plancat.c 33.4 KB
Newer Older
1 2
/*-------------------------------------------------------------------------
 *
3
 * plancat.c
4
 *	   routines for accessing the system catalogs
5 6
 *
 *
B
Bruce Momjian 已提交
7
 * Portions Copyright (c) 1996-2015, PostgreSQL Global Development Group
B
Add:  
Bruce Momjian 已提交
8
 * Portions Copyright (c) 1994, Regents of the University of California
9 10 11
 *
 *
 * IDENTIFICATION
12
 *	  src/backend/optimizer/util/plancat.c
13 14 15
 *
 *-------------------------------------------------------------------------
 */
16
#include "postgres.h"
17

18 19
#include <math.h>

B
Bruce Momjian 已提交
20 21
#include "access/genam.h"
#include "access/heapam.h"
22
#include "access/htup_details.h"
23
#include "access/nbtree.h"
24
#include "access/sysattr.h"
25
#include "access/transam.h"
26
#include "access/xlog.h"
27
#include "catalog/catalog.h"
28
#include "catalog/heap.h"
29
#include "foreign/fdwapi.h"
30
#include "miscadmin.h"
31
#include "nodes/makefuncs.h"
32
#include "optimizer/clauses.h"
33
#include "optimizer/cost.h"
34
#include "optimizer/plancat.h"
35
#include "optimizer/predtest.h"
36
#include "optimizer/prep.h"
37
#include "parser/parse_relation.h"
38
#include "parser/parsetree.h"
39
#include "rewrite/rewriteManip.h"
40
#include "storage/bufmgr.h"
41
#include "utils/lsyscache.h"
42
#include "utils/rel.h"
43
#include "utils/snapmgr.h"
44 45


46
/* GUC parameter */
47
int			constraint_exclusion = CONSTRAINT_EXCLUSION_PARTITION;
48

49 50 51
/* Hook for plugins to get control in get_relation_info() */
get_relation_info_hook_type get_relation_info_hook = NULL;

52

53
static int32 get_rel_data_width(Relation rel, int32 *attr_widths);
54 55
static List *get_relation_constraints(PlannerInfo *root,
						 Oid relationObjectId, RelOptInfo *rel,
56
						 bool include_notnull);
57 58
static List *build_index_tlist(PlannerInfo *root, IndexOptInfo *index,
				  Relation heapRelation);
59 60


61
/*
62
 * get_relation_info -
63
 *	  Retrieves catalog information for a given relation.
64 65 66 67
 *
 * Given the Oid of the relation, return the following info into fields
 * of the RelOptInfo struct:
 *
68 69
 *	min_attr	lowest valid AttrNumber
 *	max_attr	highest valid AttrNumber
70
 *	indexlist	list of IndexOptInfos for relation's indexes
71
 *	fdwroutine	if it's a foreign table, the FDW function pointers
72 73
 *	pages		number of pages
 *	tuples		number of tuples
74 75 76 77
 *
 * Also, initialize the attr_needed[] and attr_widths[] arrays.  In most
 * cases these are left as zeroes, but sometimes we need to compute attr
 * widths here, and we may as well cache the results for costsize.c.
78 79 80 81 82
 *
 * If inhparent is true, all we need to do is set up the attr arrays:
 * the RelOptInfo actually represents the appendrel formed by an inheritance
 * tree, and so the parent rel's physical size and index information isn't
 * important for it.
83 84
 */
void
85 86
get_relation_info(PlannerInfo *root, Oid relationObjectId, bool inhparent,
				  RelOptInfo *rel)
87
{
88
	Index		varno = rel->relid;
89
	Relation	relation;
90 91
	bool		hasindex;
	List	   *indexinfos = NIL;
92

93
	/*
B
Bruce Momjian 已提交
94 95 96
	 * We need not lock the relation since it was already locked, either by
	 * the rewriter or when expand_inherited_rtentry() added it to the query's
	 * rangetable.
97
	 */
98
	relation = heap_open(relationObjectId, NoLock);
99

100 101 102 103 104 105
	/* Temporary and unlogged relations are inaccessible during recovery. */
	if (!RelationNeedsWAL(relation) && RecoveryInProgress())
		ereport(ERROR,
				(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
				 errmsg("cannot access temporary or unlogged relations during recovery")));

106 107
	rel->min_attr = FirstLowInvalidHeapAttributeNumber + 1;
	rel->max_attr = RelationGetNumberOfAttributes(relation);
108
	rel->reltablespace = RelationGetForm(relation)->reltablespace;
109

110 111 112 113 114 115 116
	Assert(rel->max_attr >= rel->min_attr);
	rel->attr_needed = (Relids *)
		palloc0((rel->max_attr - rel->min_attr + 1) * sizeof(Relids));
	rel->attr_widths = (int32 *)
		palloc0((rel->max_attr - rel->min_attr + 1) * sizeof(int32));

	/*
117
	 * Estimate relation size --- unless it's an inheritance parent, in which
B
Bruce Momjian 已提交
118 119
	 * case the size will be computed later in set_append_rel_pathlist, and we
	 * must leave it zero for now to avoid bollixing the total_table_pages
120
	 * calculation.
121
	 */
122 123
	if (!inhparent)
		estimate_rel_size(relation, rel->attr_widths - rel->min_attr,
124
						  &rel->pages, &rel->tuples, &rel->allvisfrac);
125

