Merge branch 'develop' into mcol-1607

README

@@ -1,9 +1,9 @@
 This is MariaDB ColumnStore 1.2
-MariaDB ColumnStore 1.2 is the development version of MariaDB ColumnStore. 
+MariaDB ColumnStore 1.2 is the GA version of MariaDB ColumnStore. 
 It is built by porting InfiniDB 4.6.7 on MariaDB 10.2 and adding entirely 
 new features not found anywhere else.
 
-MariaDB ColumnStore 1.2 is a pre-release.
+MariaDB ColumnStore 1.2 is a GA.
 
 Additional features will be pushed in future releases. 
 A few things to notice:

README.md

@@ -1,10 +1,10 @@
 # MariaDB ColumnStore Storage/Execution engine 1.2
 
-MariaDB ColumnStore 1.2 is the development version of MariaDB ColumnStore. 
+MariaDB ColumnStore 1.2 is the GA version of MariaDB ColumnStore. 
 It is built by porting InfiniDB 4.6.7 on MariaDB 10.2 and adding entirely 
 new features not found anywhere else.
 
-# MariaDB ColumnStore 1.2 is a pre-release. 
+# MariaDB ColumnStore 1.2 is a GA release. 
 
 - Do not use pre-releases on production systems.
 

dbcon/execplan/windowfunctioncolumn.cpp

@@ -233,7 +233,8 @@ WindowFunctionColumn::WindowFunctionColumn( const WindowFunctionColumn& rhs, con
     fFunctionName(rhs.functionName()),
     fFunctionParms(rhs.functionParms()),
     fPartitions (rhs.partitions()),
-    fOrderBy (rhs.orderBy())
+    fOrderBy (rhs.orderBy()),
+    udafContext(rhs.getUDAFContext())
 {}
 
 const string WindowFunctionColumn::toString() const

dbcon/execplan/windowfunctioncolumn.h

@@ -140,6 +140,10 @@ public:
     {
         return udafContext;
     }
+    const mcsv1sdk::mcsv1Context& getUDAFContext() const
+    {
+        return udafContext;
+    }
 
 private:
     /**

dbcon/mysql/ha_calpont.cpp

@@ -1161,9 +1161,9 @@ void check_walk(const Item* item, void* arg)
     {
         case Item::FUNC_ITEM:
         {
-            Item_func* ifp = (Item_func*)item;
+            const Item_func* ifp = static_cast<const Item_func*>(item);
 
-            if ( ifp->functype() != Item_func::EQ_FUNC )
+            if ( ifp->functype() != Item_func::EQ_FUNC ) // NON-equi JOIN
             {
                 if ( ifp->argument_count() == 2 &&
                     ifp->arguments()[0]->type() == Item::FIELD_ITEM &&
@@ -1178,11 +1178,36 @@ void check_walk(const Item* item, void* arg)
                         return;
                     }
                 }
+                else // IN + correlated subquery 
+                {
+                    if ( ifp->functype() == Item_func::NOT_FUNC
+                        && ifp->arguments()[0]->type() == Item::EXPR_CACHE_ITEM )
+                    {
+                        check_walk(ifp->arguments()[0], arg);
+                    }
+                }
             }
             break;
         }
         
-        case Item::COND_ITEM:
+        case Item::EXPR_CACHE_ITEM: // IN + correlated subquery 
+        {
+            const Item_cache_wrapper* icw = static_cast<const Item_cache_wrapper*>(item);
+            if ( icw->get_orig_item()->type() == Item::FUNC_ITEM )
+            {
+                const Item_func *ifp = static_cast<const Item_func*>(icw->get_orig_item());
+                if ( ifp->argument_count() == 2 &&
+                    ( ifp->arguments()[0]->type() == Item::Item::SUBSELECT_ITEM
+                    || ifp->arguments()[1]->type() == Item::Item::SUBSELECT_ITEM ))
+                {
+                    *unsupported_feature = true;
+                    return;
+                }
+            }
+            break;
+        }
+
+        case Item::COND_ITEM: // OR in cods is unsupported yet
         {
             Item_cond* icp = (Item_cond*)item;
             if ( is_cond_or(icp) )
@@ -1198,8 +1223,6 @@ void check_walk(const Item* item, void* arg)
     }
 }
 
-#include <iostream>
-
 /*@brief  create_calpont_group_by_handler- Creates handler*/
 /***********************************************************
  * DESCRIPTION:
@@ -1221,24 +1244,25 @@ static group_by_handler*
 create_calpont_group_by_handler(THD* thd, Query* query)
 {
     ha_calpont_group_by_handler* handler = NULL;
-    LEX* lex = thd->lex;
+    // same as thd->lex->current_select
-    SELECT_LEX *select_lex = &lex->select_lex;
+    SELECT_LEX *select_lex = query->from->select_lex;
 
     // Create a handler if query is valid. See comments for details.
     if ( thd->infinidb_vtable.vtable_state == THD::INFINIDB_DISABLE_VTABLE
         && ( thd->variables.infinidb_vtable_mode == 0
             || thd->variables.infinidb_vtable_mode == 2 )
-        && ( query->group_by || thd->lex->select_lex.with_sum_func ) )
+        && ( query->group_by || select_lex->with_sum_func ) )
     {
         bool unsupported_feature = false;
         // Impossible HAVING or WHERE
-        if ( ( select_lex->having && select_lex->having_value == Item::COND_FALSE )
+        if ( ( query->having && select_lex->having_value == Item::COND_FALSE )
-            || ( select_lex->cond_value && select_lex->cond_value == Item::COND_FALSE ) )
+            || ( select_lex->cond_count > 0
+                && select_lex->cond_value == Item::COND_FALSE ) )
         {
             unsupported_feature = true;
         }
 
-        // Unsupported JOIN conditions check.
+        // Unsupported conditions check.
         if ( !unsupported_feature )
         {
             JOIN *join = select_lex->join;
@@ -1248,7 +1272,7 @@ create_calpont_group_by_handler(THD* thd, Query* query)
                 icp = reinterpret_cast<Item_cond*>(join->conds);
 
             if ( unsupported_feature == false
-                && join->table_count > 1 && icp )
+                && icp )
             {
                 icp->traverse_cond(check_walk, &unsupported_feature, Item::POSTFIX);
             }
@@ -1265,8 +1289,6 @@ create_calpont_group_by_handler(THD* thd, Query* query)
         }
     }
 
-    std::cerr << "create_calpont_group_by_handler handler " << handler << std::endl;
-
     return handler;
 }
 

dbcon/mysql/ha_calpont_execplan.cpp

@@ -3813,7 +3813,7 @@ FunctionColumn* buildCaseFunction(Item_func* item, gp_walk_info& gwi, bool& nonS
     FuncExp* funcexp = FuncExp::instance();
     string funcName = "case_simple";
 
