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

		return result;
	if (!varonleft || !IsA(other, Const))
	{
		ReleaseVariableStats(vardata);
		return result;
	}
	variable = (Node *) linitial(args);

	/*
	 * If the constant is NULL, assume operator is strict and return zero, ie,
	 * operator will never return TRUE.  (It's zero even for a negator op.)
	 */
	if (((Const *) other)->constisnull)
	{
		ReleaseVariableStats(vardata);
		return 0.0;
	}
	constval = ((Const *) other)->constvalue;
	consttype = ((Const *) other)->consttype;

	/*
	 * The right-hand const is type text or bytea for all supported operators.
	 * We do not expect to see binary-compatible types here, since
	 * const-folding should have relabeled the const to exactly match the
	 * operator's declared type.
	 */
	if (consttype != TEXTOID && consttype != BYTEAOID)
	{
		ReleaseVariableStats(vardata);
		return result;
	}

	/*
	 * Similarly, the exposed type of the left-hand side should be one of
	 * those we know.  (Do not look at vardata.atttype, which might be
	 * something binary-compatible but different.)	We can use it to choose
	 * the index opfamily from which we must draw the comparison operators.
	 *
	 * NOTE: It would be more correct to use the PATTERN opfamilies than the
	 * simple ones, but at the moment ANALYZE will not generate statistics for
	 * the PATTERN operators.  But our results are so approximate anyway that
	 * it probably hardly matters.
	 */
	vartype = vardata.vartype;

	switch (vartype)
	{
		case TEXTOID:
			opfamily = TEXT_BTREE_FAM_OID;
			break;
		case BPCHAROID:
			opfamily = BPCHAR_BTREE_FAM_OID;
			break;
		case NAMEOID:
			opfamily = NAME_BTREE_FAM_OID;
			break;
		case BYTEAOID:
			opfamily = BYTEA_BTREE_FAM_OID;
			break;
		default:
			ReleaseVariableStats(vardata);
			return result;
	}

	/* divide pattern into fixed prefix and remainder */
	patt = (Const *) other;
	pstatus = pattern_fixed_prefix(patt, ptype, &prefix, &rest);

	/*
	 * If necessary, coerce the prefix constant to the right type. (The "rest"
	 * constant need not be changed.)
	 */
	if (prefix && prefix->consttype != vartype)
	{
		char	   *prefixstr;

		switch (prefix->consttype)
		{
			case TEXTOID:
				prefixstr = DatumGetCString(DirectFunctionCall1(textout,
														prefix->constvalue));
				break;
			case BYTEAOID:
				prefixstr = DatumGetCString(DirectFunctionCall1(byteaout,
														prefix->constvalue));
				break;
			default:
				elog(ERROR, "unrecognized consttype: %u",
					 prefix->consttype);
				ReleaseVariableStats(vardata);
				return result;
		}
		prefix = string_to_const(prefixstr, vartype);
		pfree(prefixstr);
	}

	if (pstatus == Pattern_Prefix_Exact)
	{
		/*
		 * Pattern specifies an exact match, so pretend operator is '='
		 */
		Oid			eqopr = get_opfamily_member(opfamily, vartype, vartype,
												BTEqualStrategyNumber);
		List	   *eqargs;

		if (eqopr == InvalidOid)
			elog(ERROR, "no = operator for opfamily %u", opfamily);
		eqargs = list_make2(variable, prefix);
		result = DatumGetFloat8(DirectFunctionCall4(eqsel,
													PointerGetDatum(root),
													ObjectIdGetDatum(eqopr),
													PointerGetDatum(eqargs),
													Int32GetDatum(varRelid)));
	}
	else
	{
		/*
		 * Not exact-match pattern.  If we have a sufficiently large
		 * histogram, estimate selectivity for the histogram part of the
		 * population by counting matches in the histogram.  If not, estimate
		 * selectivity of the fixed prefix and remainder of pattern
		 * separately, then combine the two to get an estimate of the
		 * selectivity for the part of the column population represented by
		 * the histogram.  We then add up data for any most-common-values
		 * values; these are not in the histogram population, and we can get
		 * exact answers for them by applying the pattern operator, so there's
		 * no reason to approximate.  (If the MCVs cover a significant part of
		 * the total population, this gives us a big leg up in accuracy.)
		 */
		Selectivity selec;
		FmgrInfo	opproc;
		double		nullfrac,
					mcv_selec,
					sumcommon;

