Reformat all code to coding standard

utils/udfsdk/allnull.cpp

@@ -29,59 +29,63 @@ struct allnull_data
 mcsv1_UDAF::ReturnCode allnull::init(mcsv1Context* context,
                                      COL_TYPES& colTypes)
 {
-	context->setUserDataSize(sizeof(allnull_data));
-	if (colTypes.size() < 1)
-	{
-		// The error message will be prepended with
-		// "The storage engine for the table doesn't support "
-		context->setErrorMessage("allnull() with 0 arguments");
-		return mcsv1_UDAF::ERROR;
-	}
-	context->setResultType(CalpontSystemCatalog::TINYINT);
+    context->setUserDataSize(sizeof(allnull_data));
 
-	return mcsv1_UDAF::SUCCESS;
+    if (colTypes.size() < 1)
+    {
+        // The error message will be prepended with
+        // "The storage engine for the table doesn't support "
+        context->setErrorMessage("allnull() with 0 arguments");
+        return mcsv1_UDAF::ERROR;
+    }
+
+    context->setResultType(CalpontSystemCatalog::TINYINT);
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode allnull::reset(mcsv1Context* context)
 {
-	struct allnull_data* data = (struct allnull_data*)context->getUserData()->data;
-	data->totalQuantity = 0;
-	data->totalNulls    = 0;
-	return mcsv1_UDAF::SUCCESS;
+    struct allnull_data* data = (struct allnull_data*)context->getUserData()->data;
+    data->totalQuantity = 0;
+    data->totalNulls    = 0;
+    return mcsv1_UDAF::SUCCESS;
 }
 
-mcsv1_UDAF::ReturnCode allnull::nextValue(mcsv1Context* context, 
-										  std::vector<ColumnDatum>& valsIn)
+mcsv1_UDAF::ReturnCode allnull::nextValue(mcsv1Context* context,
+        std::vector<ColumnDatum>& valsIn)
 {
-	struct allnull_data* data = (struct allnull_data*)context->getUserData()->data;
-	
-	for (size_t i = 0; i < context->getParameterCount(); i++)
-	{
-		data->totalQuantity++;
-		if (context->isParamNull(0))
-		{
-			data->totalNulls++;
-		}
-	}
-	return mcsv1_UDAF::SUCCESS;
+    struct allnull_data* data = (struct allnull_data*)context->getUserData()->data;
+
+    for (size_t i = 0; i < context->getParameterCount(); i++)
+    {
+        data->totalQuantity++;
+
+        if (context->isParamNull(0))
+        {
+            data->totalNulls++;
+        }
+    }
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode allnull::subEvaluate(mcsv1Context* context, const UserData* userDataIn)
 {
-	struct allnull_data* outData = (struct allnull_data*)context->getUserData()->data;
-	struct allnull_data* inData = (struct allnull_data*)userDataIn->data;
-	outData->totalQuantity += inData->totalQuantity;
-	outData->totalNulls += inData->totalNulls;
-	return mcsv1_UDAF::SUCCESS;
+    struct allnull_data* outData = (struct allnull_data*)context->getUserData()->data;
+    struct allnull_data* inData = (struct allnull_data*)userDataIn->data;
+    outData->totalQuantity += inData->totalQuantity;
+    outData->totalNulls += inData->totalNulls;
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode allnull::evaluate(mcsv1Context* context, static_any::any& valOut)
 {
-	OUT_TYPE allNull;
-	struct allnull_data* data = (struct allnull_data*)context->getUserData()->data;
-	allNull = data->totalQuantity > 0 && data->totalNulls == data->totalQuantity;
-	valOut = allNull;
-	return mcsv1_UDAF::SUCCESS;
+    OUT_TYPE allNull;
+    struct allnull_data* data = (struct allnull_data*)context->getUserData()->data;
+    allNull = data->totalQuantity > 0 && data->totalNulls == data->totalQuantity;
+    valOut = allNull;
+    return mcsv1_UDAF::SUCCESS;
 }
 
 

utils/udfsdk/allnull.h

@@ -21,26 +21,26 @@
 *   mcsv1_UDAF.h
 ***********************************************************************/
 
-/** 
- * Columnstore interface for writing a User Defined Aggregate 
- * Functions (UDAF) and User Defined Analytic Functions (UDAnF) 
- * or a function that can act as either - UDA(n)F 
- *  
+/**
+ * Columnstore interface for writing a User Defined Aggregate
+ * Functions (UDAF) and User Defined Analytic Functions (UDAnF)
+ * or a function that can act as either - UDA(n)F
+ *
  * The basic steps are:
  *
- * 1. Create a the UDA(n)F function interface in some .h file. 
- * 2. Create the UDF function implementation in some .cpp file 
- * 3. Create the connector stub (MariaDB UDAF definition) for 
- * this UDF function.  
- * 4. build the dynamic library using all of the source. 
- * 5  Put the library in $COLUMNSTORE_INSTALL/lib of 
- * all modules 
- * 6. restart the Columnstore system. 
+ * 1. Create a the UDA(n)F function interface in some .h file.
+ * 2. Create the UDF function implementation in some .cpp file
+ * 3. Create the connector stub (MariaDB UDAF definition) for
+ * this UDF function.
+ * 4. build the dynamic library using all of the source.
+ * 5  Put the library in $COLUMNSTORE_INSTALL/lib of
+ * all modules
+ * 6. restart the Columnstore system.
  * 7. notify mysqld about the new functions with commands like:
- *  
+ *
  *    CREATE AGGREGATE FUNCTION allnull returns BOOL soname
  *    'libudf_mysql.so';
- *  
+ *
  */
 #ifndef HEADER_allnull
 #define HEADER_allnull
@@ -69,124 +69,124 @@ using namespace execplan;
 namespace mcsv1sdk
 {
 
-// Override mcsv1_UDAF to build your User Defined Aggregate (UDAF) and/or 
+// Override mcsv1_UDAF to build your User Defined Aggregate (UDAF) and/or
 // User Defined Analytic Function (UDAnF).
 // These will be singleton classes, so don't put any instance
 // specific data in here. All instance data is stored in mcsv1Context
 // passed to each user function and retrieved by the getUserData() method.
-// 
-// Each API function returns a ReturnCode. If ERROR is returned at any time, 
-// the query is aborted, getInterrupted() will begin to return true and the 
-// message set in config->setErrorMessage() is returned to MariaDB. 
+//
+// Each API function returns a ReturnCode. If ERROR is returned at any time,
+// the query is aborted, getInterrupted() will begin to return true and the
+// message set in config->setErrorMessage() is returned to MariaDB.
 class allnull : public  mcsv1_UDAF
 {
 public:
-	// Defaults OK
-	allnull() : mcsv1_UDAF(){};
-	virtual ~allnull(){};
+    // Defaults OK
+    allnull() : mcsv1_UDAF() {};
+    virtual ~allnull() {};
 
-	/** 
-	 * init() 
-	 *  
-	 * Mandatory. Implement this to initialize flags and instance 
-	 * data. Called once per SQL statement. You can do any sanity 
-	 * checks here. 
-	 *  
-	 * colTypes (in) - A vector of ColDataType defining the 
-	 * parameters of the UDA(n)F call. These can be used to decide 
-	 * to override the default return type. If desired, the new 
-	 * return type can be set by context->setReturnType() and 
-	 * decimal precision can be set in context-> 
-	 * setResultDecimalCharacteristics. 
-	 *  
-	 * Return mcsv1_UDAF::ERROR on any error, such as non-compatible
-	 * colTypes or wrong number of arguments. Else return 
-	 * mcsv1_UDAF::SUCCESS. 
-	 */
-	virtual ReturnCode init(mcsv1Context* context, COL_TYPES& colTypes);
+    /**
+     * init()
+     *
+     * Mandatory. Implement this to initialize flags and instance
+     * data. Called once per SQL statement. You can do any sanity
+     * checks here.
+     *
+     * colTypes (in) - A vector of ColDataType defining the
+     * parameters of the UDA(n)F call. These can be used to decide
+     * to override the default return type. If desired, the new
+     * return type can be set by context->setReturnType() and
+     * decimal precision can be set in context->
+     * setResultDecimalCharacteristics.
+     *
+     * Return mcsv1_UDAF::ERROR on any error, such as non-compatible
+     * colTypes or wrong number of arguments. Else return
+     * mcsv1_UDAF::SUCCESS.
+     */
+    virtual ReturnCode init(mcsv1Context* context, COL_TYPES& colTypes);
 
-	/** 
-	 * reset() 
-	 *  
-	 * Mandatory. Reset the UDA(n)F for a new group, partition or, 
-	 * in some cases, new Window Frame. Do not free any memory 
-	 * allocated by context->setUserDataSize(). The SDK Framework owns 
-	 * that memory and will handle that. Use this opportunity to 
-	 * reset any variables in context->getUserData() needed for the 
-	 * next aggregation. May be called multiple times if running in 
-	 * a ditributed fashion. 
-	 *  
-	 * Use this opportunity to initialize the userData.
-	 */
-	virtual ReturnCode reset(mcsv1Context* context);
+    /**
+     * reset()
+     *
+     * Mandatory. Reset the UDA(n)F for a new group, partition or,
+     * in some cases, new Window Frame. Do not free any memory
+     * allocated by context->setUserDataSize(). The SDK Framework owns
+     * that memory and will handle that. Use this opportunity to
+     * reset any variables in context->getUserData() needed for the
+     * next aggregation. May be called multiple times if running in
+     * a ditributed fashion.
+     *
+     * Use this opportunity to initialize the userData.
+     */
+    virtual ReturnCode reset(mcsv1Context* context);
 
-	/** 
-	 * nextValue() 
-	 *  
-	 * Mandatory. Handle a single row. 
-	 *  
-	 * colsIn - A vector of data structure describing the input 
-	 * data. 
-	 *  
-	 * This function is called once for every row in the filtered 
-	 * result set (before aggregation). It is very important that 
-	 * this function is efficient. 
-	 *  
-	 * If the UDAF is running in a distributed fashion, nextValue 
-	 * cannot depend on order, as it will only be called for each 
-	 * row found on the specific PM. 
-	 *  
-	 * valsIn (in) - a vector of the parameters from the row.
-	 */
-	virtual ReturnCode nextValue(mcsv1Context* context, std::vector<ColumnDatum>& valsIn);
+    /**
+     * nextValue()
+     *
+     * Mandatory. Handle a single row.
+     *
+     * colsIn - A vector of data structure describing the input
+     * data.
+     *
+     * This function is called once for every row in the filtered
+     * result set (before aggregation). It is very important that
+     * this function is efficient.
+     *
+     * If the UDAF is running in a distributed fashion, nextValue
+     * cannot depend on order, as it will only be called for each
+     * row found on the specific PM.
+     *
+     * valsIn (in) - a vector of the parameters from the row.
+     */
+    virtual ReturnCode nextValue(mcsv1Context* context, std::vector<ColumnDatum>& valsIn);
 
-	 /** 
-	  * subEvaluate() 
-	  *  
-	  * Mandatory -- Called if the UDAF is running in a distributed 
-	  * fashion. Columnstore tries to run all aggregate functions 
-	  * distributed, depending on context. 
-	  *  
-	  * Perform an aggregation on rows partially aggregated by 
-	  * nextValue. Columnstore calls nextValue for each row on a 
-	  * given PM for a group (GROUP BY). subEvaluate is called on the
-	  * UM to consolodate those values into a single instance of 
-	  * userData. Keep your aggregated totals in context's userData. 
-	  * The first time this is called for a group, reset() would have 
-	  * been called with this version of userData. 
-	  *  
-	  * Called for every partial data set in each group in GROUP BY.
-	  *  
-	  * When subEvaluate has been called for all subAggregated data 
-	  * sets, Evaluate will be called with the same context as here.
-	  *  
-	  * valIn (In) - This is a pointer to a memory block of the size 
-	  * set in setUserDataSize. It will contain the value of userData 
-	  * as seen in the last call to NextValue for a given PM.
-	  *  
-	  */
-	 virtual ReturnCode subEvaluate(mcsv1Context* context, const UserData* userDataIn);
+    /**
+     * subEvaluate()
+     *
+     * Mandatory -- Called if the UDAF is running in a distributed
+     * fashion. Columnstore tries to run all aggregate functions
+     * distributed, depending on context.
+     *
+     * Perform an aggregation on rows partially aggregated by
+     * nextValue. Columnstore calls nextValue for each row on a
+     * given PM for a group (GROUP BY). subEvaluate is called on the
+     * UM to consolodate those values into a single instance of
+     * userData. Keep your aggregated totals in context's userData.
+     * The first time this is called for a group, reset() would have
+     * been called with this version of userData.
+     *
+     * Called for every partial data set in each group in GROUP BY.
+     *
+     * When subEvaluate has been called for all subAggregated data
+     * sets, Evaluate will be called with the same context as here.
+     *
+     * valIn (In) - This is a pointer to a memory block of the size
+     * set in setUserDataSize. It will contain the value of userData
+     * as seen in the last call to NextValue for a given PM.
+     *
+     */
+    virtual ReturnCode subEvaluate(mcsv1Context* context, const UserData* userDataIn);
 
-	/** 
-	 * evaluate() 
-	 *  
-	 * Mandatory. Get the aggregated value.
-	 *  
-	 * Called for every new group if UDAF GROUP BY, UDAnF partition 
-	 * or, in some cases, new Window Frame. 
-	 *  
-	 * Set the aggregated value into valOut. The datatype is assumed
-	 * to be the same as that set in the init() function; 
-	 *  
-	 * If the UDAF is running in a distributed fashion, evaluate is 
-	 * called after a series of subEvaluate calls. 
-	 *  
-	 * valOut (out) - Set the aggregated value here. The datatype is
-	 * assumed to be the same as that set in the init() function; 
-	 *  
-	 * To return a NULL value, don't assign to valOut.
-	 */
-	virtual ReturnCode evaluate(mcsv1Context* context, static_any::any& valOut);
+    /**
+     * evaluate()
+     *
+     * Mandatory. Get the aggregated value.
+     *
+     * Called for every new group if UDAF GROUP BY, UDAnF partition
+     * or, in some cases, new Window Frame.
+     *
+     * Set the aggregated value into valOut. The datatype is assumed
+     * to be the same as that set in the init() function;
+     *
+     * If the UDAF is running in a distributed fashion, evaluate is
+     * called after a series of subEvaluate calls.
+     *
+     * valOut (out) - Set the aggregated value here. The datatype is
+     * assumed to be the same as that set in the init() function;
+     *
+     * To return a NULL value, don't assign to valOut.
+     */
+    virtual ReturnCode evaluate(mcsv1Context* context, static_any::any& valOut);
 
 protected:
 

utils/udfsdk/avg_mode.cpp

@@ -25,269 +25,282 @@
 using namespace mcsv1sdk;
 
 mcsv1_UDAF::ReturnCode avg_mode::init(mcsv1Context* context,
-									 COL_TYPES& colTypes)
+                                      COL_TYPES& colTypes)
 {
-	if (colTypes.size() < 1)
-	{
-		// The error message will be prepended with
-		// "The storage engine for the table doesn't support "
-		context->setErrorMessage("avg_mode() with 0 arguments");
-		return mcsv1_UDAF::ERROR;
-	}
-	if (colTypes.size() > 1)
-	{
-		context->setErrorMessage("avg_mode() with more than 1 argument");
-		return mcsv1_UDAF::ERROR;
-	}
+    if (colTypes.size() < 1)
+    {
+        // The error message will be prepended with
+        // "The storage engine for the table doesn't support "
+        context->setErrorMessage("avg_mode() with 0 arguments");
+        return mcsv1_UDAF::ERROR;
+    }
 
-	if (!(isNumeric(colTypes[0].second)))
-	{
-		// The error message will be prepended with
-		// "The storage engine for the table doesn't support "
-		context->setErrorMessage("avg_mode() with non-numeric argument");
-		return mcsv1_UDAF::ERROR;
-	}
+    if (colTypes.size() > 1)
+    {
+        context->setErrorMessage("avg_mode() with more than 1 argument");
+        return mcsv1_UDAF::ERROR;
+    }
 
-	context->setResultType(CalpontSystemCatalog::DOUBLE);
-	context->setColWidth(8);
-	context->setScale(context->getScale()*2);
-	context->setPrecision(19);
-	context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
-	return mcsv1_UDAF::SUCCESS;
+    if (!(isNumeric(colTypes[0].second)))
+    {
+        // The error message will be prepended with
+        // "The storage engine for the table doesn't support "
+        context->setErrorMessage("avg_mode() with non-numeric argument");
+        return mcsv1_UDAF::ERROR;
+    }
+
+    context->setResultType(CalpontSystemCatalog::DOUBLE);
+    context->setColWidth(8);
+    context->setScale(context->getScale() * 2);
+    context->setPrecision(19);
+    context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
+    return mcsv1_UDAF::SUCCESS;
 
 }
 
 mcsv1_UDAF::ReturnCode avg_mode::reset(mcsv1Context* context)
 {
-	ModeData* data = static_cast<ModeData*>(context->getUserData());
-	data->mData.clear();
-	return mcsv1_UDAF::SUCCESS;
+    ModeData* data = static_cast<ModeData*>(context->getUserData());
+    data->mData.clear();
+    return mcsv1_UDAF::SUCCESS;
 }
 
-mcsv1_UDAF::ReturnCode avg_mode::nextValue(mcsv1Context* context, 
-										  std::vector<ColumnDatum>& valsIn)
+mcsv1_UDAF::ReturnCode avg_mode::nextValue(mcsv1Context* context,
+        std::vector<ColumnDatum>& valsIn)
 {
-	static_any::any& valIn = valsIn[0].columnData;
-	MODE_DATA& data = static_cast<ModeData*>(context->getUserData())->mData;
-	DATATYPE val = 0.0;
+    static_any::any& valIn = valsIn[0].columnData;
+    MODE_DATA& data = static_cast<ModeData*>(context->getUserData())->mData;
+    DATATYPE val = 0.0;
 
-	if (valIn.empty())
-	{
-		return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
-	}
+    if (valIn.empty())
+    {
+        return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
+    }
 
-	if (valIn.compatible(charTypeId))
-	{
-		val = valIn.cast<char>();
-	}
-	else if (valIn.compatible(scharTypeId))
-	{
-		val = valIn.cast<signed char>();
-	}
-	else if (valIn.compatible(shortTypeId))
-	{
-		val = valIn.cast<short>();
-	}
-	else if (valIn.compatible(intTypeId))
-	{
-		val = valIn.cast<int>();
-	}
-	else if (valIn.compatible(longTypeId))
-	{
-		val = valIn.cast<long>();
-	}
-	else if (valIn.compatible(llTypeId))
-	{
-		val = valIn.cast<long long>();
-	}
-	else if (valIn.compatible(ucharTypeId))
-	{
-		val = valIn.cast<unsigned char>();
-	}
-	else if (valIn.compatible(ushortTypeId))
-	{
-		val = valIn.cast<unsigned short>();
-	}
-	else if (valIn.compatible(uintTypeId))
-	{
-		val = valIn.cast<unsigned int>();
-	}
-	else if (valIn.compatible(ulongTypeId))
-	{
-		val = valIn.cast<unsigned long>();
-	}
-	else if (valIn.compatible(ullTypeId))
-	{
-		val = valIn.cast<unsigned long long>();
-	}
-	else if (valIn.compatible(floatTypeId))
-	{
-		val = valIn.cast<float>();
-	}
-	else if (valIn.compatible(doubleTypeId))
-	{
-		val = valIn.cast<double>();
-	}
+    if (valIn.compatible(charTypeId))
+    {
+        val = valIn.cast<char>();
+    }
+    else if (valIn.compatible(scharTypeId))
+    {
+        val = valIn.cast<signed char>();
+    }
+    else if (valIn.compatible(shortTypeId))
+    {
+        val = valIn.cast<short>();
+    }
+    else if (valIn.compatible(intTypeId))
+    {
+        val = valIn.cast<int>();
+    }
+    else if (valIn.compatible(longTypeId))
+    {
+        val = valIn.cast<long>();
+    }
+    else if (valIn.compatible(llTypeId))
+    {
+        val = valIn.cast<long long>();
+    }
+    else if (valIn.compatible(ucharTypeId))
+    {
+        val = valIn.cast<unsigned char>();
+    }
+    else if (valIn.compatible(ushortTypeId))
+    {
+        val = valIn.cast<unsigned short>();
+    }
+    else if (valIn.compatible(uintTypeId))
+    {
+        val = valIn.cast<unsigned int>();
+    }
+    else if (valIn.compatible(ulongTypeId))
+    {
+        val = valIn.cast<unsigned long>();
+    }
+    else if (valIn.compatible(ullTypeId))
+    {
+        val = valIn.cast<unsigned long long>();
+    }
+    else if (valIn.compatible(floatTypeId))
+    {
+        val = valIn.cast<float>();
+    }
+    else if (valIn.compatible(doubleTypeId))
+    {
+        val = valIn.cast<double>();
+    }
 
-	// For decimal types, we need to move the decimal point.
-	uint32_t scale = valsIn[0].scale;
-	if (val != 0 && scale > 0)
-	{
-		val /= pow(10.0, (double)scale);
-	}
-	data[val]++;
+    // For decimal types, we need to move the decimal point.
+    uint32_t scale = valsIn[0].scale;
 
