd, TABLE *table); |
| 1245 | ~QUICK_INDEX_SORT_SELECT(); |
| 1246 | |
| 1247 | int init(); |
| 1248 | void need_sorted_output() { DBUG_ASSERT(0); /* Can't do it */ } |
| 1249 | int reset(void); |
| 1250 | bool reverse_sorted() { return false; } |
| 1251 | bool unique_key_range() { return false; } |
| 1252 | bool is_keys_used(const MY_BITMAP *fields); |
| 1253 | #ifndef DBUG_OFF |
| 1254 | void dbug_dump(int indent, bool verbose); |
| 1255 | #endif |
| 1256 | Explain_quick_select *get_explain(MEM_ROOT *alloc); |
| 1257 | |
| 1258 | bool push_quick_back(QUICK_RANGE_SELECT *quick_sel_range); |
| 1259 | |
| 1260 | /* range quick selects this index merge/intersect consists of */ |
| 1261 | List<QUICK_RANGE_SELECT> quick_selects; |
| 1262 | |
| 1263 | /* quick select that uses clustered primary key (NULL if none) */ |
| 1264 | QUICK_RANGE_SELECT* pk_quick_select; |
| 1265 | |
| 1266 | MEM_ROOT alloc; |
| 1267 | THD *thd; |
| 1268 | virtual bool is_valid() |
| 1269 | { |
| 1270 | List_iterator_fast<QUICK_RANGE_SELECT> it(quick_selects); |
| 1271 | QUICK_RANGE_SELECT *quick; |
| 1272 | bool valid= true; |
| 1273 | while ((quick= it++)) |
| 1274 | { |
| 1275 | if (!quick->is_valid()) |
| 1276 | { |
| 1277 | valid= false; |
| 1278 | break; |
| 1279 | } |
| 1280 | } |
| 1281 | return valid; |
| 1282 | } |
| 1283 | virtual int read_keys_and_merge()= 0; |
| 1284 | /* used to get rows collected in Unique */ |
| 1285 | READ_RECORD read_record; |
| 1286 | |
| 1287 | virtual void add_used_key_part_to_set(); |
| 1288 | }; |
| 1289 | |
| 1290 | |
| 1291 | |
| 1292 | class QUICK_INDEX_MERGE_SELECT : public QUICK_INDEX_SORT_SELECT |
| 1293 | { |
| 1294 | private: |
| 1295 | /* true if this select is currently doing a clustered PK scan */ |
| 1296 | bool doing_pk_scan; |
| 1297 | protected: |
| 1298 | int read_keys_and_merge(); |
| 1299 | |
| 1300 | public: |
| 1301 | QUICK_INDEX_MERGE_SELECT(THD *thd_arg, TABLE *table) |
| 1302 | :QUICK_INDEX_SORT_SELECT(thd_arg, table) {} |
| 1303 | |
| 1304 | int get_next(); |
| 1305 | int get_type() { return QS_TYPE_INDEX_MERGE; } |
| 1306 | void add_keys_and_lengths(String *key_names, String *used_lengths); |
| 1307 | }; |
| 1308 | |
| 1309 | class QUICK_INDEX_INTERSECT_SELECT : public QUICK_INDEX_SORT_SELECT |
| 1310 | { |
| 1311 | protected: |
| 1312 | int read_keys_and_merge(); |
| 1313 | |
| 1314 | public: |
| 1315 | QUICK_INDEX_INTERSECT_SELECT(THD *thd_arg, TABLE *table) |
| 1316 | :QUICK_INDEX_SORT_SELECT(thd_arg, table) {} |
| 1317 | |
| 1318 | key_map filtered_scans; |
| 1319 | int get_next(); |
| 1320 | int get_type() { return QS_TYPE_INDEX_INTERSECT; } |
| 1321 | void add_keys_and_lengths(String *key_names, String *used_lengths); |
| 1322 | Explain_quick_select *get_explain(MEM_ROOT *alloc); |
| 1323 | }; |
| 1324 | |
| 1325 | |
| 1326 | /* |
| 1327 | Rowid-Ordered Retrieval (ROR) index intersection quick select. |
| 1328 | This quick select produces intersection of row sequences returned |
| 1329 | by several QUICK_RANGE_SELECTs it "merges". |
| 1330 | |
| 1331 | All merged QUICK_RANGE_SELECTs must return rowids in rowid order. |
| 1332 | QUICK_ROR_INTERSECT_SELECT will return rows in rowid order, too. |
| 1333 | |
| 1334 | All merged quick selects retrieve {rowid, covered_fields} tuples (not full |
