clausesel.c

来自「PostgreSQL 8.1.4的源码 适用于Linux下的开源数据库系统」· C语言 代码 · 共 677 行 · 第 1/2 页

	/* No matching var found, so make a new clause-pair data structure */
	rqelem = (RangeQueryClause *) palloc(sizeof(RangeQueryClause));
	rqelem->var = var;
	if (is_lobound)
	{
		rqelem->have_lobound = true;
		rqelem->have_hibound = false;
		rqelem->lobound = s2;
	}
	else
	{
		rqelem->have_lobound = false;
		rqelem->have_hibound = true;
		rqelem->hibound = s2;
	}
	rqelem->next = *rqlist;
	*rqlist = rqelem;
}

/*
 * bms_is_subset_singleton
 *
 * Same result as bms_is_subset(s, bms_make_singleton(x)),
 * but a little faster and doesn't leak memory.
 *
 * Is this of use anywhere else?  If so move to bitmapset.c ...
 */
static bool
bms_is_subset_singleton(const Bitmapset *s, int x)
{
	switch (bms_membership(s))
	{
		case BMS_EMPTY_SET:
			return true;
		case BMS_SINGLETON:
			return bms_is_member(x, s);
		case BMS_MULTIPLE:
			return false;
	}
	/* can't get here... */
	return false;
}


/*
 * clause_selectivity -
 *	  Compute the selectivity of a general boolean expression clause.
 *
 * The clause can be either a RestrictInfo or a plain expression.  If it's
 * a RestrictInfo, we try to cache the selectivity for possible re-use,
 * so passing RestrictInfos is preferred.
 *
 * varRelid is either 0 or a rangetable index.
 *
 * When varRelid is not 0, only variables belonging to that relation are
 * considered in computing selectivity; other vars are treated as constants
 * of unknown values.  This is appropriate for estimating the selectivity of
 * a join clause that is being used as a restriction clause in a scan of a
 * nestloop join's inner relation --- varRelid should then be the ID of the
 * inner relation.
 *
 * When varRelid is 0, all variables are treated as variables.	This
 * is appropriate for ordinary join clauses and restriction clauses.
 *
 * jointype is the join type, if the clause is a join clause.  Pass JOIN_INNER
 * if the clause isn't a join clause or the context is uncertain.
 */
Selectivity
clause_selectivity(PlannerInfo *root,
				   Node *clause,
				   int varRelid,
				   JoinType jointype)
{
	Selectivity s1 = 1.0;		/* default for any unhandled clause type */
	RestrictInfo *rinfo = NULL;
	bool		cacheable = false;

	if (clause == NULL)			/* can this still happen? */
		return s1;

	if (IsA(clause, RestrictInfo))
	{
		rinfo = (RestrictInfo *) clause;

		/*
		 * If possible, cache the result of the selectivity calculation for
		 * the clause.	We can cache if varRelid is zero or the clause
		 * contains only vars of that relid --- otherwise varRelid will affect
		 * the result, so mustn't cache.  We also have to be careful about the
		 * jointype.  It's OK to cache when jointype is JOIN_INNER or one of
		 * the outer join types (any given outer-join clause should always be
		 * examined with the same jointype, so result won't change). It's not
		 * OK to cache when jointype is one of the special types associated
		 * with IN processing, because the same clause may be examined with
		 * different jointypes and the result should vary.
		 */
		if (varRelid == 0 ||
			bms_is_subset_singleton(rinfo->clause_relids, varRelid))
		{
			switch (jointype)
			{
				case JOIN_INNER:
				case JOIN_LEFT:
				case JOIN_FULL:
				case JOIN_RIGHT:
					/* Cacheable --- do we already have the result? */
					if (rinfo->this_selec >= 0)
						return rinfo->this_selec;
					cacheable = true;
					break;

				case JOIN_UNION:
					/* unimplemented anyway... */
				case JOIN_IN:
				case JOIN_REVERSE_IN:
				case JOIN_UNIQUE_OUTER:
				case JOIN_UNIQUE_INNER:
					/* unsafe to cache */
					break;
			}
		}

		/*
		 * Proceed with examination of contained clause.  If the clause is an
		 * OR-clause, we want to look at the variant with sub-RestrictInfos,
		 * so that per-subclause selectivities can be cached.
		 */
		if (rinfo->orclause)
			clause = (Node *) rinfo->orclause;
		else
			clause = (Node *) rinfo->clause;
	}

	if (IsA(clause, Var))
	{
		Var		   *var = (Var *) clause;

		/*
		 * We probably shouldn't ever see an uplevel Var here, but if we do,
		 * return the default selectivity...
		 */
		if (var->varlevelsup == 0 &&
			(varRelid == 0 || varRelid == (int) var->varno))
		{
			RangeTblEntry *rte = rt_fetch(var->varno, root->parse->rtable);

			if (rte->rtekind == RTE_SUBQUERY)
			{
				/*
				 * XXX not smart about subquery references... any way to do
				 * better?
				 */
				s1 = 0.5;
			}
			else
			{
				/*
				 * A Var at the top of a clause must be a bool Var. This is
				 * equivalent to the clause reln.attribute = 't', so we
				 * compute the selectivity as if that is what we have.
				 */
				s1 = restriction_selectivity(root,
											 BooleanEqualOperator,
											 list_make2(var,
														makeBoolConst(true,
																	  false)),
											 varRelid);
			}
		}
	}
	else if (IsA(clause, Const))
	{
		/* bool constant is pretty easy... */
		s1 = ((bool) ((Const *) clause)->constvalue) ? 1.0 : 0.0;
	}
	else if (IsA(clause, Param))
	{
		/* see if we can replace the Param */
		Node	   *subst = estimate_expression_value(clause);

		if (IsA(subst, Const))
		{
			/* bool constant is pretty easy... */
			s1 = ((bool) ((Const *) subst)->constvalue) ? 1.0 : 0.0;
		}
		else
		{
			/* XXX any way to do better? */
			s1 = (Selectivity) 0.5;
		}
	}
	else if (not_clause(clause))
	{
		/* inverse of the selectivity of the underlying clause */
		s1 = 1.0 - clause_selectivity(root,
								  (Node *) get_notclausearg((Expr *) clause),
									  varRelid,
									  jointype);
	}
	else if (and_clause(clause))
	{
		/* share code with clauselist_selectivity() */
		s1 = clauselist_selectivity(root,
									((BoolExpr *) clause)->args,
									varRelid,
									jointype);
	}
	else if (or_clause(clause))
	{
		/*
		 * Selectivities for an OR clause are computed as s1+s2 - s1*s2 to
		 * account for the probable overlap of selected tuple sets.
		 *
		 * XXX is this too conservative?
		 */
		ListCell   *arg;

		s1 = 0.0;
		foreach(arg, ((BoolExpr *) clause)->args)
		{
			Selectivity s2 = clause_selectivity(root,
												(Node *) lfirst(arg),
												varRelid,
												jointype);

			s1 = s1 + s2 - s1 * s2;
		}
	}
	else if (is_opclause(clause))
	{
		Oid			opno = ((OpExpr *) clause)->opno;
		bool		is_join_clause;

		if (varRelid != 0)
		{
			/*
			 * If we are considering a nestloop join then all clauses are
			 * restriction clauses, since we are only interested in the one
			 * relation.
			 */
			is_join_clause = false;
		}
		else
		{
			/*
			 * Otherwise, it's a join if there's more than one relation used.
			 * We can optimize this calculation if an rinfo was passed.
			 */
			if (rinfo)
				is_join_clause = (bms_membership(rinfo->clause_relids) ==
								  BMS_MULTIPLE);
			else
				is_join_clause = (NumRelids(clause) > 1);
		}

		if (is_join_clause)
		{
			/* Estimate selectivity for a join clause. */
			s1 = join_selectivity(root, opno,
								  ((OpExpr *) clause)->args,
								  jointype);
		}
		else
		{
			/* Estimate selectivity for a restriction clause. */
			s1 = restriction_selectivity(root, opno,
										 ((OpExpr *) clause)->args,
										 varRelid);
		}
	}
	else if (is_funcclause(clause))
	{
		/*
		 * This is not an operator, so we guess at the selectivity. THIS IS A
		 * HACK TO GET V4 OUT THE DOOR.  FUNCS SHOULD BE ABLE TO HAVE
		 * SELECTIVITIES THEMSELVES.	   -- JMH 7/9/92
		 */
		s1 = (Selectivity) 0.3333333;
	}
	else if (is_subplan(clause))
	{
		/*
		 * Just for the moment! FIX ME! - vadim 02/04/98
		 */
		s1 = (Selectivity) 0.5;
	}
	else if (IsA(clause, DistinctExpr) ||
			 IsA(clause, ScalarArrayOpExpr))
	{
		/* can we do better? */
		s1 = (Selectivity) 0.5;
	}
	else if (IsA(clause, NullTest))
	{
		/* Use node specific selectivity calculation function */
		s1 = nulltestsel(root,
						 ((NullTest *) clause)->nulltesttype,
						 (Node *) ((NullTest *) clause)->arg,
						 varRelid);
	}
	else if (IsA(clause, BooleanTest))
	{
		/* Use node specific selectivity calculation function */
		s1 = booltestsel(root,
						 ((BooleanTest *) clause)->booltesttype,
						 (Node *) ((BooleanTest *) clause)->arg,
						 varRelid,
						 jointype);
	}
	else if (IsA(clause, RelabelType))
	{
		/* Not sure this case is needed, but it can't hurt */
		s1 = clause_selectivity(root,
								(Node *) ((RelabelType *) clause)->arg,
								varRelid,
								jointype);
	}
	else if (IsA(clause, CoerceToDomain))
	{
		/* Not sure this case is needed, but it can't hurt */
		s1 = clause_selectivity(root,
								(Node *) ((CoerceToDomain *) clause)->arg,
								varRelid,
								jointype);
	}

	/* Cache the result if possible */
	if (cacheable)
		rinfo->this_selec = s1;

#ifdef SELECTIVITY_DEBUG
	elog(DEBUG4, "clause_selectivity: s1 %f", s1);
#endif   /* SELECTIVITY_DEBUG */

	return s1;
}