-    if (strcasecmp(((Item_func_case*)item)->case_type(), "searched") == 0)
+    if (item->functype() == Item_func::CASE_SEARCHED_FUNC)
     {
         funcName = "case_searched";
     }
@@ -3857,7 +3857,7 @@ FunctionColumn* buildCaseFunction(Item_func* item, gp_walk_info& gwi, bool& nonS
         // some cpu cycles trying to build a ReturnedColumn as below.
         // Every even numbered arg is a WHEN. In between are the THEN.
         // An odd number of args indicates an ELSE residing in the last spot.
-        if (funcName == "case_searched" &&
+        if ((item->functype() == Item_func::CASE_SEARCHED_FUNC) &&
                 (i < arg_offset))
         {
             // MCOL-1472 Nested CASE with an ISNULL predicate. We don't want the predicate 
@@ -8248,13 +8248,11 @@ int cp_get_table_plan(THD* thd, SCSEP& csep, cal_table_info& ti)
 
 int cp_get_group_plan(THD* thd, SCSEP& csep, cal_impl_if::cal_group_info& gi)
 {
-    LEX* lex = thd->lex;
-    idbassert(lex != 0);
 
-    SELECT_LEX select_lex = lex->select_lex;
+    SELECT_LEX *select_lex = gi.groupByTables->select_lex;
     gp_walk_info gwi;
     gwi.thd = thd;
-    int status = getGroupPlan(gwi, select_lex, csep, gi);
+    int status = getGroupPlan(gwi, *select_lex, csep, gi);
 
 #ifdef DEBUG_WALK_COND
     cerr << "---------------- cp_get_group_plan EXECUTION PLAN ----------------" << endl;

genii.vpw

@@ -4,12 +4,9 @@
 		<Project File="tools/configMgt/autoConfigure.vpj"/>
 		<Project File="utils/batchloader/batchloader.vpj"/>
 		<Project File="primitives/blockcache/blockcache.vpj"/>
-		<Project File="utils/boost_idb/boost_idb.vpj"/>
 		<Project File="versioning/BRM/brm.vpj"/>
 		<Project File="writeengine/bulk/bulk.vpj"/>
 		<Project File="utils/cacheutils/cacheutils.vpj"/>
-		<Project File="oamapps/calpont-console/calpontConsole.vpj"/>
-		<Project File="oamapps/calpontDB/calpontDBWrite.vpj"/>
 		<Project File="tools/dbloadxml/colxml.vpj"/>
 		<Project File="utils/common/common.vpj"/>
 		<Project File="utils/compress/compress.vpj"/>

oam/etc/Columnstore.xml

@@ -427,12 +427,9 @@
 	</DBRM_Worker10>
 	<DBBC>
 		<!-- The percentage of RAM to use for the disk block cache. Defaults to 70% -->
+        <!-- Alternatively, this can be specified in absolute terms using
+             the suffixes 'm' or 'g' to denote size in megabytes or gigabytes.-->
 		<!-- <NumBlocksPct>70</NumBlocksPct> -->
-
-        <!-- If preferred, the user can specify cache size in terms of megabytes.
-             If both Pct and MB are specified, PrimProc will use Pct. -->
-        <!-- <NumBlocksInMB>2048</NumBlocksInMB> -->
-
 		<!-- <NumThreads>16</NumThreads> --> <!-- 1-256.  Default is 16. -->
 		<NumCaches>1</NumCaches><!-- # of parallel caches to instantiate -->
 		<IOMTracing>0</IOMTracing>

oam/etc/Columnstore.xml.singleserver

@@ -419,12 +419,9 @@
 	</DBRM_Worker10>
 	<DBBC>
 		<!-- The percentage of RAM to use for the disk block cache. Defaults to 70% -->
+        <!-- Alternatively, this can be specified in absolute terms using
+             the suffixes 'm' or 'g' to denote size in megabytes or gigabytes.-->
 		<NumBlocksPct>50</NumBlocksPct>
-
-        <!-- If preferred, the user can specify cache size in terms of megabytes.
-             If both Pct and MB are specified, PrimProc will use Pct. -->
-        <!-- <NumBlocksInMB>2048</NumBlocksInMB> -->
-
 		<!-- <NumThreads>16</NumThreads> --> <!-- 1-256.  Default is 16. -->
 		<NumCaches>1</NumCaches><!-- # of parallel caches to instantiate -->
 		<IOMTracing>0</IOMTracing>

oam/install_scripts/columnstoreAlias

@@ -4,7 +4,7 @@ alias mcsmysql='/usr/local/mariadb/columnstore/mysql/bin/mysql --defaults-extra-
 alias ma=/usr/local/mariadb/columnstore/bin/mcsadmin
 alias mcsadmin=/usr/local/mariadb/columnstore/bin/mcsadmin
 alias cpimport=/usr/local/mariadb/columnstore/bin/cpimport
-alias home='cd /usr/local/mariadb/columnstore'
+alias mcshome='cd /usr/local/mariadb/columnstore'
 alias log='cd /var/log/mariadb/columnstore/'
 alias core='cd /var/log/mariadb/columnstore/corefiles'
 alias tmsg='tail -f /var/log/messages'
@@ -14,4 +14,4 @@ alias terror='tail -f /var/log/mariadb/columnstore/err.log'
 alias twarning='tail -f /var/log/mariadb/columnstore/warning.log'
 alias tcrit='tail -f /var/log/mariadb/columnstore/crit.log'
 alias dbrm='cd /usr/local/mariadb/columnstore/data1/systemFiles/dbrm'
-alias module='cat /usr/local/mariadb/columnstore/local/module'
+alias mcsmodule='cat /usr/local/mariadb/columnstore/local/module'

oam/install_scripts/post-install

@@ -315,14 +315,6 @@ if [ -z "aws" ]; then
 	$installdir/bin/MCSgetCredentials.sh >/dev/null 2>&1
 fi
 
-#log install message
-test -f $installdir/post/functions && . $installdir/post/functions
-if [ $user = "root" ]; then
-	$installdir/bin/cplogger -i 19 "***** MariaDB Columnstore Installed *****"
-else
-	LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$installdir/lib $installdir/bin/cplogger -i 19 "***** MariaDB Columnstore Installed *****"
-fi
-
 #setup hadoop
 hadoop=`which hadoop 2>/dev/null`
 if [ -z "$hadoop" ]; then

oam/install_scripts/syslogSetup.sh

@@ -11,12 +11,26 @@ prefix=/usr/local
 installdir=$prefix/mariadb/columnstore
 syslog_conf=nofile
 rsyslog7=0
+non_root_user=no
 