| 1335 | table records). |
| 1336 | QUICK_ROR_INTERSECT_SELECT retrieves full records if it is not being used |
| 1337 | by QUICK_ROR_INTERSECT_SELECT and all merged quick selects together don't |
| 1338 | cover needed all fields. |
| 1339 | |
| 1340 | If one of the merged quick selects is a Clustered PK range scan, it is |
| 1341 | used only to filter rowid sequence produced by other merged quick selects. |
| 1342 | */ |
| 1343 | |
| 1344 | class QUICK_ROR_INTERSECT_SELECT : public QUICK_SELECT_I |
| 1345 | { |
| 1346 | public: |
| 1347 | QUICK_ROR_INTERSECT_SELECT(THD *thd, TABLE *table, |
| 1348 | bool retrieve_full_rows, |
| 1349 | MEM_ROOT *parent_alloc); |
| 1350 | ~QUICK_ROR_INTERSECT_SELECT(); |
| 1351 | |
| 1352 | int init(); |
| 1353 | void need_sorted_output() { DBUG_ASSERT(0); /* Can't do it */ } |
| 1354 | int reset(void); |
| 1355 | int get_next(); |
| 1356 | bool reverse_sorted() { return false; } |
| 1357 | bool unique_key_range() { return false; } |
| 1358 | int get_type() { return QS_TYPE_ROR_INTERSECT; } |
| 1359 | void add_keys_and_lengths(String *key_names, String *used_lengths); |
| 1360 | Explain_quick_select *get_explain(MEM_ROOT *alloc); |
| 1361 | bool is_keys_used(const MY_BITMAP *fields); |
| 1362 | void add_used_key_part_to_set(); |
| 1363 | #ifndef DBUG_OFF |
| 1364 | void dbug_dump(int indent, bool verbose); |
| 1365 | #endif |
| 1366 | int init_ror_merged_scan(bool reuse_handler, MEM_ROOT *alloc); |
| 1367 | bool push_quick_back(MEM_ROOT *alloc, QUICK_RANGE_SELECT *quick_sel_range); |
| 1368 | |
| 1369 | class QUICK_SELECT_WITH_RECORD : public Sql_alloc |
| 1370 | { |
| 1371 | public: |
| 1372 | QUICK_RANGE_SELECT *quick; |
| 1373 | uchar *key_tuple; |
| 1374 | ~QUICK_SELECT_WITH_RECORD() { delete quick; } |
| 1375 | }; |
| 1376 | |
| 1377 | /* |
| 1378 | Range quick selects this intersection consists of, not including |
| 1379 | cpk_quick. |
| 1380 | */ |
| 1381 | List<QUICK_SELECT_WITH_RECORD> quick_selects; |
| 1382 | |
| 1383 | virtual bool is_valid() |
| 1384 | { |
| 1385 | List_iterator_fast<QUICK_SELECT_WITH_RECORD> it(quick_selects); |
| 1386 | QUICK_SELECT_WITH_RECORD *quick; |
| 1387 | bool valid= true; |
| 1388 | while ((quick= it++)) |
| 1389 | { |
| 1390 | if (!quick->quick->is_valid()) |
| 1391 | { |
| 1392 | valid= false; |
| 1393 | break; |
| 1394 | } |
| 1395 | } |
| 1396 | return valid; |
| 1397 | } |
| 1398 | |
| 1399 | /* |
| 1400 | Merged quick select that uses Clustered PK, if there is one. This quick |
| 1401 | select is not used for row retrieval, it is used for row retrieval. |
| 1402 | */ |
| 1403 | QUICK_RANGE_SELECT *cpk_quick; |
| 1404 | |
| 1405 | MEM_ROOT alloc; /* Memory pool for this and merged quick selects data. */ |
| 1406 | THD *thd; /* current thread */ |
| 1407 | bool need_to_fetch_row; /* if true, do retrieve full table records. */ |
| 1408 | /* in top-level quick select, true if merged scans where initialized */ |
| 1409 | bool scans_inited; |
| 1410 | }; |
| 1411 | |
| 1412 | |
| 1413 | /* |
| 1414 | Rowid-Ordered Retrieval index union select. |
| 1415 | This quick select produces union of row sequences returned by several |
| 1416 | quick select it "merges". |
| 1417 | |
| 1418 | All merged quick selects must return rowids in rowid order. |
| 1419 | QUICK_ROR_UNION_SELECT will return rows in rowid order, too. |
| 1420 | |
| 1421 | All merged quick selects are set not to retrieve full table records. |
| 1422 | ROR-union quick select always retrieves full records. |
| 1423 | |
| 1424 | */ |
| 1425 | |
| 1426 | class QUICK_ROR_UNION_SELECT : public QUICK_SELECT_I |
| 1427 | { |
| 1428 | public: |
| 1429 | QUICK_ROR_UNION_SELECT(THD *thd, TABLE *table); |
| 1430 | ~QUICK_ROR_UNION_SELECT(); |
| 1431 | |
| 1432 | int init(); |
| 1433 | void need_sorted_output() { DBUG_ASSERT(0); /* Can't do it */ } |
| 1434 | int reset(void); |
| 1435 | int get_next(); |
| 1436 | bool reverse_sorted() { return false; } |
| 1437 | bool unique_key_range() { return false; } |
| 1438 | int get_type() { return QS_TYPE_ROR_UNION; } |
| 1439 | void add_keys_and_lengths(String *key_names, String *used_lengths); |
| 1440 | Explain_quick_select *get_explain(MEM_ROOT *alloc); |
| 1441 | bool is_keys_used(const MY_BITMAP *fields); |
| 1442 | void add_used_key_part_to_set(); |
| 1443 | #ifndef DBUG_OFF |
| 1444 | void dbug_dump(int indent, bool verbose); |
| 1445 | #endif |
| 1446 | |
| 1447 | bool push_quick_back(QUICK_SELECT_I *quick_sel_range); |
| 1448 | |
| 1449 | List<QUICK_SELECT_I> quick_selects; /* Merged quick selects */ |
| 1450 | |
| 1451 | virtual bool is_valid() |
| 1452 | { |
| 1453 | List_iterator_fast<QUICK_SELECT_I> it(quick_selects); |
| 1454 | QUICK_SELECT_I *quick; |
| 1455 | bool valid= true; |
| 1456 | while ((quick= it++)) |
| 1457 | { |
| 1458 | if (!quick->is_valid()) |
| 1459 | { |
| 1460 | valid= false; |
| 1461 | break; |
| 1462 | } |
| 1463 | } |
| 1464 | return valid; |
| 1465 | } |
| 1466 | |
| 1467 | QUEUE queue; /* Priority queue for merge operation */ |
| 1468 | MEM_ROOT alloc; /* Memory pool for this and merged quick selects data. */ |
| 1469 | |
| 1470 | THD *thd; /* current thread */ |
| 1471 | uchar *cur_rowid; /* buffer used in get_next() */ |
| 1472 | uchar *prev_rowid; /* rowid of last row returned by get_next() */ |
| 1473 | bool have_prev_rowid; /* true if prev_rowid has valid data */ |
| 1474 | uint rowid_length; /* table rowid length */ |
| 1475 | private: |
| 1476 | bool scans_inited; |
| 1477 | }; |
| 1478 | |
| 1479 | |
| 1480 | /* |
| 1481 | Index scan for GROUP-BY queries with MIN/MAX aggregate functions. |
| 1482 | |
| 1483 | This class provides a specialized index access method for GROUP-BY queries |
| 1484 | of the forms: |
| 1485 | |
| 1486 | SELECT A_1,...,A_k, [B_1,...,B_m], [MIN(C)], [MAX(C)] |
| 1487 | FROM T |
| 1488 | WHERE [RNG(A_1,...,A_p ; where p <= k)] |
| 1489 | [AND EQ(B_1,...,B_m)] |
| 1490 | [AND PC(C)] |
| 1491 | [AND PA(A_i1,...,A_iq)] |
| 1492 | GROUP BY A_1,...,A_k; |
| 1493 | |
| 1494 | or |
| 1495 | |
| 1496 | SELECT DISTINCT A_i1,...,A_ik |
| 1497 | FROM T |
| 1498 | WHERE [RNG(A_1,...,A_p ; where p <= k)] |
| 1499 | [AND PA(A_i1,...,A_iq)]; |
| 1500 | |
| 1501 | where all selected fields are parts of the same index. |
| 1502 | The class of queries that can be processed by this quick select is fully |
| 1503 | specified in the description of get_best_trp_group_min_max() in opt_range.cc. |
| 1504 | |
| 1505 | The get_next() method directly produces result tuples, thus obviating the |
| 1506 | need to call end_send_group() because all grouping is already done inside |
