TTable Class Reference

Table class: Relational table with columnar data storage. More...

#include <table.h>

Public Member Functions
void	AddIntCol (const TStr &ColName)
	Adds an integer column with name ColName. More...
void	AddFltCol (const TStr &ColName)
	Adds a float column with name ColName. More...
void	AddStrCol (const TStr &ColName)
	Adds a string column with name ColName. More...
void	GroupByIntColMP (const TStr &GroupBy, THashMP< TInt, TIntV > &Grouping, TBool UsePhysicalIds=true) const
	Groups/hashes by a single column with integer values, using OpenMP multi-threading. More...
	TTable ()
	TTable (TTableContext *Context)
	TTable (const Schema &S, TTableContext *Context)
	TTable (TSIn &SIn, TTableContext *Context)
	TTable (const THash< TInt, TInt > &H, const TStr &Col1, const TStr &Col2, TTableContext *Context, const TBool IsStrKeys=false)
	Constructor to build table out of a hash table of int->int. More...
	TTable (const THash< TInt, TFlt > &H, const TStr &Col1, const TStr &Col2, TTableContext *Context, const TBool IsStrKeys=false)
	Constructor to build table out of a hash table of int->float. More...
	TTable (const TTable &Table)
	Copy constructor. More...
	TTable (const TTable &Table, const TIntV &RowIds)
void	SaveSS (const TStr &OutFNm)
	Saves table schema and content to a TSV file. More...
void	SaveBin (const TStr &OutFNm)
	Saves table schema and content to a binary file. More...
void	Save (TSOut &SOut)
	Saves table schema and content to a binary format. More...
void	Dump (FILE *OutF=stdout) const
	Prints table contents to a text file. More...
void	AddRow (const TTableRow &Row)
	Adds row with values taken from given TTableRow. More...
TTableContext *	GetContext ()
	Returns the context. More...
TTableContext *	ChangeContext (TTableContext *Context)
	Changes the current context. Moves all object items to the new context. More...
TInt	GetColIdx (const TStr &ColName) const
	Gets index of column ColName among columns of the same type in the schema. More...
TInt	GetIntVal (const TStr &ColName, const TInt &RowIdx)
	Gets the value of integer attribute ColName at row RowIdx. More...
TFlt	GetFltVal (const TStr &ColName, const TInt &RowIdx)
	Gets the value of float attribute ColName at row RowIdx. More...
TStr	GetStrVal (const TStr &ColName, const TInt &RowIdx) const
	Gets the value of string attribute ColName at row RowIdx. More...
TInt	GetStrMapById (TInt ColIdx, TInt RowIdx) const
	Gets the integer mapping of the string at column ColIdx at row RowIdx. More...
TInt	GetStrMapByName (const TStr &ColName, TInt RowIdx) const
	Gets the integer mapping of the string at column ColName at row RowIdx. More...
TStr	GetStrValById (TInt ColIdx, TInt RowIdx) const
	Gets the value of the string attribute at column ColIdx at row RowIdx. More...
TStr	GetStrValByName (const TStr &ColName, const TInt &RowIdx) const
	Gets the value of the string attribute at column ColName at row RowIdx. More...
TIntV	GetIntRowIdxByVal (const TStr &ColName, const TInt &Val) const
	Gets the rows containing Val in int column ColName. More...
TIntV	GetStrRowIdxByMap (const TStr &ColName, const TInt &Map) const
	Gets the rows containing int mapping Map in str column ColName. More...
TIntV	GetFltRowIdxByVal (const TStr &ColName, const TFlt &Val) const
	Gets the rows containing Val in flt column ColName. More...
TInt	RequestIndexInt (const TStr &ColName)
	Creates Index for Int Column ColName. More...
TInt	RequestIndexFlt (const TStr &ColName)
	Creates Index for Flt Column ColName. More...
TInt	RequestIndexStrMap (const TStr &ColName)
	Creates Index for Str Column ColName. More...
TStr	GetStr (const TInt &KeyId) const
	Gets the string with KeyId. More...
TInt	GetIntValAtRowIdx (const TInt &ColIdx, const TInt &RowIdx)
	Get the integer value at column ColIdx and row RowIdx. More...
TFlt	GetFltValAtRowIdx (const TInt &ColIdx, const TInt &RowIdx)
	Get the float value at column ColIdx and row RowIdx. More...
Schema	GetSchema ()
	Gets the schema of this table. More...
TVec< PNEANet >	ToGraphSequence (TStr SplitAttr, TAttrAggr AggrPolicy, TInt WindowSize, TInt JumpSize, TInt StartVal=TInt::Mn, TInt EndVal=TInt::Mx)
	Creates a sequence of graphs based on values of column SplitAttr and windows specified by JumpSize and WindowSize. More...
TVec< PNEANet >	ToVarGraphSequence (TStr SplitAttr, TAttrAggr AggrPolicy, TIntPrV SplitIntervals)
	Creates a sequence of graphs based on values of column SplitAttr and intervals specified by SplitIntervals. More...
TVec< PNEANet >	ToGraphPerGroup (TStr GroupAttr, TAttrAggr AggrPolicy)
	Creates a sequence of graphs based on grouping specified by GroupAttr. More...
PNEANet	ToGraphSequenceIterator (TStr SplitAttr, TAttrAggr AggrPolicy, TInt WindowSize, TInt JumpSize, TInt StartVal=TInt::Mn, TInt EndVal=TInt::Mx)
	Creates the graph sequence one at a time. More...
PNEANet	ToVarGraphSequenceIterator (TStr SplitAttr, TAttrAggr AggrPolicy, TIntPrV SplitIntervals)
	Creates the graph sequence one at a time. More...
PNEANet	ToGraphPerGroupIterator (TStr GroupAttr, TAttrAggr AggrPolicy)
	Creates the graph sequence one at a time. More...
PNEANet	NextGraphIterator ()
	Calls to this must be preceded by a call to one of the above ToGraph*Iterator functions. More...
TBool	IsLastGraphOfSequence ()
	Checks if the end of the graph sequence is reached. More...
TStr	GetSrcCol () const
	Gets the name of the column to be used as src nodes in the graph. More...
void	SetSrcCol (const TStr &Src)
	Sets the name of the column to be used as src nodes in the graph. More...
TStr	GetDstCol () const
	Gets the name of the column to be used as dst nodes in the graph. More...
void	SetDstCol (const TStr &Dst)
	Sets the name of the column to be used as dst nodes in the graph. More...
void	AddEdgeAttr (const TStr &Attr)
	Adds column to be used as graph edge attribute. More...
void	AddEdgeAttr (TStrV &Attrs)
	Adds columns to be used as graph edge attributes. More...
void	AddSrcNodeAttr (const TStr &Attr)
	Adds column to be used as src node atribute of the graph. More...
void	AddSrcNodeAttr (TStrV &Attrs)
	Adds columns to be used as src node attributes of the graph. More...
void	AddDstNodeAttr (const TStr &Attr)
	Adds column to be used as dst node atribute of the graph. More...
void	AddDstNodeAttr (TStrV &Attrs)
	Adds columns to be used as dst node attributes of the graph. More...
void	AddNodeAttr (const TStr &Attr)
	Handles the common case where src and dst both belong to the same "universe" of entities. More...
void	AddNodeAttr (TStrV &Attrs)
	Handles the common case where src and dst both belong to the same "universe" of entities. More...
void	SetCommonNodeAttrs (const TStr &SrcAttr, const TStr &DstAttr, const TStr &CommonAttrName)
	Sets the columns to be used as both src and dst node attributes. More...
TStrV	GetSrcNodeIntAttrV () const
	Gets src node int attribute name vector. More...
TStrV	GetDstNodeIntAttrV () const
	Gets dst node int attribute name vector. More...
TStrV	GetEdgeIntAttrV () const
	Gets edge int attribute name vector. More...
TStrV	GetSrcNodeFltAttrV () const
	Gets src node float attribute name vector. More...
TStrV	GetDstNodeFltAttrV () const
	Gets dst node float attribute name vector. More...
TStrV	GetEdgeFltAttrV () const
	Gets edge float attribute name vector. More...
TStrV	GetSrcNodeStrAttrV () const
	Gets src node str attribute name vector. More...
TStrV	GetDstNodeStrAttrV () const
	Gets dst node str attribute name vector. More...
TStrV	GetEdgeStrAttrV () const
	Gets edge str attribute name vector. More...
TAttrType	GetColType (const TStr &ColName) const
	Gets type of column ColName. More...
TInt	GetNumRows () const
	Gets total number of rows in this table. More...
TInt	GetNumValidRows () const
	Gets number of valid, i.e. not deleted, rows in this table. More...
THash< TInt, TInt >	GetRowIdMap () const
	Gets a map of logical to physical row ids. More...
TRowIterator	BegRI () const
	Gets iterator to the first valid row of the table. More...
TRowIterator	EndRI () const
	Gets iterator to the last valid row of the table. More...
TRowIteratorWithRemove	BegRIWR ()
	Gets iterator with reomve to the first valid row. More...
TRowIteratorWithRemove	EndRIWR ()
	Gets iterator with reomve to the last valid row. More...
void	GetPartitionRanges (TIntPrV &Partitions, TInt NumPartitions) const
	Partitions the table into NumPartitions and populate Partitions with the ranges. More...
void	Rename (const TStr &Column, const TStr &NewLabel)
	Renames a column. More...
void	Unique (const TStr &Col)
	Removes rows with duplicate values in given column. More...
void	Unique (const TStrV &Cols, TBool Ordered=true)
	Removes rows with duplicate values in given columns. More...
void	Select (TPredicate &Predicate, TIntV &SelectedRows, TBool Remove=true)
	Selects rows that satisfy given Predicate. More...
void	Select (TPredicate &Predicate)
void	Classify (TPredicate &Predicate, const TStr &LabelName, const TInt &PositiveLabel=1, const TInt &NegativeLabel=0)
void	SelectAtomic (const TStr &Col1, const TStr &Col2, TPredComp Cmp, TIntV &SelectedRows, TBool Remove=true)
	Selects rows using atomic compare operation. More...
void	SelectAtomic (const TStr &Col1, const TStr &Col2, TPredComp Cmp)
void	ClassifyAtomic (const TStr &Col1, const TStr &Col2, TPredComp Cmp, const TStr &LabelName, const TInt &PositiveLabel=1, const TInt &NegativeLabel=0)
void	SelectAtomicConst (const TStr &Col, const TPrimitive &Val, TPredComp Cmp, TIntV &SelectedRows, PTable &SelectedTable, TBool Remove=true, TBool Table=true)
	Selects rows where the value of Col matches given primitive Val. More...
template<class T >
void	SelectAtomicConst (const TStr &Col, const T &Val, TPredComp Cmp)
template<class T >
void	SelectAtomicConst (const TStr &Col, const T &Val, TPredComp Cmp, PTable &SelectedTable)
template<class T >
void	ClassifyAtomicConst (const TStr &Col, const T &Val, TPredComp Cmp, const TStr &LabelName, const TInt &PositiveLabel=1, const TInt &NegativeLabel=0)
void	SelectAtomicIntConst (const TStr &Col, const TInt &Val, TPredComp Cmp)
void	SelectAtomicIntConst (const TStr &Col, const TInt &Val, TPredComp Cmp, PTable &SelectedTable)
void	SelectAtomicStrConst (const TStr &Col, const TStr &Val, TPredComp Cmp)
void	SelectAtomicStrConst (const TStr &Col, const TStr &Val, TPredComp Cmp, PTable &SelectedTable)
void	SelectAtomicFltConst (const TStr &Col, const TFlt &Val, TPredComp Cmp)
void	SelectAtomicFltConst (const TStr &Col, const TFlt &Val, TPredComp Cmp, PTable &SelectedTable)
void	Group (const TStrV &GroupBy, const TStr &GroupColName, TBool Ordered=true, TBool UsePhysicalIds=true)
	Groups rows depending on values of GroupBy columns. More...
void	Count (const TStr &CountColName, const TStr &Col)
	Counts number of unique elements. More...
void	Order (const TStrV &OrderBy, TStr OrderColName="", TBool ResetRankByMSC=false, TBool Asc=true)
	Orders the rows according to the values in columns of OrderBy (in descending lexicographic order). More...
void	Aggregate (const TStrV &GroupByAttrs, TAttrAggr AggOp, const TStr &ValAttr, const TStr &ResAttr, TBool Ordered=true)
	Aggregates values of ValAttr after grouping with respect to GroupByAttrs. Result are stored as new attribute ResAttr. More...
void	AggregateCols (const TStrV &AggrAttrs, TAttrAggr AggOp, const TStr &ResAttr)
	Aggregates attributes in AggrAttrs across columns. More...
TVec< PTable >	SpliceByGroup (const TStrV &GroupByAttrs, TBool Ordered=true)
	Splices table into subtables according to a grouping statement. More...
PTable	Join (const TStr &Col1, const TTable &Table, const TStr &Col2)
	Performs equijoin. More...
PTable	Join (const TStr &Col1, const PTable &Table, const TStr &Col2)
PTable	ThresholdJoin (const TStr &KeyCol1, const TStr &JoinCol1, const TTable &Table, const TStr &KeyCol2, const TStr &JoinCol2, TInt Threshold, TBool PerJoinKey=false)
PTable	SelfJoin (const TStr &Col)
	Joins table with itself, on values of Col. More...
PTable	SelfSimJoin (const TStrV &Cols, const TStr &DistanceColName, const TSimType &SimType, const TFlt &Threshold)
PTable	SelfSimJoinPerGroup (const TStr &GroupAttr, const TStr &SimCol, const TStr &DistanceColName, const TSimType &SimType, const TFlt &Threshold)
	Performs join if the distance between two rows is less than the specified threshold. More...
PTable	SelfSimJoinPerGroup (const TStrV &GroupBy, const TStr &SimCol, const TStr &DistanceColName, const TSimType &SimType, const TFlt &Threshold)
	Performs join if the distance between two rows is less than the specified threshold. More...
PTable	SimJoin (const TStrV &Cols1, const TTable &Table, const TStrV &Cols2, const TStr &DistanceColName, const TSimType &SimType, const TFlt &Threshold)
	Performs join if the distance between two rows is less than the specified threshold. More...
void	SelectFirstNRows (const TInt &N)
	Selects first N rows from the table. More...
void	Defrag ()
	Releases memory of deleted rows, and defrags. More...
void	StoreIntCol (const TStr &ColName, const TIntV &ColVals)
	Adds entire int column to table. More...
void	StoreFltCol (const TStr &ColName, const TFltV &ColVals)
	Adds entire flt column to table. More...
void	StoreStrCol (const TStr &ColName, const TStrV &ColVals)
	Adds entire str column to table. More...
void	UpdateFltFromTable (const TStr &KeyAttr, const TStr &UpdateAttr, const TTable &Table, const TStr &FKeyAttr, const TStr &ReadAttr, TFlt DefaultFltVal=0.0)
void	UpdateFltFromTableMP (const TStr &KeyAttr, const TStr &UpdateAttr, const TTable &Table, const TStr &FKeyAttr, const TStr &ReadAttr, TFlt DefaultFltVal=0.0)
void	SetFltColToConstMP (TInt UpdateColIdx, TFlt DefaultFltVal)
PTable	Union (const TTable &Table)
	Returns union of this table with given Table. More...
PTable	Union (const PTable &Table)
PTable	UnionAll (const TTable &Table)
	Returns union of this table with given Table, preserving duplicates. More...
PTable	UnionAll (const PTable &Table)
void	UnionAllInPlace (const TTable &Table)
	Same as TTable::ConcatTable. More...
void	UnionAllInPlace (const PTable &Table)
PTable	Intersection (const TTable &Table)
	Returns intersection of this table with given Table. More...
PTable	Intersection (const PTable &Table)
PTable	Minus (TTable &Table)
	Returns table with rows that are present in this table but not in given Table. More...
PTable	Minus (const PTable &Table)
PTable	Project (const TStrV &ProjectCols)
	Returns table with only the columns in ProjectCols. More...
void	ProjectInPlace (const TStrV &ProjectCols)
	Keeps only the columns specified in ProjectCols. More...
void	ColGenericOp (const TStr &Attr1, const TStr &Attr2, const TStr &ResAttr, TArithOp op)
	Performs columnwise arithmetic operation. More...
void	ColGenericOpMP (TInt ArgColIdx1, TInt ArgColIdx2, TAttrType ArgType1, TAttrType ArgType2, TInt ResColIdx, TArithOp op)
void	ColAdd (const TStr &Attr1, const TStr &Attr2, const TStr &ResultAttrName="")
	Performs columnwise addition. See TTable::ColGenericOp. More...
void	ColSub (const TStr &Attr1, const TStr &Attr2, const TStr &ResultAttrName="")
	Performs columnwise subtraction. See TTable::ColGenericOp. More...
void	ColMul (const TStr &Attr1, const TStr &Attr2, const TStr &ResultAttrName="")
	Performs columnwise multiplication. See TTable::ColGenericOp. More...
void	ColDiv (const TStr &Attr1, const TStr &Attr2, const TStr &ResultAttrName="")
	Performs columnwise division. See TTable::ColGenericOp. More...
void	ColMod (const TStr &Attr1, const TStr &Attr2, const TStr &ResultAttrName="")
	Performs columnwise modulus. See TTable::ColGenericOp. More...
void	ColMin (const TStr &Attr1, const TStr &Attr2, const TStr &ResultAttrName="")
	Performs min of two columns. See TTable::ColGenericOp. More...
void	ColMax (const TStr &Attr1, const TStr &Attr2, const TStr &ResultAttrName="")
	Performs max of two columns. See TTable::ColGenericOp. More...
void	ColGenericOp (const TStr &Attr1, TTable &Table, const TStr &Attr2, const TStr &ResAttr, TArithOp op, TBool AddToFirstTable)
	Performs columnwise arithmetic operation with column of given table. More...
void	ColAdd (const TStr &Attr1, TTable &Table, const TStr &Attr2, const TStr &ResAttr="", TBool AddToFirstTable=true)
	Performs columnwise addition with column of given table. More...
void	ColSub (const TStr &Attr1, TTable &Table, const TStr &Attr2, const TStr &ResAttr="", TBool AddToFirstTable=true)
	Performs columnwise subtraction with column of given table. More...
void	ColMul (const TStr &Attr1, TTable &Table, const TStr &Attr2, const TStr &ResAttr="", TBool AddToFirstTable=true)
	Performs columnwise multiplication with column of given table. More...
void	ColDiv (const TStr &Attr1, TTable &Table, const TStr &Attr2, const TStr &ResAttr="", TBool AddToFirstTable=true)
	Performs columnwise division with column of given table. More...
void	ColMod (const TStr &Attr1, TTable &Table, const TStr &Attr2, const TStr &ResAttr="", TBool AddToFirstTable=true)
	Performs columnwise modulus with column of given table. More...
void	ColGenericOp (const TStr &Attr1, const TFlt &Num, const TStr &ResAttr, TArithOp op, const TBool floatCast)
	Performs arithmetic op of column values and given Num. More...
void	ColGenericOpMP (const TInt &ColIdx1, const TInt &ColIdx2, TAttrType ArgType, const TFlt &Num, TArithOp op, TBool ShouldCast)
void	ColAdd (const TStr &Attr1, const TFlt &Num, const TStr &ResultAttrName="", const TBool floatCast=false)
	Performs addition of column values and given Num. More...
void	ColSub (const TStr &Attr1, const TFlt &Num, const TStr &ResultAttrName="", const TBool floatCast=false)
	Performs subtraction of column values and given Num. More...
void	ColMul (const TStr &Attr1, const TFlt &Num, const TStr &ResultAttrName="", const TBool floatCast=false)
	Performs multiplication of column values and given Num. More...
void	ColDiv (const TStr &Attr1, const TFlt &Num, const TStr &ResultAttrName="", const TBool floatCast=false)
	Performs division of column values and given Num. More...
void	ColMod (const TStr &Attr1, const TFlt &Num, const TStr &ResultAttrName="", const TBool floatCast=false)
	Performs modulus of column values and given Num. More...
void	ColConcat (const TStr &Attr1, const TStr &Attr2, const TStr &Sep="", const TStr &ResAttr="")
	Concatenates two string columns. More...
void	ColConcat (const TStr &Attr1, TTable &Table, const TStr &Attr2, const TStr &Sep="", const TStr &ResAttr="", TBool AddToFirstTable=true)
	Concatenates string column with column of given table. More...
void	ColConcatConst (const TStr &Attr1, const TStr &Val, const TStr &Sep="", const TStr &ResAttr="")
	Concatenates column values with given string value. More...
void	ReadIntCol (const TStr &ColName, TIntV &Result) const
	Reads values of entire int column into Result. More...
void	ReadFltCol (const TStr &ColName, TFltV &Result) const
	Reads values of entire float column into Result. More...
void	ReadStrCol (const TStr &ColName, TStrV &Result) const
	Reads values of entire string column into Result. More...
void	InitIds ()
	Adds explicit row ids, initialize hash set mapping ids to physical rows. More...
PTable	IsNextK (const TStr &OrderCol, TInt K, const TStr &GroupBy, const TStr &RankColName="")
	Distance based filter. More...
void	PrintSize ()
void	PrintContextSize ()
TSize	GetMemUsedKB ()
	Returns approximate memory used by table in [KB]. More...
TSize	GetContextMemUsedKB ()
	Returns approximate memory used by table context in [KB]. More...