126
	/*
B
Bruce Momjian 已提交
127
	 * Make list of indexes.  Ignore indexes on system catalogs if told to.
128
	 * Don't bother with indexes for an inheritance parent, either.
129
	 */
130
	if (inhparent ||
131
		(IgnoreSystemIndexes && IsSystemRelation(relation)))
132 133 134
		hasindex = false;
	else
		hasindex = relation->rd_rel->relhasindex;
135

136
	if (hasindex)
137
	{
138 139
		List	   *indexoidlist;
		ListCell   *l;
140
		LOCKMODE	lmode;
141

142
		indexoidlist = RelationGetIndexList(relation);
143

144 145 146 147 148 149 150 151 152 153 154 155 156
		/*
		 * For each index, we get the same type of lock that the executor will
		 * need, and do not release it.  This saves a couple of trips to the
		 * shared lock manager while not creating any real loss of
		 * concurrency, because no schema changes could be happening on the
		 * index while we hold lock on the parent rel, and neither lock type
		 * blocks any other kind of index operation.
		 */
		if (rel->relid == root->parse->resultRelation)
			lmode = RowExclusiveLock;
		else
			lmode = AccessShareLock;

157
		foreach(l, indexoidlist)
158
		{
159
			Oid			indexoid = lfirst_oid(l);
160 161 162
			Relation	indexRelation;
			Form_pg_index index;
			IndexOptInfo *info;
163
			int			ncolumns;
164 165
			int			i;

166 167 168
			/*
			 * Extract info from the relation descriptor for the index.
			 */
169
			indexRelation = index_open(indexoid, lmode);
170
			index = indexRelation->rd_index;
171

172 173
			/*
			 * Ignore invalid indexes, since they can't safely be used for
B
Bruce Momjian 已提交
174 175
			 * queries.  Note that this is OK because the data structure we
			 * are constructing is only used by the planner --- the executor
176 177
			 * still needs to insert into "invalid" indexes, if they're marked
			 * IndexIsReady.
178
			 */
179
			if (!IndexIsValid(index))
180 181 182 183 184
			{
				index_close(indexRelation, NoLock);
				continue;
			}

185
			/*
B
Bruce Momjian 已提交
186 187 188
			 * If the index is valid, but cannot yet be used, ignore it; but
			 * mark the plan we are generating as transient. See
			 * src/backend/access/heap/README.HOT for discussion.
189 190 191 192 193 194 195 196 197 198
			 */
			if (index->indcheckxmin &&
				!TransactionIdPrecedes(HeapTupleHeaderGetXmin(indexRelation->rd_indextuple->t_data),
									   TransactionXmin))
			{
				root->glob->transientPlan = true;
				index_close(indexRelation, NoLock);
				continue;
			}

199 200 201
			info = makeNode(IndexOptInfo);

			info->indexoid = index->indexrelid;
202 203
			info->reltablespace =
				RelationGetForm(indexRelation)->reltablespace;
204
			info->rel = rel;
205 206
			info->ncolumns = ncolumns = index->indnatts;
			info->indexkeys = (int *) palloc(sizeof(int) * ncolumns);
207
			info->indexcollations = (Oid *) palloc(sizeof(Oid) * ncolumns);
208 209
			info->opfamily = (Oid *) palloc(sizeof(Oid) * ncolumns);
			info->opcintype = (Oid *) palloc(sizeof(Oid) * ncolumns);
210

211
			for (i = 0; i < ncolumns; i++)
212
			{
213
				info->indexkeys[i] = index->indkey.values[i];
P
Peter Eisentraut 已提交
214
				info->indexcollations[i] = indexRelation->rd_indcollation[i];
215 216
				info->opfamily[i] = indexRelation->rd_opfamily[i];
				info->opcintype[i] = indexRelation->rd_opcintype[i];
217 218 219
			}

			info->relam = indexRelation->rd_rel->relam;
220
			info->amcostestimate = indexRelation->rd_am->amcostestimate;
221
			info->canreturn = index_can_return(indexRelation);
222
			info->amcanorderbyop = indexRelation->rd_am->amcanorderbyop;
223
			info->amoptionalkey = indexRelation->rd_am->amoptionalkey;
224
			info->amsearcharray = indexRelation->rd_am->amsearcharray;
225
			info->amsearchnulls = indexRelation->rd_am->amsearchnulls;
226 227
			info->amhasgettuple = OidIsValid(indexRelation->rd_am->amgettuple);
			info->amhasgetbitmap = OidIsValid(indexRelation->rd_am->amgetbitmap);
228 229

			/*
230
			 * Fetch the ordering information for the index, if any.
231
			 */
232
			if (info->relam == BTREE_AM_OID)
233
			{
234 235 236 237 238 239 240 241 242
				/*
				 * If it's a btree index, we can use its opfamily OIDs
				 * directly as the sort ordering opfamily OIDs.
				 */
				Assert(indexRelation->rd_am->amcanorder);

				info->sortopfamily = info->opfamily;
				info->reverse_sort = (bool *) palloc(sizeof(bool) * ncolumns);
				info->nulls_first = (bool *) palloc(sizeof(bool) * ncolumns);
243

244
				for (i = 0; i < ncolumns; i++)
245
				{
B
Bruce Momjian 已提交
246
					int16		opt = indexRelation->rd_indoption[i];
247