 user=`whoami 2>/dev/null`
 
+#set default names
 groupname=adm
 username=syslog
 
+# determine username/groupname
+
+if [ -f /var/log/messages ]; then
+	  username=`stat -c "%U %G" /var/log/messages | awk '{print $1}'`
+	  groupname=`stat -c "%U %G" /var/log/messages | awk '{print $2}'`
+fi
+
+if [ -f /var/log/syslog ]; then
+	  username=`stat -c "%U %G" /var/log/syslog | awk '{print $1}'`
+	  groupname=`stat -c "%U %G" /var/log/syslog | awk '{print $2}'`
+fi
+
 for arg in "$@"; do
 	if [ `expr -- "$arg" : '--prefix='` -eq 9 ]; then
 		prefix="`echo $arg | awk -F= '{print $2}'`"
@@ -28,6 +42,7 @@ for arg in "$@"; do
 		user="`echo $arg | awk -F= '{print $2}'`"
 		groupname=$user
 		username=$user
+		non_root_user=yes
 	elif [ `expr -- "$arg" : '--..*'` -ge 3 ]; then
 		echo "ignoring unknown argument: $arg" 1>&2
 	elif [ `expr -- "$arg" : '--'` -eq 2 ]; then
@@ -157,12 +172,17 @@ fi
 }
 
 makeDir() {
-	test -d /var/log/mariadb/columnstore || mkdir -p /var/log/mariadb/columnstore >/dev/null 2>&1
+	if [ ! -d /var/log/mariadb/columnstore ]; then
+		mkdir -p /var/log/mariadb/columnstore >/dev/null 2>&1
 		test -d /var/log/mariadb/columnstore/archive || mkdir /var/log/mariadb/columnstore/archive >/dev/null 2>&1
 		test -d /var/log/mariadb/columnstore/corefiles || mkdir /var/log/mariadb/columnstore/corefiles >/dev/null 2>&1
 		test -d /var/log/mariadb/columnstore/trace || mkdir /var/log/mariadb/columnstore/trace >/dev/null 2>&1
-	chmod 777 -R /var/log/mariadb/columnstore
+		chown $username:$groupname -R /var/log/mariadb
-	chown $user:$user -R /var/log/mariadb
+	else
+		test -d /var/log/mariadb/columnstore/archive || mkdir /var/log/mariadb/columnstore/archive >/dev/null 2>&1
+		test -d /var/log/mariadb/columnstore/corefiles || mkdir /var/log/mariadb/columnstore/corefiles >/dev/null 2>&1
+		test -d /var/log/mariadb/columnstore/trace || mkdir /var/log/mariadb/columnstore/trace >/dev/null 2>&1
+	fi
 }
 
 install() {
@@ -170,15 +190,15 @@ makeDir
 checkSyslog
 if [ ! -z "$syslog_conf" ] ; then
 	$installdir/bin/setConfig -d Installation SystemLogConfigFile ${syslog_conf} >/dev/null 2>&1
-	if [ $user != "root" ]; then
+	if [ $non_root_user == "yes" ]; then
-		 chown $user:$user /home/$user/mariadb/columnstore/etc/*
+		chown $user:$user $installdir/etc/Columnstore.xml*
 	fi
 	
+	rm -f ${syslog_conf}.columnstoreSave
 	if [ "$syslog_conf" == /etc/rsyslog.d/columnstore.conf ] ||
 		[ "$syslog_conf" == /etc/rsyslog.d/49-columnstore.conf ]; then
 		i=1
 	else
-		rm -f ${syslog_conf}.columnstoreSave
 		cp ${syslog_conf} ${syslog_conf}.columnstoreSave >/dev/null 2>&1
 		sed -i '/# MariaDB/,$d' ${syslog_conf}.columnstoreSave > /dev/null 2>&1
 	fi
@@ -189,37 +209,47 @@ if [ ! -z "$syslog_conf" ] ; then
 		# remove older version incase it was installed by previous build
 		 rm -rf /etc/rsyslog.d/columnstore.conf
 		
-		# determine username/groupname
-		
-		if [ -f /var/log/messages ]; then
-		      user=`stat -c "%U %G" /var/log/messages | awk '{print $1}'`
-		      group=`stat -c "%U %G" /var/log/messages | awk '{print $2}'`
-		fi
-		
-		if [ -f /var/log/syslog ]; then
-		      user=`stat -c "%U %G" /var/log/syslog | awk '{print $1}'`
-		      group=`stat -c "%U %G" /var/log/syslog | awk '{print $2}'`
-		fi
-		
-		# set permissions
-		chown $user:$group -R /var/log/mariadb > /dev/null 2>&1
-		
 		if [ $rsyslog7 == 1 ]; then
 			 rm -f /etc/rsyslog.d/49-columnstore.conf
 			 cp  ${columnstoreSyslogFile7} ${syslog_conf}
 
 			 sed -i -e s/groupname/$groupname/g ${syslog_conf}
 			 sed -i -e s/username/$username/g ${syslog_conf}
+			 chmod 644 ${syslog_conf}
 		else		
 			 cp  ${columnstoreSyslogFile} ${syslog_conf}
 		fi
 	fi
 
+	restartSyslog
+
 	# install Columnstore Log Rotate File
+	if [ -d /etc/logrotate.d ]; then
 		cp $installdir/bin/columnstoreLogRotate /etc/logrotate.d/columnstore > /dev/null 2>&1
 		chmod 644 /etc/logrotate.d/columnstore
 
-	restartSyslog
+		#do the logrotate to start with a fresh log file during install
+		logrotate -f /etc/logrotate.d/columnstore > /dev/null 2>&1
+	fi
+	
+	#log install message and find the least permission that allows logging to work
+	CHMOD_LIST=("750" "770" "775" "777")
+	for CHMOD in "${CHMOD_LIST[@]}"; do
+		chmod $CHMOD /var/log/mariadb
+		chmod $CHMOD /var/log/mariadb/columnstore
+		chmod $CHMOD /var/log/mariadb/columnstore/archive
+		chmod $CHMOD /var/log/mariadb/columnstore/corefiles
+		chmod $CHMOD /var/log/mariadb/columnstore/trace
+
+		test -f $installdir/post/functions && . $installdir/post/functions
+		LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$installdir/lib $installdir/bin/cplogger -i 19 "***** MariaDB Columnstore Installed *****"
+		
+		if [ -f /var/log/mariadb/columnstore/info.log ]; then
+			if [ -s /var/log/mariadb/columnstore/info.log ]; then
+				exit 0
+			fi
+		fi
+	done
 fi
 
 }
@@ -233,7 +263,7 @@ if [ ! -z "$syslog_conf" ] ; then
 			if [ $? -eq 0 ]; then
 				if [ -f ${syslog_conf}.columnstoreSave ] ; then
 					#uninstall the syslog for ColumnStore logging
-					 v -f ${syslog_conf} ${syslog_conf}.ColumnStoreBackup
+					 mv -f ${syslog_conf} ${syslog_conf}.ColumnStoreBackup
 					 mv -f ${syslog_conf}.columnstoreSave ${syslog_conf} >/dev/null 2>&1
 					if [ ! -f ${syslog_conf} ] ; then
 						 cp ${syslog_conf}.ColumnStoreBackup ${syslog_conf}
@@ -252,7 +282,6 @@ if [ ! -z "$syslog_conf" ] ; then
 	rm -f /etc/logrotate.d/columnstore
 
 	restartSyslog
-
 fi
 
 }

oamapps/postConfigure/mycnfUpgrade.cpp

@@ -54,6 +54,7 @@
 using namespace std;
 using namespace oam;
 