Static Public Member Functions
static void	SetMP (TInt Value)
static TInt	GetMP ()
static TStr	NormalizeColName (const TStr &ColName)
	Adds suffix to column name if it doesn't exist. More...
static TStrV	NormalizeColNameV (const TStrV &Cols)
	Adds suffix to column name if it doesn't exist. More...
static PTable	New ()
static PTable	New (TTableContext *Context)
static PTable	New (const Schema &S, TTableContext *Context)
static PTable	New (const THash< TInt, TInt > &H, const TStr &Col1, const TStr &Col2, TTableContext *Context, const TBool IsStrKeys=false)
	Returns pointer to a table constructed from given int->int hash. More...
static PTable	New (const THash< TInt, TFlt > &H, const TStr &Col1, const TStr &Col2, TTableContext *Context, const TBool IsStrKeys=false)
	Returns pointer to a table constructed from given int->float hash. More...
static PTable	New (const PTable Table)
	Returns pointer to a new table created from given Table. More...
static void	GetSchema (const TStr &InFNm, Schema &S, const char &Separator= '\t')
	Returns pointer to a new table created from given Table, with name set to TableName. More...
static PTable	LoadSS (const Schema &S, const TStr &InFNm, TTableContext *Context, const char &Separator= '\t', TBool HasTitleLine=false)
	Loads table from spread sheet (TSV, CSV, etc). Note: HasTitleLine = true is not supported. Please comment title lines instead. More...
static PTable	LoadSS (const Schema &S, const TStr &InFNm, TTableContext *Context, const TIntV &RelevantCols, const char &Separator= '\t', TBool HasTitleLine=false)
	Loads table from spread sheet - but only load the columns specified by RelevantCols. Note: HasTitleLine = true is not supported. Please comment title lines instead. More...
static PTable	Load (TSIn &SIn, TTableContext *Context)
	Loads table from a binary format. More...
static PTable	TableFromHashMap (const THash< TInt, TInt > &H, const TStr &Col1, const TStr &Col2, TTableContext *Context, const TBool IsStrKeys=false)
	Builds table from hash table of int->int. More...
static PTable	TableFromHashMap (const THash< TInt, TFlt > &H, const TStr &Col1, const TStr &Col2, TTableContext *Context, const TBool IsStrKeys=false)
	Builds table from hash table of int->float. More...
static PTable	GetNodeTable (const PNEANet &Network, TTableContext *Context)
	Extracts node TTable from PNEANet. More...
static PTable	GetEdgeTable (const PNEANet &Network, TTableContext *Context)
	Extracts edge TTable from PNEANet. More...
static PTable	GetEdgeTablePN (const PNGraphMP &Network, TTableContext *Context)
	Extracts edge TTable from parallel graph PNGraphMP. More...
static PTable	GetFltNodePropertyTable (const PNEANet &Network, const TIntFltH &Property, const TStr &NodeAttrName, const TAttrType &NodeAttrType, const TStr &PropertyAttrName, TTableContext *Context)
	Extracts node and edge property TTables from THash. More...
static TTableIterator	GetMapPageRank (const TVec< PNEANet > &GraphSeq, TTableContext *Context, const double &C=0.85, const double &Eps=1e-4, const int &MaxIter=100)
	Gets sequence of PageRank tables from given GraphSeq. More...
static TTableIterator	GetMapHitsIterator (const TVec< PNEANet > &GraphSeq, TTableContext *Context, const int &MaxIter=20)
	Gets sequence of Hits tables from given GraphSeq. More...