-	return mcsv1_UDAF::SUCCESS;
+    if (val != 0 && scale > 0)
+    {
+        val /= pow(10.0, (double)scale);
+    }
+
+    data[val]++;
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode avg_mode::subEvaluate(mcsv1Context* context, const UserData* userDataIn)
 {
-	if (!userDataIn)
-	{
-		return mcsv1_UDAF::SUCCESS;
-	}
-	MODE_DATA& outData = static_cast<ModeData*>(context->getUserData())->mData;
-	const MODE_DATA& inData = static_cast<const ModeData*>(userDataIn)->mData;
-	MODE_DATA::const_iterator iter = inData.begin();
-	for (; iter != inData.end(); ++iter)
-	{
-		outData[iter->first] += iter->second;
-	}
-	return mcsv1_UDAF::SUCCESS;
+    if (!userDataIn)
+    {
+        return mcsv1_UDAF::SUCCESS;
+    }
+
+    MODE_DATA& outData = static_cast<ModeData*>(context->getUserData())->mData;
+    const MODE_DATA& inData = static_cast<const ModeData*>(userDataIn)->mData;
+    MODE_DATA::const_iterator iter = inData.begin();
+
+    for (; iter != inData.end(); ++iter)
+    {
+        outData[iter->first] += iter->second;
+    }
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode avg_mode::evaluate(mcsv1Context* context, static_any::any& valOut)
 {
-	uint64_t maxCnt=0;
-	MODE_DATA& data = static_cast<ModeData*>(context->getUserData())->mData;
-	if (data.size() == 0)
-	{
-		valOut = (DATATYPE)0;
-		return mcsv1_UDAF::SUCCESS;
-	}
-	MODE_DATA::iterator iter(data.begin());
-	for (; iter != data.end(); ++iter)
-	{
-		if (iter->second > maxCnt)
-		{
-			valOut = iter->first;
-			maxCnt = iter->second;
-		}
-	}
-	return mcsv1_UDAF::SUCCESS;
+    uint64_t maxCnt = 0;
+    MODE_DATA& data = static_cast<ModeData*>(context->getUserData())->mData;
+
+    if (data.size() == 0)
+    {
+        valOut = (DATATYPE)0;
+        return mcsv1_UDAF::SUCCESS;
+    }
+
+    MODE_DATA::iterator iter(data.begin());
+
+    for (; iter != data.end(); ++iter)
+    {
+        if (iter->second > maxCnt)
+        {
+            valOut = iter->first;
+            maxCnt = iter->second;
+        }
+    }
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
-mcsv1_UDAF::ReturnCode avg_mode::dropValue(mcsv1Context* context, 
-													std::vector<ColumnDatum>& valsDropped)
+mcsv1_UDAF::ReturnCode avg_mode::dropValue(mcsv1Context* context,
+        std::vector<ColumnDatum>& valsDropped)
 {
-	static_any::any& valIn = valsDropped[0].columnData;
-	MODE_DATA& data = static_cast<ModeData*>(context->getUserData())->mData;
-	DATATYPE val = 0.0;
+    static_any::any& valIn = valsDropped[0].columnData;
+    MODE_DATA& data = static_cast<ModeData*>(context->getUserData())->mData;
+    DATATYPE val = 0.0;
 
-	if (valIn.empty())
-	{
-		return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
-	}
+    if (valIn.empty())
+    {
+        return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
+    }
 
-	if (valIn.compatible(charTypeId))
-	{
-		val = valIn.cast<char>();
-	}
-	else if (valIn.compatible(scharTypeId))
-	{
-		val = valIn.cast<signed char>();
-	}
-	else if (valIn.compatible(shortTypeId))
-	{
-		val = valIn.cast<short>();
-	}
-	else if (valIn.compatible(intTypeId))
-	{
-		val = valIn.cast<int>();
-	}
-	else if (valIn.compatible(longTypeId))
-	{
-		val = valIn.cast<long>();
-	}
-	else if (valIn.compatible(llTypeId))
-	{
-		val = valIn.cast<long long>();
-	}
-	else if (valIn.compatible(ucharTypeId))
-	{
-		val = valIn.cast<unsigned char>();
-	}
-	else if (valIn.compatible(ushortTypeId))
-	{
-		val = valIn.cast<unsigned short>();
-	}
-	else if (valIn.compatible(uintTypeId))
-	{
-		val = valIn.cast<unsigned int>();
-	}
-	else if (valIn.compatible(ulongTypeId))
-	{
-		val = valIn.cast<unsigned long>();
-	}
-	else if (valIn.compatible(ullTypeId))
-	{
-		val = valIn.cast<unsigned long long>();
-	}
-	else if (valIn.compatible(floatTypeId))
-	{
-		val = valIn.cast<float>();
-	}
-	else if (valIn.compatible(doubleTypeId))
-	{
-		val = valIn.cast<double>();
-	}
+    if (valIn.compatible(charTypeId))
+    {
+        val = valIn.cast<char>();
+    }
+    else if (valIn.compatible(scharTypeId))
+    {
+        val = valIn.cast<signed char>();
+    }
+    else if (valIn.compatible(shortTypeId))
+    {
+        val = valIn.cast<short>();
+    }
+    else if (valIn.compatible(intTypeId))
+    {
+        val = valIn.cast<int>();
+    }
+    else if (valIn.compatible(longTypeId))
+    {
+        val = valIn.cast<long>();
+    }
+    else if (valIn.compatible(llTypeId))
+    {
+        val = valIn.cast<long long>();
+    }
+    else if (valIn.compatible(ucharTypeId))
+    {
+        val = valIn.cast<unsigned char>();
+    }
+    else if (valIn.compatible(ushortTypeId))
+    {
+        val = valIn.cast<unsigned short>();
+    }
+    else if (valIn.compatible(uintTypeId))
+    {
+        val = valIn.cast<unsigned int>();
+    }
+    else if (valIn.compatible(ulongTypeId))
+    {
+        val = valIn.cast<unsigned long>();
+    }
+    else if (valIn.compatible(ullTypeId))
+    {
+        val = valIn.cast<unsigned long long>();
+    }
+    else if (valIn.compatible(floatTypeId))
+    {
+        val = valIn.cast<float>();
+    }
+    else if (valIn.compatible(doubleTypeId))
+    {
+        val = valIn.cast<double>();
+    }
 
-	// For decimal types, we need to move the decimal point.
-	uint32_t scale = valsDropped[0].scale;
-	if (val != 0 && scale > 0)
-	{
-		val /= pow(10.0, (double)scale);
-	}
+    // For decimal types, we need to move the decimal point.
+    uint32_t scale = valsDropped[0].scale;
 
-	data[val]--;
+    if (val != 0 && scale > 0)
+    {
+        val /= pow(10.0, (double)scale);
+    }
 
-	return mcsv1_UDAF::SUCCESS;
+    data[val]--;
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode avg_mode::createUserData(UserData*& userData, int32_t& length)
 {
-	userData = new ModeData;
-	length = sizeof(ModeData);
-	return mcsv1_UDAF::SUCCESS;
+    userData = new ModeData;
+    length = sizeof(ModeData);
+    return mcsv1_UDAF::SUCCESS;
 }
 
 void ModeData::serialize(messageqcpp::ByteStream& bs) const
 {
-	MODE_DATA::const_iterator iter = mData.begin();
-	DATATYPE num;
-	uint32_t cnt;
-	bs << (int32_t)mData.size();
-	for (; iter != mData.end(); ++iter)
-	{
-		num = iter->first;
-		bs << num;
-		cnt = iter->second;
-		bs << cnt;
-	}
+    MODE_DATA::const_iterator iter = mData.begin();
+    DATATYPE num;
+    uint32_t cnt;
+    bs << (int32_t)mData.size();
+
+    for (; iter != mData.end(); ++iter)
+    {
+        num = iter->first;
+        bs << num;
+        cnt = iter->second;
+        bs << cnt;
+    }
 }
 
 void ModeData::unserialize(messageqcpp::ByteStream& bs)
 {
-	mData.clear();
-	int32_t sz;
-	DATATYPE num;
-	uint32_t cnt;
-	bs >> sz;
-	for (int i = 0; i < sz; ++i)
-	{
-		bs >> num;
-		bs >> cnt;
-		mData[num] = cnt;
-	}
+    mData.clear();
+    int32_t sz;
+    DATATYPE num;
+    uint32_t cnt;
+    bs >> sz;
+
+    for (int i = 0; i < sz; ++i)
+    {
+        bs >> num;
+        bs >> cnt;
+        mData[num] = cnt;
+    }
 }
 

utils/udfsdk/avg_mode.h

@@ -21,34 +21,34 @@
 *   mcsv1_UDAF.h
 ***********************************************************************/
 
-/** 
- * Columnstore interface for writing a User Defined Aggregate 
- * Functions (UDAF) and User Defined Analytic Functions (UDAnF) 
- * or a function that can act as either - UDA(n)F 
- *  
+/**
+ * Columnstore interface for writing a User Defined Aggregate
+ * Functions (UDAF) and User Defined Analytic Functions (UDAnF)
+ * or a function that can act as either - UDA(n)F
+ *
  * The basic steps are:
  *
- * 1. Create a the UDA(n)F function interface in some .h file. 
- * 2. Create the UDF function implementation in some .cpp file 
- * 3. Create the connector stub (MariaDB UDAF definition) for 
- * this UDF function.  
- * 4. build the dynamic library using all of the source. 
- * 5  Put the library in $COLUMNSTORE_INSTALL/lib of 
- * all modules 
- * 6. restart the Columnstore system. 
+ * 1. Create a the UDA(n)F function interface in some .h file.
+ * 2. Create the UDF function implementation in some .cpp file
+ * 3. Create the connector stub (MariaDB UDAF definition) for
+ * this UDF function.
+ * 4. build the dynamic library using all of the source.
+ * 5  Put the library in $COLUMNSTORE_INSTALL/lib of
+ * all modules
+ * 6. restart the Columnstore system.
  * 7. notify mysqld about the new function:
- *  
+ *
  *    CREATE AGGREGATE FUNCTION avg_mode returns REAL soname
  *    'libudf_mysql.so';
- *  
- * The UDAF functions may run distributed in the Columnstore 
- * engine. UDAnF do not run distributed. 
- *  
- * UDAF is User Defined Aggregate Function. 
- * UDAnF is User Defined Analytic Function. 
- * UDA(n)F is an acronym for a function that could be either. It 
- * is also used to describe the interface that is used for 
- * either. 
+ *
+ * The UDAF functions may run distributed in the Columnstore
+ * engine. UDAnF do not run distributed.
+ *
+ * UDAF is User Defined Aggregate Function.
+ * UDAnF is User Defined Analytic Function.
+ * UDA(n)F is an acronym for a function that could be either. It
+ * is also used to describe the interface that is used for
+ * either.
  */
 #ifndef HEADER_mode
 #define HEADER_mode
@@ -83,196 +83,196 @@ typedef std::tr1::unordered_map<DATATYPE, uint32_t> MODE_DATA;
 // Override UserData for data storage
 struct ModeData : public UserData
 {
-	ModeData() {};
+    ModeData() {};
 
-	virtual ~ModeData(){}
+    virtual ~ModeData() {}
 
-	virtual void serialize(messageqcpp::ByteStream& bs) const;
-	virtual void unserialize(messageqcpp::ByteStream& bs);
+    virtual void serialize(messageqcpp::ByteStream& bs) const;
+    virtual void unserialize(messageqcpp::ByteStream& bs);
 
-	MODE_DATA mData;
+    MODE_DATA mData;
 private:
-	// For now, copy construction is unwanted
-	ModeData(UserData&);
+    // For now, copy construction is unwanted
+    ModeData(UserData&);
 };
 
-// Override mcsv1_UDAF to build your User Defined Aggregate (UDAF) and/or 
+// Override mcsv1_UDAF to build your User Defined Aggregate (UDAF) and/or
 // User Defined Analytic Function (UDAnF).
 // These will be singleton classes, so don't put any instance
 // specific data in here. All instance data is stored in mcsv1Context
 // passed to each user function and retrieved by the getUserData() method.
-// 
-// Each API function returns a ReturnCode. If ERROR is returned at any time, 
-// the query is aborted, getInterrupted() will begin to return true and the 
-// message set in config->setErrorMessage() is returned to MariaDB. 
+//
+// Each API function returns a ReturnCode. If ERROR is returned at any time,
+// the query is aborted, getInterrupted() will begin to return true and the
+// message set in config->setErrorMessage() is returned to MariaDB.
 
 // Return the avg_mode value of the dataset
 
 class avg_mode : public  mcsv1_UDAF
 {
 public:
-	// Defaults OK
-	avg_mode() : mcsv1_UDAF(){};
-	virtual ~avg_mode(){};
+    // Defaults OK
+    avg_mode() : mcsv1_UDAF() {};
+    virtual ~avg_mode() {};
 
-	/** 
-	 * init() 
-	 *  
-	 * Mandatory. Implement this to initialize flags and instance 
-	 * data. Called once per SQL statement. You can do any sanity 
-	 * checks here. 
-	 *  
-	 * colTypes (in) - A vector of ColDataType defining the 
-	 * parameters of the UDA(n)F call. These can be used to decide 
-	 * to override the default return type. If desired, the new 
-	 * return type can be set by context->setReturnType() and 
-	 * decimal scale and precision can be set by context->setScale 
-	 * and context->setPrecision respectively. 
-	 *  
-	 * Return mcsv1_UDAF::ERROR on any error, such as non-compatible
-	 * colTypes or wrong number of arguments. Else return 
-	 * mcsv1_UDAF::SUCCESS. 
-	 */
-	virtual ReturnCode init(mcsv1Context* context,
-							COL_TYPES& colTypes);
+    /**
+     * init()
+     *
+     * Mandatory. Implement this to initialize flags and instance
+     * data. Called once per SQL statement. You can do any sanity
+     * checks here.
+     *
+     * colTypes (in) - A vector of ColDataType defining the
+     * parameters of the UDA(n)F call. These can be used to decide
+     * to override the default return type. If desired, the new
+     * return type can be set by context->setReturnType() and
+     * decimal scale and precision can be set by context->setScale
+     * and context->setPrecision respectively.
+     *
+     * Return mcsv1_UDAF::ERROR on any error, such as non-compatible
+     * colTypes or wrong number of arguments. Else return
+     * mcsv1_UDAF::SUCCESS.
+     */
+    virtual ReturnCode init(mcsv1Context* context,
+                            COL_TYPES& colTypes);
 
-	/** 
-	 * reset() 
-	 *  
-	 * Mandatory. Reset the UDA(n)F for a new group, partition or, 
-	 * in some cases, new Window Frame. Do not free any memory 
-	 * allocated by context->setUserDataSize(). The SDK Framework owns 
-	 * that memory and will handle that. Use this opportunity to 
-	 * reset any variables in context->getUserData() needed for the 
-	 * next aggregation. May be called multiple times if running in 
-	 * a ditributed fashion. 
-	 *  
-	 * Use this opportunity to initialize the userData.
-	 */
-	virtual ReturnCode reset(mcsv1Context* context);
+    /**
+     * reset()
+     *
+     * Mandatory. Reset the UDA(n)F for a new group, partition or,
+     * in some cases, new Window Frame. Do not free any memory
+     * allocated by context->setUserDataSize(). The SDK Framework owns
+     * that memory and will handle that. Use this opportunity to
+     * reset any variables in context->getUserData() needed for the
+     * next aggregation. May be called multiple times if running in
+     * a ditributed fashion.
+     *
+     * Use this opportunity to initialize the userData.
+     */
+    virtual ReturnCode reset(mcsv1Context* context);
 
-	/** 
-	 * nextValue() 
-	 *  
-	 * Mandatory. Handle a single row. 
-	 *  
-	 * colsIn - A vector of data structure describing the input 
-	 * data. 
-	 *  
-	 * This function is called once for every row in the filtered 
-	 * result set (before aggregation). It is very important that 
-	 * this function is efficient. 
-	 *  
-	 * If the UDAF is running in a distributed fashion, nextValue 
-	 * cannot depend on order, as it will only be called for each 
-	 * row found on the specific PM. 
-	 *  
-	 * valsIn (in) - a vector of the parameters from the row.
-	 */
-	virtual ReturnCode nextValue(mcsv1Context* context, 
-								 std::vector<ColumnDatum>& valsIn);
+    /**
+     * nextValue()
+     *
+     * Mandatory. Handle a single row.
+     *
+     * colsIn - A vector of data structure describing the input
+     * data.
+     *
+     * This function is called once for every row in the filtered
+     * result set (before aggregation). It is very important that
+     * this function is efficient.
+     *
+     * If the UDAF is running in a distributed fashion, nextValue
+     * cannot depend on order, as it will only be called for each
+     * row found on the specific PM.
+     *
+     * valsIn (in) - a vector of the parameters from the row.
+     */
+    virtual ReturnCode nextValue(mcsv1Context* context,
+                                 std::vector<ColumnDatum>& valsIn);
 
-	 /** 
-	  * subEvaluate() 
-	  *  
-	  * Mandatory -- Called if the UDAF is running in a distributed 
-	  * fashion. Columnstore tries to run all aggregate functions 
-	  * distributed, depending on context. 
-	  *  
-	  * Perform an aggregation on rows partially aggregated by 
-	  * nextValue. Columnstore calls nextValue for each row on a 
-	  * given PM for a group (GROUP BY). subEvaluate is called on the
-	  * UM to consolodate those values into a single instance of 
-	  * userData. Keep your aggregated totals in context's userData. 
-	  * The first time this is called for a group, reset() would have 
-	  * been called with this version of userData. 
-	  *  
-	  * Called for every partial data set in each group in GROUP BY.
-	  *  
-	  * When subEvaluate has been called for all subAggregated data 
-	  * sets, Evaluate will be called with the same context as here.
-	  *  
-	  * valIn (In) - This is a pointer to a memory block of the size 
-	  * set in setUserDataSize. It will contain the value of userData 
-	  * as seen in the last call to NextValue for a given PM.
-	  *  
-	  */
-	 virtual ReturnCode subEvaluate(mcsv1Context* context, const UserData* valIn);
+    /**
+     * subEvaluate()
+     *
+     * Mandatory -- Called if the UDAF is running in a distributed
+     * fashion. Columnstore tries to run all aggregate functions
+     * distributed, depending on context.
+     *
+     * Perform an aggregation on rows partially aggregated by
+     * nextValue. Columnstore calls nextValue for each row on a
+     * given PM for a group (GROUP BY). subEvaluate is called on the
+     * UM to consolodate those values into a single instance of
+     * userData. Keep your aggregated totals in context's userData.
+     * The first time this is called for a group, reset() would have
+     * been called with this version of userData.
+     *
+     * Called for every partial data set in each group in GROUP BY.
+     *
+     * When subEvaluate has been called for all subAggregated data
+     * sets, Evaluate will be called with the same context as here.
+     *
+     * valIn (In) - This is a pointer to a memory block of the size
+     * set in setUserDataSize. It will contain the value of userData
+     * as seen in the last call to NextValue for a given PM.
+     *
+     */
+    virtual ReturnCode subEvaluate(mcsv1Context* context, const UserData* valIn);
 
-	/** 
-	 * evaluate() 
-	 *  
-	 * Mandatory. Get the aggregated value.
-	 *  
-	 * Called for every new group if UDAF GROUP BY, UDAnF partition 
-	 * or, in some cases, new Window Frame. 
-	 *  
-	 * Set the aggregated value into valOut. The datatype is assumed
-	 * to be the same as that set in the init() function; 
-	 *  
-	 * If the UDAF is running in a distributed fashion, evaluate is 
-	 * called after a series of subEvaluate calls. 
-	 *  
-	 * valOut (out) - Set the aggregated value here. The datatype is
-	 * assumed to be the same as that set in the init() function; 
-	 *  
-	 * To return a NULL value, don't assign to valOut.
-	 */
-	virtual ReturnCode evaluate(mcsv1Context* context, static_any::any& valOut);
+    /**
+     * evaluate()
+     *
+     * Mandatory. Get the aggregated value.
+     *
+     * Called for every new group if UDAF GROUP BY, UDAnF partition
+     * or, in some cases, new Window Frame.
+     *
+     * Set the aggregated value into valOut. The datatype is assumed
+     * to be the same as that set in the init() function;
+     *
+     * If the UDAF is running in a distributed fashion, evaluate is
+     * called after a series of subEvaluate calls.
+     *
+     * valOut (out) - Set the aggregated value here. The datatype is
+     * assumed to be the same as that set in the init() function;
+     *
+     * To return a NULL value, don't assign to valOut.
+     */
+    virtual ReturnCode evaluate(mcsv1Context* context, static_any::any& valOut);
 
-	/** 
-	 * dropValue() 
-	 *  
-	 * Optional -- If defined, the server will call this instead of 
-	 * reset for UDAnF. 
-	 *  
-	 * Don't implement if a UDAnF has one or more of the following: 
-	 * The UDAnF can't be used with a Window Frame
-	 * The UDAnF is not reversable in some way 
-	 * The UDAnF is not interested in optimal performance 
-	 *  
-	 * If not implemented, reset() followed by a series of 
-	 * nextValue() will be called for each movement of the Window 
-	 * Frame. 
-	 *  
-	 * If implemented, then each movement of the Window Frame will 
-	 * result in dropValue() being called for each row falling out 
-	 * of the Frame and nextValue() being called for each new row 
-	 * coming into the Frame. 
-	 *  
-	 * valsDropped (in) - a vector of the parameters from the row 
-	 * leaving the Frame 
-	 *  
-	 * dropValue() will not be called for unbounded/current row type
-	 * frames, as those are already optimized. 
-	 */
-	virtual ReturnCode dropValue(mcsv1Context* context, 
-								 std::vector<ColumnDatum>& valsDropped);
+    /**
+     * dropValue()
+     *
+     * Optional -- If defined, the server will call this instead of
+     * reset for UDAnF.
+     *
+     * Don't implement if a UDAnF has one or more of the following:
+     * The UDAnF can't be used with a Window Frame
+     * The UDAnF is not reversable in some way
+     * The UDAnF is not interested in optimal performance
+     *
+     * If not implemented, reset() followed by a series of
+     * nextValue() will be called for each movement of the Window
+     * Frame.
+     *
+     * If implemented, then each movement of the Window Frame will
+     * result in dropValue() being called for each row falling out
+     * of the Frame and nextValue() being called for each new row
+     * coming into the Frame.
+     *
+     * valsDropped (in) - a vector of the parameters from the row
+     * leaving the Frame
+     *
+     * dropValue() will not be called for unbounded/current row type
+     * frames, as those are already optimized.
+     */
+    virtual ReturnCode dropValue(mcsv1Context* context,
+                                 std::vector<ColumnDatum>& valsDropped);
 
-	/** 
-	 * createUserData()
-	 *  
-	 * Optional -- If defined, the server will call this instead of 
-	 * createUserData on context. 
-	 *  
-	 * Create your variable length data structure via 
-	 * data = new <datatype> 
-	 *  
-	 * The data structure may contain references to containers or 
-	 * pointers to other objects. Remember that for distributed 
-	 * processing, this may be called multiple times for variaous 
-	 * computing blocks. At the least, it will be called once per PM 
-	 * that processes the data, and once more for the UM. For UDAnF, 
-	 * it may only be called once. 
-	 *  
-	 * Set length to the length of the data structure you create.
-	 *  
-	 * For each call to createUserData(), there will be a 
-	 * corresponding deleteUserData() where you must clean up. Any 
-	 * memory leaks are your fault. 
-	 *  
-	 */ 
-	virtual ReturnCode createUserData(UserData*& data, int32_t& length);
+    /**
+     * createUserData()
+     *
+     * Optional -- If defined, the server will call this instead of
+     * createUserData on context.
+     *
+     * Create your variable length data structure via
+     * data = new <datatype>
+     *
+     * The data structure may contain references to containers or
+     * pointers to other objects. Remember that for distributed
+     * processing, this may be called multiple times for variaous
+     * computing blocks. At the least, it will be called once per PM
+     * that processes the data, and once more for the UM. For UDAnF,
+     * it may only be called once.
+     *
+     * Set length to the length of the data structure you create.
+     *
+     * For each call to createUserData(), there will be a
+     * corresponding deleteUserData() where you must clean up. Any
+     * memory leaks are your fault.
+     *
+     */
+    virtual ReturnCode createUserData(UserData*& data, int32_t& length);
 protected:
 };
 

utils/udfsdk/mcsv1_udaf.cpp

@@ -25,11 +25,11 @@
 using namespace mcsv1sdk;
 /**
  * All UDA(n)F functions must be registered in the function map.
- * They will be picked up by the Columnstore modules during 
- * startup. 
- *  
- * This is a temporary kludge until we get the library loader 
- * task complete 
+ * They will be picked up by the Columnstore modules during
+ * startup.
+ *
+ * This is a temporary kludge until we get the library loader
+ * task complete
  */
 UDAF_MAP UDAFMap::fm;
 #include "allnull.h"
@@ -38,204 +38,218 @@ UDAF_MAP UDAFMap::fm;
 #include "avg_mode.h"
 UDAF_MAP& UDAFMap::getMap()
 {
-	if (fm.size() > 0)
-	{
-		return fm;
-	}
-	// first: function name
-	// second: Function pointer
-	// please use lower case for the function name. Because the names might be 
-	// case-insensitive in MySQL depending on the setting. In such case, 
-	// the function names passed to the interface is always in lower case.
-	fm["allnull"] = new allnull();
-	fm["ssq"] = new ssq();
-	fm["median"] = new median();
-	fm["avg_mode"] = new avg_mode();
-	
-	return fm;
+    if (fm.size() > 0)
+    {
+        return fm;
+    }
+
+    // first: function name
+    // second: Function pointer
+    // please use lower case for the function name. Because the names might be
+    // case-insensitive in MySQL depending on the setting. In such case,
+    // the function names passed to the interface is always in lower case.
+    fm["allnull"] = new allnull();
+    fm["ssq"] = new ssq();
+    fm["median"] = new median();
+    fm["avg_mode"] = new avg_mode();
+
+    return fm;
 }
 
 int32_t mcsv1Context::getColWidth()
 {
-	if (fColWidth > 0)
-	{
-		return fColWidth;
-	}
-	// JIT initialization for types that have a defined size.
-	switch (fResultType)
-	{
-	case CalpontSystemCatalog::BIT:
-	case CalpontSystemCatalog::TINYINT:
-	case CalpontSystemCatalog::UTINYINT:
-	case CalpontSystemCatalog::CHAR:
-		fColWidth = 1;
-		break;
-	case CalpontSystemCatalog::SMALLINT:
-	case CalpontSystemCatalog::USMALLINT:
-		fColWidth = 2;
-		break;
-	case CalpontSystemCatalog::MEDINT:
-	case CalpontSystemCatalog::INT:
-	case CalpontSystemCatalog::UMEDINT:
-	case CalpontSystemCatalog::UINT:
-	case CalpontSystemCatalog::FLOAT:
-	case CalpontSystemCatalog::UFLOAT:
-	case CalpontSystemCatalog::DATE:
-		fColWidth = 4;
-		break;
-	case CalpontSystemCatalog::BIGINT:
-	case CalpontSystemCatalog::UBIGINT:
-	case CalpontSystemCatalog::DECIMAL:
-	case CalpontSystemCatalog::UDECIMAL:
-	case CalpontSystemCatalog::DOUBLE:
-	case CalpontSystemCatalog::UDOUBLE:
-	case CalpontSystemCatalog::DATETIME:
-	case CalpontSystemCatalog::STRINT:
-		fColWidth = 8;
-		break;
-	case CalpontSystemCatalog::LONGDOUBLE:
-		fColWidth = sizeof(long double);
-		break;
-	default:
-		break;
-	}
-	return fColWidth;
+    if (fColWidth > 0)
+    {
+        return fColWidth;
+    }
+
+    // JIT initialization for types that have a defined size.
+    switch (fResultType)
+    {
+        case CalpontSystemCatalog::BIT:
+        case CalpontSystemCatalog::TINYINT:
+        case CalpontSystemCatalog::UTINYINT:
+        case CalpontSystemCatalog::CHAR:
+            fColWidth = 1;
+            break;
+
+        case CalpontSystemCatalog::SMALLINT:
+        case CalpontSystemCatalog::USMALLINT:
+            fColWidth = 2;
+            break;
+
+        case CalpontSystemCatalog::MEDINT:
+        case CalpontSystemCatalog::INT:
+        case CalpontSystemCatalog::UMEDINT:
+        case CalpontSystemCatalog::UINT:
+        case CalpontSystemCatalog::FLOAT:
+        case CalpontSystemCatalog::UFLOAT:
+        case CalpontSystemCatalog::DATE:
+            fColWidth = 4;
+            break;
+
+        case CalpontSystemCatalog::BIGINT:
+        case CalpontSystemCatalog::UBIGINT:
+        case CalpontSystemCatalog::DECIMAL:
+        case CalpontSystemCatalog::UDECIMAL:
+        case CalpontSystemCatalog::DOUBLE:
+        case CalpontSystemCatalog::UDOUBLE:
+        case CalpontSystemCatalog::DATETIME:
+        case CalpontSystemCatalog::STRINT:
+            fColWidth = 8;
+            break;
+
+        case CalpontSystemCatalog::LONGDOUBLE:
+            fColWidth = sizeof(long double);
+            break;
+
+        default:
+            break;
+    }
+
+    return fColWidth;
 }
 
 bool mcsv1Context::operator==(const mcsv1Context& c) const
 {
-	// We don't test the per row data fields. They don't determine
-	// if it's the same Context.
-	if (getName()        != c.getName()
-	 ||	fRunFlags        != c.fRunFlags
-	 || fContextFlags    != c.fContextFlags
-	 || fUserDataSize    != c.fUserDataSize
-	 || fResultType      != c.fResultType
-	 || fResultscale  != c.fResultscale
-	 || fResultPrecision != c.fResultPrecision
-	 || fStartFrame      != c.fStartFrame
-	 || fEndFrame        != c.fEndFrame
-	 || fStartConstant   != c.fStartConstant
-	 || fEndConstant     != c.fEndConstant)
-		return false;
-	return true;
+    // We don't test the per row data fields. They don't determine
+    // if it's the same Context.
+    if (getName()        != c.getName()
+            ||	fRunFlags        != c.fRunFlags
+            || fContextFlags    != c.fContextFlags
+            || fUserDataSize    != c.fUserDataSize
+            || fResultType      != c.fResultType
+            || fResultscale  != c.fResultscale
+            || fResultPrecision != c.fResultPrecision
+            || fStartFrame      != c.fStartFrame
+            || fEndFrame        != c.fEndFrame
+            || fStartConstant   != c.fStartConstant
+            || fEndConstant     != c.fEndConstant)
+        return false;
+
+    return true;
 }
 
 bool mcsv1Context::operator!=(const mcsv1Context& c) const
 {
-	return (!(*this == c));
+    return (!(*this == c));
 }
 
 const std::string mcsv1Context::toString() const
 {
-	std::ostringstream output;
-	output << "mcsv1Context: " << getName() << std::endl;
-	output << "  RunFlags=" << fRunFlags << " ContextFlags=" << fContextFlags << std::endl;
-	output << "  UserDataSize=" << fUserDataSize << " ResultType=" << colDataTypeToString(fResultType) << std::endl;
-	output << "  Resultscale=" << fResultscale << " ResultPrecision=" << fResultPrecision << std::endl;
-	output << "  ErrorMsg=" << errorMsg << std::endl;
-	output << "  bInterrupted=" << bInterrupted << std::endl;
-	output << "  StartFrame=" << fStartFrame << " EndFrame=" << fEndFrame << std::endl;
-	output << "  StartConstant=" << fStartConstant << " EndConstant=" << fEndConstant << std::endl;
-	return output.str();
+    std::ostringstream output;
+    output << "mcsv1Context: " << getName() << std::endl;
+    output << "  RunFlags=" << fRunFlags << " ContextFlags=" << fContextFlags << std::endl;
+    output << "  UserDataSize=" << fUserDataSize << " ResultType=" << colDataTypeToString(fResultType) << std::endl;
+    output << "  Resultscale=" << fResultscale << " ResultPrecision=" << fResultPrecision << std::endl;
+    output << "  ErrorMsg=" << errorMsg << std::endl;
+    output << "  bInterrupted=" << bInterrupted << std::endl;
+    output << "  StartFrame=" << fStartFrame << " EndFrame=" << fEndFrame << std::endl;
+    output << "  StartConstant=" << fStartConstant << " EndConstant=" << fEndConstant << std::endl;
+    return output.str();
 }
 
 mcsv1sdk::mcsv1_UDAF* mcsv1Context::getFunction()
 {
-	if (func)
-	{
-		return func;
-	}
+    if (func)
+    {
+        return func;
+    }
 
-	// Just in time initialization
-	if (functionName.length() == 0)
-	{
-		std::ostringstream errmsg;
-		errmsg << "mcsv1Context::getFunction: " << functionName << " is empty";
-		throw std::logic_error(errmsg.str());
-	}
-	mcsv1sdk::UDAF_MAP::iterator funcIter = mcsv1sdk::UDAFMap::getMap().find(functionName);
-	if (funcIter == mcsv1sdk::UDAFMap::getMap().end())
-	{
-		 std::ostringstream errmsg;
-		 errmsg << "mcsv1Context::getFunction: " << functionName << " is undefined";
-		 throw std::logic_error(errmsg.str());
-	}
-	func = funcIter->second;
-	return func;
+    // Just in time initialization
+    if (functionName.length() == 0)
+    {
+        std::ostringstream errmsg;
+        errmsg << "mcsv1Context::getFunction: " << functionName << " is empty";
+        throw std::logic_error(errmsg.str());
+    }
+
+    mcsv1sdk::UDAF_MAP::iterator funcIter = mcsv1sdk::UDAFMap::getMap().find(functionName);
+
+    if (funcIter == mcsv1sdk::UDAFMap::getMap().end())
+    {
+        std::ostringstream errmsg;
+        errmsg << "mcsv1Context::getFunction: " << functionName << " is undefined";
+        throw std::logic_error(errmsg.str());
+    }
+
+    func = funcIter->second;
+    return func;
 }
 
 mcsv1sdk::mcsv1_UDAF* mcsv1Context::getFunction() const
 {
-	return const_cast<mcsv1Context*>(this)->getFunction();
+    return const_cast<mcsv1Context*>(this)->getFunction();
 }
 
-void mcsv1Context::createUserData() 
+void mcsv1Context::createUserData()
 {
-	// Try the function. If not implemented, create a byte array.
-	UserData* userData = NULL;
-	mcsv1_UDAF::ReturnCode rc = getFunction()->createUserData(userData, fUserDataSize);
-	if (rc == mcsv1_UDAF::ERROR)
-	{
-		std::ostringstream errmsg;
-		errmsg << "mcsv1Context::createUserData: " << functionName << errorMsg.c_str();
-		throw std::logic_error(errmsg.str());
-	}
-	setUserData(userData);
+    // Try the function. If not implemented, create a byte array.
+    UserData* userData = NULL;
+    mcsv1_UDAF::ReturnCode rc = getFunction()->createUserData(userData, fUserDataSize);
+
+    if (rc == mcsv1_UDAF::ERROR)
+    {
+        std::ostringstream errmsg;
+        errmsg << "mcsv1Context::createUserData: " << functionName << errorMsg.c_str();
+        throw std::logic_error(errmsg.str());
+    }
+
+    setUserData(userData);
 }
 
 void mcsv1Context::serialize(messageqcpp::ByteStream& b) const
 {
-	b.needAtLeast(sizeof(mcsv1Context));
-	b << (ObjectReader::id_t) ObjectReader::MCSV1_CONTEXT;
-	b << functionName;
-	b << fRunFlags;
-	// Dont send context flags, These are set for each call
-	b << fUserDataSize;
-	b << (uint32_t)fResultType;
-	b << fResultscale;
-	b << fResultPrecision;
-	b << errorMsg;
-	// Don't send dataflags. These are set for each call
-	// bInterrupted is set internally.
-	b << (uint32_t)fStartFrame;
-	b << (uint32_t)fEndFrame;
-	b << fStartConstant;
-	b << fEndConstant;
+    b.needAtLeast(sizeof(mcsv1Context));
+    b << (ObjectReader::id_t) ObjectReader::MCSV1_CONTEXT;
+    b << functionName;
+    b << fRunFlags;
+    // Dont send context flags, These are set for each call
+    b << fUserDataSize;
+    b << (uint32_t)fResultType;
+    b << fResultscale;
+    b << fResultPrecision;
+    b << errorMsg;
+    // Don't send dataflags. These are set for each call
+    // bInterrupted is set internally.
+    b << (uint32_t)fStartFrame;
+    b << (uint32_t)fEndFrame;
+    b << fStartConstant;
+    b << fEndConstant;
 }
 
 void mcsv1Context::unserialize(messageqcpp::ByteStream& b)
 {
-	ObjectReader::checkType(b, ObjectReader::MCSV1_CONTEXT);
-	b >> functionName;
-	b >> fRunFlags;
-	b >> fUserDataSize;
-	uint32_t iResultType;
-	b >> iResultType;
-	fResultType = (CalpontSystemCatalog::ColDataType)iResultType;
-	b >> fResultscale;
-	b >> fResultPrecision;
-	b >> errorMsg;
-	uint32_t frame;
-	b >> frame;
-	fStartFrame = (WF_FRAME)frame;
-	b >> frame;
-	fEndFrame = (WF_FRAME)frame;
-	b >> fStartConstant;
-	b >> fEndConstant;
+    ObjectReader::checkType(b, ObjectReader::MCSV1_CONTEXT);
+    b >> functionName;
+    b >> fRunFlags;
+    b >> fUserDataSize;
+    uint32_t iResultType;
+    b >> iResultType;
+    fResultType = (CalpontSystemCatalog::ColDataType)iResultType;
+    b >> fResultscale;
+    b >> fResultPrecision;
+    b >> errorMsg;
+    uint32_t frame;
+    b >> frame;
+    fStartFrame = (WF_FRAME)frame;
+    b >> frame;
+    fEndFrame = (WF_FRAME)frame;
+    b >> fStartConstant;
+    b >> fEndConstant;
 }
 
 void UserData::serialize(messageqcpp::ByteStream& bs) const
 {
-	bs << size;
-	bs.append(data, size);
+    bs << size;
+    bs.append(data, size);
 }
 
 void UserData::unserialize(messageqcpp::ByteStream& bs)
 {
-	bs >> size;
-	memcpy(data, bs.buf(), size);
-	bs.advance(size);
+    bs >> size;
+    memcpy(data, bs.buf(), size);
+    bs.advance(size);
 }
 
 const std::string typeStr("");

utils/udfsdk/median.cpp

@@ -25,290 +25,307 @@
 using namespace mcsv1sdk;
 
 mcsv1_UDAF::ReturnCode median::init(mcsv1Context* context,
-									 COL_TYPES& colTypes)
+                                    COL_TYPES& colTypes)
 {
-	if (colTypes.size() < 1)
-	{
-		// The error message will be prepended with
-		// "The storage engine for the table doesn't support "
-		context->setErrorMessage("median() with 0 arguments");
-		return mcsv1_UDAF::ERROR;
-	}
-	if (colTypes.size() > 1)
-	{
-		context->setErrorMessage("median() with more than 1 argument");
-		return mcsv1_UDAF::ERROR;
-	}
+    if (colTypes.size() < 1)
+    {
+        // The error message will be prepended with
+        // "The storage engine for the table doesn't support "
+        context->setErrorMessage("median() with 0 arguments");
+        return mcsv1_UDAF::ERROR;
+    }
 
-	if (!(isNumeric(colTypes[0].second)))
-	{
-		// The error message will be prepended with
-		// "The storage engine for the table doesn't support "
-		context->setErrorMessage("median() with non-numeric argument");
-		return mcsv1_UDAF::ERROR;
-	}
+    if (colTypes.size() > 1)
+    {
+        context->setErrorMessage("median() with more than 1 argument");
+        return mcsv1_UDAF::ERROR;
+    }
 
-	context->setResultType(CalpontSystemCatalog::DOUBLE);
-	context->setColWidth(8);
-	context->setScale(context->getScale()*2);
-	context->setPrecision(19);
-	context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
-	return mcsv1_UDAF::SUCCESS;
+    if (!(isNumeric(colTypes[0].second)))
+    {
+        // The error message will be prepended with
+        // "The storage engine for the table doesn't support "
+        context->setErrorMessage("median() with non-numeric argument");
+        return mcsv1_UDAF::ERROR;
+    }
+
+    context->setResultType(CalpontSystemCatalog::DOUBLE);
+    context->setColWidth(8);
+    context->setScale(context->getScale() * 2);
+    context->setPrecision(19);
+    context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
+    return mcsv1_UDAF::SUCCESS;
 
 }
 
 mcsv1_UDAF::ReturnCode median::reset(mcsv1Context* context)
 {
-	MedianData* data = static_cast<MedianData*>(context->getUserData());
-	data->mData.clear();
-	return mcsv1_UDAF::SUCCESS;
+    MedianData* data = static_cast<MedianData*>(context->getUserData());
+    data->mData.clear();
+    return mcsv1_UDAF::SUCCESS;
 }
 
-mcsv1_UDAF::ReturnCode median::nextValue(mcsv1Context* context, 
-										  std::vector<ColumnDatum>& valsIn)
+mcsv1_UDAF::ReturnCode median::nextValue(mcsv1Context* context,
+        std::vector<ColumnDatum>& valsIn)
 {
-	static_any::any& valIn = valsIn[0].columnData;
-	MEDIAN_DATA& data = static_cast<MedianData*>(context->getUserData())->mData;
-	DATATYPE val = 0.0;
+    static_any::any& valIn = valsIn[0].columnData;
+    MEDIAN_DATA& data = static_cast<MedianData*>(context->getUserData())->mData;
+    DATATYPE val = 0.0;
 
-	if (valIn.empty())
-	{
-		return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
-	}
+    if (valIn.empty())
+    {
+        return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
+    }
 
-	if (valIn.compatible(charTypeId))
-	{
-		val = valIn.cast<char>();
-	}
-	else if (valIn.compatible(scharTypeId))
-	{
-		val = valIn.cast<signed char>();
-	}
-	else if (valIn.compatible(shortTypeId))
-	{
-		val = valIn.cast<short>();
-	}
-	else if (valIn.compatible(intTypeId))
-	{
-		val = valIn.cast<int>();
-	}
-	else if (valIn.compatible(longTypeId))
-	{
-		val = valIn.cast<long>();
-	}
-	else if (valIn.compatible(llTypeId))
-	{
-		val = valIn.cast<long long>();
-	}
-	else if (valIn.compatible(ucharTypeId))
-	{
-		val = valIn.cast<unsigned char>();
-	}
-	else if (valIn.compatible(ushortTypeId))
-	{
-		val = valIn.cast<unsigned short>();
-	}
-	else if (valIn.compatible(uintTypeId))
-	{
-		val = valIn.cast<unsigned int>();
-	}
-	else if (valIn.compatible(ulongTypeId))
-	{
-		val = valIn.cast<unsigned long>();
-	}
-	else if (valIn.compatible(ullTypeId))
-	{
-		val = valIn.cast<unsigned long long>();
-	}
-	else if (valIn.compatible(floatTypeId))
-	{
-		val = valIn.cast<float>();
-	}
-	else if (valIn.compatible(doubleTypeId))
-	{
-		val = valIn.cast<double>();
-	}
+    if (valIn.compatible(charTypeId))
+    {
+        val = valIn.cast<char>();
+    }
+    else if (valIn.compatible(scharTypeId))
+    {
+        val = valIn.cast<signed char>();
+    }
+    else if (valIn.compatible(shortTypeId))
+    {
+        val = valIn.cast<short>();
+    }
+    else if (valIn.compatible(intTypeId))
+    {
+        val = valIn.cast<int>();
+    }
+    else if (valIn.compatible(longTypeId))
+    {
+        val = valIn.cast<long>();
+    }
+    else if (valIn.compatible(llTypeId))
+    {
+        val = valIn.cast<long long>();
+    }
+    else if (valIn.compatible(ucharTypeId))
+    {
+        val = valIn.cast<unsigned char>();
+    }
+    else if (valIn.compatible(ushortTypeId))
+    {
+        val = valIn.cast<unsigned short>();
+    }
+    else if (valIn.compatible(uintTypeId))
+    {
+        val = valIn.cast<unsigned int>();
+    }
+    else if (valIn.compatible(ulongTypeId))
+    {
+        val = valIn.cast<unsigned long>();
+    }
+    else if (valIn.compatible(ullTypeId))
+    {
+        val = valIn.cast<unsigned long long>();
+    }
+    else if (valIn.compatible(floatTypeId))
+    {
+        val = valIn.cast<float>();
+    }
+    else if (valIn.compatible(doubleTypeId))
+    {
+        val = valIn.cast<double>();
+    }
 
-	// For decimal types, we need to move the decimal point.
-	uint32_t scale = valsIn[0].scale;
-	if (val != 0 && scale > 0)
-	{
-		val /= pow(10.0, (double)scale);
-	}
-	data[val]++;
+    // For decimal types, we need to move the decimal point.
+    uint32_t scale = valsIn[0].scale;
 
-	return mcsv1_UDAF::SUCCESS;
+    if (val != 0 && scale > 0)
+    {
+        val /= pow(10.0, (double)scale);
+    }
+
+    data[val]++;
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode median::subEvaluate(mcsv1Context* context, const UserData* userDataIn)
 {
-	if (!userDataIn)
-	{
-		return mcsv1_UDAF::SUCCESS;
-	}
-	MEDIAN_DATA& outData = static_cast<MedianData*>(context->getUserData())->mData;
-	const MEDIAN_DATA& inData = static_cast<const MedianData*>(userDataIn)->mData;
-	MEDIAN_DATA::const_iterator iter = inData.begin();
-	for (; iter != inData.end(); ++iter)
-	{
-		outData[iter->first] += iter->second;
-	}
-	return mcsv1_UDAF::SUCCESS;
+    if (!userDataIn)
+    {
+        return mcsv1_UDAF::SUCCESS;
+    }
+
+    MEDIAN_DATA& outData = static_cast<MedianData*>(context->getUserData())->mData;
+    const MEDIAN_DATA& inData = static_cast<const MedianData*>(userDataIn)->mData;
+    MEDIAN_DATA::const_iterator iter = inData.begin();
+
+    for (; iter != inData.end(); ++iter)
+    {
+        outData[iter->first] += iter->second;
+    }
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode median::evaluate(mcsv1Context* context, static_any::any& valOut)
 {
-	uint64_t cnt1=0, cnt2=0;
-	MEDIAN_DATA& data = static_cast<MedianData*>(context->getUserData())->mData;
-	if (data.size() == 0)
-	{
-		valOut = (DATATYPE)0;
-		return mcsv1_UDAF::SUCCESS;
-	}
-	MEDIAN_DATA::iterator iter(data.begin());
-	MEDIAN_DATA::iterator revfrom(data.end());
-	MEDIAN_DATA::reverse_iterator riter(revfrom);
-	cnt1 += iter->second;
-	cnt2 += riter->second;
-	while (iter->first < riter->first)
-	{
-		while (cnt1 < cnt2 && iter->first < riter->first)
-		{
-			++iter;
-			cnt1 += iter->second;
-		}
-		while (cnt2 < cnt1 &&iter->first < riter->first)
-		{
-			++riter;
-			cnt2 += riter->second;
-		}
-		while (cnt1 == cnt2 && iter->first < riter->first)
-		{
-			++iter;
-			cnt1 += iter->second;
-			if (iter->first > riter->first)
-			{
-				break;
-			}
-			++riter;
-			cnt2 += riter->second;
-		}
-	}
-	valOut = (iter->first + riter->first) / 2;
-	return mcsv1_UDAF::SUCCESS;
+    uint64_t cnt1 = 0, cnt2 = 0;
+    MEDIAN_DATA& data = static_cast<MedianData*>(context->getUserData())->mData;
+
+    if (data.size() == 0)
+    {
+        valOut = (DATATYPE)0;
+        return mcsv1_UDAF::SUCCESS;
+    }
+
+    MEDIAN_DATA::iterator iter(data.begin());
+    MEDIAN_DATA::iterator revfrom(data.end());
+    MEDIAN_DATA::reverse_iterator riter(revfrom);
+    cnt1 += iter->second;
+    cnt2 += riter->second;
+
+    while (iter->first < riter->first)
+    {
+        while (cnt1 < cnt2 && iter->first < riter->first)
+        {
+            ++iter;
+            cnt1 += iter->second;
+        }
+
+        while (cnt2 < cnt1 && iter->first < riter->first)
+        {
+            ++riter;
+            cnt2 += riter->second;
+        }
+
+        while (cnt1 == cnt2 && iter->first < riter->first)
+        {
+            ++iter;
+            cnt1 += iter->second;
+
+            if (iter->first > riter->first)
+            {
+                break;
+            }
+
+            ++riter;
+            cnt2 += riter->second;
+        }
+    }
+
+    valOut = (iter->first + riter->first) / 2;
+    return mcsv1_UDAF::SUCCESS;
 }
 
-mcsv1_UDAF::ReturnCode median::dropValue(mcsv1Context* context, 
-													std::vector<ColumnDatum>& valsDropped)
+mcsv1_UDAF::ReturnCode median::dropValue(mcsv1Context* context,
+        std::vector<ColumnDatum>& valsDropped)
 {
-	static_any::any& valIn = valsDropped[0].columnData;
-	MEDIAN_DATA& data = static_cast<MedianData*>(context->getUserData())->mData;
-	DATATYPE val = 0.0;
+    static_any::any& valIn = valsDropped[0].columnData;
+    MEDIAN_DATA& data = static_cast<MedianData*>(context->getUserData())->mData;
+    DATATYPE val = 0.0;
 
-	if (valIn.empty())
-	{
-		return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
-	}
+    if (valIn.empty())
+    {
+        return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
+    }
 
-	if (valIn.compatible(charTypeId))
-	{
-		val = valIn.cast<char>();
-	}
-	else if (valIn.compatible(scharTypeId))
-	{
-		val = valIn.cast<signed char>();
-	}
-	else if (valIn.compatible(shortTypeId))
-	{
-		val = valIn.cast<short>();
-	}
-	else if (valIn.compatible(intTypeId))
-	{
-		val = valIn.cast<int>();
-	}
-	else if (valIn.compatible(longTypeId))
-	{
-		val = valIn.cast<long>();
-	}
-	else if (valIn.compatible(llTypeId))
-	{
-		val = valIn.cast<long long>();
-	}
-	else if (valIn.compatible(ucharTypeId))
-	{
-		val = valIn.cast<unsigned char>();
-	}
-	else if (valIn.compatible(ushortTypeId))
-	{
-		val = valIn.cast<unsigned short>();
-	}
-	else if (valIn.compatible(uintTypeId))
-	{
-		val = valIn.cast<unsigned int>();
-	}
-	else if (valIn.compatible(ulongTypeId))
-	{
-		val = valIn.cast<unsigned long>();
-	}
-	else if (valIn.compatible(ullTypeId))
-	{
-		val = valIn.cast<unsigned long long>();
-	}
-	else if (valIn.compatible(floatTypeId))
-	{
-		val = valIn.cast<float>();
-	}
-	else if (valIn.compatible(doubleTypeId))
-	{
-		val = valIn.cast<double>();
-	}
+    if (valIn.compatible(charTypeId))
+    {
+        val = valIn.cast<char>();
+    }
+    else if (valIn.compatible(scharTypeId))
+    {
+        val = valIn.cast<signed char>();
+    }
+    else if (valIn.compatible(shortTypeId))
+    {
+        val = valIn.cast<short>();
+    }
+    else if (valIn.compatible(intTypeId))
+    {
+        val = valIn.cast<int>();
+    }
+    else if (valIn.compatible(longTypeId))
+    {
+        val = valIn.cast<long>();
+    }
+    else if (valIn.compatible(llTypeId))
+    {
+        val = valIn.cast<long long>();
+    }
+    else if (valIn.compatible(ucharTypeId))
+    {
+        val = valIn.cast<unsigned char>();
+    }
+    else if (valIn.compatible(ushortTypeId))
+    {
+        val = valIn.cast<unsigned short>();
+    }
+    else if (valIn.compatible(uintTypeId))
+    {
+        val = valIn.cast<unsigned int>();
+    }
+    else if (valIn.compatible(ulongTypeId))
+    {
+        val = valIn.cast<unsigned long>();
+    }
+    else if (valIn.compatible(ullTypeId))
+    {
+        val = valIn.cast<unsigned long long>();
+    }
+    else if (valIn.compatible(floatTypeId))
+    {
+        val = valIn.cast<float>();
+    }
+    else if (valIn.compatible(doubleTypeId))
+    {
+        val = valIn.cast<double>();
+    }
 
-	// For decimal types, we need to move the decimal point.
-	uint32_t scale = valsDropped[0].scale;
-	if (val != 0 && scale > 0)
-	{
-		val /= pow(10.0, (double)scale);
-	}
+    // For decimal types, we need to move the decimal point.
+    uint32_t scale = valsDropped[0].scale;
 
-	data[val]--;
+    if (val != 0 && scale > 0)
+    {
+        val /= pow(10.0, (double)scale);
+    }
 
-	return mcsv1_UDAF::SUCCESS;
+    data[val]--;
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode median::createUserData(UserData*& userData, int32_t& length)
 {
-	userData = new MedianData;
-	length = sizeof(MedianData);
-	return mcsv1_UDAF::SUCCESS;
+    userData = new MedianData;
+    length = sizeof(MedianData);
+    return mcsv1_UDAF::SUCCESS;
 }
 
 void MedianData::serialize(messageqcpp::ByteStream& bs) const
 {
-	MEDIAN_DATA::const_iterator iter = mData.begin();
-	DATATYPE num;
-	uint32_t cnt;
-	bs << (int32_t)mData.size();
-	for (; iter != mData.end(); ++iter)
-	{
-		num = iter->first;
-		bs << num;
-		cnt = iter->second;
-		bs << cnt;
-	}
+    MEDIAN_DATA::const_iterator iter = mData.begin();
+    DATATYPE num;
+    uint32_t cnt;
+    bs << (int32_t)mData.size();
+
+    for (; iter != mData.end(); ++iter)
+    {
+        num = iter->first;
+        bs << num;
+        cnt = iter->second;
+        bs << cnt;
+    }
 }
 
 void MedianData::unserialize(messageqcpp::ByteStream& bs)
 {
-	mData.clear();
-	int32_t sz;
-	DATATYPE num;
-	uint32_t cnt;
-	bs >> sz;
-	for (int i = 0; i < sz; ++i)
-	{
-		bs >> num;
-		bs >> cnt;
-		mData[num] = cnt;
-	}
+    mData.clear();
+    int32_t sz;
+    DATATYPE num;
+    uint32_t cnt;
+    bs >> sz;
+
+    for (int i = 0; i < sz; ++i)
+    {
+        bs >> num;
+        bs >> cnt;
+        mData[num] = cnt;
+    }
 }
 

utils/udfsdk/median.h

@@ -21,34 +21,34 @@
 *   mcsv1_UDAF.h
 ***********************************************************************/
 
-/** 
- * Columnstore interface for writing a User Defined Aggregate 
- * Functions (UDAF) and User Defined Analytic Functions (UDAnF) 
- * or a function that can act as either - UDA(n)F 
- *  
+/**
+ * Columnstore interface for writing a User Defined Aggregate
+ * Functions (UDAF) and User Defined Analytic Functions (UDAnF)
+ * or a function that can act as either - UDA(n)F
+ *
  * The basic steps are:
  *
- * 1. Create a the UDA(n)F function interface in some .h file. 
- * 2. Create the UDF function implementation in some .cpp file 
- * 3. Create the connector stub (MariaDB UDAF definition) for 
- * this UDF function.  
- * 4. build the dynamic library using all of the source. 
- * 5  Put the library in $COLUMNSTORE_INSTALL/lib of 
- * all modules 
- * 6. restart the Columnstore system. 
+ * 1. Create a the UDA(n)F function interface in some .h file.
+ * 2. Create the UDF function implementation in some .cpp file
+ * 3. Create the connector stub (MariaDB UDAF definition) for
+ * this UDF function.
+ * 4. build the dynamic library using all of the source.
+ * 5  Put the library in $COLUMNSTORE_INSTALL/lib of
+ * all modules
+ * 6. restart the Columnstore system.
  * 7. notify mysqld about the new function:
- *  
+ *
  *    CREATE AGGREGATE FUNCTION median returns REAL soname
  *    'libudf_mysql.so';
- *  
- * The UDAF functions may run distributed in the Columnstore 
- * engine. UDAnF do not run distributed. 
- *  
- * UDAF is User Defined Aggregate Function. 
- * UDAnF is User Defined Analytic Function. 
- * UDA(n)F is an acronym for a function that could be either. It 
- * is also used to describe the interface that is used for 
- * either. 
+ *
+ * The UDAF functions may run distributed in the Columnstore
+ * engine. UDAnF do not run distributed.
+ *
+ * UDAF is User Defined Aggregate Function.
+ * UDAnF is User Defined Analytic Function.
+ * UDA(n)F is an acronym for a function that could be either. It
+ * is also used to describe the interface that is used for
+ * either.
  */
 #ifndef HEADER_median
 #define HEADER_median
@@ -83,196 +83,196 @@ typedef std::map<DATATYPE, uint32_t> MEDIAN_DATA;
 // Override UserData for data storage
 struct MedianData : public UserData
 {
-	MedianData() {};
+    MedianData() {};
 
-	virtual ~MedianData(){}
+    virtual ~MedianData() {}
 
-	virtual void serialize(messageqcpp::ByteStream& bs) const;
-	virtual void unserialize(messageqcpp::ByteStream& bs);
+    virtual void serialize(messageqcpp::ByteStream& bs) const;
+    virtual void unserialize(messageqcpp::ByteStream& bs);
 
-	MEDIAN_DATA mData;
+    MEDIAN_DATA mData;
 private:
-	// For now, copy construction is unwanted
-	MedianData(UserData&);
+    // For now, copy construction is unwanted
+    MedianData(UserData&);
 };
 
-// Override mcsv1_UDAF to build your User Defined Aggregate (UDAF) and/or 
+// Override mcsv1_UDAF to build your User Defined Aggregate (UDAF) and/or
 // User Defined Analytic Function (UDAnF).
 // These will be singleton classes, so don't put any instance
 // specific data in here. All instance data is stored in mcsv1Context
 // passed to each user function and retrieved by the getUserData() method.
-// 
-// Each API function returns a ReturnCode. If ERROR is returned at any time, 
-// the query is aborted, getInterrupted() will begin to return true and the 
-// message set in config->setErrorMessage() is returned to MariaDB. 
+//
+// Each API function returns a ReturnCode. If ERROR is returned at any time,
+// the query is aborted, getInterrupted() will begin to return true and the
+// message set in config->setErrorMessage() is returned to MariaDB.
 
 // Return the median value of the dataset
 
 class median : public  mcsv1_UDAF
 {
 public:
-	// Defaults OK
-	median() : mcsv1_UDAF(){};
-	virtual ~median(){};
+    // Defaults OK
+    median() : mcsv1_UDAF() {};
+    virtual ~median() {};
 
-	/** 
-	 * init() 
-	 *  
-	 * Mandatory. Implement this to initialize flags and instance 
-	 * data. Called once per SQL statement. You can do any sanity 
-	 * checks here. 
-	 *  
-	 * colTypes (in) - A vector of ColDataType defining the 
-	 * parameters of the UDA(n)F call. These can be used to decide 
-	 * to override the default return type. If desired, the new 
-	 * return type can be set by context->setReturnType() and 
-	 * decimal scale and precision can be set by context->setScale 
-	 * and context->setPrecision respectively. 
-	 *  
-	 * Return mcsv1_UDAF::ERROR on any error, such as non-compatible
-	 * colTypes or wrong number of arguments. Else return 
-	 * mcsv1_UDAF::SUCCESS. 
-	 */
-	virtual ReturnCode init(mcsv1Context* context,
-							COL_TYPES& colTypes);
+    /**
+     * init()
+     *
+     * Mandatory. Implement this to initialize flags and instance
+     * data. Called once per SQL statement. You can do any sanity
+     * checks here.
+     *
+     * colTypes (in) - A vector of ColDataType defining the
+     * parameters of the UDA(n)F call. These can be used to decide
+     * to override the default return type. If desired, the new
+     * return type can be set by context->setReturnType() and
+     * decimal scale and precision can be set by context->setScale
+     * and context->setPrecision respectively.
+     *
+     * Return mcsv1_UDAF::ERROR on any error, such as non-compatible
+     * colTypes or wrong number of arguments. Else return
+     * mcsv1_UDAF::SUCCESS.
+     */
+    virtual ReturnCode init(mcsv1Context* context,
+                            COL_TYPES& colTypes);
 
-	/** 
-	 * reset() 
-	 *  
-	 * Mandatory. Reset the UDA(n)F for a new group, partition or, 
-	 * in some cases, new Window Frame. Do not free any memory 
-	 * allocated by context->setUserDataSize(). The SDK Framework owns 
-	 * that memory and will handle that. Use this opportunity to 
-	 * reset any variables in context->getUserData() needed for the 
-	 * next aggregation. May be called multiple times if running in 
-	 * a ditributed fashion. 
-	 *  
-	 * Use this opportunity to initialize the userData.
-	 */
-	virtual ReturnCode reset(mcsv1Context* context);
+    /**
+     * reset()
+     *
+     * Mandatory. Reset the UDA(n)F for a new group, partition or,
+     * in some cases, new Window Frame. Do not free any memory
+     * allocated by context->setUserDataSize(). The SDK Framework owns
+     * that memory and will handle that. Use this opportunity to
+     * reset any variables in context->getUserData() needed for the
+     * next aggregation. May be called multiple times if running in
+     * a ditributed fashion.
+     *
+     * Use this opportunity to initialize the userData.
+     */
+    virtual ReturnCode reset(mcsv1Context* context);
 
-	/** 
-	 * nextValue() 
-	 *  
-	 * Mandatory. Handle a single row. 
-	 *  
-	 * colsIn - A vector of data structure describing the input 
-	 * data. 
-	 *  
-	 * This function is called once for every row in the filtered 
-	 * result set (before aggregation). It is very important that 
-	 * this function is efficient. 
-	 *  
-	 * If the UDAF is running in a distributed fashion, nextValue 
-	 * cannot depend on order, as it will only be called for each 
-	 * row found on the specific PM. 
-	 *  
-	 * valsIn (in) - a vector of the parameters from the row.
-	 */
-	virtual ReturnCode nextValue(mcsv1Context* context, 
-								 std::vector<ColumnDatum>& valsIn);
+    /**
+     * nextValue()
+     *
+     * Mandatory. Handle a single row.
+     *
+     * colsIn - A vector of data structure describing the input
+     * data.
+     *
+     * This function is called once for every row in the filtered
+     * result set (before aggregation). It is very important that
+     * this function is efficient.
+     *
+     * If the UDAF is running in a distributed fashion, nextValue
+     * cannot depend on order, as it will only be called for each
+     * row found on the specific PM.
+     *
+     * valsIn (in) - a vector of the parameters from the row.
+     */
+    virtual ReturnCode nextValue(mcsv1Context* context,
+                                 std::vector<ColumnDatum>& valsIn);
 
-	 /** 
-	  * subEvaluate() 
-	  *  
-	  * Mandatory -- Called if the UDAF is running in a distributed 
-	  * fashion. Columnstore tries to run all aggregate functions 
-	  * distributed, depending on context. 
-	  *  
-	  * Perform an aggregation on rows partially aggregated by 
-	  * nextValue. Columnstore calls nextValue for each row on a 
-	  * given PM for a group (GROUP BY). subEvaluate is called on the
-	  * UM to consolodate those values into a single instance of 
-	  * userData. Keep your aggregated totals in context's userData. 
-	  * The first time this is called for a group, reset() would have 
-	  * been called with this version of userData. 
-	  *  
-	  * Called for every partial data set in each group in GROUP BY.
-	  *  
-	  * When subEvaluate has been called for all subAggregated data 
-	  * sets, Evaluate will be called with the same context as here.
-	  *  
-	  * valIn (In) - This is a pointer to a memory block of the size 
-	  * set in setUserDataSize. It will contain the value of userData 
-	  * as seen in the last call to NextValue for a given PM.
-	  *  
-	  */
-	 virtual ReturnCode subEvaluate(mcsv1Context* context, const UserData* valIn);
+    /**
+     * subEvaluate()
+     *
+     * Mandatory -- Called if the UDAF is running in a distributed
+     * fashion. Columnstore tries to run all aggregate functions
+     * distributed, depending on context.
+     *
+     * Perform an aggregation on rows partially aggregated by
+     * nextValue. Columnstore calls nextValue for each row on a
+     * given PM for a group (GROUP BY). subEvaluate is called on the
+     * UM to consolodate those values into a single instance of
+     * userData. Keep your aggregated totals in context's userData.
+     * The first time this is called for a group, reset() would have
+     * been called with this version of userData.
+     *
+     * Called for every partial data set in each group in GROUP BY.
+     *
+     * When subEvaluate has been called for all subAggregated data
+     * sets, Evaluate will be called with the same context as here.
+     *
+     * valIn (In) - This is a pointer to a memory block of the size
+     * set in setUserDataSize. It will contain the value of userData
+     * as seen in the last call to NextValue for a given PM.
+     *
+     */
+    virtual ReturnCode subEvaluate(mcsv1Context* context, const UserData* valIn);
 
-	/** 
-	 * evaluate() 
-	 *  
-	 * Mandatory. Get the aggregated value.
-	 *  
-	 * Called for every new group if UDAF GROUP BY, UDAnF partition 
-	 * or, in some cases, new Window Frame. 
-	 *  
-	 * Set the aggregated value into valOut. The datatype is assumed
-	 * to be the same as that set in the init() function; 
-	 *  
-	 * If the UDAF is running in a distributed fashion, evaluate is 
-	 * called after a series of subEvaluate calls. 
-	 *  
-	 * valOut (out) - Set the aggregated value here. The datatype is
-	 * assumed to be the same as that set in the init() function; 
-	 *  
-	 * To return a NULL value, don't assign to valOut.
-	 */
-	virtual ReturnCode evaluate(mcsv1Context* context, static_any::any& valOut);
+    /**
+     * evaluate()
+     *
+     * Mandatory. Get the aggregated value.
+     *
+     * Called for every new group if UDAF GROUP BY, UDAnF partition
+     * or, in some cases, new Window Frame.
+     *
+     * Set the aggregated value into valOut. The datatype is assumed
+     * to be the same as that set in the init() function;
+     *
+     * If the UDAF is running in a distributed fashion, evaluate is
+     * called after a series of subEvaluate calls.
+     *
+     * valOut (out) - Set the aggregated value here. The datatype is
+     * assumed to be the same as that set in the init() function;
+     *
+     * To return a NULL value, don't assign to valOut.
+     */
+    virtual ReturnCode evaluate(mcsv1Context* context, static_any::any& valOut);
 
-	/** 
-	 * dropValue() 
-	 *  
-	 * Optional -- If defined, the server will call this instead of 
-	 * reset for UDAnF. 
-	 *  
-	 * Don't implement if a UDAnF has one or more of the following: 
-	 * The UDAnF can't be used with a Window Frame
-	 * The UDAnF is not reversable in some way 
-	 * The UDAnF is not interested in optimal performance 
-	 *  
-	 * If not implemented, reset() followed by a series of 
-	 * nextValue() will be called for each movement of the Window 
-	 * Frame. 
-	 *  
-	 * If implemented, then each movement of the Window Frame will 
-	 * result in dropValue() being called for each row falling out 
-	 * of the Frame and nextValue() being called for each new row 
-	 * coming into the Frame. 
-	 *  
-	 * valsDropped (in) - a vector of the parameters from the row 
-	 * leaving the Frame 
-	 *  
-	 * dropValue() will not be called for unbounded/current row type
-	 * frames, as those are already optimized. 
-	 */
-	virtual ReturnCode dropValue(mcsv1Context* context, 
-								 std::vector<ColumnDatum>& valsDropped);
+    /**
+     * dropValue()
+     *
+     * Optional -- If defined, the server will call this instead of
+     * reset for UDAnF.
+     *
+     * Don't implement if a UDAnF has one or more of the following:
+     * The UDAnF can't be used with a Window Frame
+     * The UDAnF is not reversable in some way
+     * The UDAnF is not interested in optimal performance
+     *
+     * If not implemented, reset() followed by a series of
+     * nextValue() will be called for each movement of the Window
+     * Frame.
+     *
+     * If implemented, then each movement of the Window Frame will
+     * result in dropValue() being called for each row falling out
+     * of the Frame and nextValue() being called for each new row
+     * coming into the Frame.
+     *
+     * valsDropped (in) - a vector of the parameters from the row
+     * leaving the Frame
+     *
+     * dropValue() will not be called for unbounded/current row type
+     * frames, as those are already optimized.
+     */
+    virtual ReturnCode dropValue(mcsv1Context* context,
+                                 std::vector<ColumnDatum>& valsDropped);
 
-	/** 
-	 * createUserData()
-	 *  
-	 * Optional -- If defined, the server will call this instead of 
-	 * createUserData on context. 
-	 *  
-	 * Create your variable length data structure via 
-	 * data = new <datatype> 
-	 *  
-	 * The data structure may contain references to containers or 
-	 * pointers to other objects. Remember that for distributed 
-	 * processing, this may be called multiple times for variaous 
-	 * computing blocks. At the least, it will be called once per PM 
-	 * that processes the data, and once more for the UM. For UDAnF, 
-	 * it may only be called once. 
-	 *  
-	 * Set length to the length of the data structure you create.
-	 *  
-	 * For each call to createUserData(), there will be a 
-	 * corresponding deleteUserData() where you must clean up. Any 
-	 * memory leaks are your fault. 
-	 *  
-	 */ 
-	virtual ReturnCode createUserData(UserData*& data, int32_t& length);
+    /**
+     * createUserData()
+     *
+     * Optional -- If defined, the server will call this instead of
+     * createUserData on context.
+     *
+     * Create your variable length data structure via
+     * data = new <datatype>
+     *
+     * The data structure may contain references to containers or
+     * pointers to other objects. Remember that for distributed
+     * processing, this may be called multiple times for variaous
+     * computing blocks. At the least, it will be called once per PM
+     * that processes the data, and once more for the UM. For UDAnF,
+     * it may only be called once.
+     *
+     * Set length to the length of the data structure you create.
+     *
+     * For each call to createUserData(), there will be a
+     * corresponding deleteUserData() where you must clean up. Any
+     * memory leaks are your fault.
+     *
+     */
+    virtual ReturnCode createUserData(UserData*& data, int32_t& length);
 protected:
 };
 

utils/udfsdk/resource.h

@@ -1,14 +1,14 @@
-//{{NO_DEPENDENCIES}}
-// Microsoft Visual C++ generated include file.
-// Used by libudfsdk-ent.rc
-
-// Next default values for new objects
-// 
-#ifdef APSTUDIO_INVOKED
-#ifndef APSTUDIO_READONLY_SYMBOLS
-#define _APS_NEXT_RESOURCE_VALUE        101
-#define _APS_NEXT_COMMAND_VALUE         40001
-#define _APS_NEXT_CONTROL_VALUE         1001
-#define _APS_NEXT_SYMED_VALUE           101
-#endif
-#endif
+//{{NO_DEPENDENCIES}}
+// Microsoft Visual C++ generated include file.
+// Used by libudfsdk-ent.rc
+
+// Next default values for new objects
+//
+#ifdef APSTUDIO_INVOKED
+#ifndef APSTUDIO_READONLY_SYMBOLS
+#define _APS_NEXT_RESOURCE_VALUE        101
+#define _APS_NEXT_COMMAND_VALUE         40001
+#define _APS_NEXT_CONTROL_VALUE         1001
+#define _APS_NEXT_SYMED_VALUE           101
+#endif
+#endif

utils/udfsdk/ssq.cpp

@@ -27,225 +27,236 @@ using namespace mcsv1sdk;
 
 struct ssq_data
 {
-	uint64_t scale;
+    uint64_t scale;
     DATATYPE	sumsq;
-	ssq_data() : scale(0){}
+    ssq_data() : scale(0) {}
 };
 
 #define OUT_TYPE int64_t
 mcsv1_UDAF::ReturnCode ssq::init(mcsv1Context* context,
-									 COL_TYPES& colTypes)
+                                 COL_TYPES& colTypes)
 {
-	if (colTypes.size() < 1)
-	{
-		// The error message will be prepended with
-		// "The storage engine for the table doesn't support "
-		context->setErrorMessage("ssq() with 0 arguments");
-		return mcsv1_UDAF::ERROR;
-	}
-	if (colTypes.size() > 1)
-	{
-		context->setErrorMessage("ssq() with more than 1 argument");
-		return mcsv1_UDAF::ERROR;
-	}
+    if (colTypes.size() < 1)
+    {
+        // The error message will be prepended with
+        // "The storage engine for the table doesn't support "
+        context->setErrorMessage("ssq() with 0 arguments");
+        return mcsv1_UDAF::ERROR;
+    }
 
-	if (!(isNumeric(colTypes[0].second)))
-	{
-		// The error message will be prepended with
-		// "The storage engine for the table doesn't support "
-		context->setErrorMessage("ssq() with non-numeric argument");
-		return mcsv1_UDAF::ERROR;
-	}
+    if (colTypes.size() > 1)
+    {
+        context->setErrorMessage("ssq() with more than 1 argument");
+        return mcsv1_UDAF::ERROR;
+    }
 
-	context->setUserDataSize(sizeof(ssq_data));
-	context->setResultType(CalpontSystemCatalog::DOUBLE);
-	context->setColWidth(8);
-	context->setScale(context->getScale()*2);
-	context->setPrecision(19);
-	context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
-	return mcsv1_UDAF::SUCCESS;
+    if (!(isNumeric(colTypes[0].second)))
+    {
+        // The error message will be prepended with
+        // "The storage engine for the table doesn't support "
+        context->setErrorMessage("ssq() with non-numeric argument");
+        return mcsv1_UDAF::ERROR;
+    }
+
+    context->setUserDataSize(sizeof(ssq_data));
+    context->setResultType(CalpontSystemCatalog::DOUBLE);
+    context->setColWidth(8);
+    context->setScale(context->getScale() * 2);
+    context->setPrecision(19);
+    context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
+    return mcsv1_UDAF::SUCCESS;
 
 }
 
 mcsv1_UDAF::ReturnCode ssq::reset(mcsv1Context* context)
 {
-	struct ssq_data* data = (struct ssq_data*)context->getUserData()->data;
-	if (data)
-	{
-		data->scale = 0;
-		data->sumsq = 0;
-	}
-	return mcsv1_UDAF::SUCCESS;
+    struct ssq_data* data = (struct ssq_data*)context->getUserData()->data;
+
+    if (data)
+    {
+        data->scale = 0;
+        data->sumsq = 0;
+    }
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
-mcsv1_UDAF::ReturnCode ssq::nextValue(mcsv1Context* context, 
-										  std::vector<ColumnDatum>& valsIn)
+mcsv1_UDAF::ReturnCode ssq::nextValue(mcsv1Context* context,
+                                      std::vector<ColumnDatum>& valsIn)
 {
-	static_any::any& valIn = valsIn[0].columnData;
-	struct ssq_data* data = (struct ssq_data*)context->getUserData()->data;
-	DATATYPE val = 0.0;
+    static_any::any& valIn = valsIn[0].columnData;
+    struct ssq_data* data = (struct ssq_data*)context->getUserData()->data;
+    DATATYPE val = 0.0;
 
-	if (context->isParamNull(0) || valIn.empty())
-	{
-		return mcsv1_UDAF::SUCCESS;
-	}
+    if (context->isParamNull(0) || valIn.empty())
+    {
+        return mcsv1_UDAF::SUCCESS;
+    }
 
-	if (valIn.compatible(charTypeId))
-	{
-		val = valIn.cast<char>();
-	}
-	else if (valIn.compatible(scharTypeId))
-	{
-		val = valIn.cast<signed char>();
-	}
-	else if (valIn.compatible(shortTypeId))
-	{
-		val = valIn.cast<short>();
-	}
-	else if (valIn.compatible(intTypeId))
-	{
-		val = valIn.cast<int>();
-	}
-	else if (valIn.compatible(longTypeId))
-	{
-		val = valIn.cast<long>();
-	}
-	else if (valIn.compatible(llTypeId))
-	{
-		val = valIn.cast<long long>();
-	}
-	else if (valIn.compatible(ucharTypeId))
-	{
-		val = valIn.cast<unsigned char>();
-	}
-	else if (valIn.compatible(ushortTypeId))
-	{
-		val = valIn.cast<unsigned short>();
-	}
-	else if (valIn.compatible(uintTypeId))
-	{
-		val = valIn.cast<unsigned int>();
-	}
-	else if (valIn.compatible(ulongTypeId))
-	{
-		val = valIn.cast<unsigned long>();
-	}
-	else if (valIn.compatible(ullTypeId))
-	{
-		val = valIn.cast<unsigned long long>();
-	}
-	else if (valIn.compatible(floatTypeId))
-	{
-		val = valIn.cast<float>();
-	}
-	else if (valIn.compatible(doubleTypeId))
-	{
-		val = valIn.cast<double>();
-	}
+    if (valIn.compatible(charTypeId))
+    {
+        val = valIn.cast<char>();
+    }
+    else if (valIn.compatible(scharTypeId))
+    {
+        val = valIn.cast<signed char>();
+    }
+    else if (valIn.compatible(shortTypeId))
+    {
+        val = valIn.cast<short>();
+    }
+    else if (valIn.compatible(intTypeId))
+    {
+        val = valIn.cast<int>();
+    }
+    else if (valIn.compatible(longTypeId))
+    {
+        val = valIn.cast<long>();
+    }
+    else if (valIn.compatible(llTypeId))
+    {
+        val = valIn.cast<long long>();
+    }
+    else if (valIn.compatible(ucharTypeId))
+    {
+        val = valIn.cast<unsigned char>();
+    }
+    else if (valIn.compatible(ushortTypeId))
+    {
+        val = valIn.cast<unsigned short>();
+    }
+    else if (valIn.compatible(uintTypeId))
+    {
+        val = valIn.cast<unsigned int>();
+    }
+    else if (valIn.compatible(ulongTypeId))
+    {
+        val = valIn.cast<unsigned long>();
+    }
+    else if (valIn.compatible(ullTypeId))
+    {
+        val = valIn.cast<unsigned long long>();
+    }
+    else if (valIn.compatible(floatTypeId))
+    {
+        val = valIn.cast<float>();
+    }
+    else if (valIn.compatible(doubleTypeId))
+    {
+        val = valIn.cast<double>();
+    }
 
-	// For decimal types, we need to move the decimal point.
-	uint32_t scale = valsIn[0].scale;
-	if (val != 0 && scale > 0)
-	{
-		val /= pow(10.0, (double)scale);
-	}
-	data->sumsq += val*val;
-	return mcsv1_UDAF::SUCCESS;
+    // For decimal types, we need to move the decimal point.
+    uint32_t scale = valsIn[0].scale;
+
+    if (val != 0 && scale > 0)
+    {
+        val /= pow(10.0, (double)scale);
+    }
+
+    data->sumsq += val * val;
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode ssq::subEvaluate(mcsv1Context* context, const UserData* userDataIn)
 {
-	// If we turn off UDAF_IGNORE_NULLS in init(), then NULLS may be sent here in cases of Joins. 
-	// When a NULL value is sent here, userDataIn will be NULL, so check for NULLS.
-	if (context->isParamNull(0))
-	{
-		return mcsv1_UDAF::SUCCESS;
-	}
-	struct ssq_data* outData = (struct ssq_data*)context->getUserData()->data;
-	struct ssq_data* inData = (struct ssq_data*)userDataIn->data;
-	outData->sumsq += inData->sumsq;
-	return mcsv1_UDAF::SUCCESS;
+    // If we turn off UDAF_IGNORE_NULLS in init(), then NULLS may be sent here in cases of Joins.
+    // When a NULL value is sent here, userDataIn will be NULL, so check for NULLS.
+    if (context->isParamNull(0))
+    {
+        return mcsv1_UDAF::SUCCESS;
+    }
+
+    struct ssq_data* outData = (struct ssq_data*)context->getUserData()->data;
+
+    struct ssq_data* inData = (struct ssq_data*)userDataIn->data;
+
+    outData->sumsq += inData->sumsq;
+
+    return mcsv1_UDAF::SUCCESS;
 }
 
 mcsv1_UDAF::ReturnCode ssq::evaluate(mcsv1Context* context, static_any::any& valOut)
 {
-	struct ssq_data* data = (struct ssq_data*)context->getUserData()->data;
-	valOut = data->sumsq;
-	return mcsv1_UDAF::SUCCESS;
+    struct ssq_data* data = (struct ssq_data*)context->getUserData()->data;
+    valOut = data->sumsq;
+    return mcsv1_UDAF::SUCCESS;
 }
 
-mcsv1_UDAF::ReturnCode ssq::dropValue(mcsv1Context* context, 
-													std::vector<ColumnDatum>& valsDropped)
+mcsv1_UDAF::ReturnCode ssq::dropValue(mcsv1Context* context,
+                                      std::vector<ColumnDatum>& valsDropped)
 {
-	static_any::any& valIn = valsDropped[0].columnData;
-	struct ssq_data* data = (struct ssq_data*)context->getUserData()->data;
-	DATATYPE val = 0.0;
+    static_any::any& valIn = valsDropped[0].columnData;
+    struct ssq_data* data = (struct ssq_data*)context->getUserData()->data;
+    DATATYPE val = 0.0;
 
-	if (valIn.empty())
-	{
-		return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
-	}
+    if (valIn.empty())
+    {
+        return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
+    }
 
-	if (valIn.compatible(charTypeId))
-	{
-		val = valIn.cast<char>();
-	}
-	else if (valIn.compatible(scharTypeId))
-	{
-		val = valIn.cast<signed char>();
-	}
-	else if (valIn.compatible(shortTypeId))
-	{
-		val = valIn.cast<short>();
-	}
-	else if (valIn.compatible(intTypeId))
-	{
-		val = valIn.cast<int>();
-	}
-	else if (valIn.compatible(longTypeId))
-	{
-		val = valIn.cast<long>();
-	}
-	else if (valIn.compatible(llTypeId))
-	{
-		val = valIn.cast<long long>();
-	}
-	else if (valIn.compatible(ucharTypeId))
-	{
-		val = valIn.cast<unsigned char>();
-	}
-	else if (valIn.compatible(ushortTypeId))
-	{
-		val = valIn.cast<unsigned short>();
-	}
-	else if (valIn.compatible(uintTypeId))
-	{
-		val = valIn.cast<unsigned int>();
-	}
-	else if (valIn.compatible(ulongTypeId))
-	{
-		val = valIn.cast<unsigned long>();
-	}
-	else if (valIn.compatible(ullTypeId))
-	{
-		val = valIn.cast<unsigned long long>();
-	}
-	else if (valIn.compatible(floatTypeId))
-	{
-		val = valIn.cast<float>();
-	}
-	else if (valIn.compatible(doubleTypeId))
-	{
-		val = valIn.cast<double>();
-	}
+    if (valIn.compatible(charTypeId))
+    {
+        val = valIn.cast<char>();
+    }
+    else if (valIn.compatible(scharTypeId))
+    {
+        val = valIn.cast<signed char>();
+    }
+    else if (valIn.compatible(shortTypeId))
+    {
+        val = valIn.cast<short>();
+    }
+    else if (valIn.compatible(intTypeId))
+    {
+        val = valIn.cast<int>();
+    }
+    else if (valIn.compatible(longTypeId))
+    {
+        val = valIn.cast<long>();
+    }
+    else if (valIn.compatible(llTypeId))
+    {
+        val = valIn.cast<long long>();
+    }
+    else if (valIn.compatible(ucharTypeId))
+    {
+        val = valIn.cast<unsigned char>();
+    }
+    else if (valIn.compatible(ushortTypeId))
+    {
+        val = valIn.cast<unsigned short>();
+    }
+    else if (valIn.compatible(uintTypeId))
+    {
+        val = valIn.cast<unsigned int>();
+    }
+    else if (valIn.compatible(ulongTypeId))
+    {
+        val = valIn.cast<unsigned long>();
+    }
+    else if (valIn.compatible(ullTypeId))
+    {
+        val = valIn.cast<unsigned long long>();
+    }
+    else if (valIn.compatible(floatTypeId))
+    {
+        val = valIn.cast<float>();
+    }
+    else if (valIn.compatible(doubleTypeId))
+    {
+        val = valIn.cast<double>();
+    }
 
-	// For decimal types, we need to move the decimal point.
-	uint32_t scale = valsDropped[0].scale;
-	if (val != 0 && scale > 0)
-	{
-		val /= pow(10.0, (double)scale);
-	}
-	data->sumsq -= val*val;
-	return mcsv1_UDAF::SUCCESS;
+    // For decimal types, we need to move the decimal point.
+    uint32_t scale = valsDropped[0].scale;
+
+    if (val != 0 && scale > 0)
+    {
+        val /= pow(10.0, (double)scale);
+    }
+
+    data->sumsq -= val * val;
+    return mcsv1_UDAF::SUCCESS;
 }
 

utils/udfsdk/ssq.h

@@ -21,34 +21,34 @@
 *   mcsv1_UDAF.h
 ***********************************************************************/
 
-/** 
- * Columnstore interface for writing a User Defined Aggregate 
- * Functions (UDAF) and User Defined Analytic Functions (UDAnF) 
- * or a function that can act as either - UDA(n)F 
- *  
+/**
+ * Columnstore interface for writing a User Defined Aggregate
+ * Functions (UDAF) and User Defined Analytic Functions (UDAnF)
+ * or a function that can act as either - UDA(n)F
+ *
  * The basic steps are:
  *
- * 1. Create a the UDA(n)F function interface in some .h file. 
- * 2. Create the UDF function implementation in some .cpp file 
- * 3. Create the connector stub (MariaDB UDAF definition) for 
- * this UDF function.  
- * 4. build the dynamic library using all of the source. 
- * 5  Put the library in $COLUMNSTORE_INSTALL/lib of 
- * all modules 
- * 6. restart the Columnstore system. 
+ * 1. Create a the UDA(n)F function interface in some .h file.
+ * 2. Create the UDF function implementation in some .cpp file
+ * 3. Create the connector stub (MariaDB UDAF definition) for
+ * this UDF function.
+ * 4. build the dynamic library using all of the source.
+ * 5  Put the library in $COLUMNSTORE_INSTALL/lib of
+ * all modules
+ * 6. restart the Columnstore system.
  * 7. notify mysqld about the new function:
- *  
+ *
  *    CREATE AGGREGATE FUNCTION ssq returns REAL soname
  *    'libudf_mysql.so';
- *  
- * The UDAF function will run distributed in the Columnstore 
- * engine. UDAnF do not run distributed. 
- *  
- * UDAF is User Defined Aggregate Function. 
- * UDAnF is User Defined Analytic Function. 
- * UDA(n)F is an acronym for a function that could be either. It 
- * is also used to describe the interface that is used for 
- * either. 
+ *
+ * The UDAF function will run distributed in the Columnstore
+ * engine. UDAnF do not run distributed.
+ *
+ * UDAF is User Defined Aggregate Function.
+ * UDAnF is User Defined Analytic Function.
+ * UDA(n)F is an acronym for a function that could be either. It
+ * is also used to describe the interface that is used for
+ * either.
  */
 #ifndef HEADER_ssq
 #define HEADER_ssq
@@ -77,155 +77,155 @@ using namespace execplan;
 namespace mcsv1sdk
 {
 
-// Override mcsv1_UDAF to build your User Defined Aggregate (UDAF) and/or 
+// Override mcsv1_UDAF to build your User Defined Aggregate (UDAF) and/or
 // User Defined Analytic Function (UDAnF).
 // These will be singleton classes, so don't put any instance
 // specific data in here. All instance data is stored in mcsv1Context
 // passed to each user function and retrieved by the getUserData() method.
-// 
-// Each API function returns a ReturnCode. If ERROR is returned at any time, 
-// the query is aborted, getInterrupted() will begin to return true and the 
-// message set in config->setErrorMessage() is returned to MariaDB. 
+//
+// Each API function returns a ReturnCode. If ERROR is returned at any time,
+// the query is aborted, getInterrupted() will begin to return true and the
+// message set in config->setErrorMessage() is returned to MariaDB.
 
 // A simple aggregate to return the sum of squares
 class ssq : public  mcsv1_UDAF
 {
 public:
-	// Defaults OK
-	ssq() : mcsv1_UDAF(){};
-	virtual ~ssq(){};
+    // Defaults OK
+    ssq() : mcsv1_UDAF() {};
+    virtual ~ssq() {};
 
-	/** 
-	 * init() 
-	 *  
-	 * Mandatory. Implement this to initialize flags and instance 
-	 * data. Called once per SQL statement. You can do any sanity 
-	 * checks here. 
-	 *  
-	 * colTypes (in) - A vector of ColDataType defining the 
-	 * parameters of the UDA(n)F call. These can be used to decide 
-	 * to override the default return type. If desired, the new 
-	 * return type can be set by context->setReturnType() and 
-	 * decimal scale and precision can be set by context->setScale 
-	 * and context->setPrecision respectively. 
-	 *  
-	 * Return mcsv1_UDAF::ERROR on any error, such as non-compatible
-	 * colTypes or wrong number of arguments. Else return 
-	 * mcsv1_UDAF::SUCCESS. 
-	 */
-	virtual ReturnCode init(mcsv1Context* context,
-							COL_TYPES& colTypes);
+    /**
+     * init()
+     *
+     * Mandatory. Implement this to initialize flags and instance
+     * data. Called once per SQL statement. You can do any sanity
+     * checks here.
+     *
+     * colTypes (in) - A vector of ColDataType defining the
+     * parameters of the UDA(n)F call. These can be used to decide
+     * to override the default return type. If desired, the new
+     * return type can be set by context->setReturnType() and
+     * decimal scale and precision can be set by context->setScale
+     * and context->setPrecision respectively.
+     *
+     * Return mcsv1_UDAF::ERROR on any error, such as non-compatible
+     * colTypes or wrong number of arguments. Else return
+     * mcsv1_UDAF::SUCCESS.
+     */
+    virtual ReturnCode init(mcsv1Context* context,
+                            COL_TYPES& colTypes);
 
-	/** 
-	 * reset() 
-	 *  
-	 * Mandatory. Reset the UDA(n)F for a new group, partition or, 
-	 * in some cases, new Window Frame. Do not free any memory 
-	 * allocated by context->createUserData(). The SDK Framework 
-	 * owns that memory and will handle that. Use this opportunity 
-	 * to reset any variables in context->getUserData() needed for 
-	 * the next aggregation. May be called multiple times on 
-	 * different modules. 
-	 */
-	virtual ReturnCode reset(mcsv1Context* context);
+    /**
+     * reset()
+     *
+     * Mandatory. Reset the UDA(n)F for a new group, partition or,
+     * in some cases, new Window Frame. Do not free any memory
+     * allocated by context->createUserData(). The SDK Framework
+     * owns that memory and will handle that. Use this opportunity
+     * to reset any variables in context->getUserData() needed for
+     * the next aggregation. May be called multiple times on
+     * different modules.
+     */
+    virtual ReturnCode reset(mcsv1Context* context);
 
-	/** 
-	 * nextValue() 
-	 *  
-	 * Mandatory. Handle a single row. 
-	 *  
-	 * colsIn - A vector of data structure describing the input 
-	 * data. 
-	 *  
-	 * This function is called once for every row in the filtered 
-	 * result set (before aggregation). It is very important that 
-	 * this function is efficient. 
-	 *  
-	 * If the UDAF is running in a distributed fashion, nextValue 
-	 * cannot depend on order, as it will only be called for each 
-	 * row found on the specific PM. 
-	 *  
-	 * valsIn (in) - a vector of the parameters from the row.
-	 */
-	virtual ReturnCode nextValue(mcsv1Context* context, 
-								 std::vector<ColumnDatum>& valsIn);
+    /**
+     * nextValue()
+     *
+     * Mandatory. Handle a single row.
+     *
+     * colsIn - A vector of data structure describing the input
+     * data.
+     *
+     * This function is called once for every row in the filtered
+     * result set (before aggregation). It is very important that
+     * this function is efficient.
+     *
+     * If the UDAF is running in a distributed fashion, nextValue
+     * cannot depend on order, as it will only be called for each
+     * row found on the specific PM.
+     *
+     * valsIn (in) - a vector of the parameters from the row.
+     */
+    virtual ReturnCode nextValue(mcsv1Context* context,
+                                 std::vector<ColumnDatum>& valsIn);
 
-	 /** 
-	  * subEvaluate() 
-	  *  
-	  * Mandatory -- Called if the UDAF is running in a distributed 
-	  * fashion. Columnstore tries to run all aggregate functions 
-	  * distributed, depending on context. 
-	  *  
-	  * Perform an aggregation on rows partially aggregated by 
-	  * nextValue. Columnstore calls nextValue for each row on a 
-	  * given PM for a group (GROUP BY). subEvaluate is called on the
-	  * UM to consolodate those values into a single instance of 
-	  * userData. Keep your aggregated totals in context's userData. 
-	  * The first time this is called for a group, reset() would have 
-	  * been called with this version of userData. 
-	  *  
-	  * Called for every partial data set in each group in GROUP BY.
-	  *  
-	  * When subEvaluate has been called for all subAggregated data 
-	  * sets, Evaluate will be called.
-	  *  
-	  * valIn (In) - This is a pointer to a memory block of the size 
-	  * set in setUserDataSize. It will contain the value of userData 
-	  * as seen in the last call to NextValue for a given PM.
-	  *  
-	  */
-	 virtual ReturnCode subEvaluate(mcsv1Context* context, const UserData* userDataIn);
+    /**
+     * subEvaluate()
+     *
+     * Mandatory -- Called if the UDAF is running in a distributed
+     * fashion. Columnstore tries to run all aggregate functions
+     * distributed, depending on context.
+     *
+     * Perform an aggregation on rows partially aggregated by
+     * nextValue. Columnstore calls nextValue for each row on a
+     * given PM for a group (GROUP BY). subEvaluate is called on the
+     * UM to consolodate those values into a single instance of
+     * userData. Keep your aggregated totals in context's userData.
+     * The first time this is called for a group, reset() would have
+     * been called with this version of userData.
+     *
+     * Called for every partial data set in each group in GROUP BY.
+     *
+     * When subEvaluate has been called for all subAggregated data
+     * sets, Evaluate will be called.
+     *
+     * valIn (In) - This is a pointer to a memory block of the size
+     * set in setUserDataSize. It will contain the value of userData
+     * as seen in the last call to NextValue for a given PM.
+     *
+     */
+    virtual ReturnCode subEvaluate(mcsv1Context* context, const UserData* userDataIn);
 
-	/** 
-	 * evaluate() 
-	 *  
-	 * Mandatory. Get the aggregated value.
-	 *  
-	 * Called for every new group if UDAF GROUP BY, UDAnF partition 
-	 * or, in some cases, new Window Frame. 
-	 *  
-	 * Set the aggregated value into valOut. The datatype is assumed
-	 * to be the same as that set in the init() function; 
-	 *  
-	 * If the UDAF is running in a distributed fashion, evaluate is 
-	 * called after a series of subEvaluate calls. 
-	 *  
-	 * valOut (out) - Set the aggregated value here. The datatype is
-	 * assumed to be the same as that set in the init() function; 
-	 *  
-	 * To return a NULL value, don't assign to valOut.
-	 */
-	virtual ReturnCode evaluate(mcsv1Context* context, static_any::any& valOut);
+    /**
+     * evaluate()
+     *
+     * Mandatory. Get the aggregated value.
+     *
+     * Called for every new group if UDAF GROUP BY, UDAnF partition
+     * or, in some cases, new Window Frame.
+     *
+     * Set the aggregated value into valOut. The datatype is assumed
+     * to be the same as that set in the init() function;
+     *
+     * If the UDAF is running in a distributed fashion, evaluate is
+     * called after a series of subEvaluate calls.
+     *
+     * valOut (out) - Set the aggregated value here. The datatype is
+     * assumed to be the same as that set in the init() function;
+     *
+     * To return a NULL value, don't assign to valOut.
+     */
+    virtual ReturnCode evaluate(mcsv1Context* context, static_any::any& valOut);
 
-	/** 
-	 * dropValue() 
-	 *  
-	 * Optional -- If defined, the server will call this instead of 
-	 * reset for UDAnF. 
-	 *  
-	 * Don't implement if a UDAnF has one or more of the following: 
-	 * The UDAnF can't be used with a Window Frame
-	 * The UDAnF is not reversable in some way 
-	 * The UDAnF is not interested in optimal performance 
-	 *  
-	 * If not implemented, reset() followed by a series of 
-	 * nextValue() will be called for each movement of the Window 
-	 * Frame. 
-	 *  
-	 * If implemented, then each movement of the Window Frame will 
-	 * result in dropValue() being called for each row falling out 
-	 * of the Frame and nextValue() being called for each new row 
-	 * coming into the Frame. 
-	 *  
-	 * valsDropped (in) - a vector of the parameters from the row 
-	 * leaving the Frame 
-	 *  
-	 * dropValue() will not be called for unbounded/current row type
-	 * frames, as those are already optimized. 
-	 */
-	virtual ReturnCode dropValue(mcsv1Context* context, 
-								 std::vector<ColumnDatum>& valsDropped);
+    /**
+     * dropValue()
+     *
+     * Optional -- If defined, the server will call this instead of
+     * reset for UDAnF.
+     *
+     * Don't implement if a UDAnF has one or more of the following:
+     * The UDAnF can't be used with a Window Frame
+     * The UDAnF is not reversable in some way
+     * The UDAnF is not interested in optimal performance
+     *
+     * If not implemented, reset() followed by a series of
+     * nextValue() will be called for each movement of the Window
+     * Frame.
+     *
+     * If implemented, then each movement of the Window Frame will
+     * result in dropValue() being called for each row falling out
+     * of the Frame and nextValue() being called for each new row
+     * coming into the Frame.
+     *
+     * valsDropped (in) - a vector of the parameters from the row
+     * leaving the Frame
+     *
+     * dropValue() will not be called for unbounded/current row type
+     * frames, as those are already optimized.
+     */
+    virtual ReturnCode dropValue(mcsv1Context* context,
+                                 std::vector<ColumnDatum>& valsDropped);
 
 protected:
 };

utils/udfsdk/udfmysql.cpp

@@ -6,78 +6,94 @@ using namespace std;
 
 #include "idb_mysql.h"
 
-namespace {
+namespace
+{
 inline double cvtArgToDouble(int t, const char* v)
 {
-	double d = 0.0;
-	switch (t)
-	{
-	case INT_RESULT:
-		d = (double)(*((long long*)v));
-		break;
-	case REAL_RESULT:
-		d = *((double*)v);
-		break;
-	case DECIMAL_RESULT:
-	case STRING_RESULT:
-		d = strtod(v, 0);
-		break;
-	case ROW_RESULT:
-		break;
-	}
-	return d;
+    double d = 0.0;
+
+    switch (t)
+    {
+        case INT_RESULT:
+            d = (double)(*((long long*)v));
+            break;
+
+        case REAL_RESULT:
+            d = *((double*)v);
+            break;
+
+        case DECIMAL_RESULT:
+        case STRING_RESULT:
+            d = strtod(v, 0);
+            break;
+
+        case ROW_RESULT:
+            break;
+    }
+
+    return d;
 }
 inline long long cvtArgToInt(int t, const char* v)
 {
-	long long ll = 0;
-	switch (t)
-	{
-	case INT_RESULT:
-		ll = *((long long*)v);
-		break;
-	case REAL_RESULT:
-		ll = (long long)(*((double*)v));
-		break;
-	case DECIMAL_RESULT:
-	case STRING_RESULT:
-		ll = strtoll(v, 0, 0);
-		break;
-	case ROW_RESULT:
-		break;
-	}
-	return ll;
+    long long ll = 0;
+
+    switch (t)
+    {
+        case INT_RESULT:
+            ll = *((long long*)v);
+            break;
+
+        case REAL_RESULT:
+            ll = (long long)(*((double*)v));
+            break;
+
+        case DECIMAL_RESULT:
+        case STRING_RESULT:
+            ll = strtoll(v, 0, 0);
+            break;
+
+        case ROW_RESULT:
+            break;
+    }
+
+    return ll;
 }
 inline string cvtArgToString(int t, const char* v)
 {
-	string str;
-	switch (t)
-	{
-	case INT_RESULT:
-	{
-		long long ll;
-		ll = *((long long*)v);
-		ostringstream oss;
-		oss << ll;
-		str = oss.str();
-		break;
-	}
-	case REAL_RESULT:
-	{
-		double d;
-		d = *((double*)v);
-		ostringstream oss;
-		oss << d;
-		str = oss.str();
-		break;
-	}
-	case DECIMAL_RESULT:
-	case STRING_RESULT:
-		str = v;
-		break;
-	case ROW_RESULT:
-		break;
-	}
-	return str;
+    string str;
+
+    switch (t)
+    {
+        case INT_RESULT:
+        {
+            long long ll;
+            ll = *((long long*)v);
+            ostringstream oss;
+            oss << ll;
+            str = oss.str();
+            break;
+        }
+
+        case REAL_RESULT:
+        {
+            double d;
+            d = *((double*)v);
+            ostringstream oss;
+            oss << d;
+            str = oss.str();
+            break;
+        }
+
+        case DECIMAL_RESULT:
+        case STRING_RESULT:
+            str = v;
+            break;
+
+        case ROW_RESULT:
+            break;
+    }
+
+    return str;
 }
 }
 
@@ -95,7 +111,7 @@ inline string cvtArgToString(int t, const char* v)
  * ~/sql/udf_example.cc.
  *
  * Please note that the implementation of the function defined on the connector
- * will only be called when all the input arguments are constant. e.g., 
+ * will only be called when all the input arguments are constant. e.g.,
  * mcs_add(2,3). That way, the function does not run in a distributed fashion
  * and could be slow. If there is a need for the UDF function to run with
  * pure constant input, then one needs to put a implementation in the XXX
@@ -105,371 +121,375 @@ inline string cvtArgToString(int t, const char* v)
  */
 extern "C"
 {
-/**
- * MCS_ADD connector stub
- */
+    /**
+     * MCS_ADD connector stub
+     */
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-my_bool mcs_add_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
-{
-	if (args->arg_count != 2)
-	{
-		strcpy(message,"mcs_add() requires two argument");
-		return 1;
-	}
+    my_bool mcs_add_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
+    {
+        if (args->arg_count != 2)
+        {
+            strcpy(message, "mcs_add() requires two argument");
+            return 1;
+        }
 
-	return 0;
-}
+        return 0;
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void mcs_add_deinit(UDF_INIT* initid)
-{
-}
+    void mcs_add_deinit(UDF_INIT* initid)
+    {
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-double mcs_add(UDF_INIT *initid, UDF_ARGS *args, char *is_null, char *error)
-{
-	double op1, op2;
+    double mcs_add(UDF_INIT* initid, UDF_ARGS* args, char* is_null, char* error)
+    {
+        double op1, op2;
 
-	op1 = cvtArgToDouble(args->arg_type[0], args->args[0]);
-	op2 = cvtArgToDouble(args->arg_type[1], args->args[1]);
+        op1 = cvtArgToDouble(args->arg_type[0], args->args[0]);
+        op2 = cvtArgToDouble(args->arg_type[1], args->args[1]);
 
-	return op1+op2;
-}
+        return op1 + op2;
+    }
 
-/**
- * MCS_ISNULL connector stub
- */
- 
-#ifdef _MSC_VER
-__declspec(dllexport)
-#endif
-my_bool mcs_isnull_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
-{
-	if (args->arg_count != 1)
-	{
-		strcpy(message,"mcs_isnull() requires one argument");
-		return 1;
-	}
-
-	return 0;
-}
+    /**
+     * MCS_ISNULL connector stub
+     */
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void mcs_isnull_deinit(UDF_INIT* initid)
-{
-}	
+    my_bool mcs_isnull_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
+    {
+        if (args->arg_count != 1)
+        {
+            strcpy(message, "mcs_isnull() requires one argument");
+            return 1;
+        }
+
+        return 0;
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-long long mcs_isnull(UDF_INIT *initid, UDF_ARGS *args, char *is_null, char *error)
-{
-	return 0;
-}
+    void mcs_isnull_deinit(UDF_INIT* initid)
+    {
+    }
 
-/**
- * ALLNULL connector stub
- */
-struct allnull_data
-{
-  ulonglong	totalQuantity;
-  ulonglong	totalNulls;
-};
- 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-my_bool allnull_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
-{
-	struct allnull_data* data;
+    long long mcs_isnull(UDF_INIT* initid, UDF_ARGS* args, char* is_null, char* error)
+    {
+        return 0;
+    }
+
+    /**
+     * ALLNULL connector stub
+     */
+    struct allnull_data
+    {
+        ulonglong	totalQuantity;
+        ulonglong	totalNulls;
+    };
+
+#ifdef _MSC_VER
+    __declspec(dllexport)
+#endif
+    my_bool allnull_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
+    {
+        struct allnull_data* data;
 //	if (args->arg_count != 1)
 //	{
 //		strcpy(message,"allnull() requires one argument");
 //		return 1;
 //	}
 
-	if (!(data = (struct allnull_data*) malloc(sizeof(struct allnull_data))))
-	{
-		strmov(message,"Couldn't allocate memory");
-		return 1;
-	}
-	data->totalQuantity	= 0;
-	data->totalNulls	= 0;
+        if (!(data = (struct allnull_data*) malloc(sizeof(struct allnull_data))))
+        {
+            strmov(message, "Couldn't allocate memory");
+            return 1;
+        }
 
-	initid->ptr = (char*)data;
+        data->totalQuantity	= 0;
+        data->totalNulls	= 0;
 
-	return 0;
-}
+        initid->ptr = (char*)data;
+
+        return 0;
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void allnull_deinit(UDF_INIT* initid)
-{
-	free(initid->ptr);
-}	
+    void allnull_deinit(UDF_INIT* initid)
+    {
+        free(initid->ptr);
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-long long allnull(UDF_INIT* initid, UDF_ARGS* args __attribute__((unused)),
-				  char* is_null, char* error __attribute__((unused)))
-{
-	struct allnull_data* data = (struct allnull_data*)initid->ptr;
-	return data->totalQuantity > 0 && data->totalNulls == data->totalQuantity;
-}
+    long long allnull(UDF_INIT* initid, UDF_ARGS* args __attribute__((unused)),
+                      char* is_null, char* error __attribute__((unused)))
+    {
+        struct allnull_data* data = (struct allnull_data*)initid->ptr;
+        return data->totalQuantity > 0 && data->totalNulls == data->totalQuantity;
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void
-allnull_clear(UDF_INIT* initid, char* is_null __attribute__((unused)),
+    void
+    allnull_clear(UDF_INIT* initid, char* is_null __attribute__((unused)),
+                  char* message __attribute__((unused)))
+    {
+        struct allnull_data* data = (struct allnull_data*)initid->ptr;
+        data->totalQuantity = 0;
+        data->totalNulls    = 0;
+    }
+
+#ifdef _MSC_VER
+    __declspec(dllexport)
+#endif
+    void
+    allnull_add(UDF_INIT* initid, UDF_ARGS* args,
+                char* is_null,
+                char* message __attribute__((unused)))
+    {
+        struct allnull_data* data = (struct allnull_data*)initid->ptr;
+        const char* word = args->args[0];
+        data->totalQuantity++;
+
+        if (!word)
+        {
+            data->totalNulls++;
+        }
+    }
+
+    /**
+     * SSQ connector stub
+     */
+    struct ssq_data
+    {
+        double	sumsq;
+    };
+
+#ifdef _MSC_VER
+    __declspec(dllexport)
+#endif
+    my_bool ssq_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
+    {
+        struct ssq_data* data;
+
+        if (args->arg_count != 1)
+        {
+            strcpy(message, "ssq() requires one argument");
+            return 1;
+        }
+
+        if (!(data = (struct ssq_data*) malloc(sizeof(struct ssq_data))))
+        {
+            strmov(message, "Couldn't allocate memory");
+            return 1;
+        }
+
+        data->sumsq	= 0;
+
+        initid->ptr = (char*)data;
+        return 0;
+    }
+
+#ifdef _MSC_VER
+    __declspec(dllexport)
+#endif
+    void ssq_deinit(UDF_INIT* initid)
+    {
+        free(initid->ptr);
+    }
+
+#ifdef _MSC_VER
+    __declspec(dllexport)
+#endif
+    void
+    ssq_clear(UDF_INIT* initid, char* is_null __attribute__((unused)),
               char* message __attribute__((unused)))
-{
-	struct allnull_data* data = (struct allnull_data*)initid->ptr;
-	data->totalQuantity = 0;
-	data->totalNulls    = 0;
-}
+    {
+        struct ssq_data* data = (struct ssq_data*)initid->ptr;
+        data->sumsq = 0;
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void
-allnull_add(UDF_INIT* initid, UDF_ARGS* args,
+    void
+    ssq_add(UDF_INIT* initid, UDF_ARGS* args,
             char* is_null,
             char* message __attribute__((unused)))
-{
-	struct allnull_data* data = (struct allnull_data*)initid->ptr;
-	const char *word=args->args[0];
-	data->totalQuantity++;
-	if (!word)
-	{
-		data->totalNulls++;
-	}
-}
-
-/**
- * SSQ connector stub
- */
-struct ssq_data
-{
-  double	sumsq;
-};
- 
-#ifdef _MSC_VER
-__declspec(dllexport)
-#endif
-my_bool ssq_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
-{
-	struct ssq_data* data;
-	if (args->arg_count != 1)
-	{
-		strcpy(message,"ssq() requires one argument");
-		return 1;
-	}
-
-	if (!(data = (struct ssq_data*) malloc(sizeof(struct ssq_data))))
-	{
-		strmov(message,"Couldn't allocate memory");
-		return 1;
-	}
-	data->sumsq	= 0;
-
-	initid->ptr = (char*)data;
-	return 0;
-}
+    {
+        struct ssq_data* data = (struct ssq_data*)initid->ptr;
+        double val = cvtArgToDouble(args->arg_type[0], args->args[0]);
+        data->sumsq = val * val;
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void ssq_deinit(UDF_INIT* initid)
-{
-	free(initid->ptr);
-}	
-
-#ifdef _MSC_VER
-__declspec(dllexport)
-#endif
-void
-ssq_clear(UDF_INIT* initid, char* is_null __attribute__((unused)),
-              char* message __attribute__((unused)))
-{
-	struct ssq_data* data = (struct ssq_data*)initid->ptr;
-	data->sumsq = 0;
-}
-
-#ifdef _MSC_VER
-__declspec(dllexport)
-#endif
-void
-ssq_add(UDF_INIT* initid, UDF_ARGS* args,
-            char* is_null,
-            char* message __attribute__((unused)))
-{
-	struct ssq_data* data = (struct ssq_data*)initid->ptr;
-	double val = cvtArgToDouble(args->arg_type[0], args->args[0]);
-	data->sumsq = val*val;
-}
-
-#ifdef _MSC_VER
-__declspec(dllexport)
-#endif
-long long ssq(UDF_INIT* initid, UDF_ARGS* args __attribute__((unused)),
-				  char* is_null, char* error __attribute__((unused)))
-{
-	struct ssq_data* data = (struct ssq_data*)initid->ptr;
-	return data->sumsq;
-}
+    long long ssq(UDF_INIT* initid, UDF_ARGS* args __attribute__((unused)),
+                  char* is_null, char* error __attribute__((unused)))
+    {
+        struct ssq_data* data = (struct ssq_data*)initid->ptr;
+        return data->sumsq;
+    }
 
 //=======================================================================
 
-/**
- * MEDIAN connector stub
- */
+    /**
+     * MEDIAN connector stub
+     */
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-my_bool median_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
-{
-	if (args->arg_count != 1)
-	{
-		strcpy(message,"median() requires one argument");
-		return 1;
-	}
+    my_bool median_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
+    {
+        if (args->arg_count != 1)
+        {
+            strcpy(message, "median() requires one argument");
+            return 1;
+        }
 
-/*
-	if (!(data = (struct ssq_data*) malloc(sizeof(struct ssq_data))))
-	{
-		strmov(message,"Couldn't allocate memory");
-		return 1;
-	}
-	data->sumsq	= 0;
+        /*
+        	if (!(data = (struct ssq_data*) malloc(sizeof(struct ssq_data))))
+        	{
+        		strmov(message,"Couldn't allocate memory");
+        		return 1;
+        	}
+        	data->sumsq	= 0;
 
-	initid->ptr = (char*)data;
-*/
-	return 0;
-}
+        	initid->ptr = (char*)data;
+        */
+        return 0;
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void median_deinit(UDF_INIT* initid)
-{
+    void median_deinit(UDF_INIT* initid)
+    {
 //	free(initid->ptr);
-}	
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void
-median_clear(UDF_INIT* initid, char* is_null __attribute__((unused)),
-              char* message __attribute__((unused)))
-{
+    void
+    median_clear(UDF_INIT* initid, char* is_null __attribute__((unused)),
+                 char* message __attribute__((unused)))
+    {
 //	struct ssq_data* data = (struct ssq_data*)initid->ptr;
 //	data->sumsq = 0;
-}
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void
-median_add(UDF_INIT* initid, UDF_ARGS* args,
-            char* is_null,
-            char* message __attribute__((unused)))
-{
+    void
+    median_add(UDF_INIT* initid, UDF_ARGS* args,
+               char* is_null,
+               char* message __attribute__((unused)))
+    {
 //	struct ssq_data* data = (struct ssq_data*)initid->ptr;
 //	double val = cvtArgToDouble(args->arg_type[0], args->args[0]);
 //	data->sumsq = val*val;
-}
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-long long median(UDF_INIT* initid, UDF_ARGS* args __attribute__((unused)),
-				  char* is_null, char* error __attribute__((unused)))
-{
+    long long median(UDF_INIT* initid, UDF_ARGS* args __attribute__((unused)),
+                     char* is_null, char* error __attribute__((unused)))
+    {
 //	struct ssq_data* data = (struct ssq_data*)initid->ptr;
 //	return data->sumsq;
-	return 0;
-}
+        return 0;
+    }
 
-/**
- * avg_mode connector stub
- */
+    /**
+     * avg_mode connector stub
+     */
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-my_bool avg_mode_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
-{
-	if (args->arg_count != 1)
-	{
-		strcpy(message,"avg_mode() requires one argument");
-		return 1;
-	}
+    my_bool avg_mode_init(UDF_INIT* initid, UDF_ARGS* args, char* message)
+    {
+        if (args->arg_count != 1)
+        {
+            strcpy(message, "avg_mode() requires one argument");
+            return 1;
+        }
 
-/*
-	if (!(data = (struct ssq_data*) malloc(sizeof(struct ssq_data))))
-	{
-		strmov(message,"Couldn't allocate memory");
-		return 1;
-	}
-	data->sumsq	= 0;
+        /*
+        	if (!(data = (struct ssq_data*) malloc(sizeof(struct ssq_data))))
+        	{
+        		strmov(message,"Couldn't allocate memory");
+        		return 1;
+        	}
+        	data->sumsq	= 0;
 
-	initid->ptr = (char*)data;
-*/
-	return 0;
-}
+        	initid->ptr = (char*)data;
+        */
+        return 0;
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void avg_mode_deinit(UDF_INIT* initid)
-{
+    void avg_mode_deinit(UDF_INIT* initid)
+    {
 //	free(initid->ptr);
-}	
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void
-avg_mode_clear(UDF_INIT* initid, char* is_null __attribute__((unused)),
-              char* message __attribute__((unused)))
-{
+    void
+    avg_mode_clear(UDF_INIT* initid, char* is_null __attribute__((unused)),
+                   char* message __attribute__((unused)))
+    {
 //	struct ssq_data* data = (struct ssq_data*)initid->ptr;
 //	data->sumsq = 0;
-}
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-void
-avg_mode_add(UDF_INIT* initid, UDF_ARGS* args,
-            char* is_null,
-            char* message __attribute__((unused)))
-{
+    void
+    avg_mode_add(UDF_INIT* initid, UDF_ARGS* args,
+                 char* is_null,
+                 char* message __attribute__((unused)))
+    {
 //	struct ssq_data* data = (struct ssq_data*)initid->ptr;
 //	double val = cvtArgToDouble(args->arg_type[0], args->args[0]);
 //	data->sumsq = val*val;
-}
+    }
 
 #ifdef _MSC_VER
-__declspec(dllexport)
+    __declspec(dllexport)
 #endif
-long long avg_mode(UDF_INIT* initid, UDF_ARGS* args __attribute__((unused)),
-				  char* is_null, char* error __attribute__((unused)))
-{
+    long long avg_mode(UDF_INIT* initid, UDF_ARGS* args __attribute__((unused)),
+                       char* is_null, char* error __attribute__((unused)))
+    {
 //	struct ssq_data* data = (struct ssq_data*)initid->ptr;
 //	return data->sumsq;
-	return 0;
-}
+        return 0;
+    }
 }
 // vim:ts=4 sw=4:
 

utils/udfsdk/udfsdk.cpp

@@ -53,70 +53,71 @@ UDFSDK::~UDFSDK()
 
 /**
  * All UDF functions should be registered in the function map. They will be
- * picked up by the MariaDB ColumnStore F&E framework when the servers are started. 
+ * picked up by the MariaDB ColumnStore F&E framework when the servers are started.
  * That will make sure the UDF functions runs distributedly in ColumnStore
  * engines just like the internal ColumnStore functions.
  */
 FuncMap UDFSDK::UDFMap() const
 {
-	FuncMap fm;
-	
-	// first: function name
-	// second: Function pointer
-	// please use lower case for the function name. Because the names might be 
-	// case-insensitive in MariaDB depending on the setting. In such case,
-	// the function names passed to the interface is always in lower case.
-	fm["mcs_add"] = new MCS_add();
-	fm["mcs_isnull"] = new MCS_isnull();
-	
-	return fm;
+    FuncMap fm;
+
+    // first: function name
+    // second: Function pointer
+    // please use lower case for the function name. Because the names might be
+    // case-insensitive in MariaDB depending on the setting. In such case,
+    // the function names passed to the interface is always in lower case.
+    fm["mcs_add"] = new MCS_add();
+    fm["mcs_isnull"] = new MCS_isnull();
+
+    return fm;
 }
 
 /***************************************************************************
- * MCS_ADD implementation 
+ * MCS_ADD implementation
  *
  * OperationType() definition
  */
-CalpontSystemCatalog::ColType MCS_add::operationType (FunctionParm& fp, 
-                       CalpontSystemCatalog::ColType& resultType)
+CalpontSystemCatalog::ColType MCS_add::operationType (FunctionParm& fp,
+        CalpontSystemCatalog::ColType& resultType)
 {
-	// operation type of MCS_add is determined by the argument types
-	assert (fp.size() == 2);
-	CalpontSystemCatalog::ColType rt;
-	if (fp[0]->data()->resultType() == fp[1]->data()->resultType())
-	{
-		rt = fp[0]->data()->resultType();
-	}
-	else if (fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::CHAR ||
-		       fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::CHAR ||
-		       fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::VARCHAR ||
-		       fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::VARCHAR ||
-		       fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::DOUBLE ||
-		       fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::DOUBLE)
-	{
-		rt.colDataType = CalpontSystemCatalog::DOUBLE;
-		rt.colWidth = 8;
-	}
-	else if (fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::DATE ||
-		       fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::DATE ||
-		       fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::DATETIME ||
-		       fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::DATETIME)
-	{
-		rt.colDataType = CalpontSystemCatalog::BIGINT;
-		rt.colWidth = 8;
-	}
-	else if (fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::DECIMAL ||
+    // operation type of MCS_add is determined by the argument types
+    assert (fp.size() == 2);
+    CalpontSystemCatalog::ColType rt;
+
+    if (fp[0]->data()->resultType() == fp[1]->data()->resultType())
+    {
+        rt = fp[0]->data()->resultType();
+    }
+    else if (fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::CHAR ||
+             fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::CHAR ||
+             fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::VARCHAR ||
+             fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::VARCHAR ||
+             fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::DOUBLE ||
+             fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::DOUBLE)
+    {
+        rt.colDataType = CalpontSystemCatalog::DOUBLE;
+        rt.colWidth = 8;
+    }
+    else if (fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::DATE ||
+             fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::DATE ||
+             fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::DATETIME ||
+             fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::DATETIME)
+    {
+        rt.colDataType = CalpontSystemCatalog::BIGINT;
+        rt.colWidth = 8;
+    }
+    else if (fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::DECIMAL ||
              fp[0]->data()->resultType().colDataType == CalpontSystemCatalog::UDECIMAL ||
              fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::DECIMAL ||
              fp[1]->data()->resultType().colDataType == CalpontSystemCatalog::UDECIMAL)
-	{
-		rt.colDataType = CalpontSystemCatalog::DECIMAL;
-		rt.colWidth = 8;
-	}
-	else
-	{
+    {
+        rt.colDataType = CalpontSystemCatalog::DECIMAL;
+        rt.colWidth = 8;
+    }
+    else
+    {
         if (isUnsigned(fp[0]->data()->resultType().colDataType) ||
-            isUnsigned(fp[1]->data()->resultType().colDataType))
+                isUnsigned(fp[1]->data()->resultType().colDataType))
         {
             rt.colDataType = CalpontSystemCatalog::UBIGINT;
             rt.colWidth = 8;
@@ -126,8 +127,9 @@ CalpontSystemCatalog::ColType MCS_add::operationType (FunctionParm& fp,
             rt.colDataType = CalpontSystemCatalog::BIGINT;
             rt.colWidth = 8;
         }
-	}
-	return rt;
+    }
+
+    return rt;
 }
 
 /**
@@ -136,66 +138,73 @@ CalpontSystemCatalog::ColType MCS_add::operationType (FunctionParm& fp,
  * This API is called when an double value is needed to return from the UDF function
  */
 double MCS_add::getDoubleVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                             FunctionParm& parm,
+                             bool& isNull,
+                             CalpontSystemCatalog::ColType& op_ct)
 {
-	switch (op_ct.colDataType)
-	{
-		// The APIs for the evaluation of the function parameters are the same getXXXval()
-		// functions. However, only two arguments are passed in, the current
-		// row reference and the NULL indicator isNull.
-		case CalpontSystemCatalog::BIGINT:
-		case CalpontSystemCatalog::MEDINT:
-		case CalpontSystemCatalog::SMALLINT:
-		case CalpontSystemCatalog::TINYINT:
-			return ( parm[0]->data()->getIntVal(row, isNull) +
-			         parm[1]->data()->getIntVal(row, isNull));
+    switch (op_ct.colDataType)
+    {
+        // The APIs for the evaluation of the function parameters are the same getXXXval()
+        // functions. However, only two arguments are passed in, the current
+        // row reference and the NULL indicator isNull.
+        case CalpontSystemCatalog::BIGINT:
+        case CalpontSystemCatalog::MEDINT:
+        case CalpontSystemCatalog::SMALLINT:
+        case CalpontSystemCatalog::TINYINT:
+            return ( parm[0]->data()->getIntVal(row, isNull) +
+                     parm[1]->data()->getIntVal(row, isNull));
+
         case CalpontSystemCatalog::UBIGINT:
         case CalpontSystemCatalog::UMEDINT:
         case CalpontSystemCatalog::USMALLINT:
         case CalpontSystemCatalog::UTINYINT:
             return ( parm[0]->data()->getUintVal(row, isNull) +
                      parm[1]->data()->getUintVal(row, isNull));
-		case CalpontSystemCatalog::DOUBLE:
-			return ( parm[0]->data()->getDoubleVal(row, isNull) +
-			         parm[1]->data()->getDoubleVal(row, isNull));
-		case CalpontSystemCatalog::DECIMAL:
+
+        case CalpontSystemCatalog::DOUBLE:
+            return ( parm[0]->data()->getDoubleVal(row, isNull) +
+                     parm[1]->data()->getDoubleVal(row, isNull));
+
+        case CalpontSystemCatalog::DECIMAL:
         case CalpontSystemCatalog::UDECIMAL:
-		{
-			IDB_Decimal dec;
-			IDB_Decimal op1 = parm[0]->data()->getDecimalVal(row, isNull);
-			IDB_Decimal op2 = parm[1]->data()->getDecimalVal(row, isNull);
-			if (op1.scale == op2.scale)
-			{
-				dec.scale = op1.scale;
-			}
-			else if (op1.scale >= op2.scale)
-			{
-				dec.scale = op2.scale;
-				op1.value *= (int64_t)pow((double)10, op1.scale-op2.scale);
-			}
-			else
-			{
-				dec.scale = op1.scale;
-				op2.value *= (int64_t)pow((double)10, op2.scale-op1.scale);
-			}
-			dec.value = op1.value + op2.value;
-			return (double)(dec.value / pow((double)10, dec.scale));
-		}
-		default:
-			return ( parm[0]->data()->getDoubleVal(row, isNull) +
-			         parm[1]->data()->getDoubleVal(row, isNull));
-	}
-	return 0;
+        {
+            IDB_Decimal dec;
+            IDB_Decimal op1 = parm[0]->data()->getDecimalVal(row, isNull);
+            IDB_Decimal op2 = parm[1]->data()->getDecimalVal(row, isNull);
+
+            if (op1.scale == op2.scale)
+            {
+                dec.scale = op1.scale;
+            }
+            else if (op1.scale >= op2.scale)
+            {
+                dec.scale = op2.scale;
+                op1.value *= (int64_t)pow((double)10, op1.scale - op2.scale);
+            }
+            else
+            {
+                dec.scale = op1.scale;
+                op2.value *= (int64_t)pow((double)10, op2.scale - op1.scale);
+            }
+
+            dec.value = op1.value + op2.value;
+            return (double)(dec.value / pow((double)10, dec.scale));
+        }
+
+        default:
+            return ( parm[0]->data()->getDoubleVal(row, isNull) +
+                     parm[1]->data()->getDoubleVal(row, isNull));
+    }
+
+    return 0;
 }
 
 float MCS_add::getFloatVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                           FunctionParm& parm,
+                           bool& isNull,
+                           CalpontSystemCatalog::ColType& op_ct)
 {
-	return (float)getDoubleVal(row, parm, isNull, op_ct);
+    return (float)getDoubleVal(row, parm, isNull, op_ct);
 }
 
 /**
@@ -203,39 +212,39 @@ float MCS_add::getFloatVal(Row& row,
  *
  * This API is called when an integer value is needed to return from the UDF function
  *
- * Because the result type MCS_add is double(real), all the other API can simply call 
+ * Because the result type MCS_add is double(real), all the other API can simply call
  * getDoubleVal and apply the conversion. This method may not fit for all the UDF
  * implementation.
  */
 int64_t MCS_add::getIntVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                           FunctionParm& parm,
+                           bool& isNull,
+                           CalpontSystemCatalog::ColType& op_ct)
 {
-	return (int64_t)getDoubleVal(row, parm, isNull, op_ct);
+    return (int64_t)getDoubleVal(row, parm, isNull, op_ct);
 }
 
 string MCS_add::getStrVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                          FunctionParm& parm,
+                          bool& isNull,
+                          CalpontSystemCatalog::ColType& op_ct)
 {
-	// One will need a more efficient implementation if this API is frequently
-	// called for this UDF function. This code is for demonstration purpose.
-	ostringstream oss;
-	oss << getDoubleVal(row, parm, isNull, op_ct);
-	return oss.str();
+    // One will need a more efficient implementation if this API is frequently
+    // called for this UDF function. This code is for demonstration purpose.
+    ostringstream oss;
+    oss << getDoubleVal(row, parm, isNull, op_ct);
+    return oss.str();
 }
 
 IDB_Decimal MCS_add::getDecimalVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                                   FunctionParm& parm,
+                                   bool& isNull,
+                                   CalpontSystemCatalog::ColType& op_ct)
 {
-	IDB_Decimal dec;
-	dec.value = getIntVal(row, parm, isNull, op_ct);
-	dec.scale = 0;
-	return dec;
+    IDB_Decimal dec;
+    dec.value = getIntVal(row, parm, isNull, op_ct);
+    dec.scale = 0;
+    return dec;
 }
 
 /**
@@ -245,11 +254,11 @@ IDB_Decimal MCS_add::getDecimalVal(Row& row,
  * or throw a customized exception here.
  */
 int32_t MCS_add::getDateIntVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                               FunctionParm& parm,
+                               bool& isNull,
+                               CalpontSystemCatalog::ColType& op_ct)
 {
-	throw logic_error("Invalid API called for MCS_ADD");
+    throw logic_error("Invalid API called for MCS_ADD");
 }
 
 /**
@@ -259,32 +268,32 @@ int32_t MCS_add::getDateIntVal(Row& row,
  * or throw a customized exception here.
  */
 int64_t MCS_add::getDatetimeIntVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                                   FunctionParm& parm,
+                                   bool& isNull,
+                                   CalpontSystemCatalog::ColType& op_ct)
 {
-	return (int64_t)getDoubleVal(row, parm, isNull, op_ct);
+    return (int64_t)getDoubleVal(row, parm, isNull, op_ct);
 }
 
 bool MCS_add::getBoolVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                         FunctionParm& parm,
+                         bool& isNull,
+                         CalpontSystemCatalog::ColType& op_ct)
 {
-	return false;
+    return false;
 }
 
 /***************************************************************************
- * MCS_ISNULL implementation 
+ * MCS_ISNULL implementation
  *
  * OperationType() definition
  */
-CalpontSystemCatalog::ColType MCS_isnull::operationType (FunctionParm& fp, 
-                       CalpontSystemCatalog::ColType& resultType)
+CalpontSystemCatalog::ColType MCS_isnull::operationType (FunctionParm& fp,
+        CalpontSystemCatalog::ColType& resultType)
 {
-	// operation type of MCS_isnull should be the same as the argument type
-	assert (fp.size() == 1);
-	return fp[0]->data()->resultType();
+    // operation type of MCS_isnull should be the same as the argument type
+    assert (fp.size() == 1);
+    return fp[0]->data()->resultType();
 }
 
 /**
@@ -293,33 +302,36 @@ CalpontSystemCatalog::ColType MCS_isnull::operationType (FunctionParm& fp,
  * This would be the most commonly called API for MCS_isnull function
  */
 bool MCS_isnull::getBoolVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                            FunctionParm& parm,
+                            bool& isNull,
+                            CalpontSystemCatalog::ColType& op_ct)
 {
-	switch (op_ct.colDataType)
-	{
-		// For the purpose of this function, one does not need to get the value of
-		// the argument. One only need to know if the argument is NULL. The passed
-		// in parameter isNull will be set if the parameter is evaluated NULL. 
-		// Please note that before this function returns, isNull should be set to 
-		// false, otherwise the result of the function would be considered NULL,
-		// which is not possible for MCS_isnull().
-		case CalpontSystemCatalog::DECIMAL:
+    switch (op_ct.colDataType)
+    {
+        // For the purpose of this function, one does not need to get the value of
+        // the argument. One only need to know if the argument is NULL. The passed
+        // in parameter isNull will be set if the parameter is evaluated NULL.
+        // Please note that before this function returns, isNull should be set to
+        // false, otherwise the result of the function would be considered NULL,
+        // which is not possible for MCS_isnull().
+        case CalpontSystemCatalog::DECIMAL:
         case CalpontSystemCatalog::UDECIMAL:
-			parm[0]->data()->getDecimalVal(row, isNull);
-			break;
-		case CalpontSystemCatalog::CHAR:
-		case CalpontSystemCatalog::VARCHAR:
-			parm[0]->data()->getStrVal(row, isNull);
-			break;
-		default:
-			parm[0]->data()->getIntVal(row, isNull);
-	}
-	bool ret = isNull;
-	// It's important to reset isNull indicator.
-	isNull = false;
-	return ret;
+            parm[0]->data()->getDecimalVal(row, isNull);
+            break;
+
+        case CalpontSystemCatalog::CHAR:
+        case CalpontSystemCatalog::VARCHAR:
+            parm[0]->data()->getStrVal(row, isNull);
+            break;
+
+        default:
+            parm[0]->data()->getIntVal(row, isNull);
+    }
+
+    bool ret = isNull;
+    // It's important to reset isNull indicator.
+    isNull = false;
+    return ret;
 }
 
 /**
@@ -328,19 +340,19 @@ bool MCS_isnull::getBoolVal(Row& row,
  * This API is called when a double value is needed to return from the UDF function
  */
 double MCS_isnull::getDoubleVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                                FunctionParm& parm,
+                                bool& isNull,
+                                CalpontSystemCatalog::ColType& op_ct)
 {
-	return (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
+    return (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
 }
 
 float MCS_isnull::getFloatVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                              FunctionParm& parm,
+                              bool& isNull,
+                              CalpontSystemCatalog::ColType& op_ct)
 {
-	return (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
+    return (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
 }
 
 /**
@@ -348,54 +360,54 @@ float MCS_isnull::getFloatVal(Row& row,
  *
  * This API is called when an integer value is needed to return from the UDF function
  *
- * Because the result type MCS_add is double(real), all the other API can simply call 
+ * Because the result type MCS_add is double(real), all the other API can simply call
  * getDoubleVal and apply the conversion. This method may not fit for all the UDF
  * implementations.
  */
 int64_t MCS_isnull::getIntVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                              FunctionParm& parm,
+                              bool& isNull,
+                              CalpontSystemCatalog::ColType& op_ct)
 {
-	return (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
+    return (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
 }
 
 string MCS_isnull::getStrVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                             FunctionParm& parm,
+                             bool& isNull,
+                             CalpontSystemCatalog::ColType& op_ct)
 {
-	// This needs to be more efficient if this API will be frequently
-	// called for this UDF function. 
-		return (getBoolVal(row, parm, isNull, op_ct) ? "1" : "0");
+    // This needs to be more efficient if this API will be frequently
+    // called for this UDF function.
+    return (getBoolVal(row, parm, isNull, op_ct) ? "1" : "0");
 
 }
 
 IDB_Decimal MCS_isnull::getDecimalVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                                      FunctionParm& parm,
+                                      bool& isNull,
+                                      CalpontSystemCatalog::ColType& op_ct)
 {
-	IDB_Decimal dec;
-	dec.value = (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
-	dec.scale = 0;
-	return dec;
+    IDB_Decimal dec;
+    dec.value = (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
+    dec.scale = 0;
+    return dec;
 }
 
 int32_t MCS_isnull::getDateIntVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                                  FunctionParm& parm,
+                                  bool& isNull,
+                                  CalpontSystemCatalog::ColType& op_ct)
 {
-	return (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
+    return (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
 }
 
 int64_t MCS_isnull::getDatetimeIntVal(Row& row,
-							FunctionParm& parm,
-							bool& isNull,
-							CalpontSystemCatalog::ColType& op_ct)
+                                      FunctionParm& parm,
+                                      bool& isNull,
+                                      CalpontSystemCatalog::ColType& op_ct)
 {
-	return (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
+    return (getBoolVal(row, parm, isNull, op_ct) ? 1 : 0);
 }
 
 }

utils/udfsdk/udfsdk.h

@@ -22,7 +22,7 @@
 *
 ***********************************************************************/
 
-/** 
+/**
  * MariaDB ColumnStore interface for writing a user defined function (UDF).
  *
  * The basic steps are:
@@ -62,18 +62,18 @@ namespace udfsdk
 class UDFSDK
 {
 public:
-	EXPORT UDFSDK();
+    EXPORT UDFSDK();
 
-	EXPORT ~UDFSDK();
+    EXPORT ~UDFSDK();
 
-	EXPORT funcexp::FuncMap UDFMap() const;
+    EXPORT funcexp::FuncMap UDFMap() const;
 
 protected:
 
 private:
-	//defaults okay
-	//UDFSDK(const UDFSDK& rhs);
-	//UDFSDK& operator=(const UDFSDK& rhs);
+    //defaults okay
+    //UDFSDK(const UDFSDK& rhs);
+    //UDFSDK& operator=(const UDFSDK& rhs);
 
 };
 
@@ -85,18 +85,18 @@ private:
  * The function interface is defined here. All UDF functions are derived from
  * class funcexp::Func. A set of getXXXval interface APIs are declared in the
  * parent class Func, which will be called by IDB function and expression (F&E)
- * framwork when evaluating the function. Which API to be called depends on 
- * the context of the function in the SQL query, i.e., the result type that 
+ * framwork when evaluating the function. Which API to be called depends on
+ * the context of the function in the SQL query, i.e., the result type that
  * the function is expected to return.
  *
  * For example, given the following two queries, different APIs will be called
  * to evaluate the function MCS_add.
  *
- * select MCS_add(int1, int2) from t1; 
+ * select MCS_add(int1, int2) from t1;
  * getDoubleVal() is called, because the result type of MCS_add is DOUBLE(real).
  *
  * select substr(string1, int1, MCS_add(int1+int2));
- * getIntVal() will be called, because MCS_add() is passed as the third argument 
+ * getIntVal() will be called, because MCS_add() is passed as the third argument
  * to substr function, and an integer result is expected.
  *
  * If one API is not implemented but called for a function, IDB-5001 error will
@@ -105,126 +105,126 @@ private:
 class MCS_add : public funcexp::Func
 {
 public:
-	/*
-	 * Constructor. Pass the function name to the base constructor. 
-	 */
-	MCS_add() : Func("mcs_add") {}
-	
-	/*
-	 * Destructor. MCS_add does not need to do anything here to clean up.
-	 */
-	virtual ~MCS_add() {}
-	
-	/**
-	 * Decide on the function's operation type
-	 * 
-	 * Operation type decides which API needs to be called for each function
-	 * parameter. Sometimes it is obvious. e.g. for function substr (c1, c2, c3),
-	 * one knows that getStrVal(), getIntVal() and getIntVal() should be called for
-	 * the three parameters in sequence. In that case, a dummy type can be returned
-	 * because it won't be used in the function implementation. Sometimes the 
-	 * operation type is decided by the data type of the function parameters. 
-	 * e.g., isnull(c1) function, one should call the corresponding getXXXval()
-	 * function that in compatible with the result type of c1. 
-	 *
-	 * @parm fp vector of function parameters
-	 *       Each element is a boost::shared_ptr of execplan::ParseTree. class 
-	 *       ParseTree is defined in ~/dbcon/execplan/parsetree.h
-	 * @parm resultType result type of this function 
-	 *       Sometimes it may affect the operation type, but most of the time it
-	 *       can be ignored. Struct ColType is defined in ~/dbcon/execplan/calpontsystemcatalog.h
-	 * @return operation type for this function
-	 * 
-	 * This function is called only one from the connector. Once it's determined, it
-	 * will be passed to the getXXXval() APIs during function evaluation.
-	 */
-	execplan::CalpontSystemCatalog::ColType operationType(funcexp::FunctionParm& fp, execplan::CalpontSystemCatalog::ColType& resultType);
+    /*
+     * Constructor. Pass the function name to the base constructor.
+     */
+    MCS_add() : Func("mcs_add") {}
 
-	/**
-	 * Returns an integer result of this function. 
-	 * All the getXXXvalue APIs take the same arguments. They will be called
-	 * for every row in the result set when the function is being evaluated.
-	 * So these functions needs to be efficient.
-	 *
-	 * @parm row reference of the current row
-	 * @parm fp function parameters
-	 * @parm isNull NULL indicator throughout this function evaluation.
-	 *       the same reference is passed to all the function argument
-	 *       evaluations. One always need to know if any argument is NULL
-	 *       to decide the result of the function. It's explained in detail
-	 *       in MCS_isnull() function example.
-	 * @parm op_ct the operation type that is determined in operationType().
-	 * 
-	 */
-	virtual int64_t getIntVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns a double result of this function.
-	 */
+    /*
+     * Destructor. MCS_add does not need to do anything here to clean up.
+     */
+    virtual ~MCS_add() {}
 
-	virtual double getDoubleVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
+    /**
+     * Decide on the function's operation type
+     *
+     * Operation type decides which API needs to be called for each function
+     * parameter. Sometimes it is obvious. e.g. for function substr (c1, c2, c3),
+     * one knows that getStrVal(), getIntVal() and getIntVal() should be called for
+     * the three parameters in sequence. In that case, a dummy type can be returned
+     * because it won't be used in the function implementation. Sometimes the
+     * operation type is decided by the data type of the function parameters.
+     * e.g., isnull(c1) function, one should call the corresponding getXXXval()
+     * function that in compatible with the result type of c1.
+     *
+     * @parm fp vector of function parameters
+     *       Each element is a boost::shared_ptr of execplan::ParseTree. class
+     *       ParseTree is defined in ~/dbcon/execplan/parsetree.h
+     * @parm resultType result type of this function
+     *       Sometimes it may affect the operation type, but most of the time it
+     *       can be ignored. Struct ColType is defined in ~/dbcon/execplan/calpontsystemcatalog.h
+     * @return operation type for this function
+     *
+     * This function is called only one from the connector. Once it's determined, it
+     * will be passed to the getXXXval() APIs during function evaluation.
+     */
+    execplan::CalpontSystemCatalog::ColType operationType(funcexp::FunctionParm& fp, execplan::CalpontSystemCatalog::ColType& resultType);
 
-	/**
-	 * Returns a float result of this function.
-	 */
-	virtual float getFloatVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns a string result of this function.
-	 */
-	virtual std::string getStrVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns a bool result of this function.
-	 */
-	virtual bool getBoolVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns a decimal result of this function.
-	 *
-	 * IDB_Decimal is defined in ~/execplan/treenode.h
-	 */
-	virtual execplan::IDB_Decimal getDecimalVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns an integer representation of a date result of the function.
-	 *
-	 * Check the date/time functions in ~/utils/funcexp for implementation
-	 * example of this API.
-	 */
-	virtual int32_t getDateIntVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns an integer representation of a datetime result of the function.
-	 *
-	 * Check the date/time functions in ~/utils/funcexp for implementation
-	 * example of this API.
-	 */
-	virtual int64_t getDatetimeIntVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
+    /**
+     * Returns an integer result of this function.
+     * All the getXXXvalue APIs take the same arguments. They will be called
+     * for every row in the result set when the function is being evaluated.
+     * So these functions needs to be efficient.
+     *
+     * @parm row reference of the current row
+     * @parm fp function parameters
+     * @parm isNull NULL indicator throughout this function evaluation.
+     *       the same reference is passed to all the function argument
+     *       evaluations. One always need to know if any argument is NULL
+     *       to decide the result of the function. It's explained in detail
+     *       in MCS_isnull() function example.
+     * @parm op_ct the operation type that is determined in operationType().
+     *
+     */
+    virtual int64_t getIntVal(rowgroup::Row& row,
+                              funcexp::FunctionParm& fp,
+                              bool& isNull,
+                              execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns a double result of this function.
+     */
+
+    virtual double getDoubleVal(rowgroup::Row& row,
+                                funcexp::FunctionParm& fp,
+                                bool& isNull,
+                                execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns a float result of this function.
+     */
+    virtual float getFloatVal(rowgroup::Row& row,
+                              funcexp::FunctionParm& fp,
+                              bool& isNull,
+                              execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns a string result of this function.
+     */
+    virtual std::string getStrVal(rowgroup::Row& row,
+                                  funcexp::FunctionParm& fp,
+                                  bool& isNull,
+                                  execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns a bool result of this function.
+     */
+    virtual bool getBoolVal(rowgroup::Row& row,
+                            funcexp::FunctionParm& fp,
+                            bool& isNull,
+                            execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns a decimal result of this function.
+     *
+     * IDB_Decimal is defined in ~/execplan/treenode.h
+     */
+    virtual execplan::IDB_Decimal getDecimalVal(rowgroup::Row& row,
+            funcexp::FunctionParm& fp,
+            bool& isNull,
+            execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns an integer representation of a date result of the function.
+     *
+     * Check the date/time functions in ~/utils/funcexp for implementation
+     * example of this API.
+     */
+    virtual int32_t getDateIntVal(rowgroup::Row& row,
+                                  funcexp::FunctionParm& fp,
+                                  bool& isNull,
+                                  execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns an integer representation of a datetime result of the function.
+     *
+     * Check the date/time functions in ~/utils/funcexp for implementation
+     * example of this API.
+     */
+    virtual int64_t getDatetimeIntVal(rowgroup::Row& row,
+                                      funcexp::FunctionParm& fp,
+                                      bool& isNull,
+                                      execplan::CalpontSystemCatalog::ColType& op_ct);
 };
 
 /**
@@ -235,92 +235,92 @@ public:
 class MCS_isnull : public funcexp::Func
 {
 public:
-	/*
-	 * Constructor. Pass the function name to the base constructor. 
-	 */
-	MCS_isnull() : Func("mcs_isnull") {}
-	
-	/*
-	 * Destructor. MCS_add does not need to do anything here to clean up.
-	 */
-	virtual ~MCS_isnull() {}
-	
-	/**
-	 * Decide on the function's operation type
-	 */
-	execplan::CalpontSystemCatalog::ColType operationType(funcexp::FunctionParm& fp, execplan::CalpontSystemCatalog::ColType& resultType);
+    /*
+     * Constructor. Pass the function name to the base constructor.
+     */
+    MCS_isnull() : Func("mcs_isnull") {}
 
-	/**
-	 * Returns an integer result of this function.  
-	 */
-	virtual int64_t getIntVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns a double result of this function.
-	 */
-	virtual double getDoubleVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
+    /*
+     * Destructor. MCS_add does not need to do anything here to clean up.
+     */
+    virtual ~MCS_isnull() {}
 
-	/**
-	 * Returns a float result of this function.
-	 */
-	virtual float getFloatVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns a string result of this function.
-	 */
-	virtual std::string getStrVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns a bool result of this function.
-	 */
-	virtual bool getBoolVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns a decimal result of this function.
-	 *
-	 * IDB_Decimal is defined in ~/execplan/treenode.h
-	 */
-	virtual execplan::IDB_Decimal getDecimalVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
+    /**
+     * Decide on the function's operation type
+     */
+    execplan::CalpontSystemCatalog::ColType operationType(funcexp::FunctionParm& fp, execplan::CalpontSystemCatalog::ColType& resultType);
 
-	/**
-	 * Returns an integer representation of a date result of the function.
-	 *
-	 * Check the date/time functions in ~/utils/funcexp for implementation
-	 * example of this API.
-	 */
-	virtual int32_t getDateIntVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);
-	
-	/**
-	 * Returns an integer representation of a datetime result of the function.
-	 *
-	 * Check the date/time functions in ~/utils/funcexp for implementation
-	 * example of this API.
-	 */
-	virtual int64_t getDatetimeIntVal(rowgroup::Row& row,
-								funcexp::FunctionParm& fp,
-								bool& isNull,
-								execplan::CalpontSystemCatalog::ColType& op_ct);									
+    /**
+     * Returns an integer result of this function.
+     */
+    virtual int64_t getIntVal(rowgroup::Row& row,
+                              funcexp::FunctionParm& fp,
+                              bool& isNull,
+                              execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns a double result of this function.
+     */
+    virtual double getDoubleVal(rowgroup::Row& row,
+                                funcexp::FunctionParm& fp,
+                                bool& isNull,
+                                execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns a float result of this function.
+     */
+    virtual float getFloatVal(rowgroup::Row& row,
+                              funcexp::FunctionParm& fp,
+                              bool& isNull,
+                              execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns a string result of this function.
+     */
+    virtual std::string getStrVal(rowgroup::Row& row,
+                                  funcexp::FunctionParm& fp,
+                                  bool& isNull,
+                                  execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns a bool result of this function.
+     */
+    virtual bool getBoolVal(rowgroup::Row& row,
+                            funcexp::FunctionParm& fp,
+                            bool& isNull,
+                            execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns a decimal result of this function.
+     *
+     * IDB_Decimal is defined in ~/execplan/treenode.h
+     */
+    virtual execplan::IDB_Decimal getDecimalVal(rowgroup::Row& row,
+            funcexp::FunctionParm& fp,
+            bool& isNull,
+            execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns an integer representation of a date result of the function.
+     *
+     * Check the date/time functions in ~/utils/funcexp for implementation
+     * example of this API.
+     */
+    virtual int32_t getDateIntVal(rowgroup::Row& row,
+                                  funcexp::FunctionParm& fp,
+                                  bool& isNull,
+                                  execplan::CalpontSystemCatalog::ColType& op_ct);
+
+    /**
+     * Returns an integer representation of a datetime result of the function.
+     *
+     * Check the date/time functions in ~/utils/funcexp for implementation
+     * example of this API.
+     */
+    virtual int64_t getDatetimeIntVal(rowgroup::Row& row,
+                                      funcexp::FunctionParm& fp,
+                                      bool& isNull,
+                                      execplan::CalpontSystemCatalog::ColType& op_ct);
 };
 
 }