		/* Try to use the histogram entries to get selectivity */
		fmgr_info(get_opcode(operator), &opproc);

		selec = histogram_selectivity(&vardata, &opproc, constval, true,
									  100, 1);
		if (selec < 0)
		{
			/* Nope, so fake it with the heuristic method */
			Selectivity prefixsel;
			Selectivity restsel;

			if (pstatus == Pattern_Prefix_Partial)
				prefixsel = prefix_selectivity(&vardata, vartype,
											   opfamily, prefix);
			else
				prefixsel = 1.0;
			restsel = pattern_selectivity(rest, ptype);
			selec = prefixsel * restsel;
		}
		else
		{
			/* Yes, but don't believe extremely small or large estimates. */
			if (selec < 0.0001)
				selec = 0.0001;
			else if (selec > 0.9999)
				selec = 0.9999;
		}

		/*
		 * If we have most-common-values info, add up the fractions of the MCV
		 * entries that satisfy MCV OP PATTERN.  These fractions contribute
		 * directly to the result selectivity.	Also add up the total fraction
		 * represented by MCV entries.
		 */
		mcv_selec = mcv_selectivity(&vardata, &opproc, constval, true,
									&sumcommon);

		if (HeapTupleIsValid(vardata.statsTuple))
			nullfrac = ((Form_pg_statistic) GETSTRUCT(vardata.statsTuple))->stanullfrac;
		else
			nullfrac = 0.0;

		/*
		 * Now merge the results from the MCV and histogram calculations,
		 * realizing that the histogram covers only the non-null values that
		 * are not listed in MCV.
		 */
		selec *= 1.0 - nullfrac - sumcommon;
		selec += mcv_selec;

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

	if (prefix)
	{
		pfree(DatumGetPointer(prefix->constvalue));
		pfree(prefix);
	}

	ReleaseVariableStats(vardata);

	return negate ? (1.0 - result) : result;
}

/*
 *		regexeqsel		- Selectivity of regular-expression pattern match.
 */
Datum
regexeqsel(PG_FUNCTION_ARGS)
{
	PG_RETURN_FLOAT8(patternsel(fcinfo, Pattern_Type_Regex, false));
}

/*
 *		icregexeqsel	- Selectivity of case-insensitive regex match.
 */
Datum
icregexeqsel(PG_FUNCTION_ARGS)
{
	PG_RETURN_FLOAT8(patternsel(fcinfo, Pattern_Type_Regex_IC, false));
}

/*
 *		likesel			- Selectivity of LIKE pattern match.
 */
Datum
likesel(PG_FUNCTION_ARGS)
{
	PG_RETURN_FLOAT8(patternsel(fcinfo, Pattern_Type_Like, false));
}

/*
 *		iclikesel			- Selectivity of ILIKE pattern match.
 */
Datum
iclikesel(PG_FUNCTION_ARGS)
{
	PG_RETURN_FLOAT8(patternsel(fcinfo, Pattern_Type_Like_IC, false));
}

/*
 *		regexnesel		- Selectivity of regular-expression pattern non-match.
 */
Datum
regexnesel(PG_FUNCTION_ARGS)
{
	PG_RETURN_FLOAT8(patternsel(fcinfo, Pattern_Type_Regex, true));
}

/*
 *		icregexnesel	- Selectivity of case-insensitive regex non-match.
 */
Datum
icregexnesel(PG_FUNCTION_ARGS)
{
	PG_RETURN_FLOAT8(patternsel(fcinfo, Pattern_Type_Regex_IC, true));
}

/*
 *		nlikesel		- Selectivity of LIKE pattern non-match.
 */
Datum
nlikesel(PG_FUNCTION_ARGS)
{
	PG_RETURN_FLOAT8(patternsel(fcinfo, Pattern_Type_Like, true));
}

/*
 *		icnlikesel		- Selectivity of ILIKE pattern non-match.
 */
Datum
icnlikesel(PG_FUNCTION_ARGS)
{
	PG_RETURN_FLOAT8(patternsel(fcinfo, Pattern_Type_Like_IC, true));
}

/*
 *		booltestsel		- Selectivity of BooleanTest Node.
 */
Selectivity
booltestsel(PlannerInfo *root, BoolTestType booltesttype, Node *arg,
			int varRelid, JoinType jointype)
{
	VariableStatData vardata;
	double		selec;

	examine_variable(root, arg, varRelid, &vardata);

	if (HeapTupleIsValid(vardata.statsTuple))
	{
		Form_pg_statistic stats;
		double		freq_null;
		Datum	   *values;
		int			nvalues;
		float4	   *numbers;
		int			nnumbers;

		stats = (Form_pg_statistic) GETSTRUCT(vardata.statsTuple);
		freq_null = stats->stanullfrac;

		if (get_attstatsslot(vardata.statsTuple,
							 vardata.atttype, vardata.atttypmod,
							 STATISTIC_KIND_MCV, InvalidOid,
							 &values, &nvalues,
							 &numbers, &nnumbers)
			&& nnumbers > 0)
		{
			double		freq_true;
			double		freq_false;