Protected Member Functions
void	InvalidatePhysicalGroupings ()
void	InvalidateAffectedGroupings (const TStr &Attr)
void	IncrementNext ()
	Increments the next vector and set last, NumRows and NumValidRows. More...
void	ClassifyAux (const TIntV &SelectedRows, const TStr &LabelName, const TInt &PositiveLabel=1, const TInt &NegativeLabel=0)
	Adds a label attribute with positive labels on selected rows and negative labels on the rest. More...
const char *	GetContextKey (TInt Val) const
	Gets the Key of the Context StringVals pool. Used by ToGraph method in conv.cpp. More...
TStr	GetStrVal (TInt ColIdx, TInt RowIdx) const
	Gets the value in column with id ColIdx at row RowIdx. More...
void	AddStrVal (const TInt &ColIdx, const TStr &Val)
	Adds Val in column with id ColIdx. More...
void	AddStrVal (const TStr &Col, const TStr &Val)
	Adds Val in column with name Col. More...
TStr	GetIdColName () const
	Gets name of the id column of this table. More...
TStr	GetSchemaColName (TInt Idx) const
	Gets name of the column with index Idx in the schema. More...
TAttrType	GetSchemaColType (TInt Idx) const
	Gets type of the column with index Idx in the schema. More...
void	AddSchemaCol (const TStr &ColName, TAttrType ColType)
	Adds column with name ColName and type ColType to the schema. More...
TBool	IsColName (const TStr &ColName) const
void	AddColType (const TStr &ColName, TPair< TAttrType, TInt > ColType)
	Adds column with name ColName and type ColType to the ColTypeMap. More...
void	AddColType (const TStr &ColName, TAttrType ColType, TInt Index)
	Adds column with name ColName and type ColType to the ColTypeMap. More...
void	DelColType (const TStr &ColName)
	Adds column with name ColName and type ColType to the ColTypeMap. More...
TPair< TAttrType, TInt >	GetColTypeMap (const TStr &ColName) const
	Gets column type and index of ColName. More...
TStr	RenumberColName (const TStr &ColName) const
	Returns a re-numbered column name based on number of existing columns with conflicting names. More...
TStr	DenormalizeColName (const TStr &ColName) const
	Removes suffix to column name if exists. More...
Schema	DenormalizeSchema () const
	Removes suffix to column names in the Schema. More...
TBool	IsAttr (const TStr &Attr)
	Checks if Attr is an attribute of this table schema. More...
void	AddTable (const TTable &T)
	Adds all the rows of the input table. Allows duplicate rows (not a union). More...
void	ConcatTable (const PTable &T)
	Appends all rows of T to this table, and recalculate indices. More...
void	AddRow (const TRowIterator &RI)
	Adds row corresponding to RI. More...
void	AddRow (const TIntV &IntVals, const TFltV &FltVals, const TStrV &StrVals)
	Adds row with values corresponding to the given vectors by type. More...
void	AddGraphAttribute (const TStr &Attr, TBool IsEdge, TBool IsSrc, TBool IsDst)
	Adds names of columns to be used as graph attributes. More...
void	AddGraphAttributeV (TStrV &Attrs, TBool IsEdge, TBool IsSrc, TBool IsDst)
	Adds vector of names of columns to be used as graph attributes. More...
void	CheckAndAddIntNode (PNEANet Graph, THashSet< TInt > &NodeVals, TInt NodeId)
	Checks if given NodeId is seen earlier; if not, add it to Graph and hashmap NodeVals. More...
template<class T >
TInt	CheckAndAddFltNode (T Graph, THash< TFlt, TInt > &NodeVals, TFlt FNodeVal)
	Checks if given NodeVal is seen earlier; if not, add it to Graph and hashmap NodeVals. More...
void	AddEdgeAttributes (PNEANet &Graph, int RowId)
	Adds attributes of edge corresponding to RowId to the Graph. More...
void	AddNodeAttributes (TInt NId, TStrV NodeAttrV, TInt RowId, THash< TInt, TStrIntVH > &NodeIntAttrs, THash< TInt, TStrFltVH > &NodeFltAttrs, THash< TInt, TStrStrVH > &NodeStrAttrs)
	Takes as parameters, and updates, maps NodeXAttrs: Node Id –> (attribute name –> Vector of attribute values). More...
PNEANet	BuildGraph (const TIntV &RowIds, TAttrAggr AggrPolicy)
	Makes a single pass over the rows in the given row id set, and creates nodes, edges, assigns node and edge attributes. More...
void	InitRowIdBuckets (int NumBuckets)
	Initializes the RowIdBuckets vector which will be used for the graph sequence creation. More...
void	FillBucketsByWindow (TStr SplitAttr, TInt JumpSize, TInt WindowSize, TInt StartVal, TInt EndVal)
	Fills RowIdBuckets with sets of row ids. More...
void	FillBucketsByInterval (TStr SplitAttr, TIntPrV SplitIntervals)
	Fills RowIdBuckets with sets of row ids. More...
TVec< PNEANet >	GetGraphsFromSequence (TAttrAggr AggrPolicy)
	Returns a sequence of graphs. More...
PNEANet	GetFirstGraphFromSequence (TAttrAggr AggrPolicy)
	Returns the first graph of the sequence. More...
PNEANet	GetNextGraphFromSequence ()
	Returns the next graph in sequence corresponding to RowIdBuckets. More...
template<class T >
T	AggregateVector (TVec< T > &V, TAttrAggr Policy)
	Aggregates vector into a single scalar value according to a policy. More...
void	GroupingSanityCheck (const TStr &GroupBy, const TAttrType &AttrType) const
	Checks if grouping key exists and matches given attr type. More...
template<class T >
void	GroupByIntCol (const TStr &GroupBy, T &Grouping, const TIntV &IndexSet, TBool All, TBool UsePhysicalIds=true) const
	Groups/hashes by a single column with integer values. More...
template<class T >
void	GroupByFltCol (const TStr &GroupBy, T &Grouping, const TIntV &IndexSet, TBool All, TBool UsePhysicalIds=true) const
	Groups/hashes by a single column with float values. Returns hash table with grouping. More...
template<class T >
void	GroupByStrCol (const TStr &GroupBy, T &Grouping, const TIntV &IndexSet, TBool All, TBool UsePhysicalIds=true) const
	Groups/hashes by a single column with string values. Returns hash table with grouping. More...
template<class T >
void	UpdateGrouping (THash< T, TIntV > &Grouping, T Key, TInt Val) const
	Template for utility function to update a grouping hash map. More...
template<class T >
void	UpdateGrouping (THashMP< T, TIntV > &Grouping, T Key, TInt Val) const
	Template for utility function to update a parallel grouping hash map. More...
void	PrintGrouping (const THash< TGroupKey, TIntV > &Grouping) const
TInt	CompareRows (TInt R1, TInt R2, const TAttrType &CompareByType, const TInt &CompareByIndex, TBool Asc=true)
	Returns positive value if R1 is bigger, negative value if R2 is bigger, and 0 if they are equal (strcmp semantics). More...
TInt	CompareRows (TInt R1, TInt R2, const TVec< TAttrType > &CompareByTypes, const TIntV &CompareByIndices, TBool Asc=true)
	Returns positive value if R1 is bigger, negative value if R2 is bigger, and 0 if they are equal (strcmp semantics). More...
TInt	GetPivot (TIntV &V, TInt StartIdx, TInt EndIdx, const TVec< TAttrType > &SortByTypes, const TIntV &SortByIndices, TBool Asc)
	Gets pivot element for QSort. More...
TInt	Partition (TIntV &V, TInt StartIdx, TInt EndIdx, const TVec< TAttrType > &SortByTypes, const TIntV &SortByIndices, TBool Asc)
	Partitions vector for QSort. More...
void	ISort (TIntV &V, TInt StartIdx, TInt EndIdx, const TVec< TAttrType > &SortByTypes, const TIntV &SortByIndices, TBool Asc=true)
	Performs insertion sort on given vector V. More...
void	QSort (TIntV &V, TInt StartIdx, TInt EndIdx, const TVec< TAttrType > &SortByTypes, const TIntV &SortByIndices, TBool Asc=true)
	Performs QSort on given vector V. More...
void	Merge (TIntV &V, TInt Idx1, TInt Idx2, TInt Idx3, const TVec< TAttrType > &SortByTypes, const TIntV &SortByIndices, TBool Asc=true)
	Helper function for parallel QSort. More...
void	QSortPar (TIntV &V, const TVec< TAttrType > &SortByTypes, const TIntV &SortByIndices, TBool Asc=true)
	Performs QSort in parallel on given vector V. More...
bool	IsRowValid (TInt RowIdx) const
	Checks if RowIdx corresponds to a valid (i.e. not deleted) row. More...
TInt	GetLastValidRowIdx ()
	Gets the id of the last valid row of the table. More...
void	RemoveFirstRow ()
	Removes first valid row of the table. More...
void	RemoveRow (TInt RowIdx, TInt PrevRowIdx)
	Removes row with id RowIdx. More...
void	KeepSortedRows (const TIntV &KeepV)
	Removes all rows that are not mentioned in the SORTED vector KeepV. More...
void	SetFirstValidRow ()
	Sets the first valid row of the TTable. More...
PTable	InitializeJointTable (const TTable &Table)
	Initializes an empty table for the join of this table with the given table. More...
void	AddJointRow (const TTable &T1, const TTable &T2, TInt RowIdx1, TInt RowIdx2)
	Adds joint row T1[RowIdx1]<=>T2[RowIdx2]. More...
void	ThresholdJoinInputCorrectness (const TStr &KeyCol1, const TStr &JoinCol1, const TTable &Table, const TStr &KeyCol2, const TStr &JoinCol2)
void	ThresholdJoinCountCollisions (const TTable &TB, const TTable &TS, const TIntIntVH &T, TInt JoinColIdxB, TInt KeyColIdxB, TInt KeyColIdxS, THash< TIntPr, TIntTr > &Counters, TBool ThisIsSmaller, TAttrType JoinColType, TAttrType KeyType)
PTable	ThresholdJoinOutputTable (const THash< TIntPr, TIntTr > &Counters, TInt Threshold, const TTable &Table)
void	ThresholdJoinCountPerJoinKeyCollisions (const TTable &TB, const TTable &TS, const TIntIntVH &T, TInt JoinColIdxB, TInt KeyColIdxB, TInt KeyColIdxS, THash< TIntTr, TIntTr > &Counters, TBool ThisIsSmaller, TAttrType JoinColType, TAttrType KeyType)
PTable	ThresholdJoinPerJoinKeyOutputTable (const THash< TIntTr, TIntTr > &Counters, TInt Threshold, const TTable &Table)
void	ResizeTable (int RowCount)
	Resizes the table to hold RowCount rows. More...
int	GetEmptyRowsStart (int NewRows)
	Gets the start index to a chunk of empty rows of size NewRows. More...
void	AddSelectedRows (const TTable &Table, const TIntV &RowIDs)
	Adds rows from Table that correspond to ids in RowIDs. More...
void	AddNRows (int NewRows, const TVec< TIntV > &IntColsP, const TVec< TFltV > &FltColsP, const TVec< TIntV > &StrColMapsP)
	Adds NewRows rows from the given vectors for each column type. More...
void	AddNJointRowsMP (const TTable &T1, const TTable &T2, const TVec< TIntPrV > &JointRowIDSet)
	Adds rows from T1 and T2 to this table in a parallel manner. Used by Join. More...
void	UpdateTableForNewRow ()
	Updates table state after adding one or more rows. More...
void	GroupAux (const TStrV &GroupBy, THash< TGroupKey, TPair< TInt, TIntV > > &Grouping, TBool Ordered, const TStr &GroupColName, TBool KeepUnique, TIntV &UniqueVec, TBool UsePhysicalIds=true)
	Helper function for grouping. More...
void	StoreGroupCol (const TStr &GroupColName, const TVec< TPair< TInt, TInt > > &GroupAndRowIds)
	Parallel helper function for grouping. - we currently don't support such parallel grouping by complex keys. More...
void	Reindex ()
	Reinitializes row ids. More...
void	AddIdColumn (const TStr &IdColName)
	Adds a column of explicit integer identifiers to the rows. More...
void	GetCollidingRows (const TTable &T, THashSet< TInt > &Collisions)
	Gets set of row ids of rows common with table T. More...

Static Protected Member Functions
static void	LoadSSPar (PTable &NewTable, const Schema &S, const TStr &InFNm, const TIntV &RelevantCols, const char &Separator, TBool HasTitleLine)
	Parallelly loads data from input file at InFNm into NewTable. Only work when NewTable has no string columns. More...
static void	LoadSSSeq (PTable &NewTable, const Schema &S, const TStr &InFNm, const TIntV &RelevantCols, const char &Separator, TBool HasTitleLine)
	Sequentially loads data from input file at InFNm into NewTable. More...
static TInt	CompareKeyVal (const TInt &K1, const TInt &V1, const TInt &K2, const TInt &V2)
static TInt	CheckSortedKeyVal (TIntV &Key, TIntV &Val, TInt Start, TInt End)
static void	ISortKeyVal (TIntV &Key, TIntV &Val, TInt Start, TInt End)
static TInt	GetPivotKeyVal (TIntV &Key, TIntV &Val, TInt Start, TInt End)
static TInt	PartitionKeyVal (TIntV &Key, TIntV &Val, TInt Start, TInt End)
static void	QSortKeyVal (TIntV &Key, TIntV &Val, TInt Start, TInt End)

Protected Attributes
TTableContext *	Context
	Execution Context. More...
Schema	Sch
	Table Schema. More...
TCRef	CRef
TInt	NumRows
	Number of rows in the table (valid and invalid). More...
TInt	NumValidRows
	Number of valid rows in the table (i.e. rows that were not logically removed). More...
TInt	FirstValidRow
	Physical index of first valid row. More...
TInt	LastValidRow
	Physical index of last valid row. More...
TIntV	Next
	A vector describing the logical order of the rows. More...
TVec< TIntV >	IntCols
	Next[i] is the successor of row i. Table iterators follow the order dictated by Next More...
TVec< TFltV >	FltCols
	Data columns of floating point attributes. More...
TVec< TIntV >	StrColMaps
	Data columns of integer mappings of string attributes. More...
THash< TStr, TPair< TAttrType, TInt > >	ColTypeMap
TStr	IdColName
	A mapping from column name to column type and column index among columns of the same type. More...
TIntIntH	RowIdMap
	Mapping of permanent row ids to physical id. More...
THash< TStr, THash< TInt, TIntV > >	IntColIndexes
	Indexes for Int Columns. More...
THash< TStr, THash< TInt, TIntV > >	StrMapColIndexes
	Indexes for String Columns. More...
THash< TStr, THash< TFlt, TIntV > >	FltColIndexes
	Indexes for Float Columns. More...
THash< TStr, GroupStmt >	GroupStmtNames
	Maps user-given grouping statement names to their group-by attributes. More...
THash< GroupStmt, THash< TInt, TGroupKey > >	GroupIDMapping
	Maps grouping statements to their (group id –> group-by key) mapping. More...
THash< GroupStmt, THash < TGroupKey, TIntV > >	GroupMapping
	Maps grouping statements to their (group-by key –> group id) mapping. More...
TStr	SrcCol
	Column (attribute) to serve as src nodes when constructing the graph. More...
TStr	DstCol
	Column (attribute) to serve as dst nodes when constructing the graph. More...
TStrV	EdgeAttrV
	List of columns (attributes) to serve as edge attributes. More...
TStrV	SrcNodeAttrV
	List of columns (attributes) to serve as source node attributes. More...
TStrV	DstNodeAttrV
	List of columns (attributes) to serve as destination node attributes. More...
TStrTrV	CommonNodeAttrs
	List of attribute pairs with values common to source and destination and their common given name. More...
TVec< TIntV >	RowIdBuckets
	Partitioning of row ids into buckets corresponding to different graph objects when generating a sequence of graphs. More...
TInt	CurrBucket
	Current row id bucket - used when generating a sequence of graphs using an iterator. More...
TAttrAggr	AggrPolicy
	Aggregation policy used for solving conflicts between different values of an attribute of the same node. More...
TInt	IsNextDirty
	Flag to signify whether the rows are stored in logical sequence or reordered. Used for optimizing GetPartitionRanges. More...