+
 /* MCOL-1844.  On an upgrade, the user may have customized options in their old 
  * myCnf-include-args.text file.  Merge it with the packaged version, and then process as we
  * have before.
@@ -218,7 +219,7 @@ int main(int argc, char* argv[])
                     ofstream newFile (mycnfFile.c_str());
 
                     //create new file
-                    int fd = open(mycnfFile.c_str(), O_RDWR | O_CREAT, 0666);
+					int fd = open(mycnfFile.c_str(), O_RDWR|O_CREAT, 0644);
 
                     copy(lines.begin(), lines.end(), ostream_iterator<string>(newFile, "\n"));
                     newFile.close();

primitives/primproc/primproc.cpp

@@ -391,7 +391,7 @@ int main(int argc, char* argv[])
     int serverQueueSize = 10;
     int processorWeight = 8 * 1024;
     int processorQueueSize = 10 * 1024;
-    int BRPBlocksPct = 70;
+    int64_t BRPBlocksPct = 70;
     uint32_t BRPBlocks = 1887437;
     int BRPThreads = 16;
     int cacheCount = 1;
@@ -503,16 +503,10 @@ int main(int argc, char* argv[])
     }
 
     string strBlockPct = cf->getConfig(dbbc, "NumBlocksPct");
-    string strBlockAbs = cf->getConfig(dbbc, "NumBlocksInMB");
-    bool usePct = !(strBlockPct.empty());   // which to use.  Prefer Pct if both are specified.
-
-    if (usePct)
     temp = atoi(strBlockPct.c_str());
-    else
-        temp = atoi(strBlockAbs.c_str());
 
 #ifdef _MSC_VER
-    /* TODO: implement handling for NumBlocksInMB */
+    /* TODO: implement handling for the 'm' or 'g' chars in NumBlocksPct */
     if (temp > 0)
         BRPBlocksPct = temp;
 
@@ -532,19 +526,21 @@ int main(int argc, char* argv[])
     }
 
 #else
-    if (usePct)
+    bool absCache = false;
-    {
+    if (temp > 0) {
-        if (temp > 0)
         BRPBlocksPct = temp;
+        /* MCOL-1847.  Did the user specify this as an absolute? */
+        int len = strBlockPct.length();
+        if ((strBlockPct[len-1] >= 'a' && strBlockPct[len-1] <= 'z') ||
+          (strBlockPct[len-1] >= 'A' && strBlockPct[len-1] <= 'Z')) {
+            absCache = true;
+            BRPBlocksPct = Config::fromText(strBlockPct);
+        }
+    }
+    if (absCache)
+        BRPBlocks = BRPBlocksPct / 8192;
+    else
         BRPBlocks = ((BRPBlocksPct / 100.0) * (double) cg.getTotalMemory()) / 8192;
-    }
-    else
-    {
-        if (temp > 0)
-            BRPBlocks = temp * 128;   // 128 blocks per MB.
-        else
-            BRPBlocks = 131072;   // 1GB, why not.
-    }
 #endif
 #if 0
     temp = toInt(cf->getConfig(dbbc, "NumThreads"));

procmgr/processmanager.cpp

@@ -9176,7 +9176,11 @@ int ProcessManager::getDBRMData(messageqcpp::IOSocket fIos, std::string moduleNa
 
         char line[200];
         oldFile.getline(line, 200);
+        // MCOL-1558.  Handle absolute and relative paths.
+        if (line[0] == '/')
             currentDbrmFile = line;
+        else
+            currentDbrmFile = DBRMroot.substr(0, DBRMroot.find_last_of('/') + 1) + line;
     }
     else
     {

utils/clusterTester/columnstoreClusterTester.sh

@@ -334,7 +334,6 @@ checkSSH()
 
 checkRemoteDir()
 {
-
   if [ "$USER" != "root" ]; then
     # Non-root User directory permissions check
     #
@@ -358,7 +357,6 @@ checkRemoteDir()
 		pass=false
 		REPORTPASS=false
     fi
-    done
     
     if ! $pass; then
       checkContinue

utils/funcexp/func_concat_ws.cpp

@@ -97,7 +97,7 @@ string Func_concat_ws::getStrVal(Row& row,
     string tmp;
     for ( uint32_t i = 1 ; i < parm.size() ; i++)
     {
-		string(stringValue(parm[i], row, isNull).c_str(), tmp);
+		stringValue(parm[i], row, isNull, tmp);
         str += tmp;
 
         if (isNull)

utils/regr/corr.cpp

@@ -61,8 +61,8 @@ mcsv1_UDAF::ReturnCode corr::init(mcsv1Context* context,
     context->setUserDataSize(sizeof(corr_data));
     context->setResultType(CalpontSystemCatalog::DOUBLE);
     context->setColWidth(8);
-    context->setScale(colTypes[0].scale + 8);
+    context->setScale(DECIMAL_NOT_SPECIFIED);
-    context->setPrecision(19);
+    context->setPrecision(0);
     context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
     return mcsv1_UDAF::SUCCESS;
 

utils/regr/covar_pop.cpp

@@ -59,8 +59,8 @@ mcsv1_UDAF::ReturnCode covar_pop::init(mcsv1Context* context,
     context->setUserDataSize(sizeof(covar_pop_data));
     context->setResultType(CalpontSystemCatalog::DOUBLE);
     context->setColWidth(8);
-    context->setScale(colTypes[0].scale + 8);
+    context->setScale(DECIMAL_NOT_SPECIFIED);
-    context->setPrecision(19);
+    context->setPrecision(0);
     context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
     return mcsv1_UDAF::SUCCESS;
 

utils/regr/covar_samp.cpp

@@ -59,8 +59,8 @@ mcsv1_UDAF::ReturnCode covar_samp::init(mcsv1Context* context,
     context->setUserDataSize(sizeof(covar_samp_data));
     context->setResultType(CalpontSystemCatalog::DOUBLE);
     context->setColWidth(8);
-    context->setScale(colTypes[0].scale + 8);
+    context->setScale(DECIMAL_NOT_SPECIFIED);
-    context->setPrecision(19);
+    context->setPrecision(0);
     context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
     return mcsv1_UDAF::SUCCESS;
 
@@ -136,7 +136,7 @@ mcsv1_UDAF::ReturnCode covar_samp::evaluate(mcsv1Context* context, static_any::a
 {
     struct covar_samp_data* data = (struct covar_samp_data*)context->getUserData()->data;
     double N = data->cnt;
-    if (N > 0)
+    if (N > 1)
     {
         double sumx = data->sumx;
         double sumy = data->sumy;
@@ -145,6 +145,11 @@ mcsv1_UDAF::ReturnCode covar_samp::evaluate(mcsv1Context* context, static_any::a
         double covar_samp = (sumxy - ((sumx * sumy) / N)) / (N - 1);
         valOut = covar_samp;
     }
+    else
+    if (N == 1)
+    {
+        valOut = 0;
+    }
     return mcsv1_UDAF::SUCCESS;
 }
 

utils/regr/regr_intercept.cpp

@@ -60,8 +60,8 @@ mcsv1_UDAF::ReturnCode regr_intercept::init(mcsv1Context* context,
     context->setUserDataSize(sizeof(regr_intercept_data));
     context->setResultType(CalpontSystemCatalog::DOUBLE);
     context->setColWidth(8);
-    context->setScale(colTypes[0].scale + 8);
+    context->setScale(DECIMAL_NOT_SPECIFIED);
-    context->setPrecision(19);
+    context->setPrecision(0);
     context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
     return mcsv1_UDAF::SUCCESS;
 
@@ -145,12 +145,11 @@ mcsv1_UDAF::ReturnCode regr_intercept::evaluate(mcsv1Context* context, static_an
         double sumy = data->sumy;
         double sumx2 = data->sumx2;
         double sumxy = data->sumxy;
-        double slope = 0.0;
+        double numerator = sumy * sumx2 - sumx * sumxy;
-        double variance = (N * sumx2) - (sumx * sumx);
+        double var_pop = (N * sumx2) - (sumx * sumx);
-        if (variance != 0)
+        if (var_pop != 0)
         {
-            slope = ((N * sumxy) - (sumx * sumy)) / variance;
+            valOut = numerator / var_pop;
-            valOut = (sumy - (slope * sumx)) / N;
         }
     }
     return mcsv1_UDAF::SUCCESS;

utils/regr/regr_r2.cpp

@@ -61,8 +61,8 @@ mcsv1_UDAF::ReturnCode regr_r2::init(mcsv1Context* context,
     context->setUserDataSize(sizeof(regr_r2_data));
     context->setResultType(CalpontSystemCatalog::DOUBLE);
     context->setColWidth(8);
-    context->setScale(colTypes[0].scale + 8);
+    context->setScale(DECIMAL_NOT_SPECIFIED);
-    context->setPrecision(19);
+    context->setPrecision(0);
     context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
     return mcsv1_UDAF::SUCCESS;
 

utils/regr/regr_slope.cpp

@@ -60,8 +60,8 @@ mcsv1_UDAF::ReturnCode regr_slope::init(mcsv1Context* context,
     context->setUserDataSize(sizeof(regr_slope_data));
     context->setResultType(CalpontSystemCatalog::DOUBLE);
     context->setColWidth(8);
-    context->setScale(colTypes[0].scale + 8);
+    context->setScale(DECIMAL_NOT_SPECIFIED);
-    context->setPrecision(19);
+    context->setPrecision(0);
     context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
     return mcsv1_UDAF::SUCCESS;
 
@@ -141,14 +141,16 @@ mcsv1_UDAF::ReturnCode regr_slope::evaluate(mcsv1Context* context, static_any::a
     double N = data->cnt;
     if (N > 0)
     {
+        // COVAR_POP(y, x) / VAR_POP(x)
         double sumx = data->sumx;
         double sumy = data->sumy;
         double sumx2 = data->sumx2;
         double sumxy = data->sumxy;
-        double variance = (N * sumx2) - (sumx * sumx);
+        double covar_pop = N * sumxy - sumx * sumy;
-        if (variance != 0)
+        double var_pop = N * sumx2 - sumx * sumx;
+        if (var_pop != 0)
         {
-            double slope = ((N * sumxy) - (sumx * sumy)) / variance;
+            double slope = covar_pop / var_pop;
             valOut = slope;
         }
     }

utils/regr/regr_sxx.cpp

@@ -58,8 +58,8 @@ mcsv1_UDAF::ReturnCode regr_sxx::init(mcsv1Context* context,
     context->setUserDataSize(sizeof(regr_sxx_data));
     context->setResultType(CalpontSystemCatalog::DOUBLE);
     context->setColWidth(8);
-    context->setScale(colTypes[0].scale + 8);
+    context->setScale(DECIMAL_NOT_SPECIFIED);
-    context->setPrecision(19);
+    context->setPrecision(0);
     context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
     return mcsv1_UDAF::SUCCESS;
 

utils/regr/regr_sxy.cpp

@@ -59,8 +59,8 @@ mcsv1_UDAF::ReturnCode regr_sxy::init(mcsv1Context* context,
     context->setUserDataSize(sizeof(regr_sxy_data));
     context->setResultType(CalpontSystemCatalog::DOUBLE);
     context->setColWidth(8);
-    context->setScale(colTypes[0].scale + 8);
+    context->setScale(DECIMAL_NOT_SPECIFIED);
-    context->setPrecision(19);
+    context->setPrecision(0);
     context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
     return mcsv1_UDAF::SUCCESS;
 

utils/regr/regr_syy.cpp

@@ -58,8 +58,8 @@ mcsv1_UDAF::ReturnCode regr_syy::init(mcsv1Context* context,
     context->setUserDataSize(sizeof(regr_syy_data));
     context->setResultType(CalpontSystemCatalog::DOUBLE);
     context->setColWidth(8);
-    context->setScale(colTypes[0].scale + 8);
+    context->setScale(DECIMAL_NOT_SPECIFIED);
-    context->setPrecision(19);
+    context->setPrecision(0);
     context->setRunFlag(mcsv1sdk::UDAF_IGNORE_NULLS);
     return mcsv1_UDAF::SUCCESS;
 

utils/regr/regrmysql.cpp

@@ -581,12 +581,13 @@ extern "C"
             double sumy = data->sumy;
             double sumx2 = data->sumx2;
             double sumxy = data->sumxy;
+            double slope = 0;
             double variance = (N * sumx2) - (sumx * sumx);
             if (variance)
             {
-                double slope = ((N * sumxy) - (sumx * sumy)) / variance;
+                slope = ((N * sumxy) - (sumx * sumy)) / variance;
-                return (sumy - (slope * sumx)) / N;
             }
+            return (sumy - (slope * sumx)) / N;
         }
         *is_null = 1;
     	return 0;

utils/rowgroup/rowaggregation.cpp

@@ -1673,11 +1673,10 @@ void RowAggregation::updateEntry(const Row& rowIn)
 {
     for (uint64_t i = 0; i < fFunctionCols.size(); i++)
     {
-        SP_ROWAGG_FUNC_t pFunctionCol = fFunctionCols[i];
+        int64_t colIn  = fFunctionCols[i]->fInputColumnIndex;
-        int64_t colIn  = pFunctionCol->fInputColumnIndex;
+        int64_t colOut = fFunctionCols[i]->fOutputColumnIndex;
-        int64_t colOut = pFunctionCol->fOutputColumnIndex;
 
-        switch (pFunctionCol->fAggFunction)
+        switch (fFunctionCols[i]->fAggFunction)
         {
             case ROWAGG_COUNT_COL_NAME:
 
@@ -1691,7 +1690,7 @@ void RowAggregation::updateEntry(const Row& rowIn)
             case ROWAGG_MIN:
             case ROWAGG_MAX:
             case ROWAGG_SUM:
-                doMinMaxSum(rowIn, colIn, colOut, pFunctionCol->fAggFunction);
+                doMinMaxSum(rowIn, colIn, colOut, fFunctionCols[i]->fAggFunction);
                 break;
 
             case ROWAGG_AVG:
@@ -1708,7 +1707,7 @@ void RowAggregation::updateEntry(const Row& rowIn)
             case ROWAGG_BIT_OR:
             case ROWAGG_BIT_XOR:
             {
-                doBitOp(rowIn, colIn, colOut, pFunctionCol->fAggFunction);
+                doBitOp(rowIn, colIn, colOut, fFunctionCols[i]->fAggFunction);
                 break;
             }
 
@@ -1731,7 +1730,7 @@ void RowAggregation::updateEntry(const Row& rowIn)
             {
                 std::ostringstream errmsg;
                 errmsg << "RowAggregation: function (id = " <<
-                       (uint64_t) pFunctionCol->fAggFunction << ") is not supported.";
+                       (uint64_t) fFunctionCols[i]->fAggFunction << ") is not supported.";
                 cerr << errmsg.str() << endl;
                 throw logging::QueryDataExcept(errmsg.str(), logging::aggregateFuncErr);
                 break;
@@ -2015,7 +2014,6 @@ void RowAggregation::doUDAF(const Row& rowIn, int64_t colIn, int64_t colOut,
 
     for (uint32_t i = 0; i < paramCount; ++i)
     {
-        SP_ROWAGG_FUNC_t pFunctionCol = fFunctionCols[funcColsIdx];
         mcsv1sdk::ColumnDatum& datum = valsIn[i];
         // Turn on NULL flags based on the data
         dataFlags[i] = 0;
@@ -2024,9 +2022,9 @@ void RowAggregation::doUDAF(const Row& rowIn, int64_t colIn, int64_t colOut,
         // to acces the constant value rather than a row value.
         cc = NULL;
 
-        if (pFunctionCol->fpConstCol)
+        if (fFunctionCols[funcColsIdx]->fpConstCol)
         {
-            cc = dynamic_cast<ConstantColumn*>(pFunctionCol->fpConstCol.get());
+            cc = dynamic_cast<ConstantColumn*>(fFunctionCols[funcColsIdx]->fpConstCol.get());
         }
 
         if ((cc && cc->type() == ConstantColumn::NULLDATA)
@@ -2243,9 +2241,8 @@ void RowAggregation::doUDAF(const Row& rowIn, int64_t colIn, int64_t colOut,
                 &&  fFunctionCols[funcColsIdx + 1]->fAggFunction == ROWAGG_MULTI_PARM)
         {
             ++funcColsIdx;
-            SP_ROWAGG_FUNC_t pFunctionCol = fFunctionCols[funcColsIdx];
+            colIn  = fFunctionCols[funcColsIdx]->fInputColumnIndex;
-            colIn  = pFunctionCol->fInputColumnIndex;
+            colOut = fFunctionCols[funcColsIdx]->fOutputColumnIndex;
-            colOut = pFunctionCol->fOutputColumnIndex;
         }
         else
         {

utils/udfsdk/avg_mode.cpp

@@ -69,65 +69,13 @@ mcsv1_UDAF::ReturnCode avg_mode::nextValue(mcsv1Context* context, ColumnDatum* v
 {
     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.compatible(charTypeId))
+    DATATYPE val = convertAnyTo<double>(valIn);
-    {
-        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;
@@ -190,65 +138,13 @@ mcsv1_UDAF::ReturnCode avg_mode::dropValue(mcsv1Context* context, ColumnDatum* v
 {
     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.compatible(charTypeId))
+    DATATYPE val = convertAnyTo<double>(valIn);
-    {
-        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;

utils/udfsdk/avgx.cpp

@@ -75,69 +75,13 @@ mcsv1_UDAF::ReturnCode avgx::nextValue(mcsv1Context* context, ColumnDatum* valsI
 {
     static_any::any& valIn_x = valsIn[0].columnData;
     struct  avgx_data* data = (struct avgx_data*)context->getUserData()->data;
-    DATATYPE val = 0.0;
 
     if (valIn_x.empty())
     {
         return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
     }
 
-    if (valIn_x.compatible(longTypeId))
+    DATATYPE val = convertAnyTo<double>(valIn_x);
-    {
-        val = valIn_x.cast<long>();
-    }
-    else if (valIn_x.compatible(charTypeId))
-    {
-        val = valIn_x.cast<char>();
-    }
-    else if (valIn_x.compatible(scharTypeId))
-    {
-        val = valIn_x.cast<signed char>();
-    }
-    else if (valIn_x.compatible(shortTypeId))
-    {
-        val = valIn_x.cast<short>();
-    }
-    else if (valIn_x.compatible(intTypeId))
-    {
-        val = valIn_x.cast<int>();
-    }
-    else if (valIn_x.compatible(longTypeId))
-    {
-        val = valIn_x.cast<long>();
-    }
-    else if (valIn_x.compatible(llTypeId))
-    {
-        val = valIn_x.cast<long long>();
-    }
-    else if (valIn_x.compatible(ucharTypeId))
-    {
-        val = valIn_x.cast<unsigned char>();
-    }
-    else if (valIn_x.compatible(ushortTypeId))
-    {
-        val = valIn_x.cast<unsigned short>();
-    }
-    else if (valIn_x.compatible(uintTypeId))
-    {
-        val = valIn_x.cast<unsigned int>();
-    }
-    else if (valIn_x.compatible(ulongTypeId))
-    {
-        val = valIn_x.cast<unsigned long>();
-    }
-    else if (valIn_x.compatible(ullTypeId))
-    {
-        val = valIn_x.cast<unsigned long long>();
-    }
-    else if (valIn_x.compatible(floatTypeId))
-    {
-        val = valIn_x.cast<float>();
-    }
-    else if (valIn_x.compatible(doubleTypeId))
-    {
-        val = valIn_x.cast<double>();
-    }
 
     // For decimal types, we need to move the decimal point.
     uint32_t scale = valsIn[0].scale;
@@ -183,65 +127,13 @@ mcsv1_UDAF::ReturnCode avgx::dropValue(mcsv1Context* context, ColumnDatum* valsD
 {
     static_any::any& valIn_x = valsDropped[0].columnData;
     struct avgx_data* data = (struct avgx_data*)context->getUserData()->data;
-    DATATYPE val = 0.0;
 
     if (valIn_x.empty())
     {
         return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
     }
 
-    if (valIn_x.compatible(charTypeId))
+    DATATYPE val = convertAnyTo<double>(valIn_x);
-    {
-        val = valIn_x.cast<char>();
-    }
-    else if (valIn_x.compatible(scharTypeId))
-    {
-        val = valIn_x.cast<signed char>();
-    }
-    else if (valIn_x.compatible(shortTypeId))
-    {
-        val = valIn_x.cast<short>();
-    }
-    else if (valIn_x.compatible(intTypeId))
-    {
-        val = valIn_x.cast<int>();
-    }
-    else if (valIn_x.compatible(longTypeId))
-    {
-        val = valIn_x.cast<long>();
-    }
-    else if (valIn_x.compatible(llTypeId))
-    {
-        val = valIn_x.cast<long long>();
-    }
-    else if (valIn_x.compatible(ucharTypeId))
-    {
-        val = valIn_x.cast<unsigned char>();
-    }
-    else if (valIn_x.compatible(ushortTypeId))
-    {
-        val = valIn_x.cast<unsigned short>();
-    }
-    else if (valIn_x.compatible(uintTypeId))
-    {
-        val = valIn_x.cast<unsigned int>();
-    }
-    else if (valIn_x.compatible(ulongTypeId))
-    {
-        val = valIn_x.cast<unsigned long>();
-    }
-    else if (valIn_x.compatible(ullTypeId))
-    {
-        val = valIn_x.cast<unsigned long long>();
-    }
-    else if (valIn_x.compatible(floatTypeId))
-    {
-        val = valIn_x.cast<float>();
-    }
-    else if (valIn_x.compatible(doubleTypeId))
-    {
-        val = valIn_x.cast<double>();
-    }
 
     // For decimal types, we need to move the decimal point.
     uint32_t scale = valsDropped[0].scale;

utils/udfsdk/docs/source/usage/sourcefile.rst

@@ -124,9 +124,9 @@ nextValue()
 
 nextValue() is called from the PM for aggregate usage and the UM for Analytic usage.
 
-valsIn contains a vector of all the parameters from the function call in the SQL query (In Columndtore 1.1, this will always contain exactly one entry).
+valsIn contains a vector of all the parameters from the function call in the SQL query.
 
-Depending on your function, you may wish to be able to handle many different types of input. A good way to handle this is to have a series of if..else..if statements comparing the input type and dealing with each separately. For instace, if you want to handle multiple numeric types, you might use::
+Depending on your function, you may wish to be able to handle many different types of input. There's a helper template function convertAnyTo() which will convert the input static:any value to the designated type. For Example, if your internal accumulater is of type double, you might use::
 
 	static_any::any& valIn = valsDropped[0].columnData;
 	AVGData& data = static_cast<MedianData*>(context->getUserData())->mData;
@@ -137,21 +137,7 @@ Depending on your function, you may wish to be able to handle many different typ
 		return mcsv1_UDAF::SUCCESS; // Ought not happen when UDAF_IGNORE_NULLS is on.
 	}
 
-	if (valIn.compatible(charTypeId))
+	val = convertAnyTo<double>(valIn);
-	{
-		val = valIn.cast<char>();
-	}
-	else if (valIn.compatible(scharTypeId))
-	{
-		val = valIn.cast<signed char>();
-	}
-	else if (valIn.compatible(shortTypeId))
-	{
-		val = valIn.cast<short>();
-	}
-	.
-	.
-	.
 
 Once you've gotten your data in a format you like, then do your aggregation. For AVG, you might see::
 

utils/udfsdk/median.cpp

@@ -69,65 +69,13 @@ mcsv1_UDAF::ReturnCode median::nextValue(mcsv1Context* context, ColumnDatum* val
 {
     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.compatible(charTypeId))
+    DATATYPE val = convertAnyTo<double>(valIn);
-    {
-        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;
@@ -215,65 +163,13 @@ mcsv1_UDAF::ReturnCode median::dropValue(mcsv1Context* context, ColumnDatum* val
 {
     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.compatible(charTypeId))
+    DATATYPE val = convertAnyTo<double>(valIn);
-    {
-        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;

utils/udfsdk/ssq.cpp

@@ -85,65 +85,13 @@ mcsv1_UDAF::ReturnCode ssq::nextValue(mcsv1Context* context, ColumnDatum* valsIn
 {
     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 (valIn.compatible(charTypeId))
+    DATATYPE val = convertAnyTo<double>(valIn);
-    {
-        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;
@@ -186,65 +134,13 @@ mcsv1_UDAF::ReturnCode ssq::dropValue(mcsv1Context* context, ColumnDatum* valsDr
 {
     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.compatible(charTypeId))
+    DATATYPE val = convertAnyTo<double>(valIn);
-    {
-        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;

utils/udfsdk/udfsdk.vpj

@@ -205,6 +205,7 @@
             <F N="allnull.cpp"/>
             <F N="avg_mode.cpp"/>
             <F N="avgx.cpp"/>
+            <F N="distinct_count.cpp"/>
             <F N="mcsv1_udaf.cpp"/>
             <F N="median.cpp"/>
             <F N="ssq.cpp"/>
@@ -217,6 +218,7 @@
             <F N="allnull.h"/>
             <F N="avg_mode.h"/>
             <F N="avgx.h"/>
+            <F N="distinct_count.h"/>
             <F N="mcsv1_udaf.h"/>
             <F N="median.h"/>
             <F N="ssq.h"/>

utils/windowfunction/wf_udaf.cpp

@@ -481,6 +481,7 @@ bool WF_udaf::dropValues(int64_t b, int64_t e)
         }
     }
 
+    WindowFunctionType::resetData();
     return true;
 }
 

utils/windowfunction/windowfunction.cpp

@@ -192,20 +192,35 @@ void WindowFunction::operator()()
                     // values leaving the window and nextValue for those entering, rather
                     // than a resetData() and then iterating over the entire window.
                     // Built-in functions may have this functionality added in the future.
+                    // If b > e then the frame is entirely outside of the partition
+                    // and there's no values to drop
+                    if (b <= e)
+                    {
+                        if (!firstTime)
+                        {
                             if (fFunctionType->dropValues(prevFrame.first, w.first))
                             {
-                        b = firstTime ? w.first : prevFrame.second + 1;
+                                // Adjust the beginning of the frame for nextValue
+                                // to start where the previous frame left off.
+                                b = prevFrame.second + 1;
+                            }
+                            else
+                            {
+                                // dropValues failed or doesn't exist
+                                // so calculate the entire frame.
+                                fFunctionType->resetData();
+                            }
                         }
                         else
                         {
                             fFunctionType->resetData();
-                    }
-
-                    fFunctionType->operator()(b, e, i);
-                    prevFrame = w;
                             firstTime = false;
                         }
                     }
+                    fFunctionType->operator()(b, e, i); // UDAnF: Calls nextValue and evaluate
+                    prevFrame = w;
+                }
+            }
         }
     }
     catch (IDBExcept& iex)
@@ -218,7 +233,7 @@ void WindowFunction::operator()()
     }
     catch (...)
     {
-        fStep->handleException("unknow exception", logging::ERR_EXECUTE_WINDOW_FUNCTION);
+        fStep->handleException("unknown exception", logging::ERR_EXECUTE_WINDOW_FUNCTION);
     }
 }
 

writeengine/bulk/we_bulkloadbuffer.cpp

@@ -31,8 +31,6 @@
 #include <ctype.h>
 #include <cfloat>
 
-#include <boost/algorithm/string/predicate.hpp>
-
 #include "we_bulkload.h"
 #include "we_bulkloadbuffer.h"
 #include "we_brm.h"
@@ -370,18 +368,11 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 {
                     errno = 0;
 
-                    if (iequals(field, "true"))
-                    {
-                        fVal = 1;
-                    }
-                    else
-                    {
 #ifdef _MSC_VER
                     fVal = (float)strtod( field, 0 );
 #else
                     fVal = strtof( field, 0 );
 #endif
-                    }
 
                     if (errno == ERANGE)
                     {
@@ -414,6 +405,11 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                             fVal = minFltSat;
                             bufStats.satCount++;
                         }
+                        if ( fVal == 0
+                        && isTrueWord(const_cast<const char*>(field), fieldLength) )
+                        {
+                            fVal = 1;
+                        }
                     }
                 }
 
@@ -474,14 +470,7 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 {
                     errno = 0;
 
-                    if (iequals(field, "true"))
-                    {
-                        dVal = 1;
-                    }
-                    else
-                    {
                     dVal = strtod(field, 0);
-                    }
 
                     if (errno == ERANGE)
                     {
@@ -514,6 +503,11 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                             dVal = column.fMinDblSat;
                             bufStats.satCount++;
                         }
+                        else if (dVal == 0
+                            && isTrueWord(const_cast<const char*>(field), fieldLength))
+                        {
+                            dVal = 1;
+                        }
                     }
                 }
 
@@ -629,12 +623,6 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 }
                 else
                 {
-                    if (iequals(field, "true"))
-                    {
-                        strcpy(field, "1");
-                        fieldLength = 1;
-                    }
-
                     if ( (column.dataType == CalpontSystemCatalog::DECIMAL ) ||
                             (column.dataType == CalpontSystemCatalog::UDECIMAL) )
                     {
@@ -664,6 +652,11 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 origVal = static_cast<int64_t>(column.fMaxIntSat);
                 bSatVal = true;
             }
+            else if ( origVal == 0
+                && isTrueWord(const_cast<const char*>(field), fieldLength) )
+            {
+                origVal = 1;
+            }
 
             if (bSatVal)
                 bufStats.satCount++;
@@ -722,14 +715,7 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 {
                     errno   = 0;
 
-                    if (iequals(field, "true"))
-                    {
-                        origVal = 1;
-                    }
-                    else
-                    {
                     origVal = strtoll(field, 0, 10);
-                    }
                     
                     if (errno == ERANGE)
                         bSatVal = true;
@@ -747,6 +733,11 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 origVal = static_cast<int64_t>(column.fMaxIntSat);
                 bSatVal = true;
             }
+            else if ( origVal == 0
+                && isTrueWord(const_cast<const char*>(field), fieldLength) )
+            {
+                origVal = 1;
+            }
 
             if (bSatVal)
                 bufStats.satCount++;
@@ -805,7 +796,7 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 }
                 else
                 {
-                    if (iequals(field, "true"))
+                    if (isTrueWord(const_cast<const char*>(field), fieldLength))
                     {
                         strcpy(field, "1");
                         fieldLength = 1;
@@ -898,14 +889,7 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 {
                     errno   = 0;
 
-                    if (iequals(field, "true"))
-                    {
-                        origVal = 1;
-                    }
-                    else
-                    {
                     origVal = strtoll(field, 0, 10);
-                    }
 
                     if (errno == ERANGE)
                         bSatVal = true;
@@ -923,6 +907,11 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 origVal = static_cast<int64_t>(column.fMaxIntSat);
                 bSatVal = true;
             }
+            else if ( origVal == 0
+                && isTrueWord(const_cast<const char*>(field), fieldLength) )
+            {
+                origVal = 1;
+            }
 
             if (bSatVal)
                 bufStats.satCount++;
@@ -981,7 +970,7 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                     }
                     else
                     {
-                        if (iequals(field, "true"))
+                        if (isTrueWord(const_cast<const char*>(field), fieldLength))
                         {
                             strcpy(field, "1");
                             fieldLength = 1;
@@ -1018,6 +1007,7 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                     bSatVal = true;
                 }
 
+
                 if (bSatVal)
                     bufStats.satCount++;
 
@@ -1199,14 +1189,7 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                     {
                         errno = 0;
 
-                        if (iequals(field, "true"))
-                        {
-                            ullVal = 1;
-                        }
-                        else
-                        {
                         ullVal = strtoull(field, 0, 10);
-                        }
 
                         if (errno == ERANGE)
                             bSatVal = true;
@@ -1220,6 +1203,11 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 ullVal = column.fMaxIntSat;
                 bSatVal = true;
             }
+            else if ( ullVal == 0
+                && isTrueWord(const_cast<const char*>(field), fieldLength) )
+            {
+                ullVal = 1;
+            }
 
             if (bSatVal)
                 bufStats.satCount++;
@@ -1276,14 +1264,7 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 {
                     errno   = 0;
 
-                    if (iequals(field, "true"))
-                    {
-                        origVal = 1;
-                    }
-                    else
-                    {
                     origVal = strtoll(field, 0, 10);
-                    }
 
                     if (errno == ERANGE)
                         bSatVal = true;
@@ -1301,6 +1282,11 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                 origVal = static_cast<int64_t>(column.fMaxIntSat);
                 bSatVal = true;
             }
+            else if ( origVal == 0
+                && isTrueWord(const_cast<const char*>(field), fieldLength) )
+            {
+                origVal = 1;
+            }
 
             if (bSatVal)
                 bufStats.satCount++;
@@ -1361,7 +1347,7 @@ void BulkLoadBuffer::convert(char* field, int fieldLength,
                     }
                     else
                     {
-                        if (iequals(field, "true"))
+                        if (isTrueWord(const_cast<const char*>(field), fieldLength))
                         {
                             strcpy(field, "1");
                             fieldLength = 1;

writeengine/bulk/we_bulkloadbuffer.h

@@ -383,5 +383,14 @@ public:
     }
 };
 
+inline bool isTrueWord(const char *field, int fieldLength)
+{
+    //return false;
+    return fieldLength == 4 && ( field[0] == 'T' || field[0] == 't' )
+        && ( field[1] == 'R' || field[1] == 'r' )
+        && ( field[2] == 'U' || field[2] == 'u' )
+        && ( field[3] == 'E' || field[3] == 'e' );
+}
+
 }
 #endif