selfuncs.c

来自「postgresql8.3.4源码,开源数据库」· C语言 代码 · 共 2,362 行 · 第 1/5 页

				/*
				 * Use freq_null directly.
				 */
				selec = freq_null;
				break;
			case IS_NOT_NULL:

				/*
				 * Select not unknown (not null) values. Calculate from
				 * freq_null.
				 */
				selec = 1.0 - freq_null;
				break;
			default:
				elog(ERROR, "unrecognized nulltesttype: %d",
					 (int) nulltesttype);
				return (Selectivity) 0; /* keep compiler quiet */
		}
	}
	else
	{
		/*
		 * No ANALYZE stats available, so make a guess
		 */
		switch (nulltesttype)
		{
			case IS_NULL:
				selec = DEFAULT_UNK_SEL;
				break;
			case IS_NOT_NULL:
				selec = DEFAULT_NOT_UNK_SEL;
				break;
			default:
				elog(ERROR, "unrecognized nulltesttype: %d",
					 (int) nulltesttype);
				return (Selectivity) 0; /* keep compiler quiet */
		}
	}

	ReleaseVariableStats(vardata);

	/* result should be in range, but make sure... */
	CLAMP_PROBABILITY(selec);

	return (Selectivity) selec;
}

/*
 * strip_array_coercion - strip binary-compatible relabeling from an array expr
 *
 * For array values, the parser normally generates ArrayCoerceExpr conversions,
 * but it seems possible that RelabelType might show up.  Also, the planner
 * is not currently tense about collapsing stacked ArrayCoerceExpr nodes,
 * so we need to be ready to deal with more than one level.
 */
static Node *
strip_array_coercion(Node *node)
{
	for (;;)
	{
		if (node && IsA(node, ArrayCoerceExpr) &&
			((ArrayCoerceExpr *) node)->elemfuncid == InvalidOid)
		{
			node = (Node *) ((ArrayCoerceExpr *) node)->arg;
		}
		else if (node && IsA(node, RelabelType))
		{
			/* We don't really expect this case, but may as well cope */
			node = (Node *) ((RelabelType *) node)->arg;
		}
		else
			break;
	}
	return node;
}

/*
 *		scalararraysel		- Selectivity of ScalarArrayOpExpr Node.
 */
Selectivity
scalararraysel(PlannerInfo *root,
			   ScalarArrayOpExpr *clause,
			   bool is_join_clause,
			   int varRelid, JoinType jointype)
{
	Oid			operator = clause->opno;
	bool		useOr = clause->useOr;
	Node	   *leftop;
	Node	   *rightop;
	Oid			nominal_element_type;
	RegProcedure oprsel;
	FmgrInfo	oprselproc;
	Datum		selarg4;
	Selectivity s1;

	/*
	 * First, look up the underlying operator's selectivity estimator. Punt if
	 * it hasn't got one.
	 */
	if (is_join_clause)
	{
		oprsel = get_oprjoin(operator);
		selarg4 = Int16GetDatum(jointype);
	}
	else
	{
		oprsel = get_oprrest(operator);
		selarg4 = Int32GetDatum(varRelid);
	}
	if (!oprsel)
		return (Selectivity) 0.5;
	fmgr_info(oprsel, &oprselproc);

	/* deconstruct the expression */
	Assert(list_length(clause->args) == 2);
	leftop = (Node *) linitial(clause->args);
	rightop = (Node *) lsecond(clause->args);

	/* get nominal (after relabeling) element type of rightop */
	nominal_element_type = get_element_type(exprType(rightop));
	if (!OidIsValid(nominal_element_type))
		return (Selectivity) 0.5;		/* probably shouldn't happen */

	/* look through any binary-compatible relabeling of rightop */
	rightop = strip_array_coercion(rightop);

	/*
	 * We consider three cases:
	 *
	 * 1. rightop is an Array constant: deconstruct the array, apply the
	 * operator's selectivity function for each array element, and merge the
	 * results in the same way that clausesel.c does for AND/OR combinations.
	 *
	 * 2. rightop is an ARRAY[] construct: apply the operator's selectivity
	 * function for each element of the ARRAY[] construct, and merge.
	 *
	 * 3. otherwise, make a guess ...
	 */
	if (rightop && IsA(rightop, Const))
	{
		Datum		arraydatum = ((Const *) rightop)->constvalue;
		bool		arrayisnull = ((Const *) rightop)->constisnull;
		ArrayType  *arrayval;
		int16		elmlen;
		bool		elmbyval;
		char		elmalign;
		int			num_elems;
		Datum	   *elem_values;
		bool	   *elem_nulls;
		int			i;

		if (arrayisnull)		/* qual can't succeed if null array */
			return (Selectivity) 0.0;
		arrayval = DatumGetArrayTypeP(arraydatum);
		get_typlenbyvalalign(ARR_ELEMTYPE(arrayval),
							 &elmlen, &elmbyval, &elmalign);
		deconstruct_array(arrayval,
						  ARR_ELEMTYPE(arrayval),
						  elmlen, elmbyval, elmalign,
						  &elem_values, &elem_nulls, &num_elems);
		s1 = useOr ? 0.0 : 1.0;
		for (i = 0; i < num_elems; i++)
		{
			List	   *args;
			Selectivity s2;

			args = list_make2(leftop,
							  makeConst(nominal_element_type,
										-1,
										elmlen,
										elem_values[i],
										elem_nulls[i],
										elmbyval));
			s2 = DatumGetFloat8(FunctionCall4(&oprselproc,
											  PointerGetDatum(root),
											  ObjectIdGetDatum(operator),
											  PointerGetDatum(args),
											  selarg4));
			if (useOr)
				s1 = s1 + s2 - s1 * s2;
			else
				s1 = s1 * s2;
		}
	}
	else if (rightop && IsA(rightop, ArrayExpr) &&
			 !((ArrayExpr *) rightop)->multidims)
	{
		ArrayExpr  *arrayexpr = (ArrayExpr *) rightop;
		int16		elmlen;
		bool		elmbyval;
		ListCell   *l;

		get_typlenbyval(arrayexpr->element_typeid,
						&elmlen, &elmbyval);
		s1 = useOr ? 0.0 : 1.0;
		foreach(l, arrayexpr->elements)
		{
			Node	   *elem = (Node *) lfirst(l);
			List	   *args;
			Selectivity s2;

			/*
			 * Theoretically, if elem isn't of nominal_element_type we should
			 * insert a RelabelType, but it seems unlikely that any operator
			 * estimation function would really care ...
			 */
			args = list_make2(leftop, elem);
			s2 = DatumGetFloat8(FunctionCall4(&oprselproc,
											  PointerGetDatum(root),
											  ObjectIdGetDatum(operator),
											  PointerGetDatum(args),
											  selarg4));
			if (useOr)
				s1 = s1 + s2 - s1 * s2;
			else
				s1 = s1 * s2;
		}
	}
	else
	{
		CaseTestExpr *dummyexpr;
		List	   *args;
		Selectivity s2;
		int			i;

		/*
		 * We need a dummy rightop to pass to the operator selectivity
		 * routine.  It can be pretty much anything that doesn't look like a
		 * constant; CaseTestExpr is a convenient choice.
		 */
		dummyexpr = makeNode(CaseTestExpr);
		dummyexpr->typeId = nominal_element_type;
		dummyexpr->typeMod = -1;
		args = list_make2(leftop, dummyexpr);
		s2 = DatumGetFloat8(FunctionCall4(&oprselproc,
										  PointerGetDatum(root),
										  ObjectIdGetDatum(operator),
										  PointerGetDatum(args),
										  selarg4));
		s1 = useOr ? 0.0 : 1.0;

		/*
		 * Arbitrarily assume 10 elements in the eventual array value (see
		 * also estimate_array_length)
		 */
		for (i = 0; i < 10; i++)
		{
			if (useOr)
				s1 = s1 + s2 - s1 * s2;
			else
				s1 = s1 * s2;
		}
	}

	/* result should be in range, but make sure... */
	CLAMP_PROBABILITY(s1);

	return s1;
}

/*
 * Estimate number of elements in the array yielded by an expression.
 *
 * It's important that this agree with scalararraysel.
 */
int
estimate_array_length(Node *arrayexpr)
{
	/* look through any binary-compatible relabeling of arrayexpr */
	arrayexpr = strip_array_coercion(arrayexpr);

	if (arrayexpr && IsA(arrayexpr, Const))
	{
		Datum		arraydatum = ((Const *) arrayexpr)->constvalue;
		bool		arrayisnull = ((Const *) arrayexpr)->constisnull;
		ArrayType  *arrayval;

		if (arrayisnull)
			return 0;
		arrayval = DatumGetArrayTypeP(arraydatum);
		return ArrayGetNItems(ARR_NDIM(arrayval), ARR_DIMS(arrayval));
	}
	else if (arrayexpr && IsA(arrayexpr, ArrayExpr) &&
			 !((ArrayExpr *) arrayexpr)->multidims)
	{
		return list_length(((ArrayExpr *) arrayexpr)->elements);
	}
	else
	{
		/* default guess --- see also scalararraysel */
		return 10;
	}
}

/*
 *		rowcomparesel		- Selectivity of RowCompareExpr Node.
 *
 * We estimate RowCompare selectivity by considering just the first (high
 * order) columns, which makes it equivalent to an ordinary OpExpr.  While
 * this estimate could be refined by considering additional columns, it
 * seems unlikely that we could do a lot better without multi-column
 * statistics.
 */
Selectivity
rowcomparesel(PlannerInfo *root,
			  RowCompareExpr *clause,
			  int varRelid, JoinType jointype)
{
	Selectivity s1;
	Oid			opno = linitial_oid(clause->opnos);
	List	   *opargs;
	bool		is_join_clause;

	/* Build equivalent arg list for single operator */
	opargs = list_make2(linitial(clause->largs), linitial(clause->rargs));