			/*
			 * Get first MCV frequency and derive frequency for true.
			 */
			if (DatumGetBool(values[0]))
				freq_true = numbers[0];
			else
				freq_true = 1.0 - numbers[0] - freq_null;

			/*
			 * Next derive frequency for false. Then use these as appropriate
			 * to derive frequency for each case.
			 */
			freq_false = 1.0 - freq_true - freq_null;

			switch (booltesttype)
			{
				case IS_UNKNOWN:
					/* select only NULL values */
					selec = freq_null;
					break;
				case IS_NOT_UNKNOWN:
					/* select non-NULL values */
					selec = 1.0 - freq_null;
					break;
				case IS_TRUE:
					/* select only TRUE values */
					selec = freq_true;
					break;
				case IS_NOT_TRUE:
					/* select non-TRUE values */
					selec = 1.0 - freq_true;
					break;
				case IS_FALSE:
					/* select only FALSE values */
					selec = freq_false;
					break;
				case IS_NOT_FALSE:
					/* select non-FALSE values */
					selec = 1.0 - freq_false;
					break;
				default:
					elog(ERROR, "unrecognized booltesttype: %d",
						 (int) booltesttype);
					selec = 0.0;	/* Keep compiler quiet */
					break;
			}

			free_attstatsslot(vardata.atttype, values, nvalues,
							  numbers, nnumbers);
		}
		else
		{
			/*
			 * No most-common-value info available. Still have null fraction
			 * information, so use it for IS [NOT] UNKNOWN. Otherwise adjust
			 * for null fraction and assume an even split for boolean tests.
			 */
			switch (booltesttype)
			{
				case IS_UNKNOWN:

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

					/*
					 * Select not unknown (not null) values. Calculate from
					 * freq_null.
					 */
					selec = 1.0 - freq_null;
					break;
				case IS_TRUE:
				case IS_NOT_TRUE:
				case IS_FALSE:
				case IS_NOT_FALSE:
					selec = (1.0 - freq_null) / 2.0;
					break;
				default:
					elog(ERROR, "unrecognized booltesttype: %d",
						 (int) booltesttype);
					selec = 0.0;	/* Keep compiler quiet */
					break;
			}
		}
	}
	else
	{
		/*
		 * If we can't get variable statistics for the argument, perhaps
		 * clause_selectivity can do something with it.  We ignore the
		 * possibility of a NULL value when using clause_selectivity, and just
		 * assume the value is either TRUE or FALSE.
		 */
		switch (booltesttype)
		{
			case IS_UNKNOWN:
				selec = DEFAULT_UNK_SEL;
				break;
			case IS_NOT_UNKNOWN:
				selec = DEFAULT_NOT_UNK_SEL;
				break;
			case IS_TRUE:
			case IS_NOT_FALSE:
				selec = (double) clause_selectivity(root, arg,
													varRelid, jointype);
				break;
			case IS_FALSE:
			case IS_NOT_TRUE:
				selec = 1.0 - (double) clause_selectivity(root, arg,
														  varRelid, jointype);
				break;
			default:
				elog(ERROR, "unrecognized booltesttype: %d",
					 (int) booltesttype);
				selec = 0.0;	/* Keep compiler quiet */
				break;
		}
	}

	ReleaseVariableStats(vardata);

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

	return (Selectivity) selec;
}

/*
 *		nulltestsel		- Selectivity of NullTest Node.
 */
Selectivity
nulltestsel(PlannerInfo *root, NullTestType nulltesttype,
			Node *arg, int varRelid, JoinType jointype)
{
	VariableStatData vardata;
	double		selec;

	/*
	 * Special hack: an IS NULL test being applied at an outer join should not
	 * be taken at face value, since it's very likely being used to select the
	 * outer-side rows that don't have a match, and thus its selectivity has
	 * nothing whatever to do with the statistics of the original table
	 * column.	We do not have nearly enough context here to determine its
	 * true selectivity, so for the moment punt and guess at 0.5.  Eventually
	 * the planner should be made to provide enough info about the clause's
	 * context to let us do better.
	 */
	if (IS_OUTER_JOIN(jointype) && nulltesttype == IS_NULL)
		return (Selectivity) 0.5;

	examine_variable(root, arg, varRelid, &vardata);

	if (HeapTupleIsValid(vardata.statsTuple))
	{
		Form_pg_statistic stats;
		double		freq_null;

		stats = (Form_pg_statistic) GETSTRUCT(vardata.statsTuple);
		freq_null = stats->stanullfrac;

		switch (nulltesttype)
		{
			case IS_NULL: