/*
 * CopyShardInterval copies fields from the specified source ShardInterval
 * into the fields of the provided destination ShardInterval.
 */
void
CopyShardInterval(ShardInterval *srcInterval, ShardInterval *destInterval)
{
	destInterval->type = srcInterval->type;
	destInterval->relationId = srcInterval->relationId;
	destInterval->storageType = srcInterval->storageType;
	destInterval->valueTypeId = srcInterval->valueTypeId;
	destInterval->valueTypeLen = srcInterval->valueTypeLen;
	destInterval->valueByVal = srcInterval->valueByVal;
	destInterval->minValueExists = srcInterval->minValueExists;
	destInterval->maxValueExists = srcInterval->maxValueExists;
	destInterval->shardId = srcInterval->shardId;

	destInterval->minValue = 0;
	if (destInterval->minValueExists)
	{
		destInterval->minValue = datumCopy(srcInterval->minValue,
										   srcInterval->valueByVal,
										   srcInterval->valueTypeLen);
	}

	destInterval->maxValue = 0;
	if (destInterval->maxValueExists)
	{
		destInterval->maxValue = datumCopy(srcInterval->maxValue,
										   srcInterval->valueByVal,
										   srcInterval->valueTypeLen);
	}
}
Beispiel #2
0
void setHstoreFromDatum(lua_State *L, Datum datum)
{
    Lua_Hstore * strg;

    BEGINLUA;

    strg = (Lua_Hstore *)lua_newuserdata(L, sizeof(Lua_Hstore));
    MTOLUA(L);
    datum = datumCopy(datum, hs_type.byval, hs_type.len);
    MTOPG;

    strg->hstore = DatumGetHStoreP(datum);
    strg->datum = datum;
    strg->issync = 1;

    strg->havetodel = 1;

    lua_pushlightuserdata(L, strg);
    lua_newtable(L);
    lua_settable(L, LUA_REGISTRYINDEX);

    luaP_getfield(L, hstore_type_name);
    lua_setmetatable(L, -2);
    ENDLUAV(1);
}
Beispiel #3
0
/*
 * Copy a ParamListInfo structure.
 *
 * The result is allocated in CurrentMemoryContext.
 */
ParamListInfo
copyParamList(ParamListInfo from)
{
	ParamListInfo retval;
	Size		size;
	int			i;

	if (from == NULL || from->numParams <= 0)
		return NULL;

	/* sizeof(ParamListInfoData) includes the first array element */
	size = sizeof(ParamListInfoData) +
		(from->numParams - 1) *sizeof(ParamExternData);

	retval = (ParamListInfo) palloc(size);
	memcpy(retval, from, size);

	/*
	 * Flat-copy is not good enough for pass-by-ref data values, so make a
	 * pass over the array to copy those.
	 */
	for (i = 0; i < retval->numParams; i++)
	{
		ParamExternData *prm = &retval->params[i];
		int16		typLen;
		bool		typByVal;

		if (prm->isnull || !OidIsValid(prm->ptype))
			continue;
		get_typlenbyval(prm->ptype, &typLen, &typByVal);
		prm->value = datumCopy(prm->value, typByVal, typLen);
	}

	return retval;
}
Beispiel #4
0
/*-------------------------------------------------------------------------
 * datumTransfer
 *
 * Transfer a non-NULL datum into the current memory context.
 *
 * This is equivalent to datumCopy() except when the datum is a read-write
 * pointer to an expanded object.  In that case we merely reparent the object
 * into the current context, and return its standard R/W pointer (in case the
 * given one is a transient pointer of shorter lifespan).
 *-------------------------------------------------------------------------
 */
Datum
datumTransfer(Datum value, bool typByVal, int typLen)
{
	if (!typByVal && typLen == -1 &&
		VARATT_IS_EXTERNAL_EXPANDED_RW(DatumGetPointer(value)))
		value = TransferExpandedObject(value, CurrentMemoryContext);
	else
		value = datumCopy(value, typByVal, typLen);
	return value;
}
Beispiel #5
0
/*
 * This is basically the same as datumCopy(), but extended to count
 * palloc'd space in accum->allocatedMemory.
 */
static Datum
getDatumCopy(BuildAccumulator *accum, OffsetNumber attnum, Datum value)
{
	Form_pg_attribute att = accum->ginstate->origTupdesc->attrs[attnum - 1];
	Datum		res;

	if (att->attbyval)
		res = value;
	else
	{
		res = datumCopy(value, false, att->attlen);
		accum->allocatedMemory += GetMemoryChunkSpace(DatumGetPointer(res));
	}
	return res;
}
Beispiel #6
0
Datum anyold_transfn(PG_FUNCTION_ARGS)
{
	Oid           type;
	Datum         state;
	MemoryContext aggcontext,
	              oldcontext;
	int16         typlen;
	bool          typbyval;
	char          typalign;
	
	if (!AggCheckCallContext(fcinfo, &aggcontext))
	{
		/* cannot be called directly because of internal-type argument */
		elog(ERROR, "anyold_transfn called in non-aggregate context");
	}
	
	if (PG_ARGISNULL(0))
	{
		if (PG_ARGISNULL(1))
			PG_RETURN_NULL();
		/* First non-null value --- initialize */
		oldcontext = MemoryContextSwitchTo(aggcontext);
		
		type = get_fn_expr_argtype(fcinfo->flinfo, 1);
		if (type == InvalidOid)
			ereport(ERROR,
					(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
					 errmsg("could not determine input data type")));
		get_typlenbyvalalign(type,
							&typlen,
							&typbyval,
							&typalign);

		/* Copy initial value */
		if (typlen == -1)
			state = PointerGetDatum(PG_DETOAST_DATUM_COPY(PG_GETARG_DATUM(1)));
		else
			state = datumCopy(PG_GETARG_DATUM(1), typbyval, typlen);

		MemoryContextSwitchTo(oldcontext);
	}
	else
	{
		state = PG_GETARG_DATUM(0);
	}
	
	PG_RETURN_DATUM(state);
}
/*
 * Get datum representations of the attoptions field in pg_attribute_encoding
 * for the given relation.
 */
Datum *
get_rel_attoptions(Oid relid, AttrNumber max_attno)
{
	Form_pg_attribute attform;
	HeapTuple		tuple;
	cqContext		cqc;
	cqContext	   *pcqCtx;
	Datum		   *dats;
	Relation 		pgae = heap_open(AttributeEncodingRelationId,
									 AccessShareLock);

	/* used for attbyval and len below */
	attform = pgae->rd_att->attrs[Anum_pg_attribute_encoding_attoptions - 1];

	dats = palloc0(max_attno * sizeof(Datum));

	pcqCtx = caql_beginscan(
			caql_addrel(cqclr(&cqc), pgae),
			cql("SELECT * FROM pg_attribute_encoding "
				" WHERE attrelid = :1 ",
				ObjectIdGetDatum(relid)));

	while (HeapTupleIsValid(tuple = caql_getnext(pcqCtx)))
	{
		Form_pg_attribute_encoding a = 
			(Form_pg_attribute_encoding)GETSTRUCT(tuple);
		int16 attnum = a->attnum;
		Datum attoptions;
		bool isnull;

		Insist(attnum > 0 && attnum <= max_attno);

		attoptions = heap_getattr(tuple, Anum_pg_attribute_encoding_attoptions,
								  RelationGetDescr(pgae), &isnull);
		Insist(!isnull);

		dats[attnum - 1] = datumCopy(attoptions,
									 attform->attbyval,
									 attform->attlen);
	}

	caql_endscan(pcqCtx);

	heap_close(pgae, AccessShareLock);

	return dats;

}
Beispiel #8
0
static void 
argm_copy_datum(bool is_null, Datum src, ArgmDatumWithType *dest, bool free)
{
	if (free && !dest->typbyval && !dest->is_null)
		pfree(DatumGetPointer(dest->value));
	
	if (is_null)
		dest->is_null = true;
	else {
		dest->is_null = false;
		if (dest->typlen == -1)
			dest->value = PointerGetDatum(PG_DETOAST_DATUM_COPY(src));
		else
			dest->value = datumCopy(src, dest->typbyval, dest->typlen);
	}
}
Beispiel #9
0
/*
 * Copy a ParamListInfo structure.
 *
 * The result is allocated in CurrentMemoryContext.
 *
 * Note: the intent of this function is to make a static, self-contained
 * set of parameter values.  If dynamic parameter hooks are present, we
 * intentionally do not copy them into the result.	Rather, we forcibly
 * instantiate all available parameter values and copy the datum values.
 */
ParamListInfo
copyParamList(ParamListInfo from)
{
	ParamListInfo retval;
	Size		size;
	int			i;

	if (from == NULL || from->numParams <= 0)
		return NULL;

	/* sizeof(ParamListInfoData) includes the first array element */
	size = sizeof(ParamListInfoData) +
		(from->numParams - 1) * sizeof(ParamExternData);

	retval = (ParamListInfo) palloc(size);
	retval->paramFetch = NULL;
	retval->paramFetchArg = NULL;
	retval->parserSetup = NULL;
	retval->parserSetupArg = NULL;
	retval->numParams = from->numParams;

	for (i = 0; i < from->numParams; i++)
	{
		ParamExternData *oprm = &from->params[i];
		ParamExternData *nprm = &retval->params[i];
		int16		typLen;
		bool		typByVal;

		/* give hook a chance in case parameter is dynamic */
		if (!OidIsValid(oprm->ptype) && from->paramFetch != NULL)
			(*from->paramFetch) (from, i + 1);

		/* flat-copy the parameter info */
		*nprm = *oprm;

		/* need datumCopy in case it's a pass-by-reference datatype */
		if (nprm->isnull || !OidIsValid(nprm->ptype))
			continue;
		get_typlenbyval(nprm->ptype, &typLen, &typByVal);
		nprm->value = datumCopy(nprm->value, typByVal, typLen);
	}

	return retval;
}
Beispiel #10
0
Datum
jsonb_numeric(PG_FUNCTION_ARGS)
{
	Jsonb	   *j = PG_GETARG_JSONB(0);

	if (JB_ROOT_IS_SCALAR(j))
	{
		JsonbValue *jv;

		jv = getIthJsonbValueFromContainer(&j->root, 0);
		if (jv->type == jbvNumeric)
			PG_RETURN_DATUM(datumCopy(NumericGetDatum(jv->val.numeric), false,
													  -1));
	}

	ereport(ERROR,
			(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
			 errmsg("%s cannot be converted to numeric",
					JsonbToCString(NULL, &j->root, VARSIZE(j)))));
	PG_RETURN_NULL();
}
Beispiel #11
0
Datum
spg_text_inner_consistent(PG_FUNCTION_ARGS)
{
    spgInnerConsistentIn *in = (spgInnerConsistentIn *) PG_GETARG_POINTER(0);
    spgInnerConsistentOut *out = (spgInnerConsistentOut *) PG_GETARG_POINTER(1);
    StrategyNumber strategy = in->strategy;
    text	   *inText;
    int			inSize;
    int			i;
    text	   *reconstrText = NULL;
    int			maxReconstrLen = 0;
    text	   *prefixText = NULL;
    int			prefixSize = 0;

    /*
     * If it's a collation-aware operator, but the collation is C, we can
     * treat it as non-collation-aware.
     */
    if (strategy > 10 &&
            lc_collate_is_c(PG_GET_COLLATION()))
        strategy -= 10;

    inText = DatumGetTextPP(in->query);
    inSize = VARSIZE_ANY_EXHDR(inText);

    /*
     * Reconstruct values represented at this tuple, including parent data,
     * prefix of this tuple if any, and the node label if any.  in->level
     * should be the length of the previously reconstructed value, and the
     * number of bytes added here is prefixSize or prefixSize + 1.
     *
     * Note: we assume that in->reconstructedValue isn't toasted and doesn't
     * have a short varlena header.  This is okay because it must have been
     * created by a previous invocation of this routine, and we always emit
     * long-format reconstructed values.
     */
    Assert(in->level == 0 ? DatumGetPointer(in->reconstructedValue) == NULL :
           VARSIZE_ANY_EXHDR(DatumGetPointer(in->reconstructedValue)) == in->level);

    maxReconstrLen = in->level + 1;
    if (in->hasPrefix)
    {
        prefixText = DatumGetTextPP(in->prefixDatum);
        prefixSize = VARSIZE_ANY_EXHDR(prefixText);
        maxReconstrLen += prefixSize;
    }

    reconstrText = palloc(VARHDRSZ + maxReconstrLen);
    SET_VARSIZE(reconstrText, VARHDRSZ + maxReconstrLen);

    if (in->level)
        memcpy(VARDATA(reconstrText),
               VARDATA(DatumGetPointer(in->reconstructedValue)),
               in->level);
    if (prefixSize)
        memcpy(((char *) VARDATA(reconstrText)) + in->level,
               VARDATA_ANY(prefixText),
               prefixSize);
    /* last byte of reconstrText will be filled in below */

    /*
     * Scan the child nodes.  For each one, complete the reconstructed value
     * and see if it's consistent with the query.  If so, emit an entry into
     * the output arrays.
     */
    out->nodeNumbers = (int *) palloc(sizeof(int) * in->nNodes);
    out->levelAdds = (int *) palloc(sizeof(int) * in->nNodes);
    out->reconstructedValues = (Datum *) palloc(sizeof(Datum) * in->nNodes);
    out->nNodes = 0;

    for (i = 0; i < in->nNodes; i++)
    {
        uint8		nodeChar = DatumGetUInt8(in->nodeLabels[i]);
        int			thisLen;
        int			r;
        bool		res = false;

        /* If nodeChar is zero, don't include it in data */
        if (nodeChar == '\0')
            thisLen = maxReconstrLen - 1;
        else
        {
            ((char *) VARDATA(reconstrText))[maxReconstrLen - 1] = nodeChar;
            thisLen = maxReconstrLen;
        }

        r = memcmp(VARDATA(reconstrText), VARDATA_ANY(inText),
                   Min(inSize, thisLen));

        switch (strategy)
        {
        case BTLessStrategyNumber:
        case BTLessEqualStrategyNumber:
            if (r <= 0)
                res = true;
            break;
        case BTEqualStrategyNumber:
            if (r == 0 && inSize >= thisLen)
                res = true;
            break;
        case BTGreaterEqualStrategyNumber:
        case BTGreaterStrategyNumber:
            if (r >= 0)
                res = true;
            break;
        case BTLessStrategyNumber + 10:
        case BTLessEqualStrategyNumber + 10:
        case BTGreaterEqualStrategyNumber + 10:
        case BTGreaterStrategyNumber + 10:
            /*
             * with non-C collation we need to traverse whole tree :-(
             */
            res = true;
            break;
        default:
            elog(ERROR, "unrecognized strategy number: %d",
                 in->strategy);
            break;
        }

        if (res)
        {
            out->nodeNumbers[out->nNodes] = i;
            out->levelAdds[out->nNodes] = thisLen - in->level;
            SET_VARSIZE(reconstrText, VARHDRSZ + thisLen);
            out->reconstructedValues[out->nNodes] =
                datumCopy(PointerGetDatum(reconstrText), false, -1);
            out->nNodes++;
        }
    }

    PG_RETURN_VOID();
}
Beispiel #12
0
/*
 * SPI_cursor_open()
 *
 *	Open a prepared SPI plan as a portal
 */
Portal
SPI_cursor_open(const char *name, void *plan,
				Datum *Values, const char *Nulls,
				bool read_only)
{
	_SPI_plan  *spiplan = (_SPI_plan *) plan;
	List	   *qtlist = spiplan->qtlist;
	List	   *ptlist = spiplan->ptlist;
	Query	   *queryTree;
	Plan	   *planTree;
	ParamListInfo paramLI;
	Snapshot	snapshot;
	MemoryContext oldcontext;
	Portal		portal;
	int			k;

	/* Ensure that the plan contains only one query */
	if (list_length(ptlist) != 1 || list_length(qtlist) != 1)
		ereport(ERROR,
				(errcode(ERRCODE_INVALID_CURSOR_DEFINITION),
				 errmsg("cannot open multi-query plan as cursor")));
	queryTree = (Query *) linitial((List *) linitial(qtlist));
	planTree = (Plan *) linitial(ptlist);

	/* Must be a query that returns tuples */
	switch (queryTree->commandType)
	{
		case CMD_SELECT:
			if (queryTree->into != NULL)
				ereport(ERROR,
						(errcode(ERRCODE_INVALID_CURSOR_DEFINITION),
						 errmsg("cannot open SELECT INTO query as cursor")));
			break;
		case CMD_UTILITY:
			if (!UtilityReturnsTuples(queryTree->utilityStmt))
				ereport(ERROR,
						(errcode(ERRCODE_INVALID_CURSOR_DEFINITION),
						 errmsg("cannot open non-SELECT query as cursor")));
			break;
		default:
			ereport(ERROR,
					(errcode(ERRCODE_INVALID_CURSOR_DEFINITION),
					 errmsg("cannot open non-SELECT query as cursor")));
			break;
	}

	/* Reset SPI result */
	SPI_processed = 0;
	SPI_tuptable = NULL;
	_SPI_current->processed = 0;
	_SPI_current->tuptable = NULL;

	/* Create the portal */
	if (name == NULL || name[0] == '\0')
	{
		/* Use a random nonconflicting name */
		portal = CreateNewPortal();
	}
	else
	{
		/* In this path, error if portal of same name already exists */
		portal = CreatePortal(name, false, false);
	}

	/* Switch to portals memory and copy the parsetree and plan to there */
	oldcontext = MemoryContextSwitchTo(PortalGetHeapMemory(portal));
	queryTree = copyObject(queryTree);
	planTree = copyObject(planTree);

	/* If the plan has parameters, set them up */
	if (spiplan->nargs > 0)
	{
		paramLI = (ParamListInfo) palloc0((spiplan->nargs + 1) *
										  sizeof(ParamListInfoData));

		for (k = 0; k < spiplan->nargs; k++)
		{
			paramLI[k].kind = PARAM_NUM;
			paramLI[k].id = k + 1;
			paramLI[k].ptype = spiplan->argtypes[k];
			paramLI[k].isnull = (Nulls && Nulls[k] == 'n');
			if (paramLI[k].isnull)
			{
				/* nulls just copy */
				paramLI[k].value = Values[k];
			}
			else
			{
				/* pass-by-ref values must be copied into portal context */
				int16		paramTypLen;
				bool		paramTypByVal;

				get_typlenbyval(spiplan->argtypes[k],
								&paramTypLen, &paramTypByVal);
				paramLI[k].value = datumCopy(Values[k],
											 paramTypByVal, paramTypLen);
			}
		}
		paramLI[k].kind = PARAM_INVALID;
	}
	else
		paramLI = NULL;

	/*
	 * Set up the portal.
	 */
	PortalDefineQuery(portal,
					  NULL,		/* unfortunately don't have sourceText */
					  "SELECT", /* nor the raw parse tree... */
					  list_make1(queryTree),
					  list_make1(planTree),
					  PortalGetHeapMemory(portal));

	MemoryContextSwitchTo(oldcontext);

	/*
	 * Set up options for portal.
	 */
	portal->cursorOptions &= ~(CURSOR_OPT_SCROLL | CURSOR_OPT_NO_SCROLL);
	if (planTree == NULL || ExecSupportsBackwardScan(planTree))
		portal->cursorOptions |= CURSOR_OPT_SCROLL;
	else
		portal->cursorOptions |= CURSOR_OPT_NO_SCROLL;

	/*
	 * Set up the snapshot to use.  (PortalStart will do CopySnapshot,
	 * so we skip that here.)
	 */
	if (read_only)
		snapshot = ActiveSnapshot;
	else
	{
		CommandCounterIncrement();
		snapshot = GetTransactionSnapshot();
	}

	/*
	 * Start portal execution.
	 */
	PortalStart(portal, paramLI, snapshot);

	Assert(portal->strategy == PORTAL_ONE_SELECT ||
		   portal->strategy == PORTAL_UTIL_SELECT);

	/* Return the created portal */
	return portal;
}
Beispiel #13
0
/*
 * get_attstatsslot
 *
 *		Extract the contents of a "slot" of a pg_statistic tuple.
 *		Returns TRUE if requested slot type was found, else FALSE.
 *
 * Unlike other routines in this file, this takes a pointer to an
 * already-looked-up tuple in the pg_statistic cache.  We do this since
 * most callers will want to extract more than one value from the cache
 * entry, and we don't want to repeat the cache lookup unnecessarily.
 *
 * statstuple: pg_statistics tuple to be examined.
 * atttype: type OID of attribute (can be InvalidOid if values == NULL).
 * atttypmod: typmod of attribute (can be 0 if values == NULL).
 * reqkind: STAKIND code for desired statistics slot kind.
 * reqop: STAOP value wanted, or InvalidOid if don't care.
 * values, nvalues: if not NULL, the slot's stavalues are extracted.
 * numbers, nnumbers: if not NULL, the slot's stanumbers are extracted.
 *
 * If assigned, values and numbers are set to point to palloc'd arrays.
 * If the attribute type is pass-by-reference, the values referenced by
 * the values array are themselves palloc'd.  The palloc'd stuff can be
 * freed by calling free_attstatsslot.
 */
bool
get_attstatsslot(HeapTuple statstuple,
				 Oid atttype, int32 atttypmod,
				 int reqkind, Oid reqop,
				 Datum **values, int *nvalues,
				 float4 **numbers, int *nnumbers)
{
	Form_pg_statistic stats = (Form_pg_statistic) GETSTRUCT(statstuple);
	int			i,
				j;
	Datum		val;
	bool		isnull;
	ArrayType  *statarray;
	int			narrayelem;
	HeapTuple	typeTuple;
	Form_pg_type typeForm;

	for (i = 0; i < STATISTIC_NUM_SLOTS; i++)
	{
		if ((&stats->stakind1)[i] == reqkind &&
			(reqop == InvalidOid || (&stats->staop1)[i] == reqop))
			break;
	}
	if (i >= STATISTIC_NUM_SLOTS)
		return false;			/* not there */

	if (values)
	{
		val = SysCacheGetAttr(STATRELATT, statstuple,
							  Anum_pg_statistic_stavalues1 + i,
							  &isnull);
		if (isnull)
			elog(ERROR, "stavalues is null");
		statarray = DatumGetArrayTypeP(val);

		/* Need to get info about the array element type */
		typeTuple = SearchSysCache(TYPEOID,
								   ObjectIdGetDatum(atttype),
								   0, 0, 0);
		if (!HeapTupleIsValid(typeTuple))
			elog(ERROR, "cache lookup failed for type %u", atttype);
		typeForm = (Form_pg_type) GETSTRUCT(typeTuple);

		/* Deconstruct array into Datum elements; NULLs not expected */
		deconstruct_array(statarray,
						  atttype,
						  typeForm->typlen,
						  typeForm->typbyval,
						  typeForm->typalign,
						  values, NULL, nvalues);

		/*
		 * If the element type is pass-by-reference, we now have a bunch of
		 * Datums that are pointers into the syscache value.  Copy them to
		 * avoid problems if syscache decides to drop the entry.
		 */
		if (!typeForm->typbyval)
		{
			for (j = 0; j < *nvalues; j++)
			{
				(*values)[j] = datumCopy((*values)[j],
										 typeForm->typbyval,
										 typeForm->typlen);
			}
		}

		ReleaseSysCache(typeTuple);

		/*
		 * Free statarray if it's a detoasted copy.
		 */
		if ((Pointer) statarray != DatumGetPointer(val))
			pfree(statarray);
	}

	if (numbers)
	{
		val = SysCacheGetAttr(STATRELATT, statstuple,
							  Anum_pg_statistic_stanumbers1 + i,
							  &isnull);
		if (isnull)
			elog(ERROR, "stanumbers is null");
		statarray = DatumGetArrayTypeP(val);

		/*
		 * We expect the array to be a 1-D float4 array; verify that. We don't
		 * need to use deconstruct_array() since the array data is just going
		 * to look like a C array of float4 values.
		 */
		narrayelem = ARR_DIMS(statarray)[0];
		if (ARR_NDIM(statarray) != 1 || narrayelem <= 0 ||
			ARR_HASNULL(statarray) ||
			ARR_ELEMTYPE(statarray) != FLOAT4OID)
			elog(ERROR, "stanumbers is not a 1-D float4 array");
		*numbers = (float4 *) palloc(narrayelem * sizeof(float4));
		memcpy(*numbers, ARR_DATA_PTR(statarray), narrayelem * sizeof(float4));
		*nnumbers = narrayelem;

		/*
		 * Free statarray if it's a detoasted copy.
		 */
		if ((Pointer) statarray != DatumGetPointer(val))
			pfree(statarray);
	}

	return true;
}
Datum
spg_text_inner_consistent(PG_FUNCTION_ARGS)
{
	spgInnerConsistentIn *in = (spgInnerConsistentIn *) PG_GETARG_POINTER(0);
	spgInnerConsistentOut *out = (spgInnerConsistentOut *) PG_GETARG_POINTER(1);
	bool		collate_is_c = lc_collate_is_c(PG_GET_COLLATION());
	text	   *reconstrText = NULL;
	int			maxReconstrLen = 0;
	text	   *prefixText = NULL;
	int			prefixSize = 0;
	int			i;

	/*
	 * Reconstruct values represented at this tuple, including parent data,
	 * prefix of this tuple if any, and the node label if it's non-dummy.
	 * in->level should be the length of the previously reconstructed value,
	 * and the number of bytes added here is prefixSize or prefixSize + 1.
	 *
	 * Note: we assume that in->reconstructedValue isn't toasted and doesn't
	 * have a short varlena header.  This is okay because it must have been
	 * created by a previous invocation of this routine, and we always emit
	 * long-format reconstructed values.
	 */
	Assert(in->level == 0 ? DatumGetPointer(in->reconstructedValue) == NULL :
	VARSIZE_ANY_EXHDR(DatumGetPointer(in->reconstructedValue)) == in->level);

	maxReconstrLen = in->level + 1;
	if (in->hasPrefix)
	{
		prefixText = DatumGetTextPP(in->prefixDatum);
		prefixSize = VARSIZE_ANY_EXHDR(prefixText);
		maxReconstrLen += prefixSize;
	}

	reconstrText = palloc(VARHDRSZ + maxReconstrLen);
	SET_VARSIZE(reconstrText, VARHDRSZ + maxReconstrLen);

	if (in->level)
		memcpy(VARDATA(reconstrText),
			   VARDATA(DatumGetPointer(in->reconstructedValue)),
			   in->level);
	if (prefixSize)
		memcpy(((char *) VARDATA(reconstrText)) + in->level,
			   VARDATA_ANY(prefixText),
			   prefixSize);
	/* last byte of reconstrText will be filled in below */

	/*
	 * Scan the child nodes.  For each one, complete the reconstructed value
	 * and see if it's consistent with the query.  If so, emit an entry into
	 * the output arrays.
	 */
	out->nodeNumbers = (int *) palloc(sizeof(int) * in->nNodes);
	out->levelAdds = (int *) palloc(sizeof(int) * in->nNodes);
	out->reconstructedValues = (Datum *) palloc(sizeof(Datum) * in->nNodes);
	out->nNodes = 0;

	for (i = 0; i < in->nNodes; i++)
	{
		int16		nodeChar = DatumGetInt16(in->nodeLabels[i]);
		int			thisLen;
		bool		res = true;
		int			j;

		/* If nodeChar is a dummy value, don't include it in data */
		if (nodeChar <= 0)
			thisLen = maxReconstrLen - 1;
		else
		{
			((unsigned char *) VARDATA(reconstrText))[maxReconstrLen - 1] = nodeChar;
			thisLen = maxReconstrLen;
		}

		for (j = 0; j < in->nkeys; j++)
		{
			StrategyNumber strategy = in->scankeys[j].sk_strategy;
			text	   *inText;
			int			inSize;
			int			r;

			/*
			 * If it's a collation-aware operator, but the collation is C, we
			 * can treat it as non-collation-aware.  With non-C collation we
			 * need to traverse whole tree :-( so there's no point in making
			 * any check here.  (Note also that our reconstructed value may
			 * well end with a partial multibyte character, so that applying
			 * any encoding-sensitive test to it would be risky anyhow.)
			 */
			if (strategy > 10)
			{
				if (collate_is_c)
					strategy -= 10;
				else
					continue;
			}

			inText = DatumGetTextPP(in->scankeys[j].sk_argument);
			inSize = VARSIZE_ANY_EXHDR(inText);

			r = memcmp(VARDATA(reconstrText), VARDATA_ANY(inText),
					   Min(inSize, thisLen));

			switch (strategy)
			{
				case BTLessStrategyNumber:
				case BTLessEqualStrategyNumber:
					if (r > 0)
						res = false;
					break;
				case BTEqualStrategyNumber:
					if (r != 0 || inSize < thisLen)
						res = false;
					break;
				case BTGreaterEqualStrategyNumber:
				case BTGreaterStrategyNumber:
					if (r < 0)
						res = false;
					break;
				default:
					elog(ERROR, "unrecognized strategy number: %d",
						 in->scankeys[j].sk_strategy);
					break;
			}

			if (!res)
				break;			/* no need to consider remaining conditions */
		}

		if (res)
		{
			out->nodeNumbers[out->nNodes] = i;
			out->levelAdds[out->nNodes] = thisLen - in->level;
			SET_VARSIZE(reconstrText, VARHDRSZ + thisLen);
			out->reconstructedValues[out->nNodes] =
				datumCopy(PointerGetDatum(reconstrText), false, -1);
			out->nNodes++;
		}
	}

	PG_RETURN_VOID();
}
Beispiel #15
0
/*
 * Examine the given index tuple (which contains partial status of a certain
 * page range) by comparing it to the given value that comes from another heap
 * tuple.  If the new value is outside the min/max range specified by the
 * existing tuple values, update the index tuple and return true.  Otherwise,
 * return false and do not modify in this case.
 */
Datum
brin_minmax_add_value(PG_FUNCTION_ARGS)
{
	BrinDesc   *bdesc = (BrinDesc *) PG_GETARG_POINTER(0);
	BrinValues *column = (BrinValues *) PG_GETARG_POINTER(1);
	Datum		newval = PG_GETARG_DATUM(2);
	bool		isnull = PG_GETARG_DATUM(3);
	Oid			colloid = PG_GET_COLLATION();
	FmgrInfo   *cmpFn;
	Datum		compar;
	bool		updated = false;
	Form_pg_attribute attr;
	AttrNumber	attno;

	/*
	 * If the new value is null, we record that we saw it if it's the first
	 * one; otherwise, there's nothing to do.
	 */
	if (isnull)
	{
		if (column->bv_hasnulls)
			PG_RETURN_BOOL(false);

		column->bv_hasnulls = true;
		PG_RETURN_BOOL(true);
	}

	attno = column->bv_attno;
	attr = bdesc->bd_tupdesc->attrs[attno - 1];

	/*
	 * If the recorded value is null, store the new value (which we know to be
	 * not null) as both minimum and maximum, and we're done.
	 */
	if (column->bv_allnulls)
	{
		column->bv_values[0] = datumCopy(newval, attr->attbyval, attr->attlen);
		column->bv_values[1] = datumCopy(newval, attr->attbyval, attr->attlen);
		column->bv_allnulls = false;
		PG_RETURN_BOOL(true);
	}

	/*
	 * Otherwise, need to compare the new value with the existing boundaries
	 * and update them accordingly.  First check if it's less than the
	 * existing minimum.
	 */
	cmpFn = minmax_get_strategy_procinfo(bdesc, attno, attr->atttypid,
										 BTLessStrategyNumber);
	compar = FunctionCall2Coll(cmpFn, colloid, newval, column->bv_values[0]);
	if (DatumGetBool(compar))
	{
		if (!attr->attbyval)
			pfree(DatumGetPointer(column->bv_values[0]));
		column->bv_values[0] = datumCopy(newval, attr->attbyval, attr->attlen);
		updated = true;
	}

	/*
	 * And now compare it to the existing maximum.
	 */
	cmpFn = minmax_get_strategy_procinfo(bdesc, attno, attr->atttypid,
										 BTGreaterStrategyNumber);
	compar = FunctionCall2Coll(cmpFn, colloid, newval, column->bv_values[1]);
	if (DatumGetBool(compar))
	{
		if (!attr->attbyval)
			pfree(DatumGetPointer(column->bv_values[1]));
		column->bv_values[1] = datumCopy(newval, attr->attbyval, attr->attlen);
		updated = true;
	}

	PG_RETURN_BOOL(updated);
}
Beispiel #16
0
/*
 * Given two BrinValues, update the first of them as a union of the summary
 * values contained in both.  The second one is untouched.
 */
Datum
brin_minmax_union(PG_FUNCTION_ARGS)
{
	BrinDesc   *bdesc = (BrinDesc *) PG_GETARG_POINTER(0);
	BrinValues *col_a = (BrinValues *) PG_GETARG_POINTER(1);
	BrinValues *col_b = (BrinValues *) PG_GETARG_POINTER(2);
	Oid			colloid = PG_GET_COLLATION();
	AttrNumber	attno;
	Form_pg_attribute attr;
	FmgrInfo   *finfo;
	bool		needsadj;

	Assert(col_a->bv_attno == col_b->bv_attno);

	/* Adjust "hasnulls" */
	if (!col_a->bv_hasnulls && col_b->bv_hasnulls)
		col_a->bv_hasnulls = true;

	/* If there are no values in B, there's nothing left to do */
	if (col_b->bv_allnulls)
		PG_RETURN_VOID();

	attno = col_a->bv_attno;
	attr = bdesc->bd_tupdesc->attrs[attno - 1];

	/*
	 * Adjust "allnulls".  If A doesn't have values, just copy the values
	 * from B into A, and we're done.  We cannot run the operators in this
	 * case, because values in A might contain garbage.  Note we already
	 * established that B contains values.
	 */
	if (col_a->bv_allnulls)
	{
		col_a->bv_allnulls = false;
		col_a->bv_values[0] = datumCopy(col_b->bv_values[0],
										attr->attbyval, attr->attlen);
		col_a->bv_values[1] = datumCopy(col_b->bv_values[1],
										attr->attbyval, attr->attlen);
		PG_RETURN_VOID();
	}

	/* Adjust minimum, if B's min is less than A's min */
	finfo = minmax_get_strategy_procinfo(bdesc, attno, attr->atttypid,
										 BTLessStrategyNumber);
	needsadj = FunctionCall2Coll(finfo, colloid, col_b->bv_values[0],
								 col_a->bv_values[0]);
	if (needsadj)
	{
		if (!attr->attbyval)
			pfree(DatumGetPointer(col_a->bv_values[0]));
		col_a->bv_values[0] = datumCopy(col_b->bv_values[0],
										attr->attbyval, attr->attlen);
	}

	/* Adjust maximum, if B's max is greater than A's max */
	finfo = minmax_get_strategy_procinfo(bdesc, attno, attr->atttypid,
										 BTGreaterStrategyNumber);
	needsadj = FunctionCall2Coll(finfo, colloid, col_b->bv_values[1],
								 col_a->bv_values[1]);
	if (needsadj)
	{
		if (!attr->attbyval)
			pfree(DatumGetPointer(col_a->bv_values[1]));
		col_a->bv_values[1] = datumCopy(col_b->bv_values[1],
										attr->attbyval, attr->attlen);
	}

	PG_RETURN_VOID();
}
/*
 * compute_array_stats() -- compute statistics for an array column
 *
 * This function computes statistics useful for determining selectivity of
 * the array operators <@, &&, and @>.  It is invoked by ANALYZE via the
 * compute_stats hook after sample rows have been collected.
 *
 * We also invoke the standard compute_stats function, which will compute
 * "scalar" statistics relevant to the btree-style array comparison operators.
 * However, exact duplicates of an entire array may be rare despite many
 * arrays sharing individual elements.  This especially afflicts long arrays,
 * which are also liable to lack all scalar statistics due to the low
 * WIDTH_THRESHOLD used in analyze.c.  So, in addition to the standard stats,
 * we find the most common array elements and compute a histogram of distinct
 * element counts.
 *
 * The algorithm used is Lossy Counting, as proposed in the paper "Approximate
 * frequency counts over data streams" by G. S. Manku and R. Motwani, in
 * Proceedings of the 28th International Conference on Very Large Data Bases,
 * Hong Kong, China, August 2002, section 4.2. The paper is available at
 * http://www.vldb.org/conf/2002/S10P03.pdf
 *
 * The Lossy Counting (aka LC) algorithm goes like this:
 * Let s be the threshold frequency for an item (the minimum frequency we
 * are interested in) and epsilon the error margin for the frequency. Let D
 * be a set of triples (e, f, delta), where e is an element value, f is that
 * element's frequency (actually, its current occurrence count) and delta is
 * the maximum error in f. We start with D empty and process the elements in
 * batches of size w. (The batch size is also known as "bucket size" and is
 * equal to 1/epsilon.) Let the current batch number be b_current, starting
 * with 1. For each element e we either increment its f count, if it's
 * already in D, or insert a new___ triple into D with values (e, 1, b_current
 * - 1). After processing each batch we prune D, by removing from it all
 * elements with f + delta <= b_current.  After the algorithm finishes we
 * suppress all elements from D that do not satisfy f >= (s - epsilon) * N,
 * where N is the total number of elements in the input.  We emit the
 * remaining elements with estimated frequency f/N.  The LC paper proves
 * that this algorithm finds all elements with true frequency at least s,
 * and that no frequency is overestimated or is underestimated by more than
 * epsilon.  Furthermore, given reasonable assumptions about the input
 * distribution, the required table size is no more than about 7 times w.
 *
 * In the absence of a principled basis for other particular values, we
 * follow ts_typanalyze() and use parameters s = 0.07/K, epsilon = s/10.
 * But we leave out the correction for stopwords, which do not apply to
 * arrays.  These parameters give bucket width w = K/0.007 and maximum
 * expected hashtable size of about 1000 * K.
 *
 * Elements may repeat within an array.  Since duplicates do not change the
 * behavior of <@, && or @>, we want to count each element only once per
 * array.  Therefore, we store in the finished pg_statistic entry each
 * element's frequency as the fraction of all non-null rows that contain it.
 * We divide the raw counts by nonnull_cnt to get those figures.
 */
static void
compute_array_stats(VacAttrStats *stats, AnalyzeAttrFetchFunc fetchfunc,
					int samplerows, double totalrows)
{
	ArrayAnalyzeExtraData *extra_data;
	int			num_mcelem;
	int			null_cnt = 0;
	int			null_elem_cnt = 0;
	int			analyzed_rows = 0;

	/* This is D from the LC algorithm. */
	HTAB	   *elements_tab;
	HASHCTL		elem_hash_ctl;
	HASH_SEQ_STATUS scan_status;

	/* This is the current bucket number from the LC algorithm */
	int			b_current;

	/* This is 'w' from the LC algorithm */
	int			bucket_width;
	int			array_no;
	int64		element_no;
	TrackItem  *item;
	int			slot_idx;
	HTAB	   *count_tab;
	HASHCTL		count_hash_ctl;
	DECountItem *count_item;

	extra_data = (ArrayAnalyzeExtraData *) stats->extra_data;

	/*
	 * Invoke analyze.c's standard analysis function to create scalar-style
	 * stats for the column.  It will expect its own extra_data pointer, so
	 * temporarily install that.
	 */
	stats->extra_data = extra_data->std_extra_data;
	(*extra_data->std_compute_stats) (stats, fetchfunc, samplerows, totalrows);
	stats->extra_data = extra_data;

	/*
	 * Set up static pointer for use by subroutines.  We wait till here in
	 * case std_compute_stats somehow recursively invokes us (probably not
	 * possible, but ...)
	 */
	array_extra_data = extra_data;

	/*
	 * We want statistics_target * 10 elements in the MCELEM array. This
	 * multiplier is pretty arbitrary, but is meant to reflect the fact that
	 * the number of individual elements tracked in pg_statistic ought to be
	 * more than the number of values for a simple scalar column.
	 */
	num_mcelem = stats->attr->attstattarget * 10;

	/*
	 * We set bucket width equal to num_mcelem / 0.007 as per the comment
	 * above.
	 */
	bucket_width = num_mcelem * 1000 / 7;

	/*
	 * Create the hashtable. It will be in local memory, so we don't need to
	 * worry about overflowing the initial size. Also we don't need to pay any
	 * attention to locking and memory management.
	 */
	MemSet(&elem_hash_ctl, 0, sizeof(elem_hash_ctl));
	elem_hash_ctl.keysize = sizeof(Datum);
	elem_hash_ctl.entrysize = sizeof(TrackItem);
	elem_hash_ctl.hash = element_hash;
	elem_hash_ctl.match = element_match;
	elem_hash_ctl.hcxt = CurrentMemoryContext;
	elements_tab = hash_create("Analyzed elements table",
							   num_mcelem,
							   &elem_hash_ctl,
					HASH_ELEM | HASH_FUNCTION | HASH_COMPARE | HASH_CONTEXT);

	/* hashtable for array distinct elements counts */
	MemSet(&count_hash_ctl, 0, sizeof(count_hash_ctl));
	count_hash_ctl.keysize = sizeof(int);
	count_hash_ctl.entrysize = sizeof(DECountItem);
	count_hash_ctl.hcxt = CurrentMemoryContext;
	count_tab = hash_create("Array distinct element count table",
							64,
							&count_hash_ctl,
							HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);

	/* Initialize counters. */
	b_current = 1;
	element_no = 0;

	/* Loop over the arrays. */
	for (array_no = 0; array_no < samplerows; array_no++)
	{
		Datum		value;
		bool		isnull;
		ArrayType  *array;
		int			num_elems;
		Datum	   *elem_values;
		bool	   *elem_nulls;
		bool		null_present;
		int			j;
		int64		prev_element_no = element_no;
		int			distinct_count;
		bool		count_item_found;

		vacuum_delay_point();

		value = fetchfunc(stats, array_no, &isnull);
		if (isnull)
		{
			/* array is null, just count that */
			null_cnt++;
			continue;
		}

		/* Skip too-large values. */
		if (toast_raw_datum_size(value) > ARRAY_WIDTH_THRESHOLD)
			continue;
		else
			analyzed_rows++;

		/*
		 * Now detoast the array if needed, and deconstruct into datums.
		 */
		array = DatumGetArrayTypeP(value);

		Assert(ARR_ELEMTYPE(array) == extra_data->type_id);
		deconstruct_array(array,
						  extra_data->type_id,
						  extra_data->typlen,
						  extra_data->typbyval,
						  extra_data->typalign,
						  &elem_values, &elem_nulls, &num_elems);

		/*
		 * We loop through the elements in the array and add them to our
		 * tracking hashtable.
		 */
		null_present = false;
		for (j = 0; j < num_elems; j++)
		{
			Datum		elem_value;
			bool		found;

			/* No null element processing other than flag setting here */
			if (elem_nulls[j])
			{
				null_present = true;
				continue;
			}

			/* Lookup current element in hashtable, adding it if new___ */
			elem_value = elem_values[j];
			item = (TrackItem *) hash_search(elements_tab,
											 (const void *) &elem_value,
											 HASH_ENTER, &found);

			if (found)
			{
				/* The element value is already on the tracking list */

				/*
				 * The operators we assist ignore duplicate array elements, so
				 * count a given distinct element only once per array.
				 */
				if (item->last_container == array_no)
					continue;

				item->frequency++;
				item->last_container = array_no;
			}
			else
			{
				/* Initialize new___ tracking list element */

				/*
				 * If element type is pass-by-reference, we must copy it into
				 * palloc'd space, so that we can release the array below. (We
				 * do this so that the space needed for element values is
				 * limited by the size of the hashtable; if we kept all the
				 * array values around, it could be much more.)
				 */
				item->key = datumCopy(elem_value,
									  extra_data->typbyval,
									  extra_data->typlen);

				item->frequency = 1;
				item->delta = b_current - 1;
				item->last_container = array_no;
			}

			/* element_no is the number of elements processed (ie N) */
			element_no++;

			/* We prune the D structure after processing each bucket */
			if (element_no % bucket_width == 0)
			{
				prune_element_hashtable(elements_tab, b_current);
				b_current++;
			}
		}

		/* Count null element presence once per array. */
		if (null_present)
			null_elem_cnt++;

		/* Update frequency of the particular array distinct element count. */
		distinct_count = (int) (element_no - prev_element_no);
		count_item = (DECountItem *) hash_search(count_tab, &distinct_count,
												 HASH_ENTER,
												 &count_item_found);

		if (count_item_found)
			count_item->frequency++;
		else
			count_item->frequency = 1;

		/* Free memory allocated while detoasting. */
		if (PointerGetDatum(array) != value)
			pfree(array);
		pfree(elem_values);
		pfree(elem_nulls);
	}

	/* Skip pg_statistic slots occupied by standard statistics */
	slot_idx = 0;
	while (slot_idx < STATISTIC_NUM_SLOTS && stats->stakind[slot_idx] != 0)
		slot_idx++;
	if (slot_idx > STATISTIC_NUM_SLOTS - 2)
		elog(ERROR, "insufficient pg_statistic slots for array stats");

	/* We can only compute real stats if we found some non-null values. */
	if (analyzed_rows > 0)
	{
		int			nonnull_cnt = analyzed_rows;
		int			count_items_count;
		int			i;
		TrackItem **sort_table;
		int			track_len;
		int64		cutoff_freq;
		int64		minfreq,
					maxfreq;

		/*
		 * We assume the standard stats code already took care of setting
		 * stats_valid, stanullfrac, stawidth, stadistinct.  We'd have to
		 * re-compute those values if we wanted to not store the standard
		 * stats.
		 */

		/*
		 * Construct an array of the interesting hashtable items, that is,
		 * those meeting the cutoff frequency (s - epsilon)*N.  Also identify
		 * the minimum and maximum frequencies among these items.
		 *
		 * Since epsilon = s/10 and bucket_width = 1/epsilon, the cutoff
		 * frequency is 9*N / bucket_width.
		 */
		cutoff_freq = 9 * element_no / bucket_width;

		i = hash_get_num_entries(elements_tab); /* surely enough space */
		sort_table = (TrackItem **) palloc(sizeof(TrackItem *) * i);

		hash_seq_init(&scan_status, elements_tab);
		track_len = 0;
		minfreq = element_no;
		maxfreq = 0;
		while ((item = (TrackItem *) hash_seq_search(&scan_status)) != NULL)
		{
			if (item->frequency > cutoff_freq)
			{
				sort_table[track_len++] = item;
				minfreq = Min(minfreq, item->frequency);
				maxfreq = Max(maxfreq, item->frequency);
			}
		}
		Assert(track_len <= i);

		/* emit some statistics for debug purposes */
		elog(DEBUG3, "compute_array_stats: target # mces = %d, "
			 "bucket width = %d, "
			 "# elements = " INT64_FORMAT ", hashtable size = %d, "
			 "usable entries = %d",
			 num_mcelem, bucket_width, element_no, i, track_len);

		/*
		 * If we obtained more elements than we really want, get rid of those
		 * with least frequencies.  The easiest way is to qsort the array into
		 * descending frequency order and truncate the array.
		 */
		if (num_mcelem < track_len)
		{
			qsort(sort_table, track_len, sizeof(TrackItem *),
				  trackitem_compare_frequencies_desc);
			/* reset minfreq to the smallest frequency we're keeping */
			minfreq = sort_table[num_mcelem - 1]->frequency;
		}
		else
			num_mcelem = track_len;

		/* Generate MCELEM slot entry */
		if (num_mcelem > 0)
		{
			MemoryContext old_context;
			Datum	   *mcelem_values;
			float4	   *mcelem_freqs;

			/*
			 * We want to store statistics sorted on the element value using
			 * the element type's default comparison function.  This permits
			 * fast binary searches in selectivity estimation functions.
			 */
			qsort(sort_table, num_mcelem, sizeof(TrackItem *),
				  trackitem_compare_element);

			/* Must copy the target values into anl_context */
			old_context = MemoryContextSwitchTo(stats->anl_context);

			/*
			 * We sorted statistics on the element value, but we want to be
			 * able to find the minimal and maximal frequencies without going
			 * through all the values.  We also want the frequency of null
			 * elements.  Store these three values at the end of mcelem_freqs.
			 */
			mcelem_values = (Datum *) palloc(num_mcelem * sizeof(Datum));
			mcelem_freqs = (float4 *) palloc((num_mcelem + 3) * sizeof(float4));

			/*
			 * See comments above about use of nonnull_cnt as the divisor for
			 * the final frequency estimates.
			 */
			for (i = 0; i < num_mcelem; i++)
			{
				TrackItem  *item = sort_table[i];

				mcelem_values[i] = datumCopy(item->key,
											 extra_data->typbyval,
											 extra_data->typlen);
				mcelem_freqs[i] = (double) item->frequency /
					(double) nonnull_cnt;
			}
			mcelem_freqs[i++] = (double) minfreq / (double) nonnull_cnt;
			mcelem_freqs[i++] = (double) maxfreq / (double) nonnull_cnt;
			mcelem_freqs[i++] = (double) null_elem_cnt / (double) nonnull_cnt;

			MemoryContextSwitchTo(old_context);

			stats->stakind[slot_idx] = STATISTIC_KIND_MCELEM;
			stats->staop[slot_idx] = extra_data->eq_opr;
			stats->stanumbers[slot_idx] = mcelem_freqs;
			/* See above comment about extra stanumber entries */
			stats->numnumbers[slot_idx] = num_mcelem + 3;
			stats->stavalues[slot_idx] = mcelem_values;
			stats->numvalues[slot_idx] = num_mcelem;
			/* We are storing values of element type */
			stats->statypid[slot_idx] = extra_data->type_id;
			stats->statyplen[slot_idx] = extra_data->typlen;
			stats->statypbyval[slot_idx] = extra_data->typbyval;
			stats->statypalign[slot_idx] = extra_data->typalign;
			slot_idx++;
		}

		/* Generate DECHIST slot entry */
		count_items_count = hash_get_num_entries(count_tab);
		if (count_items_count > 0)
		{
			int			num_hist = stats->attr->attstattarget;
			DECountItem **sorted_count_items;
			int			j;
			int			delta;
			int64		frac;
			float4	   *hist;

			/* num_hist must be at least 2 for the loop below to work */
			num_hist = Max(num_hist, 2);

			/*
			 * Create an array of DECountItem pointers, and sort them into
			 * increasing count order.
			 */
			sorted_count_items = (DECountItem **)
				palloc(sizeof(DECountItem *) * count_items_count);
			hash_seq_init(&scan_status, count_tab);
			j = 0;
			while ((count_item = (DECountItem *) hash_seq_search(&scan_status)) != NULL)
			{
				sorted_count_items[j++] = count_item;
			}
			qsort(sorted_count_items, count_items_count,
				  sizeof(DECountItem *), countitem_compare_count);

			/*
			 * Prepare to fill stanumbers with the histogram, followed by the
			 * average count.  This array must be stored in anl_context.
			 */
			hist = (float4 *)
				MemoryContextAlloc(stats->anl_context,
								   sizeof(float4) * (num_hist + 1));
			hist[num_hist] = (double) element_no / (double) nonnull_cnt;

			/*----------
			 * Construct the histogram of distinct-element counts (DECs).
			 *
			 * The object of this loop is to copy the min and max DECs to
			 * hist[0] and hist[num_hist - 1], along with evenly-spaced DECs
			 * in between (where "evenly-spaced" is with reference to the
			 * whole input population of arrays).  If we had a complete sorted
			 * array of DECs, one per analyzed row, the i'th hist value would
			 * come from DECs[i * (analyzed_rows - 1) / (num_hist - 1)]
			 * (compare the histogram-making loop in compute_scalar_stats()).
			 * But instead of that we have the sorted_count_items[] array,
			 * which holds unique DEC values with their frequencies (that is,
			 * a run-length-compressed version of the full array).  So we
			 * control advancing through sorted_count_items[] with the
			 * variable "frac", which is defined as (x - y) * (num_hist - 1),
			 * where x is the index in the notional DECs array corresponding
			 * to the start of the next sorted_count_items[] element's run,
			 * and y is the index in DECs from which we should take the next
			 * histogram value.  We have to advance whenever x <= y, that is
			 * frac <= 0.  The x component is the sum of the frequencies seen
			 * so far (up through the current sorted_count_items[] element),
			 * and of course y * (num_hist - 1) = i * (analyzed_rows - 1),
			 * per the subscript calculation above.  (The subscript calculation
			 * implies dropping any fractional part of y; in this formulation
			 * that's handled by not advancing until frac reaches 1.)
			 *
			 * Even though frac has a bounded range, it could overflow int32
			 * when working with very large statistics targets, so we do that
			 * math in int64.
			 *----------
			 */
			delta = analyzed_rows - 1;
			j = 0;				/* current index in sorted_count_items */
			/* Initialize frac for sorted_count_items[0]; y is initially 0 */
			frac = (int64) sorted_count_items[0]->frequency * (num_hist - 1);
			for (i = 0; i < num_hist; i++)
			{
				while (frac <= 0)
				{
					/* Advance, and update x component of frac */
					j++;
					frac += (int64) sorted_count_items[j]->frequency * (num_hist - 1);
				}
				hist[i] = sorted_count_items[j]->count;
				frac -= delta;	/* update y for upcoming i increment */
			}
			Assert(j == count_items_count - 1);

			stats->stakind[slot_idx] = STATISTIC_KIND_DECHIST;
			stats->staop[slot_idx] = extra_data->eq_opr;
			stats->stanumbers[slot_idx] = hist;
			stats->numnumbers[slot_idx] = num_hist + 1;
			slot_idx++;
		}
	}

	/*
	 * We don't need to bother cleaning up any of our temporary palloc's. The
	 * hashtable should also go away, as it used a child memory context.
	 */
}
Beispiel #18
0
/*
 * Convert a BrinTuple back to a BrinMemTuple.  This is the reverse of
 * brin_form_tuple.
 *
 * As an optimization, the caller can pass a previously allocated 'dMemtuple'.
 * This avoids having to allocate it here, which can be useful when this
 * function is called many times in a loop.  It is caller's responsibility
 * that the given BrinMemTuple matches what we need here.
 *
 * Note we don't need the "on disk tupdesc" here; we rely on our own routine to
 * deconstruct the tuple from the on-disk format.
 */
BrinMemTuple *
brin_deform_tuple(BrinDesc *brdesc, BrinTuple *tuple, BrinMemTuple *dMemtuple)
{
	BrinMemTuple *dtup;
	Datum	   *values;
	bool	   *allnulls;
	bool	   *hasnulls;
	char	   *tp;
	bits8	   *nullbits;
	int			keyno;
	int			valueno;
	MemoryContext oldcxt;

	dtup = dMemtuple ? brin_memtuple_initialize(dMemtuple, brdesc) :
		brin_new_memtuple(brdesc);

	if (BrinTupleIsPlaceholder(tuple))
		dtup->bt_placeholder = true;
	dtup->bt_blkno = tuple->bt_blkno;

	values = dtup->bt_values;
	allnulls = dtup->bt_allnulls;
	hasnulls = dtup->bt_hasnulls;

	tp = (char *) tuple + BrinTupleDataOffset(tuple);

	if (BrinTupleHasNulls(tuple))
		nullbits = (bits8 *) ((char *) tuple + SizeOfBrinTuple);
	else
		nullbits = NULL;
	brin_deconstruct_tuple(brdesc,
						   tp, nullbits, BrinTupleHasNulls(tuple),
						   values, allnulls, hasnulls);

	/*
	 * Iterate to assign each of the values to the corresponding item in the
	 * values array of each column.  The copies occur in the tuple's context.
	 */
	oldcxt = MemoryContextSwitchTo(dtup->bt_context);
	for (valueno = 0, keyno = 0; keyno < brdesc->bd_tupdesc->natts; keyno++)
	{
		int			i;

		if (allnulls[keyno])
		{
			valueno += brdesc->bd_info[keyno]->oi_nstored;
			continue;
		}

		/*
		 * We would like to skip datumCopy'ing the values datum in some cases,
		 * caller permitting ...
		 */
		for (i = 0; i < brdesc->bd_info[keyno]->oi_nstored; i++)
			dtup->bt_columns[keyno].bv_values[i] =
				datumCopy(values[valueno++],
						  brdesc->bd_info[keyno]->oi_typcache[i]->typbyval,
						  brdesc->bd_info[keyno]->oi_typcache[i]->typlen);

		dtup->bt_columns[keyno].bv_hasnulls = hasnulls[keyno];
		dtup->bt_columns[keyno].bv_allnulls = false;
	}

	MemoryContextSwitchTo(oldcxt);

	return dtup;
}
Beispiel #19
0
/*
 * Copied from Postgres' src/backend/optimizer/util/clauses.c
 *
 * evaluate_expr: pre-evaluate a constant expression
 *
 * We use the executor's routine ExecEvalExpr() to avoid duplication of
 * code and ensure we get the same result as the executor would get.
 */
Expr *
evaluate_expr(Expr *expr, Oid result_type, int32 result_typmod,
              Oid result_collation)
{
    EState	   *estate;
    ExprState  *exprstate;
    MemoryContext oldcontext;
    Datum		const_val;
    bool		const_is_null;
    int16		resultTypLen;
    bool		resultTypByVal;

    /*
     * To use the executor, we need an EState.
     */
    estate = CreateExecutorState();

    /* We can use the estate's working context to avoid memory leaks. */
    oldcontext = MemoryContextSwitchTo(estate->es_query_cxt);

    /* Make sure any opfuncids are filled in. */
    fix_opfuncids((Node *) expr);

    /*
     * Prepare expr for execution.  (Note: we can't use ExecPrepareExpr
     * because it'd result in recursively invoking eval_const_expressions.)
     */
    exprstate = ExecInitExpr(expr, NULL);

    /*
     * And evaluate it.
     *
     * It is OK to use a default econtext because none of the ExecEvalExpr()
     * code used in this situation will use econtext.  That might seem
     * fortuitous, but it's not so unreasonable --- a constant expression does
     * not depend on context, by definition, n'est ce pas?
     */
    const_val = ExecEvalExprSwitchContext(exprstate,
                                          GetPerTupleExprContext(estate),
                                          &const_is_null, NULL);

    /* Get info needed about result datatype */
    get_typlenbyval(result_type, &resultTypLen, &resultTypByVal);

    /* Get back to outer memory context */
    MemoryContextSwitchTo(oldcontext);

    /*
     * Must copy result out of sub-context used by expression eval.
     *
     * Also, if it's varlena, forcibly detoast it.  This protects us against
     * storing TOAST pointers into plans that might outlive the referenced
     * data.  (makeConst would handle detoasting anyway, but it's worth a few
     * extra lines here so that we can do the copy and detoast in one step.)
     */
    if (!const_is_null)
    {
        if (resultTypLen == -1)
            const_val = PointerGetDatum(PG_DETOAST_DATUM_COPY(const_val));
        else
            const_val = datumCopy(const_val, resultTypByVal, resultTypLen);
    }

    /* Release all the junk we just created */
    FreeExecutorState(estate);

    /*
     * Make the constant result node.
     */
    return (Expr *) makeConst(result_type, result_typmod, result_collation,
                              resultTypLen,
                              const_val, const_is_null,
                              resultTypByVal);
}