| 1507 | get_next(). |
| 1508 | |
| 1509 | Since one of the requirements is that all select fields are part of the same |
| 1510 | index, this class produces only index keys, and not complete records. |
| 1511 | */ |
| 1512 | |
| 1513 | class QUICK_GROUP_MIN_MAX_SELECT : public QUICK_SELECT_I |
| 1514 | { |
| 1515 | private: |
| 1516 | handler * const file; /* The handler used to get data. */ |
| 1517 | JOIN *join; /* Descriptor of the current query */ |
| 1518 | KEY *index_info; /* The index chosen for data access */ |
| 1519 | uchar *record; /* Buffer where the next record is returned. */ |
| 1520 | uchar *tmp_record; /* Temporary storage for next_min(), next_max(). */ |
| 1521 | uchar *group_prefix; /* Key prefix consisting of the GROUP fields. */ |
| 1522 | const uint group_prefix_len; /* Length of the group prefix. */ |
| 1523 | uint group_key_parts; /* A number of keyparts in the group prefix */ |
| 1524 | uchar *last_prefix; /* Prefix of the last group for detecting EOF. */ |
| 1525 | bool have_min; /* Specify whether we are computing */ |
| 1526 | bool have_max; /* a MIN, a MAX, or both. */ |
| 1527 | bool have_agg_distinct;/* aggregate_function(DISTINCT ...). */ |
| 1528 | bool seen_first_key; /* Denotes whether the first key was retrieved.*/ |
| 1529 | bool doing_key_read; /* true if we enabled key only reads */ |
| 1530 | |
| 1531 | KEY_PART_INFO *min_max_arg_part; /* The keypart of the only argument field */ |
| 1532 | /* of all MIN/MAX functions. */ |
| 1533 | uint min_max_arg_len; /* The length of the MIN/MAX argument field */ |
| 1534 | uchar *key_infix; /* Infix of constants from equality predicates. */ |
| 1535 | uint key_infix_len; |
| 1536 | DYNAMIC_ARRAY min_max_ranges; /* Array of range ptrs for the MIN/MAX field. */ |
| 1537 | uint real_prefix_len; /* Length of key prefix extended with key_infix. */ |
| 1538 | uint real_key_parts; /* A number of keyparts in the above value. */ |
| 1539 | List<Item_sum> *min_functions; |
| 1540 | List<Item_sum> *max_functions; |
| 1541 | List_iterator<Item_sum> *min_functions_it; |
| 1542 | List_iterator<Item_sum> *max_functions_it; |
| 1543 | /* |
| 1544 | Use index scan to get the next different key instead of jumping into it |
| 1545 | through index read |
| 1546 | */ |
| 1547 | bool is_index_scan; |
| 1548 | public: |
| 1549 | /* |
| 1550 | The following two members are public to allow easy access from |
| 1551 | TRP_GROUP_MIN_MAX::make_quick() |
| 1552 | */ |
| 1553 | MEM_ROOT alloc; /* Memory pool for this and quick_prefix_select data. */ |
| 1554 | QUICK_RANGE_SELECT *quick_prefix_select;/* For retrieval of group prefixes. */ |
| 1555 | private: |
| 1556 | int next_prefix(); |
| 1557 | int next_min_in_range(); |
| 1558 | int next_max_in_range(); |
| 1559 | int next_min(); |
| 1560 | int next_max(); |
| 1561 | void update_min_result(); |
| 1562 | void update_max_result(); |
| 1563 | int cmp_min_max_key(const uchar *key, uint16 length); |
| 1564 | public: |
| 1565 | QUICK_GROUP_MIN_MAX_SELECT(TABLE *table, JOIN *join, bool have_min, |
| 1566 | bool have_max, bool have_agg_distinct, |
| 1567 | KEY_PART_INFO *min_max_arg_part, |
| 1568 | uint group_prefix_len, uint group_key_parts, |
| 1569 | uint used_key_parts, KEY *index_info, uint |
| 1570 | use_index, double read_cost, ha_rows records, uint |
| 1571 | key_infix_len, uchar *key_infix, MEM_ROOT |
| 1572 | *parent_alloc, bool is_index_scan); |
| 1573 | ~QUICK_GROUP_MIN_MAX_SELECT(); |
| 1574 | bool add_range(SEL_ARG *sel_range); |
| 1575 | void update_key_stat(); |
| 1576 | void adjust_prefix_ranges(); |
| 1577 | bool alloc_buffers(); |
| 1578 | int init(); |
| 1579 | void need_sorted_output() { /* always do it */ } |
| 1580 | int reset(); |
| 1581 | int get_next(); |
| 1582 | bool reverse_sorted() { return false; } |
| 1583 | bool unique_key_range() { return false; } |
| 1584 | int get_type() { return QS_TYPE_GROUP_MIN_MAX; } |
| 1585 | void add_keys_and_lengths(String *key_names, String *used_lengths); |
| 1586 | void add_used_key_part_to_set(); |
| 1587 | #ifndef DBUG_OFF |
| 1588 | void dbug_dump(int indent, bool verbose); |
| 1589 | #endif |
| 1590 | bool is_agg_distinct() { return have_agg_distinct; } |
| 1591 | bool loose_scan_is_scanning() { return is_index_scan; } |
| 1592 | Explain_quick_select *get_explain(MEM_ROOT *alloc); |
| 1593 | }; |
| 1594 | |
| 1595 | |
| 1596 | class QUICK_SELECT_DESC: public QUICK_RANGE_SELECT |
| 1597 | { |
| 1598 | public: |
| 1599 | QUICK_SELECT_DESC(QUICK_RANGE_SELECT *q, uint used_key_parts); |
| 1600 | virtual QUICK_RANGE_SELECT *clone(bool *create_error) |
| 1601 | { DBUG_ASSERT(0); return new QUICK_SELECT_DESC(this, used_key_parts); } |
| 1602 | int get_next(); |
| 1603 | bool reverse_sorted() { return 1; } |
| 1604 | int get_type() { return QS_TYPE_RANGE_DESC; } |
| 1605 | QUICK_SELECT_I *make_reverse(uint used_key_parts_arg) |
| 1606 | { |
| 1607 | return this; // is already reverse sorted |
| 1608 | } |
| 1609 | private: |
| 1610 | bool range_reads_after_key(QUICK_RANGE *range); |
| 1611 | int reset(void) { rev_it.rewind(); return QUICK_RANGE_SELECT::reset(); } |
| 1612 | List<QUICK_RANGE> rev_ranges; |
| 1613 | List_iterator<QUICK_RANGE> rev_it; |
| 1614 | uint used_key_parts; |
| 1615 | }; |
| 1616 | |
| 1617 | |
| 1618 | class SQL_SELECT :public Sql_alloc { |
| 1619 | public: |
| 1620 | QUICK_SELECT_I *quick; // If quick-select used |
| 1621 | COND *cond; // where condition |
| 1622 | |
| 1623 | /* |
| 1624 | When using Index Condition Pushdown: condition that we've had before |
| 1625 | extracting and pushing index condition. |
| 1626 | In other cases, NULL. |
| 1627 | */ |
| 1628 | Item *pre_idx_push_select_cond; |
| 1629 | TABLE *head; |
| 1630 | IO_CACHE file; // Positions to used records |
| 1631 | ha_rows records; // Records in use if read from file |
| 1632 | double read_time; // Time to read rows |
| 1633 | key_map quick_keys; // Possible quick keys |
| 1634 | key_map needed_reg; // Possible quick keys after prev tables. |
| 1635 | table_map const_tables,read_tables; |
| 1636 | /* See PARAM::possible_keys */ |
| 1637 | key_map possible_keys; |
| 1638 | bool free_cond; /* Currently not used and always FALSE */ |
| 1639 | |
| 1640 | SQL_SELECT(); |
| 1641 | ~SQL_SELECT(); |
| 1642 | void cleanup(); |
| 1643 | void set_quick(QUICK_SELECT_I *new_quick) { delete quick; quick= new_quick; } |
| 1644 | bool check_quick(THD *thd, bool force_quick_range, ha_rows limit) |
| 1645 | { |
| 1646 | key_map tmp; |
| 1647 | tmp.set_all(); |
| 1648 | return test_quick_select(thd, tmp, 0, limit, force_quick_range, FALSE, FALSE) < 0; |
| 1649 | } |
| 1650 | /* |
| 1651 | RETURN |
| 1652 | 0 if record must be skipped <-> (cond && cond->val_int() == 0) |
| 1653 | -1 if error |
| 1654 | 1 otherwise |
| 1655 | */ |
| 1656 | inline int skip_record(THD *thd) |
| 1657 | { |
| 1658 | int rc= MY_TEST(!cond || cond->val_int()); |
| 1659 | if (thd->is_error()) |
| 1660 | rc= -1; |
| 1661 | return rc; |
| 1662 | } |
| 1663 | int test_quick_select(THD *thd, key_map keys, table_map prev_tables, |
| 1664 | ha_rows limit, bool force_quick_range, |
| 1665 | bool ordered_output, bool remove_false_parts_of_where); |
| 1666 | }; |
| 1667 | |
| 1668 | |
| 1669 | class SQL_SELECT_auto |
| 1670 | { |
| 1671 | SQL_SELECT *select; |
| 1672 | public: |
| 1673 | SQL_SELECT_auto(): select(NULL) |
| 1674 | {} |
| 1675 | ~SQL_SELECT_auto() |
| 1676 | { |
| 1677 | delete select; |
| 1678 | } |
| 1679 | SQL_SELECT_auto& |
| 1680 | operator= (SQL_SELECT *_select) |
| 1681 | { |
| 1682 | select= _select; |
| 1683 | return *this; |
| 1684 | } |
| 1685 | operator SQL_SELECT * () const |
| 1686 | { |
| 1687 | return select; |
| 1688 | } |
| 1689 | SQL_SELECT * |
| 1690 | operator-> () const |
| 1691 | { |
| 1692 | return select; |
| 1693 | } |
| 1694 | operator bool () const |
| 1695 | { |
| 1696 | return select; |
| 1697 | } |
| 1698 | }; |
| 1699 | |
| 1700 | |
| 1701 | class FT_SELECT: public QUICK_RANGE_SELECT |
| 1702 | { |
| 1703 | public: |
| 1704 | FT_SELECT(THD *thd, TABLE *table, uint key, bool *create_err) : |
| 1705 | QUICK_RANGE_SELECT (thd, table, key, 1, NULL, create_err) |
| 1706 | { (void) init(); } |
| 1707 | ~FT_SELECT() { file->ft_end(); } |
| 1708 | virtual QUICK_RANGE_SELECT *clone(bool *create_error) |
| 1709 | { DBUG_ASSERT(0); return new FT_SELECT(thd, head, index, create_error); } |
| 1710 | int init() { return file->ft_init(); } |
| 1711 | int reset() { return 0; } |
| 1712 | int get_next() { return file->ha_ft_read(record); } |
| 1713 | int get_type() { return QS_TYPE_FULLTEXT; } |
| 1714 | }; |
| 1715 | |
| 1716 | FT_SELECT *get_ft_select(THD *thd, TABLE *table, uint key); |
| 1717 | QUICK_RANGE_SELECT *get_quick_select_for_ref(THD *thd, TABLE *table, |
| 1718 | struct st_table_ref *ref, |
| 1719 | ha_rows records); |
| 1720 | SQL_SELECT *make_select(TABLE *head, table_map const_tables, |
| 1721 | table_map read_tables, COND *conds, |
| 1722 | SORT_INFO* filesort, |
| 1723 | bool allow_null_cond, int *error); |
| 1724 | |
| 1725 | bool calculate_cond_selectivity_for_table(THD *thd, TABLE *table, Item **cond); |
| 1726 | |
| 1727 | #ifdef WITH_PARTITION_STORAGE_ENGINE |
| 1728 | bool prune_partitions(THD *thd, TABLE *table, Item *pprune_cond); |
| 1729 | #endif |
| 1730 | void store_key_image_to_rec(Field *field, uchar *ptr, uint len); |
| 1731 | |
| 1732 | extern String null_string; |
| 1733 | |
| 1734 | /* check this number of rows (default value) */ |
| 1735 | #define SELECTIVITY_SAMPLING_LIMIT 100 |
| 1736 | /* but no more then this part of table (10%) */ |
| 1737 | #define SELECTIVITY_SAMPLING_SHARE 0.10 |
| 1738 | /* do not check if we are going check less then this number of records */ |
| 1739 | #define SELECTIVITY_SAMPLING_THRESHOLD 10 |
| 1740 | |
| 1741 | #endif |
| 1742 | |