248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
					info->reverse_sort[i] = (opt & INDOPTION_DESC) != 0;
					info->nulls_first[i] = (opt & INDOPTION_NULLS_FIRST) != 0;
				}
			}
			else if (indexRelation->rd_am->amcanorder)
			{
				/*
				 * Otherwise, identify the corresponding btree opfamilies by
				 * trying to map this index's "<" operators into btree.  Since
				 * "<" uniquely defines the behavior of a sort order, this is
				 * a sufficient test.
				 *
				 * XXX This method is rather slow and also requires the
				 * undesirable assumption that the other index AM numbers its
				 * strategies the same as btree.  It'd be better to have a way
				 * to explicitly declare the corresponding btree opfamily for
				 * each opfamily of the other index type.  But given the lack
				 * of current or foreseeable amcanorder index types, it's not
				 * worth expending more effort on now.
				 */
				info->sortopfamily = (Oid *) palloc(sizeof(Oid) * ncolumns);
				info->reverse_sort = (bool *) palloc(sizeof(bool) * ncolumns);
				info->nulls_first = (bool *) palloc(sizeof(bool) * ncolumns);

				for (i = 0; i < ncolumns; i++)
				{
					int16		opt = indexRelation->rd_indoption[i];
					Oid			ltopr;
					Oid			btopfamily;
					Oid			btopcintype;
					int16		btstrategy;
B
Bruce Momjian 已提交
279

280 281 282 283 284 285 286 287 288 289 290 291 292 293
					info->reverse_sort[i] = (opt & INDOPTION_DESC) != 0;
					info->nulls_first[i] = (opt & INDOPTION_NULLS_FIRST) != 0;

					ltopr = get_opfamily_member(info->opfamily[i],
												info->opcintype[i],
												info->opcintype[i],
												BTLessStrategyNumber);
					if (OidIsValid(ltopr) &&
						get_ordering_op_properties(ltopr,
												   &btopfamily,
												   &btopcintype,
												   &btstrategy) &&
						btopcintype == info->opcintype[i] &&
						btstrategy == BTLessStrategyNumber)
294
					{
295 296
						/* Successful mapping */
						info->sortopfamily[i] = btopfamily;
297
					}
298
					else
299
					{
300 301 302 303 304
						/* Fail ... quietly treat index as unordered */
						info->sortopfamily = NULL;
						info->reverse_sort = NULL;
						info->nulls_first = NULL;
						break;
305
					}
306 307
				}
			}
308 309 310 311 312 313
			else
			{
				info->sortopfamily = NULL;
				info->reverse_sort = NULL;
				info->nulls_first = NULL;
			}
314

315 316 317
			/*
			 * Fetch the index expressions and predicate, if any.  We must
			 * modify the copies we obtain from the relcache to have the
B
Bruce Momjian 已提交
318 319
			 * correct varno for the parent relation, so that they match up
			 * correctly against qual clauses.
320 321 322 323 324 325 326
			 */
			info->indexprs = RelationGetIndexExpressions(indexRelation);
			info->indpred = RelationGetIndexPredicate(indexRelation);
			if (info->indexprs && varno != 1)
				ChangeVarNodes((Node *) info->indexprs, 1, varno, 0);
			if (info->indpred && varno != 1)
				ChangeVarNodes((Node *) info->indpred, 1, varno, 0);
327 328 329 330

			/* Build targetlist using the completed indexprs data */
			info->indextlist = build_index_tlist(root, info, relation);

B
Bruce Momjian 已提交
331
			info->predOK = false;		/* set later in indxpath.c */
332
			info->unique = index->indisunique;
333
			info->immediate = index->indimmediate;
334
			info->hypothetical = false;
335

336
			/*
B
Bruce Momjian 已提交
337 338 339 340 341
			 * Estimate the index size.  If it's not a partial index, we lock
			 * the number-of-tuples estimate to equal the parent table; if it
			 * is partial then we have to use the same methods as we would for
			 * a table, except we can be sure that the index is not larger
			 * than the table.
342 343 344 345 346 347 348 349
			 */
			if (info->indpred == NIL)
			{
				info->pages = RelationGetNumberOfBlocks(indexRelation);
				info->tuples = rel->tuples;
			}
			else
			{
350
				double		allvisfrac; /* dummy */
351

352
				estimate_rel_size(indexRelation, NULL,
353
								  &info->pages, &info->tuples, &allvisfrac);
354 355 356 357
				if (info->tuples > rel->tuples)
					info->tuples = rel->tuples;
			}

358 359 360 361 362 363 364 365 366 367 368
			if (info->relam == BTREE_AM_OID)
			{
				/* For btrees, get tree height while we have the index open */
				info->tree_height = _bt_getrootheight(indexRelation);
			}
			else
			{
				/* For other index types, just set it to "unknown" for now */
				info->tree_height = -1;
			}

369
			index_close(indexRelation, NoLock);
370 371 372 373

			indexinfos = lcons(info, indexinfos);
		}

374
		list_free(indexoidlist);
375
	}
376

377 378
	rel->indexlist = indexinfos;

379 380 381 382 383 384
	/* Grab the fdwroutine info using the relcache, while we have it */
	if (relation->rd_rel->relkind == RELKIND_FOREIGN_TABLE)
		rel->fdwroutine = GetFdwRoutineForRelation(relation, true);
	else
		rel->fdwroutine = NULL;

385
	heap_close(relation, NoLock);
386 387 388 389 390 391 392 393

	/*
	 * Allow a plugin to editorialize on the info we obtained from the
	 * catalogs.  Actions might include altering the assumed relation size,
	 * removing an index, or adding a hypothetical index to the indexlist.
	 */
	if (get_relation_info_hook)
		(*get_relation_info_hook) (root, relationObjectId, inhparent, rel);
394 395
}

396 397 398
/*
 * estimate_rel_size - estimate # pages and # tuples in a table or index
 *
399 400 401
 * We also estimate the fraction of the pages that are marked all-visible in
 * the visibility map, for use in estimation of index-only scans.
 *
402
 * If attr_widths isn't NULL, it points to the zero-index entry of the
403
 * relation's attr_widths[] cache; we fill this in if we have need to compute
404 405
 * the attribute widths for estimation purposes.
 */
406
void
407
estimate_rel_size(Relation rel, int32 *attr_widths,
408
				  BlockNumber *pages, double *tuples, double *allvisfrac)
409
{
B
Bruce Momjian 已提交
410 411
	BlockNumber curpages;
	BlockNumber relpages;
412
	double		reltuples;
413
	BlockNumber relallvisible;
414 415 416 417 418 419
	double		density;

	switch (rel->rd_rel->relkind)
	{
		case RELKIND_RELATION:
		case RELKIND_INDEX:
420
		case RELKIND_MATVIEW:
421 422
		case RELKIND_TOASTVALUE:
			/* it has storage, ok to call the smgr */
423 424 425 426
			curpages = RelationGetNumberOfBlocks(rel);

			/*
			 * HACK: if the relation has never yet been vacuumed, use a
427 428 429
			 * minimum size estimate of 10 pages.  The idea here is to avoid
			 * assuming a newly-created table is really small, even if it
			 * currently is, because that may not be true once some data gets
B
Bruce Momjian 已提交
430
			 * loaded into it.  Once a vacuum or analyze cycle has been done
431 432 433 434
			 * on it, it's more reasonable to believe the size is somewhat
			 * stable.
			 *
			 * (Note that this is only an issue if the plan gets cached and
B
Bruce Momjian 已提交
435
			 * used again after the table has been filled.  What we're trying
436 437 438 439 440 441
			 * to avoid is using a nestloop-type plan on a table that has
			 * grown substantially since the plan was made.  Normally,
			 * autovacuum/autoanalyze will occur once enough inserts have
			 * happened and cause cached-plan invalidation; but that doesn't
			 * happen instantaneously, and it won't happen at all for cases
			 * such as temporary tables.)
442
			 *
443
			 * We approximate "never vacuumed" by "has relpages = 0", which
B
Bruce Momjian 已提交
444
			 * means this will also fire on genuinely empty relations.  Not
445 446 447
			 * great, but fortunately that's a seldom-seen case in the real
			 * world, and it shouldn't degrade the quality of the plan too
			 * much anyway to err in this direction.
448 449 450 451 452 453
			 *
			 * There are two exceptions wherein we don't apply this heuristic.
			 * One is if the table has inheritance children.  Totally empty
			 * parent tables are quite common, so we should be willing to
			 * believe that they are empty.  Also, we don't apply the 10-page
			 * minimum to indexes.
454
			 */
455 456 457 458
			if (curpages < 10 &&
				rel->rd_rel->relpages == 0 &&
				!rel->rd_rel->relhassubclass &&
				rel->rd_rel->relkind != RELKIND_INDEX)
459 460 461 462
				curpages = 10;

			/* report estimated # pages */
			*pages = curpages;
463 464 465 466
			/* quick exit if rel is clearly empty */
			if (curpages == 0)
			{
				*tuples = 0;
467
				*allvisfrac = 0;
468 469 470 471 472
				break;
			}
			/* coerce values in pg_class to more desirable types */
			relpages = (BlockNumber) rel->rd_rel->relpages;
			reltuples = (double) rel->rd_rel->reltuples;
473
			relallvisible = (BlockNumber) rel->rd_rel->relallvisible;
B
Bruce Momjian 已提交
474

475
			/*
476 477 478 479
			 * If it's an index, discount the metapage while estimating the
			 * number of tuples.  This is a kluge because it assumes more than
			 * it ought to about index structure.  Currently it's OK for
			 * btree, hash, and GIN indexes but suspect for GiST indexes.
480 481 482 483 484 485 486
			 */
			if (rel->rd_rel->relkind == RELKIND_INDEX &&
				relpages > 0)
			{
				curpages--;
				relpages--;
			}
487

488 489 490 491 492 493 494 495 496
			/* estimate number of tuples from previous tuple density */
			if (relpages > 0)
				density = reltuples / (double) relpages;
			else
			{
				/*
				 * When we have no data because the relation was truncated,
				 * estimate tuple width from attribute datatypes.  We assume
				 * here that the pages are completely full, which is OK for
B
Bruce Momjian 已提交
497 498
				 * tables (since they've presumably not been VACUUMed yet) but
				 * is probably an overestimate for indexes.  Fortunately
499 500
				 * get_relation_info() can clamp the overestimate to the
				 * parent table's size.
501 502
				 *
				 * Note: this code intentionally disregards alignment
B
Bruce Momjian 已提交
503 504 505 506
				 * considerations, because (a) that would be gilding the lily
				 * considering how crude the estimate is, and (b) it creates
				 * platform dependencies in the default plans which are kind
				 * of a headache for regression testing.
507
				 */
508
				int32		tuple_width;
509

510
				tuple_width = get_rel_data_width(rel, attr_widths);
511
				tuple_width += sizeof(HeapTupleHeaderData);
512
				tuple_width += sizeof(ItemIdData);
513
				/* note: integer division is intentional here */
514
				density = (BLCKSZ - SizeOfPageHeaderData) / tuple_width;
515 516
			}
			*tuples = rint(density * (double) curpages);
517 518 519 520 521 522 523 524 525 526 527 528 529

			/*
			 * We use relallvisible as-is, rather than scaling it up like we
			 * do for the pages and tuples counts, on the theory that any
			 * pages added since the last VACUUM are most likely not marked
			 * all-visible.  But costsize.c wants it converted to a fraction.
			 */
			if (relallvisible == 0 || curpages <= 0)
				*allvisfrac = 0;
			else if ((double) relallvisible >= curpages)
				*allvisfrac = 1;
			else
				*allvisfrac = (double) relallvisible / curpages;
530 531 532 533 534
			break;
		case RELKIND_SEQUENCE:
			/* Sequences always have a known size */
			*pages = 1;
			*tuples = 1;
535
			*allvisfrac = 0;
536
			break;
537 538 539 540
		case RELKIND_FOREIGN_TABLE:
			/* Just use whatever's in pg_class */
			*pages = rel->rd_rel->relpages;
			*tuples = rel->rd_rel->reltuples;
541
			*allvisfrac = 0;
542
			break;
543 544 545 546
		default:
			/* else it has no disk storage; probably shouldn't get here? */
			*pages = 0;
			*tuples = 0;
547
			*allvisfrac = 0;
548 549 550 551
			break;
	}
}

552

553 554 555 556
/*
 * get_rel_data_width
 *
 * Estimate the average width of (the data part of) the relation's tuples.
557 558 559
 *
 * If attr_widths isn't NULL, it points to the zero-index entry of the
 * relation's attr_widths[] cache; use and update that cache as appropriate.
560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578
 *
 * Currently we ignore dropped columns.  Ideally those should be included
 * in the result, but we haven't got any way to get info about them; and
 * since they might be mostly NULLs, treating them as zero-width is not
 * necessarily the wrong thing anyway.
 */
static int32
get_rel_data_width(Relation rel, int32 *attr_widths)
{
	int32		tuple_width = 0;
	int			i;

	for (i = 1; i <= RelationGetNumberOfAttributes(rel); i++)
	{
		Form_pg_attribute att = rel->rd_att->attrs[i - 1];
		int32		item_width;

		if (att->attisdropped)
			continue;
579 580 581 582 583 584 585 586

		/* use previously cached data, if any */
		if (attr_widths != NULL && attr_widths[i] > 0)
		{
			tuple_width += attr_widths[i];
			continue;
		}

587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604
		/* This should match set_rel_width() in costsize.c */
		item_width = get_attavgwidth(RelationGetRelid(rel), i);
		if (item_width <= 0)
		{
			item_width = get_typavgwidth(att->atttypid, att->atttypmod);
			Assert(item_width > 0);
		}
		if (attr_widths != NULL)
			attr_widths[i] = item_width;
		tuple_width += item_width;
	}

	return tuple_width;
}

/*
 * get_relation_data_width
 *
605 606
 * External API for get_rel_data_width: same behavior except we have to
 * open the relcache entry.
607 608
 */
int32
609
get_relation_data_width(Oid relid, int32 *attr_widths)
610 611 612 613 614 615 616
{
	int32		result;
	Relation	relation;

	/* As above, assume relation is already locked */
	relation = heap_open(relid, NoLock);

617
	result = get_rel_data_width(relation, attr_widths);
618 619 620 621 622 623 624

	heap_close(relation, NoLock);

	return result;
}


625 626 627
/*
 * get_relation_constraints
 *
628
 * Retrieve the validated CHECK constraint expressions of the given relation.
629 630 631 632 633 634
 *
 * Returns a List (possibly empty) of constraint expressions.  Each one
 * has been canonicalized, and its Vars are changed to have the varno
 * indicated by rel->relid.  This allows the expressions to be easily
 * compared to expressions taken from WHERE.
 *
635 636 637
 * If include_notnull is true, "col IS NOT NULL" expressions are generated
 * and added to the result for each column that's marked attnotnull.
 *
638 639 640 641
 * Note: at present this is invoked at most once per relation per planner
 * run, and in many cases it won't be invoked at all, so there seems no
 * point in caching the data in RelOptInfo.
 */
642
static List *
643 644
get_relation_constraints(PlannerInfo *root,
						 Oid relationObjectId, RelOptInfo *rel,
645
						 bool include_notnull)
646 647 648 649 650 651 652 653 654 655 656 657 658 659
{
	List	   *result = NIL;
	Index		varno = rel->relid;
	Relation	relation;
	TupleConstr *constr;

	/*
	 * We assume the relation has already been safely locked.
	 */
	relation = heap_open(relationObjectId, NoLock);

	constr = relation->rd_att->constr;
	if (constr != NULL)
	{
B
Bruce Momjian 已提交
660 661
		int			num_check = constr->num_check;
		int			i;
662 663 664

		for (i = 0; i < num_check; i++)
		{
B
Bruce Momjian 已提交
665
			Node	   *cexpr;
666

667 668 669 670 671 672 673
			/*
			 * If this constraint hasn't been fully validated yet, we must
			 * ignore it here.
			 */
			if (!constr->check[i].ccvalid)
				continue;

674 675 676 677 678 679 680 681 682 683 684 685
			cexpr = stringToNode(constr->check[i].ccbin);