	/* Decide if it's a join clause, same as for OpExpr */
	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.
		 * Notice we ignore the low-order columns here.
		 */
		is_join_clause = (NumRelids((Node *) opargs) > 1);
	}

	if (is_join_clause)
	{
		/* Estimate selectivity for a join clause. */
		s1 = join_selectivity(root, opno,
							  opargs,
							  jointype);
	}
	else
	{
		/* Estimate selectivity for a restriction clause. */
		s1 = restriction_selectivity(root, opno,
									 opargs,
									 varRelid);
	}

	return s1;
}

/*
 *		eqjoinsel		- Join selectivity of "="
 */
Datum
eqjoinsel(PG_FUNCTION_ARGS)
{
	PlannerInfo *root = (PlannerInfo *) PG_GETARG_POINTER(0);
	Oid			operator = PG_GETARG_OID(1);
	List	   *args = (List *) PG_GETARG_POINTER(2);
	JoinType	jointype = (JoinType) PG_GETARG_INT16(3);
	double		selec;
	VariableStatData vardata1;
	VariableStatData vardata2;
	double		nd1;
	double		nd2;
	Form_pg_statistic stats1 = NULL;
	Form_pg_statistic stats2 = NULL;
	bool		have_mcvs1 = false;
	Datum	   *values1 = NULL;
	int			nvalues1 = 0;
	float4	   *numbers1 = NULL;
	int			nnumbers1 = 0;
	bool		have_mcvs2 = false;
	Datum	   *values2 = NULL;
	int			nvalues2 = 0;
	float4	   *numbers2 = NULL;
	int			nnumbers2 = 0;

	get_join_variables(root, args, &vardata1, &vardata2);

	nd1 = get_variable_numdistinct(&vardata1);
	nd2 = get_variable_numdistinct(&vardata2);

	if (HeapTupleIsValid(vardata1.statsTuple))
	{
		stats1 = (Form_pg_statistic) GETSTRUCT(vardata1.statsTuple);
		have_mcvs1 = get_attstatsslot(vardata1.statsTuple,
									  vardata1.atttype,
									  vardata1.atttypmod,
									  STATISTIC_KIND_MCV,
									  InvalidOid,
									  &values1, &nvalues1,
									  &numbers1, &nnumbers1);
	}

	if (HeapTupleIsValid(vardata2.statsTuple))
	{
		stats2 = (Form_pg_statistic) GETSTRUCT(vardata2.statsTuple);
		have_mcvs2 = get_attstatsslot(vardata2.statsTuple,
									  vardata2.atttype,
									  vardata2.atttypmod,
									  STATISTIC_KIND_MCV,
									  InvalidOid,
									  &values2, &nvalues2,
									  &numbers2, &nnumbers2);
	}

	if (have_mcvs1 && have_mcvs2)
	{
		/*
		 * We have most-common-value lists for both relations.	Run through
		 * the lists to see which MCVs actually join to each other with the
		 * given operator.	This allows us to determine the exact join
		 * selectivity for the portion of the relations represented by the MCV
		 * lists.  We still have to estimate for the remaining population, but
		 * in a skewed distribution this gives us a big leg up in accuracy.
		 * For motivation see the analysis in Y. Ioannidis and S.
		 * Christodoulakis, "On the propagation of errors in the size of join
		 * results", Technical Report 1018, Computer Science Dept., University
		 * of Wisconsin, Madison, March 1991 (available from ftp.cs.wisc.edu).
		 */
		FmgrInfo	eqproc;
		bool	   *hasmatch1;
		bool	   *hasmatch2;
		double		nullfrac1 = stats1->stanullfrac;
		double		nullfrac2 = stats2->stanullfrac;
		double		matchprodfreq,
					matchfreq1,
					matchfreq2,
					unmatchfreq1,
					unmatchfreq2,
					otherfreq1,
					otherfreq2,
					totalsel1,
					totalsel2;
		int			i,
					nmatches;

		fmgr_info(get_opcode(operator), &eqproc);
		hasmatch1 = (bool *) palloc0(nvalues1 * sizeof(bool));
		hasmatch2 = (bool *) palloc0(nvalues2 * sizeof(bool));

		/*
		 * If we are doing any variant of JOIN_IN, pretend all the values of
		 * the righthand relation are unique (ie, act as if it's been
		 * DISTINCT'd).
		 *
		 * NOTE: it might seem that we should unique-ify the lefthand input
		 * when considering JOIN_REVERSE_IN.  But this is not so, because the
		 * join clause we've been handed has not been commuted from the way
		 * the parser originally wrote it.	We know that the unique side of
		 * the IN clause is *always* on the right.
		 *
		 * NOTE: it would be dangerous to try to be smart about JOIN_LEFT or
		 * JOIN_RIGHT here, because we do not have enough information to
		 * determine which var is really on which side of the join. Perhaps
		 * someday we should pass in more information.
		 */
		if (jointype == JOIN_IN ||
			jointype == JOIN_REVERSE_IN ||
			jointype == JOIN_UNIQUE_INNER ||
			jointype == JOIN_UNIQUE_OUTER)
		{
			float4		oneovern = 1.0 / nd2;

			for (i = 0; i < nvalues2; i++)
				numbers2[i] = oneovern;
			nullfrac2 = oneovern;
		}