			/*
			 * Run each expression through const-simplification and
			 * canonicalization.  This is not just an optimization, but is
			 * necessary, because we will be comparing it to
			 * similarly-processed qual clauses, and may fail to detect valid
			 * matches without this.  This must match the processing done to
			 * qual clauses in preprocess_expression()!  (We can skip the
			 * stuff involving subqueries, however, since we don't allow any
			 * in check constraints.)
			 */
686
			cexpr = eval_const_expressions(root, cexpr);
687 688 689 690 691 692 693 694

			cexpr = (Node *) canonicalize_qual((Expr *) cexpr);

			/* Fix Vars to have the desired varno */
			if (varno != 1)
				ChangeVarNodes(cexpr, 1, varno, 0);

			/*
B
Bruce Momjian 已提交
695 696
			 * Finally, convert to implicit-AND format (that is, a List) and
			 * append the resulting item(s) to our output list.
697 698 699 700
			 */
			result = list_concat(result,
								 make_ands_implicit((Expr *) cexpr));
		}
701 702 703 704

		/* Add NOT NULL constraints in expression form, if requested */
		if (include_notnull && constr->has_not_null)
		{
705
			int			natts = relation->rd_att->natts;
706 707 708 709 710 711 712

			for (i = 1; i <= natts; i++)
			{
				Form_pg_attribute att = relation->rd_att->attrs[i - 1];

				if (att->attnotnull && !att->attisdropped)
				{
713
					NullTest   *ntest = makeNode(NullTest);
714 715 716 717 718

					ntest->arg = (Expr *) makeVar(varno,
												  i,
												  att->atttypid,
												  att->atttypmod,
P
Peter Eisentraut 已提交
719
												  att->attcollation,
720 721
												  0);
					ntest->nulltesttype = IS_NOT_NULL;
722
					ntest->argisrow = type_is_rowtype(att->atttypid);
723 724 725 726
					result = lappend(result, ntest);
				}
			}
		}
727 728 729 730 731 732 733 734
	}

	heap_close(relation, NoLock);

	return result;
}


735 736 737
/*
 * relation_excluded_by_constraints
 *
738 739
 * Detect whether the relation need not be scanned because it has either
 * self-inconsistent restrictions, or restrictions inconsistent with the
740
 * relation's validated CHECK constraints.
741
 *
742 743 744
 * Note: this examines only rel->relid, rel->reloptkind, and
 * rel->baserestrictinfo; therefore it can be called before filling in
 * other fields of the RelOptInfo.
745 746
 */
bool
747 748
relation_excluded_by_constraints(PlannerInfo *root,
								 RelOptInfo *rel, RangeTblEntry *rte)
749
{
750
	List	   *safe_restrictions;
751
	List	   *constraint_pred;
752 753
	List	   *safe_constraints;
	ListCell   *lc;
754

755 756 757
	/* Skip the test if constraint exclusion is disabled for the rel */
	if (constraint_exclusion == CONSTRAINT_EXCLUSION_OFF ||
		(constraint_exclusion == CONSTRAINT_EXCLUSION_PARTITION &&
758 759 760 761
		 !(rel->reloptkind == RELOPT_OTHER_MEMBER_REL ||
		   (root->hasInheritedTarget &&
			rel->reloptkind == RELOPT_BASEREL &&
			rel->relid == root->parse->resultRelation))))
762 763
		return false;

764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783
	/*
	 * Check for self-contradictory restriction clauses.  We dare not make
	 * deductions with non-immutable functions, but any immutable clauses that
	 * are self-contradictory allow us to conclude the scan is unnecessary.
	 *
	 * Note: strip off RestrictInfo because predicate_refuted_by() isn't
	 * expecting to see any in its predicate argument.
	 */
	safe_restrictions = NIL;
	foreach(lc, rel->baserestrictinfo)
	{
		RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc);

		if (!contain_mutable_functions((Node *) rinfo->clause))
			safe_restrictions = lappend(safe_restrictions, rinfo->clause);
	}

	if (predicate_refuted_by(safe_restrictions, safe_restrictions))
		return true;

784 785 786 787
	/* Only plain relations have constraints */
	if (rte->rtekind != RTE_RELATION || rte->inh)
		return false;

788
	/*
B
Bruce Momjian 已提交
789
	 * OK to fetch the constraint expressions.  Include "col IS NOT NULL"
790 791
	 * expressions for attnotnull columns, in case we can refute those.
	 */
792
	constraint_pred = get_relation_constraints(root, rte->relid, rel, true);
793 794 795 796

	/*
	 * We do not currently enforce that CHECK constraints contain only
	 * immutable functions, so it's necessary to check here. We daren't draw
B
Bruce Momjian 已提交
797 798 799
	 * conclusions from plan-time evaluation of non-immutable functions. Since
	 * they're ANDed, we can just ignore any mutable constraints in the list,
	 * and reason about the rest.
800
	 */
801 802 803
	safe_constraints = NIL;
	foreach(lc, constraint_pred)
	{
B
Bruce Momjian 已提交
804
		Node	   *pred = (Node *) lfirst(lc);
805 806 807 808

		if (!contain_mutable_functions(pred))
			safe_constraints = lappend(safe_constraints, pred);
	}
809 810 811 812 813

	/*
	 * The constraints are effectively ANDed together, so we can just try to
	 * refute the entire collection at once.  This may allow us to make proofs
	 * that would fail if we took them individually.
814
	 *
B
Bruce Momjian 已提交
815 816 817
	 * Note: we use rel->baserestrictinfo, not safe_restrictions as might seem
	 * an obvious optimization.  Some of the clauses might be OR clauses that
	 * have volatile and nonvolatile subclauses, and it's OK to make
818
	 * deductions with the nonvolatile parts.
819
	 */
820
	if (predicate_refuted_by(safe_constraints, rel->baserestrictinfo))
821 822 823 824 825 826
		return true;

	return false;
}


827 828 829 830 831 832 833 834 835 836
/*
 * build_physical_tlist
 *
 * Build a targetlist consisting of exactly the relation's user attributes,
 * in order.  The executor can special-case such tlists to avoid a projection
 * step at runtime, so we use such tlists preferentially for scan nodes.
 *
 * Exception: if there are any dropped columns, we punt and return NIL.
 * Ideally we would like to handle the dropped-column case too.  However this
 * creates problems for ExecTypeFromTL, which may be asked to build a tupdesc
B
Bruce Momjian 已提交
837
 * for a tlist that includes vars of no-longer-existent types.  In theory we
838 839 840 841
 * could dig out the required info from the pg_attribute entries of the
 * relation, but that data is not readily available to ExecTypeFromTL.
 * For now, we don't apply the physical-tlist optimization when there are
 * dropped cols.
842
 *
843
 * We also support building a "physical" tlist for subqueries, functions,
844 845
 * values lists, and CTEs, since the same optimization can occur in
 * SubqueryScan, FunctionScan, ValuesScan, CteScan, and WorkTableScan nodes.
846 847
 */
List *
848
build_physical_tlist(PlannerInfo *root, RelOptInfo *rel)
849
{
850
	List	   *tlist = NIL;
851
	Index		varno = rel->relid;
852
	RangeTblEntry *rte = planner_rt_fetch(varno, root);
853
	Relation	relation;
854 855 856
	Query	   *subquery;
	Var		   *var;
	ListCell   *l;
857 858
	int			attrno,
				numattrs;
859
	List	   *colvars;
860

861 862 863
	switch (rte->rtekind)
	{
		case RTE_RELATION:
864 865
			/* Assume we already have adequate lock */
			relation = heap_open(rte->relid, NoLock);
866

867 868 869 870
			numattrs = RelationGetNumberOfAttributes(relation);
			for (attrno = 1; attrno <= numattrs; attrno++)
			{
				Form_pg_attribute att_tup = relation->rd_att->attrs[attrno - 1];
871

872 873 874 875 876 877
				if (att_tup->attisdropped)
				{
					/* found a dropped col, so punt */
					tlist = NIL;
					break;
				}
878

879 880 881 882
				var = makeVar(varno,
							  attrno,
							  att_tup->atttypid,
							  att_tup->atttypmod,
P
Peter Eisentraut 已提交
883
							  att_tup->attcollation,
884 885 886 887 888 889 890 891
							  0);

				tlist = lappend(tlist,
								makeTargetEntry((Expr *) var,
												attrno,
												NULL,
												false));
			}
892

893
			heap_close(relation, NoLock);
894 895
			break;

896 897 898 899 900 901
		case RTE_SUBQUERY:
			subquery = rte->subquery;
			foreach(l, subquery->targetList)
			{
				TargetEntry *tle = (TargetEntry *) lfirst(l);

902 903 904 905
				/*
				 * A resjunk column of the subquery can be reflected as
				 * resjunk in the physical tlist; we need not punt.
				 */
906
				var = makeVarFromTargetEntry(varno, tle);
907 908 909 910 911 912 913 914

				tlist = lappend(tlist,
								makeTargetEntry((Expr *) var,
												tle->resno,
												NULL,
												tle->resjunk));
			}
			break;
915

916
		case RTE_FUNCTION:
917 918 919
		case RTE_VALUES:
		case RTE_CTE:
			/* Not all of these can have dropped cols, but share code anyway */
920
			expandRTE(rte, varno, 0, -1, true /* include dropped */ ,
921 922 923 924
					  NULL, &colvars);
			foreach(l, colvars)
			{
				var = (Var *) lfirst(l);
B
Bruce Momjian 已提交
925

926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943
				/*
				 * A non-Var in expandRTE's output means a dropped column;
				 * must punt.
				 */
				if (!IsA(var, Var))
				{
					tlist = NIL;
					break;
				}

				tlist = lappend(tlist,
								makeTargetEntry((Expr *) var,
												var->varattno,
												NULL,
												false));
			}
			break;

944 945 946 947 948 949
		default:
			/* caller error */
			elog(ERROR, "unsupported RTE kind %d in build_physical_tlist",
				 (int) rte->rtekind);
			break;
	}
950

951
	return tlist;
952 953
}

954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985
/*
 * build_index_tlist
 *
 * Build a targetlist representing the columns of the specified index.
 * Each column is represented by a Var for the corresponding base-relation
 * column, or an expression in base-relation Vars, as appropriate.
 *
 * There are never any dropped columns in indexes, so unlike
 * build_physical_tlist, we need no failure case.
 */
static List *
build_index_tlist(PlannerInfo *root, IndexOptInfo *index,
				  Relation heapRelation)
{
	List	   *tlist = NIL;
	Index		varno = index->rel->relid;
	ListCell   *indexpr_item;
	int			i;

	indexpr_item = list_head(index->indexprs);
	for (i = 0; i < index->ncolumns; i++)
	{
		int			indexkey = index->indexkeys[i];
		Expr	   *indexvar;

		if (indexkey != 0)
		{
			/* simple column */
			Form_pg_attribute att_tup;

			if (indexkey < 0)
				att_tup = SystemAttributeDefinition(indexkey,
986
										   heapRelation->rd_rel->relhasoids);
987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
			else
				att_tup = heapRelation->rd_att->attrs[indexkey - 1];

			indexvar = (Expr *) makeVar(varno,
										indexkey,
										att_tup->atttypid,
										att_tup->atttypmod,
										att_tup->attcollation,
										0);
		}
		else
		{
			/* expression column */
			if (indexpr_item == NULL)
				elog(ERROR, "wrong number of index expressions");
			indexvar = (Expr *) lfirst(indexpr_item);
			indexpr_item = lnext(indexpr_item);
		}

		tlist = lappend(tlist,
						makeTargetEntry(indexvar,
										i + 1,
										NULL,
										false));
	}
	if (indexpr_item != NULL)
		elog(ERROR, "wrong number of index expressions");

	return tlist;
}

1018
/*
1019
 * restriction_selectivity
1020
 *
1021
 * Returns the selectivity of a specified restriction operator clause.
1022 1023 1024
 * This code executes registered procedures stored in the
 * operator relation, by calling the function manager.
 *
1025
 * See clause_selectivity() for the meaning of the additional parameters.
1026
 */
1027
Selectivity
1028
restriction_selectivity(PlannerInfo *root,
1029
						Oid operatorid,
1030
						List *args,
1031
						Oid inputcollid,
1032
						int varRelid)
1033
{
1034
	RegProcedure oprrest = get_oprrest(operatorid);
1035 1036
	float8		result;

1037
	/*
1038 1039
	 * if the oprrest procedure is missing for whatever reason, use a
	 * selectivity of 0.5
1040 1041 1042 1043
	 */
	if (!oprrest)
		return (Selectivity) 0.5;

1044 1045 1046 1047 1048 1049
	result = DatumGetFloat8(OidFunctionCall4Coll(oprrest,
												 inputcollid,
												 PointerGetDatum(root),
												 ObjectIdGetDatum(operatorid),
												 PointerGetDatum(args),
												 Int32GetDatum(varRelid)));
1050 1051

	if (result < 0.0 || result > 1.0)
1052
		elog(ERROR, "invalid restriction selectivity: %f", result);
1053 1054

	return (Selectivity) result;
1055 1056 1057
}

/*
1058
 * join_selectivity
1059
 *
1060 1061 1062
 * Returns the selectivity of a specified join operator clause.
 * This code executes registered procedures stored in the
 * operator relation, by calling the function manager.
1063
 */
1064
Selectivity
1065
join_selectivity(PlannerInfo *root,
1066
				 Oid operatorid,
1067
				 List *args,
1068
				 Oid inputcollid,
1069 1070
				 JoinType jointype,
				 SpecialJoinInfo *sjinfo)
1071
{
1072
	RegProcedure oprjoin = get_oprjoin(operatorid);
1073 1074
	float8		result;

1075
	/*
1076 1077
	 * if the oprjoin procedure is missing for whatever reason, use a
	 * selectivity of 0.5
1078 1079 1080 1081
	 */
	if (!oprjoin)
		return (Selectivity) 0.5;

1082 1083 1084 1085 1086 1087 1088
	result = DatumGetFloat8(OidFunctionCall5Coll(oprjoin,
												 inputcollid,
												 PointerGetDatum(root),
												 ObjectIdGetDatum(operatorid),
												 PointerGetDatum(args),
												 Int16GetDatum(jointype),
												 PointerGetDatum(sjinfo)));
1089 1090

	if (result < 0.0 || result > 1.0)
1091
		elog(ERROR, "invalid join selectivity: %f", result);
1092 1093

	return (Selectivity) result;
1094 1095
}

1096 1097 1098 1099 1100 1101
/*
 * has_unique_index
 *
 * Detect whether there is a unique index on the specified attribute
 * of the specified relation, thus allowing us to conclude that all
 * the (non-null) values of the attribute are distinct.
1102 1103 1104 1105 1106
 *
 * This function does not check the index's indimmediate property, which
 * means that uniqueness may transiently fail to hold intra-transaction.
 * That's appropriate when we are making statistical estimates, but beware
 * of using this for any correctness proofs.
1107 1108 1109 1110
 */
bool
has_unique_index(RelOptInfo *rel, AttrNumber attno)
{
1111
	ListCell   *ilist;
1112 1113 1114 1115 1116 1117

	foreach(ilist, rel->indexlist)
	{
		IndexOptInfo *index = (IndexOptInfo *) lfirst(ilist);

		/*
B
Bruce Momjian 已提交
1118
		 * Note: ignore partial indexes, since they don't allow us to conclude
1119
		 * that all attr values are distinct, *unless* they are marked predOK
1120 1121 1122 1123
		 * which means we know the index's predicate is satisfied by the
		 * query. We don't take any interest in expressional indexes either.
		 * Also, a multicolumn unique index doesn't allow us to conclude that
		 * just the specified attr is unique.
1124 1125
		 */
		if (index->unique &&
1126
			index->ncolumns == 1 &&
1127
			index->indexkeys[0] == attno &&
1128
			(index->indpred == NIL || index->predOK))
1129 1130 1131 1132
			return true;
	}
	return false;
}