Static Protected Attributes
static const TInt	Last = -1
	Special value for Next vector entry - last row in table. More...
static const TInt	Invalid = -2
	Special value for Next vector entry - logically removed row. More...
static TInt	UseMP = 1
	Global switch for choosing multi-threaded versions of TTable functions. More...

Friends
class	TPt< TTable >
class	TRowIterator
class	TRowIteratorWithRemove
template<class PGraph >
PGraph	TSnap::ToGraph (PTable Table, const TStr &SrcCol, const TStr &DstCol, TAttrAggr AggrPolicy)
template<class PGraph >
PGraph	TSnap::ToNetwork (PTable Table, const TStr &SrcCol, const TStr &DstCol, TStrV &SrcAttrs, TStrV &DstAttrs, TStrV &EdgeAttrs, TAttrAggr AggrPolicy)
template<class PGraph >
PGraph	TSnap::ToNetwork (PTable Table, const TStr &SrcCol, const TStr &DstCol, TStrV &EdgeAttrV, TAttrAggr AggrPolicy)
template<class PGraph >
PGraph	TSnap::ToNetwork (PTable Table, const TStr &SrcCol, const TStr &DstCol, TAttrAggr AggrPolicy)
template<class PGraph >
PGraph	TSnap::ToNetwork (PTable Table, const TStr &SrcCol, const TStr &DstCol, TStrV &EdgeAttrV, PTable NodeTable, const TStr &NodeCol, TStrV &NodeAttrV, TAttrAggr AggrPolicy)
int	TSnap::LoadCrossNet (TCrossNet &Graph, PTable Table, const TStr &SrcCol, const TStr &DstCol, TStrV &EdgeAttrV)
int	TSnap::LoadMode (TModeNet &Graph, PTable Table, const TStr &NCol, TStrV &NodeAttrV)
template<class PGraphMP >
PGraphMP	TSnap::ToGraphMP (PTable Table, const TStr &SrcCol, const TStr &DstCol)
template<class PGraphMP >
PGraphMP	TSnap::ToGraphMP3 (PTable Table, const TStr &SrcCol, const TStr &DstCol)
template<class PGraphMP >
PGraphMP	TSnap::ToNetworkMP (PTable Table, const TStr &SrcCol, const TStr &DstCol, TStrV &SrcAttrs, TStrV &DstAttrs, TStrV &EdgeAttrs, TAttrAggr AggrPolicy)
template<class PGraphMP >
PGraphMP	TSnap::ToNetworkMP2 (PTable Table, const TStr &SrcCol, const TStr &DstCol, TStrV &SrcAttrs, TStrV &DstAttrs, TStrV &EdgeAttrs, TAttrAggr AggrPolicy)
template<class PGraphMP >
PGraphMP	TSnap::ToNetworkMP (PTable Table, const TStr &SrcCol, const TStr &DstCol, TStrV &EdgeAttrV, TAttrAggr AggrPolicy)
template<class PGraphMP >
PGraphMP	TSnap::ToNetworkMP (PTable Table, const TStr &SrcCol, const TStr &DstCol, TAttrAggr AggrPolicy)
template<class PGraphMP >
PGraphMP	TSnap::ToNetworkMP (PTable Table, const TStr &SrcCol, const TStr &DstCol, TStrV &EdgeAttrV, PTable NodeTable, const TStr &NodeCol, TStrV &NodeAttrV, TAttrAggr AggrPolicy)

Detailed Description

Table class: Relational table with columnar data storage.

Definition at line 495 of file table.h.

Constructor & Destructor Documentation

TTable::TTable

(

)

Definition at line 302 of file table.cpp.

              : Context(new TTableContext), NumRows(0), NumValidRows(0),
  FirstValidRow(0), LastValidRow(-1) {}

TTable::TTable

(

TTableContext *

Context

)

Definition at line 305 of file table.cpp.

                                    : Context(Context), NumRows(0),
  NumValidRows(0), FirstValidRow(0), LastValidRow(-1) {}

TTable::TTable	(	const Schema &	S,
		TTableContext *	Context
	)

Definition at line 308 of file table.cpp.

                                                               : Context(Context), 
  NumRows(0), NumValidRows(0), FirstValidRow(0), LastValidRow(-1), IsNextDirty(0) {
  TInt IntColCnt = 0;
  TInt FltColCnt = 0;
  TInt StrColCnt = 0;
  for (TInt i = 0; i < TableSchema.Len(); i++) {
    TStr ColName = TableSchema[i].Val1;
    TAttrType ColType = TableSchema[i].Val2;
    AddSchemaCol(ColName, ColType);
    switch (ColType) {
      case atInt:
        AddColType(ColName, atInt, IntColCnt);
        IntColCnt++;
        break;
      case atFlt:
        AddColType(ColName, atFlt, FltColCnt);
        FltColCnt++;
        break;
      case atStr:
        AddColType(ColName, atStr, StrColCnt);
        StrColCnt++;
        break;
    }
  }
  IntCols = TVec<TIntV>(IntColCnt);
  FltCols = TVec<TFltV>(FltColCnt);
  StrColMaps = TVec<TIntV>(StrColCnt);
}

TTable::TTable	(	TSIn &	SIn,
		TTableContext *	Context
	)

Definition at line 337 of file table.cpp.

                                               : Context(Context), NumRows(SIn),
  NumValidRows(SIn), FirstValidRow(SIn), LastValidRow(SIn), Next(SIn), IntCols(SIn),
  FltCols(SIn), StrColMaps(SIn) {
  THash<TStr,TPair<TInt,TInt> > ColTypeIntMap(SIn);

  ColTypeMap.Clr();
  Sch.Clr();
  for (THash<TStr,TPair<TInt,TInt> >::TIter it = ColTypeIntMap.BegI(); it < ColTypeIntMap.EndI(); it++) {
    TPair<TInt,TInt> dat = it.GetDat();
    switch (dat.GetVal1()) {
      case 0:
        AddColType(it.GetKey(), atInt, dat.GetVal2());
        AddSchemaCol(it.GetKey(), atInt);
        break;
      case 1:
        AddColType(it.GetKey(), atFlt, dat.GetVal2());
        AddSchemaCol(it.GetKey(), atFlt);
        break;
      case 2:
        AddColType(it.GetKey(), atStr, dat.GetVal2());
        AddSchemaCol(it.GetKey(), atStr);
        break;
    }
  }

  IsNextDirty = 0;
}

TTable::TTable	(	const THash< TInt, TInt > &	H,
		const TStr &	Col1,
		const TStr &	Col2,
		TTableContext *	Context,
		const TBool	IsStrKeys = false
	)

Constructor to build table out of a hash table of int->int.

Definition at line 365 of file table.cpp.

                                                : Context(Context), NumRows(H.Len()),
  NumValidRows(H.Len()), FirstValidRow(0), LastValidRow(H.Len()-1) {
    TAttrType KeyType = IsStrKeys ? atStr : atInt;
    AddSchemaCol(Col1, KeyType);
    AddSchemaCol(Col2, atInt);
    AddColType(Col1, KeyType, 0);
    AddColType(Col2, atInt, 1);
    if (IsStrKeys) {
      StrColMaps = TVec<TIntV>(1);
      IntCols = TVec<TIntV>(1);
      H.GetKeyV(StrColMaps[0]);
      H.GetDatV(IntCols[0]);
    } else {
      IntCols = TVec<TIntV>(2);
      H.GetKeyV(IntCols[0]);
      H.GetDatV(IntCols[1]);
    }
    Next = TIntV(NumRows);
    for (TInt i = 0; i < NumRows; i++) {
      Next[i] = i+1;
    }
    Next[NumRows-1] = Last;
    IsNextDirty = 0;
    InitIds();
}

TTable::TTable	(	const THash< TInt, TFlt > &	H,
		const TStr &	Col1,
		const TStr &	Col2,
		TTableContext *	Context,
		const TBool	IsStrKeys = false
	)

Constructor to build table out of a hash table of int->float.

Definition at line 392 of file table.cpp.

                                                : Context(Context),
  NumRows(H.Len()), NumValidRows(H.Len()), FirstValidRow(0), LastValidRow(H.Len()-1) {
  TAttrType KeyType = IsStrKeys ? atStr : atInt;
  AddSchemaCol(Col1, KeyType);
  AddSchemaCol(Col2, atFlt);
  AddColType(Col1, KeyType, 0);
  AddColType(Col2, atFlt, 0);
  if (IsStrKeys) {
    StrColMaps = TVec<TIntV>(1);
    H.GetKeyV(StrColMaps[0]);
  } else {
    IntCols = TVec<TIntV>(1);
    H.GetKeyV(IntCols[0]);
  }
  FltCols = TVec<TFltV>(1);
  H.GetDatV(FltCols[0]);
  Next = TIntV(NumRows);
  for (TInt i = 0; i < NumRows; i++) {
    Next[i] = i+1;
  }
  Next[NumRows-1] = Last;
  IsNextDirty = 0;
  InitIds();
}

TTable::TTable ( const TTable &  Table )

inline

Copy constructor.

Definition at line 918 of file table.h.

                             : Context(Table.Context), Sch(Table.Sch),
    NumRows(Table.NumRows), NumValidRows(Table.NumValidRows), FirstValidRow(Table.FirstValidRow),
    LastValidRow(Table.LastValidRow), Next(Table.Next), IntCols(Table.IntCols),
    FltCols(Table.FltCols), StrColMaps(Table.StrColMaps), ColTypeMap(Table.ColTypeMap),
    IdColName(Table.IdColName), RowIdMap(Table.RowIdMap), GroupStmtNames(Table.GroupStmtNames),
    GroupIDMapping(Table.GroupIDMapping), GroupMapping(Table.GroupMapping),
    SrcCol(Table.SrcCol), DstCol(Table.DstCol),
    EdgeAttrV(Table.EdgeAttrV), SrcNodeAttrV(Table.SrcNodeAttrV),
    DstNodeAttrV(Table.DstNodeAttrV), CommonNodeAttrs(Table.CommonNodeAttrs),
    IsNextDirty(Table.IsNextDirty) {}

TTable::TTable	(	const TTable &	Table,
		const TIntV &	RowIds
	)

Definition at line 418 of file table.cpp.

                                                       : Context(Table.Context),
  Sch(Table.Sch), SrcCol(Table.SrcCol), DstCol(Table.DstCol), EdgeAttrV(Table.EdgeAttrV),
  SrcNodeAttrV(Table.SrcNodeAttrV), DstNodeAttrV(Table.DstNodeAttrV),
  CommonNodeAttrs(Table.CommonNodeAttrs) {
  ColTypeMap = Table.ColTypeMap;
  IntCols = TVec<TIntV>(Table.IntCols.Len());
  FltCols = TVec<TFltV>(Table.FltCols.Len());
  StrColMaps = TVec<TIntV>(Table.StrColMaps.Len());
  FirstValidRow = 0;
  LastValidRow = -1;
  NumRows = 0;
  NumValidRows = 0;
  AddSelectedRows(Table, RowIDs);
  IsNextDirty = 0;
  InitIds();
}

Member Function Documentation

void TTable::AddColType ( const TStr &  ColName, TPair< TAttrType, TInt >  ColType  )

inlineprotected

Adds column with name ColName and type ColType to the ColTypeMap.

Definition at line 661 of file table.h.

                                                                      {
    TStr NColName = NormalizeColName(ColName);
    ColTypeMap.AddDat(NColName, ColType);
  }

void TTable::AddColType ( const TStr &  ColName, TAttrType  ColType, TInt  Index  )

inlineprotected

Adds column with name ColName and type ColType to the ColTypeMap.

Definition at line 666 of file table.h.

                                                                      {
    TStr NColName = NormalizeColName(ColName);
    AddColType(NColName, TPair<TAttrType,TInt>(ColType, Index));
  }

void TTable::AddDstNodeAttr ( const TStr &  Attr )

inline

Adds column to be used as dst node atribute of the graph.

Definition at line 1171 of file table.h.

{ AddGraphAttribute(Attr, false, false, true); }

void TTable::AddDstNodeAttr ( TStrV &  Attrs )

inline

Adds columns to be used as dst node attributes of the graph.

Definition at line 1173 of file table.h.

{ AddGraphAttributeV(Attrs, false, false, true); }

void TTable::AddEdgeAttr ( const TStr &  Attr )

inline

Adds column to be used as graph edge attribute.

Definition at line 1163 of file table.h.

{ AddGraphAttribute(Attr, true, false, false); }

void TTable::AddEdgeAttr ( TStrV &  Attrs )

inline

Adds columns to be used as graph edge attributes.

Definition at line 1165 of file table.h.

{ AddGraphAttributeV(Attrs, true, false, false); }

void TTable::AddEdgeAttributes ( PNEANet &  Graph, int  RowId  )

inlineprotected

Adds attributes of edge corresponding to RowId to the Graph.

Definition at line 3375 of file table.cpp.

                                                               {
  for (TInt i = 0; i < EdgeAttrV.Len(); i++) {
    TStr ColName = EdgeAttrV[i];
    TAttrType T = GetColType(ColName);
    TInt Index = GetColIdx(ColName);
    switch (T) {
      case atInt:
        Graph->AddIntAttrDatE(RowId, IntCols[Index][RowId], ColName);
        break;
      case atFlt:
        Graph->AddFltAttrDatE(RowId, FltCols[Index][RowId], ColName);
        break;
      case atStr:
        Graph->AddStrAttrDatE(RowId, GetStrVal(Index, RowId), ColName);
        break;
    }
  }
}

void TTable::AddFltCol

(

const TStr &

ColName

)

Adds a float column with name ColName.

Definition at line 4657 of file table.cpp.

                                          {
  AddSchemaCol(ColName, atFlt);
  FltCols.Add(TFltV(NumRows));
  TInt L = FltCols.Len();
  AddColType(ColName, atFlt, L-1);
}

void TTable::AddGraphAttribute ( const TStr &  Attr, TBool  IsEdge, TBool  IsSrc, TBool  IsDst  )

protected

Adds names of columns to be used as graph attributes.

Definition at line 965 of file table.cpp.

                                                                                       {
  if (!IsColName(Attr)) { TExcept::Throw(Attr + ": No such column"); }
  if (IsEdge) { EdgeAttrV.Add(NormalizeColName(Attr)); }
  if (IsSrc) { SrcNodeAttrV.Add(NormalizeColName(Attr)); }
  if (IsDst) { DstNodeAttrV.Add(NormalizeColName(Attr)); }
}

void TTable::AddGraphAttributeV ( TStrV &  Attrs, TBool  IsEdge, TBool  IsSrc, TBool  IsDst  )

protected

Adds vector of names of columns to be used as graph attributes.

Definition at line 972 of file table.cpp.

                                                                                    {
  for (TInt i = 0; i < Attrs.Len(); i++) {
    if (!IsColName(Attrs[i])) {
      TExcept::Throw(Attrs[i] + ": no such column");
    }
  }
  for (TInt i = 0; i < Attrs.Len(); i++) {
    if (IsEdge) { EdgeAttrV.Add(NormalizeColName(Attrs[i])); }
    if (IsSrc) { SrcNodeAttrV.Add(NormalizeColName(Attrs[i])); }
    if (IsDst) { DstNodeAttrV.Add(NormalizeColName(Attrs[i])); }
  }
}

void TTable::AddIdColumn ( const TStr &  IdColName )

protected

Adds a column of explicit integer identifiers to the rows.

Definition at line 1880 of file table.cpp.

                                            {
  //printf("NumRows: %d\n", NumRows.Val);
  TInt IdCol = IntCols.Add();
  IntCols[IdCol].Reserve(NumRows, NumRows);
  //printf("IdCol Reserved\n");
  TInt IdCnt = 0;
  RowIdMap.Clr();
  for (TRowIterator RI = BegRI(); RI < EndRI(); RI++) {
    IntCols[IdCol][RI.GetRowIdx()] = IdCnt;
    RowIdMap.AddDat(IdCnt, RI.GetRowIdx());
    IdCnt++;
  }
  AddSchemaCol(ColName, atInt);
  AddColType(ColName, atInt, IntCols.Len()-1);
}

void TTable::AddIntCol

(

const TStr &

ColName

)

Adds an integer column with name ColName.

Definition at line 4650 of file table.cpp.

                                          {
  AddSchemaCol(ColName, atInt);
  IntCols.Add(TIntV(NumRows));
  TInt L = IntCols.Len();
  AddColType(ColName, atInt, L-1);
}

void TTable::AddJointRow ( const TTable &  T1, const TTable &  T2, TInt  RowIdx1, TInt  RowIdx2  )

protected

Adds joint row T1[RowIdx1]<=>T2[RowIdx2].

Definition at line 1937 of file table.cpp.

                                                                                       {
  for (TInt i = 0; i < T1.IntCols.Len(); i++) {
    IntCols[i].Add(T1.IntCols[i][RowIdx1]);
  }
  for (TInt i = 0; i < T1.FltCols.Len(); i++) {
    FltCols[i].Add(T1.FltCols[i][RowIdx1]);
  }
  for (TInt i = 0; i < T1.StrColMaps.Len(); i++) {
    StrColMaps[i].Add(T1.StrColMaps[i][RowIdx1]);
  }
  TInt IntOffset = T1.IntCols.Len();
  TInt FltOffset = T1.FltCols.Len();
  TInt StrOffset = T1.StrColMaps.Len();
  for (TInt i = 0; i < T2.IntCols.Len(); i++) {
    IntCols[i+IntOffset].Add(T2.IntCols[i][RowIdx2]);
  }
  for (TInt i = 0; i < T2.FltCols.Len(); i++) {
    FltCols[i+FltOffset].Add(T2.FltCols[i][RowIdx2]);
  }
  for (TInt i = 0; i < T2.StrColMaps.Len(); i++) {
    StrColMaps[i+StrOffset].Add(T2.StrColMaps[i][RowIdx2]);
  }
  TInt IdOffset = IntOffset + T2.IntCols.Len(); 
  NumRows++;
  NumValidRows++;
  if (!Next.Empty()) {
    Next[Next.Len()-1] = NumValidRows-1;
    LastValidRow = NumValidRows-1;
  }
  Next.Add(Last);
  RowIdMap.AddDat(NumRows-1,NumRows-1);
  IntCols[IdOffset].Add(NumRows-1);
}

void TTable::AddNJointRowsMP ( const TTable &  T1, const TTable &  T2, const TVec< TIntPrV > &  JointRowIDSet  )

protected

Adds rows from T1 and T2 to this table in a parallel manner. Used by Join.

Definition at line 4419 of file table.cpp.

                                                                                                   {
  //double startFn = omp_get_wtime();
  int JointTableSize = 0;
  TIntV StartOffsets(JointRowIDSet.Len());
  for (int i = 0; i < JointRowIDSet.Len(); i++) {
    StartOffsets[i] = JointTableSize;
    JointTableSize += JointRowIDSet[i].Len();
  }
  if (JointTableSize == 0) {
    TExcept::Throw("Joint table is empty");
  }
  //double endOffsets = omp_get_wtime();
  //printf("Offsets time = %f\n",endOffsets-startFn);
  ResizeTable(JointTableSize);
  //double endResize = omp_get_wtime();
  //printf("Resize time = %f\n",endResize-endOffsets);
  NumRows = JointTableSize;
  NumValidRows = JointTableSize;
  Assert(NumRows <= Next.Len());

  TInt IntOffset = T1.IntCols.Len();
  TInt FltOffset = T1.FltCols.Len();
  TInt StrOffset = T1.StrColMaps.Len();

  TInt IdOffset = IntOffset + T2.IntCols.Len();
  RowIdMap.Clr();
  for (TInt IdCnt = 0; IdCnt < JointTableSize; IdCnt++) {
    RowIdMap.AddDat(IdCnt, IdCnt);
  }

  #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD) 
  for (int j = 0; j < JointRowIDSet.Len(); j++) {
    const TIntPrV& RowIDs = JointRowIDSet[j];
    int start = StartOffsets[j];
    int NewRows = RowIDs.Len();
    if (NewRows == 0) {continue;}
    for (TInt r = 0; r < NewRows; r++){
      TIntPr CurrRowIdPr = RowIDs[r]; 
      for(TInt i = 0; i < T1.IntCols.Len(); i++){
        IntCols[i][start+r] = T1.IntCols[i][CurrRowIdPr.GetVal1()];
      }
      for(TInt i = 0; i < T1.FltCols.Len(); i++){
        FltCols[i][start+r] = T1.FltCols[i][CurrRowIdPr.GetVal1()];
      }
      for(TInt i = 0; i < T1.StrColMaps.Len(); i++){
        StrColMaps[i][start+r] = T1.StrColMaps[i][CurrRowIdPr.GetVal1()];
      }
      for(TInt i = 0; i < T2.IntCols.Len(); i++){
        IntCols[i+IntOffset][start+r] = T2.IntCols[i][CurrRowIdPr.GetVal2()];
      }
      for(TInt i = 0; i < T2.FltCols.Len(); i++){
        FltCols[i+FltOffset][start+r] = T2.FltCols[i][CurrRowIdPr.GetVal2()];
      }
      for(TInt i = 0; i < T2.StrColMaps.Len(); i++){
        StrColMaps[i+StrOffset][start+r] = T2.StrColMaps[i][CurrRowIdPr.GetVal2()];
      }
      IntCols[IdOffset][start+r] = start+r;
    }
    for(TInt r = 0; r < NewRows; r++){
      Next[start+r] = start+r+1;
    }
  }      
  LastValidRow = JointTableSize-1;
  Next[LastValidRow] = Last;
  //double endIterate = omp_get_wtime();
  //printf("Iterate time = %f\n",endIterate-endResize);
}

void TTable::AddNodeAttr ( const TStr &  Attr )

inline

Handles the common case where src and dst both belong to the same "universe" of entities.

Definition at line 1175 of file table.h.

{ AddSrcNodeAttr(Attr); AddDstNodeAttr(Attr); }

void TTable::AddNodeAttr ( TStrV &  Attrs )

inline

Handles the common case where src and dst both belong to the same "universe" of entities.

Definition at line 1177 of file table.h.

{ AddSrcNodeAttr(Attrs); AddDstNodeAttr(Attrs); }

void TTable::AddNodeAttributes ( TInt  NId, TStrV  NodeAttrV, TInt  RowId, THash< TInt, TStrIntVH > &  NodeIntAttrs, THash< TInt, TStrFltVH > &  NodeFltAttrs, THash< TInt, TStrStrVH > &  NodeStrAttrs  )

inlineprotected

Takes as parameters, and updates, maps NodeXAttrs: Node Id –> (attribute name –> Vector of attribute values).

Definition at line 3394 of file table.cpp.

                                                                              {
  for (TInt i = 0; i < NodeAttrV.Len(); i++) {
    TStr ColAttr = NodeAttrV[i];
    TAttrType CT = GetColType(ColAttr);
    int ColId = GetColIdx(ColAttr);
    // check if this is a common src-dst attribute
    for (TInt i = 0; i < CommonNodeAttrs.Len(); i++) {
      if (CommonNodeAttrs[i].Val1 == ColAttr || CommonNodeAttrs[i].Val2 == ColAttr) {
        ColAttr = CommonNodeAttrs[i].Val3;
        break;
      }
    }
    if (CT == atInt) {
      if (!NodeIntAttrs.IsKey(NId)) { NodeIntAttrs.AddKey(NId); }
      if (!NodeIntAttrs.GetDat(NId).IsKey(ColAttr)) { NodeIntAttrs.GetDat(NId).AddKey(ColAttr); }
      NodeIntAttrs.GetDat(NId).GetDat(ColAttr).Add(IntCols[ColId][RowId]);
    } else if (CT == atFlt) {
      if (!NodeFltAttrs.IsKey(NId)) { NodeFltAttrs.AddKey(NId); }
      if (!NodeFltAttrs.GetDat(NId).IsKey(ColAttr)) { NodeFltAttrs.GetDat(NId).AddKey(ColAttr); }
      NodeFltAttrs.GetDat(NId).GetDat(ColAttr).Add(FltCols[ColId][RowId]);
    } else {
      if (!NodeStrAttrs.IsKey(NId)) { NodeStrAttrs.AddKey(NId); }
      if (!NodeStrAttrs.GetDat(NId).IsKey(ColAttr)) { NodeStrAttrs.GetDat(NId).AddKey(ColAttr); }
      NodeStrAttrs.GetDat(NId).GetDat(ColAttr).Add(GetStrVal(ColId, RowId));
    }
  }
}

void TTable::AddNRows ( int  NewRows, const TVec< TIntV > &  IntColsP, const TVec< TFltV > &  FltColsP, const TVec< TIntV > &  StrColMapsP  )

protected

Adds NewRows rows from the given vectors for each column type.

Definition at line 4398 of file table.cpp.

                                                                                                                           {
  if (NewRows == 0) { return; }
  // this call should be thread-safe
  int start = GetEmptyRowsStart(NewRows);
  for (TInt r = 0; r < NewRows; r++) {
    for (TInt i = 0; i < IntColsP.Len(); i++) {
      IntCols[i][start+r] = IntColsP[i][r];
    }
    for (TInt i = 0; i < FltColsP.Len(); i++) {
      FltCols[i][start+r] = FltColsP[i][r];
    }
    for (TInt i = 0; i < StrColMapsP.Len(); i++) {
      StrColMaps[i][start+r] = StrColMapsP[i][r];
    }
  }
  for (TInt r = 0; r < NewRows-1; r++) {
    Next[start+r] = start+r+1;
  }
}

void TTable::AddRow ( const TRowIterator &  RI )

protected

Adds row corresponding to RI.

Definition at line 4272 of file table.cpp.

                                          {
  for (TInt c = 0; c < Sch.Len(); c++) {
    TStr ColName = GetSchemaColName(c);
    if (ColName == IdColName) { continue; }

    TInt ColIdx = GetColIdx(ColName);

    switch (GetColType(ColName)) {
    case atInt:
       IntCols[ColIdx].Add(RI.GetIntAttr(ColName));
       break;
    case atFlt:
       FltCols[ColIdx].Add(RI.GetFltAttr(ColName));
       break;
    case atStr:
       StrColMaps[ColIdx].Add(RI.GetStrMapByName(ColName));
       break;
    }
  }
  UpdateTableForNewRow();
}

void TTable::AddRow ( const TIntV &  IntVals, const TFltV &  FltVals, const TStrV &  StrVals  )

protected

Adds row with values corresponding to the given vectors by type.

Definition at line 4294 of file table.cpp.

                                                                                    {
  for (TInt c = 0; c < IntVals.Len(); c++) {
    IntCols[c].Add(IntVals[c]);
  }
  for (TInt c = 0; c < FltVals.Len(); c++) {
    FltCols[c].Add(FltVals[c]);
  }
  for (TInt c = 0; c < StrVals.Len(); c++) {
    AddStrVal(c, StrVals[c]);
  }
  UpdateTableForNewRow();
}

void TTable::AddRow ( const TTableRow &  Row )

inline

Adds row with values taken from given TTableRow.

Definition at line 993 of file table.h.

{ AddRow(Row.GetIntVals(), Row.GetFltVals(), Row.GetStrVals()); };

void TTable::AddSchemaCol ( const TStr &  ColName, TAttrType  ColType  )

inlineprotected

Adds column with name ColName and type ColType to the schema.

Definition at line 652 of file table.h.

                                                            {
    TStr NColName = NormalizeColName(ColName);
    Sch.Add(TPair<TStr,TAttrType>(NColName, ColType));
  }

void TTable::AddSelectedRows ( const TTable &  Table, const TIntV &  RowIDs  )

protected

Adds rows from Table that correspond to ids in RowIDs.

Definition at line 4376 of file table.cpp.

                                                                     {
  int NewRows = RowIDs.Len();
  if (NewRows == 0) { return; }
  // this call should be thread-safe
  int start = GetEmptyRowsStart(NewRows);
  for (TInt r = 0; r < NewRows; r++) {
    TInt CurrRowIdx = RowIDs[r];
    for (TInt i = 0; i < Table.IntCols.Len(); i++) {
      IntCols[i][start+r] = Table.IntCols[i][CurrRowIdx];
    }
    for (TInt i = 0; i < Table.FltCols.Len(); i++) {
      FltCols[i][start+r] = Table.FltCols[i][CurrRowIdx];
    }
    for (TInt i = 0; i < Table.StrColMaps.Len(); i++) {
      StrColMaps[i][start+r] = Table.StrColMaps[i][CurrRowIdx];
    }
  }
  for (TInt r = 0; r < NewRows-1; r++) {
    Next[start+r] = start+r+1;
  }
}  

void TTable::AddSrcNodeAttr ( const TStr &  Attr )

inline

Adds column to be used as src node atribute of the graph.

Definition at line 1167 of file table.h.

{ AddGraphAttribute(Attr, false, true, false); }

void TTable::AddSrcNodeAttr ( TStrV &  Attrs )

inline

Adds columns to be used as src node attributes of the graph.

Definition at line 1169 of file table.h.

{ AddGraphAttributeV(Attrs, false, true, false); }

void TTable::AddStrCol

(

const TStr &

ColName

)

Adds a string column with name ColName.

Definition at line 4664 of file table.cpp.

                                          {
  AddSchemaCol(ColName, atStr);
  StrColMaps.Add(TIntV(NumRows));
  TInt L = StrColMaps.Len();
  AddColType(ColName, atStr, L-1);
}

void TTable::AddStrVal ( const TInt &  ColIdx, const TStr &  Val  )

protected

Adds Val in column with id ColIdx.

Definition at line 951 of file table.cpp.

                                                          {
  TInt KeyId = TInt(Context->StringVals.AddKey(Key));
  //printf("TTable::AddStrVal2  %d  .%s.  %d\n", ColIdx.Val, Key.CStr(), KeyId.Val);
  StrColMaps[ColIdx].Add(KeyId);
}

void TTable::AddStrVal ( const TStr &  Col, const TStr &  Val  )

protected

Adds Val in column with name Col.

Definition at line 957 of file table.cpp.

                                                       {
  if (GetColType(Col) != atStr) {
    TExcept::Throw(Col + " is not a string valued column");
  }
  //printf("TTable::AddStrVal1  .%s.  .%s.\n", Col.CStr(), Key.CStr());
  AddStrVal(GetColIdx(Col), Key);
}

void TTable::AddTable ( const TTable &  T )

protected

Adds all the rows of the input table. Allows duplicate rows (not a union).

Definition at line 3952 of file table.cpp.

                                     {
  //for (TInt c = 0; c < S.Len(); c++) {
  //  if (S[c] != T.S[c]) { printf("(%s,%d) != (%s,%d)\n", S[c].Val1.CStr(), S[c].Val2, T.S[c].Val1.CStr(), T.S[c].Val2); TExcept::Throw("when adding tables, their schemas must match!"); }
  //}
  for (TInt c = 0; c < Sch.Len(); c++) {
    TStr ColName = GetSchemaColName(c);
    TInt ColIdx = GetColIdx(ColName);
    TInt TColIdx = ColName == IdColName ? T.GetColIdx(T.IdColName) : T.GetColIdx(ColName);
    if (TColIdx < 0) { TExcept::Throw("when adding a table, it must contain all columns of source table!"); }
    switch (GetColType(ColName)) { 
    case atInt:
       IntCols[ColIdx].AddV(T.IntCols[TColIdx]);
       break;
    case atFlt:
       FltCols[ColIdx].AddV(T.FltCols[TColIdx]);
       break;
    case atStr:
       StrColMaps[ColIdx].AddV(T.StrColMaps[TColIdx]);
       break;
    }
  }

  TIntV TNext(T.Next);
  for (TInt i = 0; i < TNext.Len(); i++) {
    if (TNext[i] != Last && TNext[i] != Invalid) { TNext[i] += NumRows; }
  }

  Next.AddV(TNext);
  // checks if table is empty 
  if (LastValidRow >= 0) {
    Next[LastValidRow] = NumRows + T.FirstValidRow;
  }
  LastValidRow = NumRows + T.LastValidRow;
  NumRows += T.NumRows;
  NumValidRows += T.NumValidRows;
}

void TTable::Aggregate	(	const TStrV &	GroupByAttrs,
		TAttrAggr	AggOp,
		const TStr &	ValAttr,
		const TStr &	ResAttr,
		TBool	Ordered = true
	)

Aggregates values of ValAttr after grouping with respect to GroupByAttrs. Result are stored as new attribute ResAttr.

Definition at line 1565 of file table.cpp.

                                                          {
 
   for (TInt c = 0; c < GroupByAttrs.Len(); c++) {
    if (!IsColName(GroupByAttrs[c])) { 
      TExcept::Throw("no such column " + GroupByAttrs[c]); 
    }
   }
    
  // double startFn = omp_get_wtime();
  TStrV NGroupByAttrs = NormalizeColNameV(GroupByAttrs);
  TBool UsePhysicalIds = (GetColIdx(IdColName) < 0);
  
  THash<TInt,TIntV> GroupByIntMapping;
  THash<TFlt,TIntV> GroupByFltMapping;
  THash<TInt,TIntV> GroupByStrMapping;
  THash<TGroupKey,TIntV> Mapping;
#ifdef GCC_ATOMIC
  THashMP<TInt,TIntV> GroupByIntMapping_MP(NumValidRows);
  TIntV GroupByIntMPKeys(NumValidRows);
#endif
  TInt NumOfGroups = 0;
  TInt GroupingCase = 0;

  // check if grouping already exists
  GroupStmt Stmt(NGroupByAttrs, Ordered, UsePhysicalIds);
  if (GroupMapping.IsKey(Stmt)) {
    Mapping = GroupMapping.GetDat(Stmt);
  } else{
        if(NGroupByAttrs.Len() == 1){
                switch(GetColType(NGroupByAttrs[0])){
                        case atInt:
#ifdef GCC_ATOMIC
                                if(GetMP()){
                                        GroupByIntColMP(NGroupByAttrs[0], GroupByIntMapping_MP, UsePhysicalIds);
                                        int x = 0;
                                        for(THashMP<TInt,TIntV>::TIter it = GroupByIntMapping_MP.BegI(); it < GroupByIntMapping_MP.EndI(); it++){
                                                GroupByIntMPKeys[x] = it.GetKey();
                                                x++;
                                                /*
                                                printf("%d --> ", it.GetKey().Val);
                                                TIntV& V = it.GetDat();
                                                for(int i = 0; i < V.Len(); i++){
                                                        printf(" %d", V[i].Val); 
                                                }
                                                printf("\n");
                                                */
                                        }
                                        NumOfGroups = x;
                                        GroupingCase = 4;
                                        //printf("Number of groups: %d\n", NumOfGroups.Val);
                                        break;
                                }
#endif // GCC_ATOMIC
                                GroupByIntCol(NGroupByAttrs[0], GroupByIntMapping, TIntV(), true, UsePhysicalIds);
                                NumOfGroups = GroupByIntMapping.Len();
                                GroupingCase = 1;
                                break;
                        case atFlt:
                                GroupByFltCol(NGroupByAttrs[0], GroupByFltMapping, TIntV(), true, UsePhysicalIds);
                                NumOfGroups = GroupByFltMapping.Len();
                                GroupingCase = 2;
                                break;
                        case atStr:
                                GroupByStrCol(NGroupByAttrs[0], GroupByStrMapping, TIntV(), true, UsePhysicalIds);
                                NumOfGroups = GroupByStrMapping.Len();
                                GroupingCase = 3;
                                break;
                }
        }
        else{
                TIntV UniqueVector;
                THash<TGroupKey, TPair<TInt, TIntV> > Mapping_aux;
                GroupAux(NGroupByAttrs, Mapping_aux, Ordered, "", false, UniqueVector, UsePhysicalIds);
                for(THash<TGroupKey, TPair<TInt, TIntV> >::TIter it = Mapping_aux.BegI(); it < Mapping_aux.EndI(); it++){
                        Mapping.AddDat(it.GetKey(), it.GetDat().Val2);
                }
                NumOfGroups = Mapping.Len();
        }
  }
    
  // double endGroup = omp_get_wtime();
  // printf("Group time = %f\n", endGroup-startFn);

  TAttrType T = GetColType(ValAttr);

  // add column corresponding to result attribute type
  if (AggOp == aaCount) { AddIntCol(ResAttr); } 
  else {
    if (T == atInt) { AddIntCol(ResAttr); }
    else if (T == atFlt) { AddFltCol(ResAttr); }
    else {
      // Count is the only aggregation operation handled for Str
      TExcept::Throw("Invalid aggregation for Str type!");
    }
  }
  TInt ColIdx = GetColIdx(ResAttr);
  TInt AggrColIdx = GetColIdx(ValAttr);

  // double endAdd = omp_get_wtime();
  // printf("AddCol time = %f\n", endAdd-endGroup);

#ifdef USE_OPENMP
  #pragma omp parallel for schedule(dynamic)
#endif 
  for (int g = 0; g < NumOfGroups; g++) {
        TIntV* GroupRows = NULL;
        switch(GroupingCase){
                case 0:
                        GroupRows = & Mapping.GetDat(Mapping.GetKey(g));
                        break;
                case 1:
                        GroupRows = & GroupByIntMapping.GetDat(GroupByIntMapping.GetKey(g));
                        break;
                case 2:
                        GroupRows = & GroupByIntMapping.GetDat(GroupByIntMapping.GetKey(g));
                        break;
            case 3:
                        GroupRows = & GroupByStrMapping.GetDat(GroupByStrMapping.GetKey(g));
                        break;
                case 4:
#ifdef GCC_ATOMIC
                        GroupRows = & GroupByIntMapping_MP.GetDat(GroupByIntMPKeys[g]);
#endif
                        break;
        }

    // find valid rows of group
    /*
    TIntV ValidRows;
    for (TInt i = 0; i < GroupRows.Len(); i++) {
      // TODO: This should not be necessary
      if (!RowIdMap.IsKey(GroupRows[i])) { continue; }
      TInt RowId = RowIdMap.GetDat(GroupRows[i]);
      // GroupRows has physical row indices
      if (RowId != Invalid) { ValidRows.Add(RowId); }
    }
    */
        TIntV& ValidRows = *GroupRows;
    TInt sz = ValidRows.Len();
    if (sz <= 0) continue;
    // Count is handled separately (other operations have aggregation policies defined in a template)
    if (AggOp == aaCount) {
      for (TInt i = 0; i < sz; i++) { IntCols[ColIdx][ValidRows[i]] = sz; }
    } else {
      // aggregate based on column type
      if (T == atInt) {
        TIntV V;
        for (TInt i = 0; i < sz; i++) { V.Add(IntCols[AggrColIdx][ValidRows[i]]); }
        TInt Res = AggregateVector<TInt>(V, AggOp);
        if (AggOp == aaMean) { Res = Res / sz; }
        for (TInt i = 0; i < sz; i++) { IntCols[ColIdx][ValidRows[i]] = Res; }
      } else {
        TFltV V;
        for (TInt i = 0; i < sz; i++) { V.Add(FltCols[AggrColIdx][ValidRows[i]]); }
        TFlt Res = AggregateVector<TFlt>(V, AggOp);
        if (AggOp == aaMean) { Res /= sz; }
        for (TInt i = 0; i < sz; i++) { FltCols[ColIdx][ValidRows[i]] = Res; }
      }
    }
  }
  // double endIter = omp_get_wtime();
  // printf("Iter time = %f\n", endIter-endAdd);
}

void TTable::AggregateCols	(	const TStrV &	AggrAttrs,
		TAttrAggr	AggOp,
		const TStr &	ResAttr
	)

Aggregates attributes in AggrAttrs across columns.

Definition at line 1730 of file table.cpp.

                                                                                       {
  TVec<TPair<TAttrType, TInt> >Info;
  for (TInt i = 0; i < AggrAttrs.Len(); i++) {
    Info.Add(GetColTypeMap(AggrAttrs[i]));
    if (Info[i].Val1 != Info[0].Val1) {
      TExcept::Throw("AggregateCols: Aggregation attributes must have the same type");
    }
  }

  if (Info[0].Val1 == atInt) {
    AddIntCol(ResAttr);
    TInt ResIdx = GetColIdx(ResAttr);

    for (TRowIterator RI = BegRI(); RI < EndRI(); RI++) {
      TInt RowIdx = RI.GetRowIdx();
      TIntV V;
      for (TInt i = 0; i < AggrAttrs.Len(); i++) {
        V.Add(IntCols[Info[i].Val2][RowIdx]);
      }
      IntCols[ResIdx][RowIdx] = AggregateVector<TInt>(V, AggOp);
    }
  } else if (Info[0].Val1 == atFlt) {
    AddFltCol(ResAttr);
    TInt ResIdx = GetColIdx(ResAttr);

    for (TRowIterator RI = BegRI(); RI < EndRI(); RI++) {
      TInt RowIdx = RI.GetRowIdx();
      TFltV V;
      for (TInt i = 0; i < AggrAttrs.Len(); i++) {
        V.Add(FltCols[Info[i].Val2][RowIdx]);
      }
      FltCols[ResIdx][RowIdx] = AggregateVector<TFlt>(V, AggOp);
    }
  } else {
    TExcept::Throw("AggregateCols: Only Int and Flt aggregation supported right now");
  }
}

template<class T >

T TTable::AggregateVector ( TVec< T > &  V, TAttrAggr  Policy  )

protected

Aggregates vector into a single scalar value according to a policy.

Aggregate vector into a single scalar value according to a policy. Used for choosing an attribute value for a node when this node appears in several records and has conflicting attribute values

Definition at line 1551 of file table.h.

                                                      {
  switch (Policy) {
    case aaMin: {
      T Res = V[0];
      for (TInt i = 1; i < V.Len(); i++) {
        if (V[i] < Res) { Res = V[i]; }
      }
      return Res;
    }
    case aaMax: {
      T Res = V[0];
      for (TInt i = 1; i < V.Len(); i++) {
        if (V[i] > Res) { Res = V[i]; }
      }
      return Res;
    }
    case aaFirst: {
      return V[0];
    }
    case aaLast:{
      return V[V.Len()-1];
    }
    case aaSum: {
      T Res = V[0];
      for (TInt i = 1; i < V.Len(); i++) {
        Res = Res + V[i];
      }
      return Res;
    }
    case aaMean: {
      T Res = V[0];
      for (TInt i = 1; i < V.Len(); i++) {
        Res = Res + V[i];
      }
      //Res = Res / V.Len(); // TODO: Handle Str case separately?
      return Res;
    }
    case aaMedian: {
      V.Sort();
      return V[V.Len()/2];
    }
    case aaCount: {
      // NOTE: Code should never reach here
      // I had to put this here to avoid a compiler warning.
      // Is there a better way to do this?
      return V[0];
    }
  }
  // Added to remove a compiler warning.
  T ShouldNotComeHere;
  return ShouldNotComeHere;
}

TRowIterator TTable::BegRI ( ) const

inline

Gets iterator to the first valid row of the table.

Definition at line 1232 of file table.h.

{ return TRowIterator(FirstValidRow, this);}

TRowIteratorWithRemove TTable::BegRIWR ( )

inline

Gets iterator with reomve to the first valid row.

Definition at line 1236 of file table.h.

{ return TRowIteratorWithRemove(FirstValidRow, this);}

PNEANet TTable::BuildGraph ( const TIntV &  RowIds, TAttrAggr  AggrPolicy  )

protected

Makes a single pass over the rows in the given row id set, and creates nodes, edges, assigns node and edge attributes.

Definition at line 3425 of file table.cpp.

                                                                    {
  PNEANet Graph = TNEANet::New();
  
  const TAttrType NodeType = GetColType(SrcCol);
  Assert(NodeType == GetColType(DstCol));
  const TInt SrcColIdx = GetColIdx(SrcCol);
  const TInt DstColIdx = GetColIdx(DstCol);
  
  // node values - i.e. the unique values of src/dst col
  //THashSet<TInt> IntNodeVals; // for both int and string node attr types.
  THash<TFlt, TInt> FltNodeVals;

  // node attributes
  THash<TInt, TStrIntVH> NodeIntAttrs;
  THash<TInt, TStrFltVH> NodeFltAttrs;
  THash<TInt, TStrStrVH> NodeStrAttrs;

  // make single pass over all rows in given row id set
  for (TVec<TInt>::TIter it = RowIds.BegI(); it < RowIds.EndI(); it++) {
    TInt CurrRowIdx = *it;

    // add src and dst nodes to graph if they are not seen earlier
    TInt SVal, DVal;
    if (NodeType == atFlt) {
      TFlt FSVal = FltCols[SrcColIdx][CurrRowIdx];
      SVal = CheckAndAddFltNode(Graph, FltNodeVals, FSVal);
      TFlt FDVal = FltCols[SrcColIdx][CurrRowIdx];
      DVal = CheckAndAddFltNode(Graph, FltNodeVals, FDVal);
    } else if (NodeType == atInt || NodeType == atStr) {
      if (NodeType == atInt) {
        SVal = IntCols[SrcColIdx][CurrRowIdx];
        DVal = IntCols[DstColIdx][CurrRowIdx];
      } else {
        SVal = StrColMaps[SrcColIdx][CurrRowIdx];
        if (strlen(Context->StringVals.GetKey(SVal)) == 0) { continue; }  //illegal value
        DVal = StrColMaps[DstColIdx][CurrRowIdx];
        if (strlen(Context->StringVals.GetKey(DVal)) == 0) { continue; }  //illegal value
      }
      if (!Graph->IsNode(SVal)) { Graph->AddNode(SVal); }
      if (!Graph->IsNode(DVal)) { Graph->AddNode(DVal); }
      //CheckAndAddIntNode(Graph, IntNodeVals, SVal);
      //CheckAndAddIntNode(Graph, IntNodeVals, DVal);
    } 

    // add edge and edge attributes 
    Graph->AddEdge(SVal, DVal, CurrRowIdx);
    if (EdgeAttrV.Len() > 0) { AddEdgeAttributes(Graph, CurrRowIdx); }

    // get src and dst node attributes into hashmaps
    if (SrcNodeAttrV.Len() > 0) { 
      AddNodeAttributes(SVal, SrcNodeAttrV, CurrRowIdx, NodeIntAttrs, NodeFltAttrs, NodeStrAttrs);
    }
    if (DstNodeAttrV.Len() > 0) {
      AddNodeAttributes(DVal, DstNodeAttrV, CurrRowIdx, NodeIntAttrs, NodeFltAttrs, NodeStrAttrs);
    }
  }

  // aggregate node attributes and add to graph
  if (SrcNodeAttrV.Len() > 0 || DstNodeAttrV.Len() > 0) {
    for (TNEANet::TNodeI NodeI = Graph->BegNI(); NodeI < Graph->EndNI(); NodeI++) {
      TInt NId = NodeI.GetId();
      if (NodeIntAttrs.IsKey(NId)) {
        TStrIntVH IntAttrVals = NodeIntAttrs.GetDat(NId);
        for (TStrIntVH::TIter it = IntAttrVals.BegI(); it < IntAttrVals.EndI(); it++) {
          TInt AttrVal = AggregateVector<TInt>(it.GetDat(), AggrPolicy);
          Graph->AddIntAttrDatN(NId, AttrVal, it.GetKey());
        }
      }
      if (NodeFltAttrs.IsKey(NId)) {
        TStrFltVH FltAttrVals = NodeFltAttrs.GetDat(NId);
        for (TStrFltVH::TIter it = FltAttrVals.BegI(); it < FltAttrVals.EndI(); it++) {
          TFlt AttrVal = AggregateVector<TFlt>(it.GetDat(), AggrPolicy);
          Graph->AddFltAttrDatN(NId, AttrVal, it.GetKey());
        }
      }
      if (NodeStrAttrs.IsKey(NId)) {
        TStrStrVH StrAttrVals = NodeStrAttrs.GetDat(NId);
        for (TStrStrVH::TIter it = StrAttrVals.BegI(); it < StrAttrVals.EndI(); it++) {
          TStr AttrVal = AggregateVector<TStr>(it.GetDat(), AggrPolicy);
          Graph->AddStrAttrDatN(NId, AttrVal, it.GetKey());
        }
      }
    }
  }

  return Graph;
}

TTableContext * TTable::ChangeContext

(

TTableContext *

Context

)

Changes the current context. Moves all object items to the new context.

Definition at line 901 of file table.cpp.

                                                              {
  TInt L = Sch.Len();

#if 0
  // print table on the input, iterate over all columns
  for (TInt i = 0; i < L; i++) {
    // skip non-string columns
    if (GetSchemaColType(i) != atStr) {
      continue;
    }

    TInt ColIdx = GetColIdx(GetSchemaColName(i));

    // iterate over all rows
    for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
      TInt RowIdx = RowI.GetRowIdx();
      TInt KeyId = StrColMaps[ColIdx][RowIdx];
      printf("ChangeContext in  %d  %d  %d  .%s.\n",
          ColIdx.Val, RowIdx.Val, KeyId.Val, GetStrVal(ColIdx, RowIdx).CStr());
    }
  }
#endif

  // add strings to the new context, change values
  // iterate over all columns
  for (TInt i = 0; i < L; i++) {
    // skip non-string columns
    if (GetSchemaColType(i) != atStr) {
      continue;
    }

    TInt ColIdx = GetColIdx(GetSchemaColName(i));

    // iterate over all rows
    for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
      TInt RowIdx = RowI.GetRowIdx();
      // get the string
      TStr Key = GetStrVal(ColIdx, RowIdx);
      // add the string to the new context
      TInt KeyId = TInt(NewContext->StringVals.AddKey(Key));
      // change the value in the table
      StrColMaps[ColIdx][RowIdx] = KeyId;
    }
  }

  // set the new context
  Context = NewContext;
  return Context;
}

template<class T >

TInt TTable::CheckAndAddFltNode ( T  Graph, THash< TFlt, TInt > &  NodeVals, TFlt  FNodeVal  )

protected

Checks if given NodeVal is seen earlier; if not, add it to Graph and hashmap NodeVals.

Definition at line 1540 of file table.h.

                                                                                   {
  if (!NodeVals.IsKey(FNodeVal)) {
    TInt NodeVal = NodeVals.Len();
    Graph->AddNode(NodeVal);
    NodeVals.AddKey(FNodeVal);
    NodeVals.AddDat(FNodeVal, NodeVal);
    return NodeVal;
  } else { return NodeVals.GetDat(FNodeVal); }
}

void TTable::CheckAndAddIntNode ( PNEANet  Graph, THashSet< TInt > &  NodeVals, TInt  NodeId  )

inlineprotected

Checks if given NodeId is seen earlier; if not, add it to Graph and hashmap NodeVals.

Definition at line 3368 of file table.cpp.

                                                                                           {
  if (!NodeVals.IsKey(NodeId)) {
    Graph->AddNode(NodeId);
    NodeVals.AddKey(NodeId);
  }
}

TInt TTable::CheckSortedKeyVal ( TIntV &  Key, TIntV &  Val, TInt  Start, TInt  End  )

staticprotected

Definition at line 5287 of file table.cpp.

                                                                           {
  TInt j;
  for (j = Start; j < End; j++) {
    if (CompareKeyVal(Key[j], Val[j], Key[j+1], Val[j+1]) > 0) {
      break;
    }
  }
  if (j >= End) { return 0; }
  else { return 1; }
}

void TTable::Classify	(	TPredicate &	Predicate,
		const TStr &	LabelName,
		const TInt &	PositiveLabel = 1,
		const TInt &	NegativeLabel = 0
	)

Definition at line 2785 of file table.cpp.

                                                                                                                        {
  TIntV SelectedRows;
  Select(Predicate, SelectedRows, false);
  ClassifyAux(SelectedRows, LabelName, PositiveLabel, NegativeLabel);
}

void TTable::ClassifyAtomic	(	const TStr &	Col1,
		const TStr &	Col2,
		TPredComp	Cmp,
		const TStr &	LabelName,
		const TInt &	PositiveLabel = 1,
		const TInt &	NegativeLabel = 0
	)

Definition at line 2846 of file table.cpp.

                                                                               {
  TIntV SelectedRows;
  SelectAtomic(Col1, Col2, Cmp, SelectedRows, false);
  ClassifyAux(SelectedRows, LabelName, PositiveLabel, NegativeLabel);
}

template<class T >

void TTable::ClassifyAtomicConst ( const TStr &  Col, const T &  Val, TPredComp  Cmp, const TStr &  LabelName, const TInt &  PositiveLabel = 1, const TInt &  NegativeLabel = 0  )

inline

Definition at line 1292 of file table.h.

                                                                                        {
    TIntV SelectedRows;
    PTable SelectedTable;
    SelectAtomicConst(Col, TPrimitive(Val), Cmp, SelectedRows, SelectedTable, false, false);
    ClassifyAux(SelectedRows, LabelName, PositiveLabel, NegativeLabel);
  }

void TTable::ClassifyAux ( const TIntV &  SelectedRows, const TStr &  LabelName, const TInt &  PositiveLabel = 1, const TInt &  NegativeLabel = 0  )

protected

Adds a label attribute with positive labels on selected rows and negative labels on the rest.

Definition at line 4671 of file table.cpp.

                                                                                                                               {
  AddSchemaCol(LabelName, atInt);
  TInt LabelColIdx = IntCols.Len();
  AddColType(LabelName, atInt, LabelColIdx);
  IntCols.Add(TIntV(NumRows));
  for (TInt i = 0; i < NumRows; i++) {
    IntCols[LabelColIdx][i] = NegativeLabel;
  }
  for (TInt i = 0; i < SelectedRows.Len(); i++) {
    IntCols[LabelColIdx][SelectedRows[i]] = PositiveLabel;
  }
}

void TTable::ColAdd	(	const TStr &	Attr1,
		const TStr &	Attr2,
		const TStr &	ResultAttrName = ""
	)

Performs columnwise addition. See TTable::ColGenericOp.

Definition at line 4793 of file table.cpp.

                                                                                    {
  ColGenericOp(Attr1, Attr2, ResultAttrName, aoAdd);
}

void TTable::ColAdd	(	const TStr &	Attr1,
		TTable &	Table,
		const TStr &	Attr2,
		const TStr &	ResAttr = "",
		TBool	AddToFirstTable = true
	)

Performs columnwise addition with column of given table.

Definition at line 4926 of file table.cpp.

                                                    {
  ColGenericOp(Attr1, Table, Attr2, ResultAttrName, aoAdd, AddToFirstTable);
}

void TTable::ColAdd	(	const TStr &	Attr1,
		const TFlt &	Num,
		const TStr &	ResultAttrName = "",
		const TBool	floatCast = false
	)

Performs addition of column values and given Num.

Definition at line 5040 of file table.cpp.

                                                                                                         {
  ColGenericOp(Attr1, Num, ResultAttrName, aoAdd, floatCast);
}

void TTable::ColConcat	(	const TStr &	Attr1,
		const TStr &	Attr2,
		const TStr &	Sep = "",
		const TStr &	ResAttr = ""
	)

Concatenates two string columns.

Definition at line 5060 of file table.cpp.

                                                                                                 {
  // check if attributes are valid
  if (!IsAttr(Attr1)) TExcept::Throw("No attribute present: " + Attr1);
  if (!IsAttr(Attr2)) TExcept::Throw("No attribute present: " + Attr2);

  TPair<TAttrType, TInt> Info1 = GetColTypeMap(Attr1);
  TPair<TAttrType, TInt> Info2 = GetColTypeMap(Attr2);

  if (Info1.Val1 != atStr || Info2.Val1 != atStr) {
    TExcept::Throw("Only string columns supported in concat.");
  }

  // source column indices
  TInt ColIdx1 = Info1.Val2;
  TInt ColIdx2 = Info2.Val2;

  // destination column index
  TInt ColIdx3 = ColIdx1;

  // Create empty result column with type that of first attribute
  if (ResAttr != "") {
      AddStrCol(ResAttr);
      ColIdx3 = GetColIdx(ResAttr);
  }

  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    TStr CurVal1 = RowI.GetStrAttr(ColIdx1);
    TStr CurVal2 = RowI.GetStrAttr(ColIdx2);
    TStr NewVal = CurVal1 + Sep + CurVal2;
    TInt Key = TInt(Context->StringVals.AddKey(NewVal));
    StrColMaps[ColIdx3][RowI.GetRowIdx()] = Key;
  }
}

void TTable::ColConcat	(	const TStr &	Attr1,
		TTable &	Table,
		const TStr &	Attr2,
		const TStr &	Sep = "",
		const TStr &	ResAttr = "",
		TBool	AddToFirstTable = true
	)

Concatenates string column with column of given table.

Definition at line 5094 of file table.cpp.

                                             {
  // check if attributes are valid
  if (!IsAttr(Attr1)) { TExcept::Throw("No attribute present: " + Attr1); }
  if (!Table.IsAttr(Attr2)) { TExcept::Throw("No attribute present: " + Attr2); }

  if (NumValidRows != Table.NumValidRows) {
    TExcept::Throw("Tables do not have equal number of rows");
  }

  TPair<TAttrType, TInt> Info1 = GetColTypeMap(Attr1);
  TPair<TAttrType, TInt> Info2 = Table.GetColTypeMap(Attr2);

  if (Info1.Val1 != atStr || Info2.Val1 != atStr) {
    TExcept::Throw("Only string columns supported in concat.");
  }

  // source column indices
  TInt ColIdx1 = Info1.Val2;
  TInt ColIdx2 = Info2.Val2;

  // destination column index
  TInt ColIdx3 = ColIdx1;

  if (!AddToFirstTable) {
    ColIdx3 = ColIdx2;
  }

  // Create empty result column in appropriate table with type that of first attribute
  if (ResAttr != "") {
    if (AddToFirstTable) {
      AddStrCol(ResAttr);
      ColIdx3 = GetColIdx(ResAttr);
    }
    else {
      Table.AddStrCol(ResAttr);
      ColIdx3 = Table.GetColIdx(ResAttr);
    }
  }

  TRowIterator RI1, RI2;

  RI1 = BegRI();
  RI2 = Table.BegRI();

  while (RI1 < EndRI() && RI2 < Table.EndRI()) {
    TStr CurVal1 = RI1.GetStrAttr(ColIdx1);
    TStr CurVal2 = RI2.GetStrAttr(ColIdx2);
    TStr NewVal = CurVal1 + Sep + CurVal2;
    TInt Key = TInt(Context->StringVals.AddKey(NewVal));
    if (AddToFirstTable) {
      StrColMaps[ColIdx3][RI1.GetRowIdx()] = Key;
    }
    else {
      Table.StrColMaps[ColIdx3][RI2.GetRowIdx()] = Key;
    }
    RI1++;
    RI2++;
  }

  if (RI1 != EndRI() || RI2 != Table.EndRI()) {
    TExcept::Throw("ColGenericOp: Iteration error");
  }
}

void TTable::ColConcatConst	(	const TStr &	Attr1,
		const TStr &	Val,
		const TStr &	Sep = "",
		const TStr &	ResAttr = ""
	)

Concatenates column values with given string value.

Definition at line 5159 of file table.cpp.

                                                                                                    {
  // check if attribute is valid
  if (!IsAttr(Attr1)) { TExcept::Throw("No attribute present: " + Attr1); }

  TPair<TAttrType, TInt> Info1 = GetColTypeMap(Attr1);

  if (Info1.Val1 != atStr) {
    TExcept::Throw("Only string columns supported in concat.");
  }

  // source column index
  TInt ColIdx1 = Info1.Val2;

  // destination column index
  TInt ColIdx2 = ColIdx1;

  // Create empty result column with type that of first attribute
  if (ResAttr != "") {
    AddStrCol(ResAttr);
    ColIdx2 = GetColIdx(ResAttr);
  }

  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    TStr CurVal = RowI.GetStrAttr(ColIdx1);
    TStr NewVal = CurVal + Sep + Val;
    TInt Key = TInt(Context->StringVals.AddKey(NewVal));
    StrColMaps[ColIdx2][RowI.GetRowIdx()] = Key;
  }  
}

void TTable::ColDiv	(	const TStr &	Attr1,
		const TStr &	Attr2,
		const TStr &	ResultAttrName = ""
	)

Performs columnwise division. See TTable::ColGenericOp.

Definition at line 4805 of file table.cpp.

                                                                                    {
  ColGenericOp(Attr1, Attr2, ResultAttrName, aoDiv);
}

void TTable::ColDiv	(	const TStr &	Attr1,
		TTable &	Table,
		const TStr &	Attr2,
		const TStr &	ResAttr = "",
		TBool	AddToFirstTable = true
	)

Performs columnwise division with column of given table.

Definition at line 4941 of file table.cpp.

                                                    {
  ColGenericOp(Attr1, Table, Attr2, ResultAttrName, aoDiv, AddToFirstTable);
}

void TTable::ColDiv	(	const TStr &	Attr1,
		const TFlt &	Num,
		const TStr &	ResultAttrName = "",
		const TBool	floatCast = false
	)

Performs division of column values and given Num.

Definition at line 5052 of file table.cpp.

                                                                                                         {
  ColGenericOp(Attr1, Num, ResultAttrName, aoDiv, floatCast);
}

void TTable::ColGenericOp	(	const TStr &	Attr1,
		const TStr &	Attr2,
		const TStr &	ResAttr,
		TArithOp	op
	)

Performs columnwise arithmetic operation.

Performs Attr1 OP Attr2 and stores it in Attr1 If ResAttr != "", result is stored in a new column ResAttr

Definition at line 4729 of file table.cpp.

                                                                                                {
  // check if attributes are valid
  if (!IsAttr(Attr1)) TExcept::Throw("No attribute present: " + Attr1);
  if (!IsAttr(Attr2)) TExcept::Throw("No attribute present: " + Attr2);
  TPair<TAttrType, TInt> Info1 = GetColTypeMap(Attr1);
  TPair<TAttrType, TInt> Info2 = GetColTypeMap(Attr2);
  TAttrType Arg1Type = Info1.Val1;
  TAttrType Arg2Type = Info2.Val1;
  if (Arg1Type == atStr || Arg2Type == atStr) {
    TExcept::Throw("Only numeric columns supported in arithmetic operations.");
  }
  if(Arg1Type == atInt && Arg2Type == atFlt && ResAttr == ""){
        TExcept::Throw("Trying to write float values to an existing int-typed column");
  }
  // source column indices
  TInt ColIdx1 = Info1.Val2;
  TInt ColIdx2 = Info2.Val2;
  
  // destination column index
  TInt ColIdx3 = ColIdx1;
  // Create empty result column with type that of first attribute
  if (ResAttr != "") {
      if (Arg1Type == atInt && Arg2Type == atInt) {
          AddIntCol(ResAttr);
      }
      else {
          AddFltCol(ResAttr);
      }
      ColIdx3 = GetColIdx(ResAttr);
  }
#ifdef USE_OPENMP
  if(GetMP()){
        ColGenericOpMP(ColIdx1, ColIdx2, Arg1Type, Arg2Type, ColIdx3, op);
        return;
  }
#endif  //USE_OPENMP
  TAttrType ResType = atFlt;
  if(Arg1Type == atInt && Arg2Type == atInt){ printf("hooray!\n"); ResType = atInt;}
  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
        //printf("%d %d %d %d\n", ColIdx1.Val, ColIdx2.Val, ColIdx3.Val, RowI.GetRowIdx().Val);
                if(ResType == atInt){
                        TInt V1 = RowI.GetIntAttr(ColIdx1);
                        TInt V2 = RowI.GetIntAttr(ColIdx2);
                        if (op == aoAdd) { IntCols[ColIdx3][RowI.GetRowIdx()] = V1 + V2; }
                if (op == aoSub) { IntCols[ColIdx3][RowI.GetRowIdx()] = V1 - V2; }
                if (op == aoMul) { IntCols[ColIdx3][RowI.GetRowIdx()] = V1 * V2; }
                if (op == aoDiv) { IntCols[ColIdx3][RowI.GetRowIdx()] = V1 / V2; }
                if (op == aoMod) { IntCols[ColIdx3][RowI.GetRowIdx()] = V1 % V2; }
                if (op == aoMin) { IntCols[ColIdx3][RowI.GetRowIdx()] = (V1 < V2) ? V1 : V2;}
                if (op == aoMax) { IntCols[ColIdx3][RowI.GetRowIdx()] = (V1 > V2) ? V1 : V2;}
                } else{
                        TFlt V1 = (Arg1Type == atInt) ? (TFlt)RowI.GetIntAttr(ColIdx1) : RowI.GetFltAttr(ColIdx1);
                        TFlt V2 = (Arg2Type == atInt) ? (TFlt)RowI.GetIntAttr(ColIdx2) : RowI.GetFltAttr(ColIdx2);
                        if (op == aoAdd) { FltCols[ColIdx3][RowI.GetRowIdx()] = V1 + V2; }
                if (op == aoSub) { FltCols[ColIdx3][RowI.GetRowIdx()] = V1 - V2; }
                if (op == aoMul) { FltCols[ColIdx3][RowI.GetRowIdx()] = V1 * V2; }
                if (op == aoDiv) { FltCols[ColIdx3][RowI.GetRowIdx()] = V1 / V2; }
                if (op == aoMod) { TExcept::Throw("Cannot find modulo for float columns");  }
                if (op == aoMin) { FltCols[ColIdx3][RowI.GetRowIdx()] = (V1 < V2) ? V1 : V2;}
                if (op == aoMax) { FltCols[ColIdx3][RowI.GetRowIdx()] = (V1 > V2) ? V1 : V2;}
                }
  }
}

void TTable::ColGenericOp	(	const TStr &	Attr1,
		TTable &	Table,
		const TStr &	Attr2,
		const TStr &	ResAttr,
		TArithOp	op,
		TBool	AddToFirstTable
	)

Performs columnwise arithmetic operation with column of given table.

Definition at line 4821 of file table.cpp.

                                     {
  // check if attributes are valid
  if (!IsAttr(Attr1)) { TExcept::Throw("No attribute present: " + Attr1); }
  if (!Table.IsAttr(Attr2)) { TExcept::Throw("No attribute present: " + Attr2); }

  if (NumValidRows != Table.NumValidRows) {
    TExcept::Throw("Tables do not have equal number of rows");
  }

  TPair<TAttrType, TInt> Info1 = GetColTypeMap(Attr1);
  TPair<TAttrType, TInt> Info2 = Table.GetColTypeMap(Attr2);
  TAttrType Arg1Type = Info1.Val1;
  TAttrType Arg2Type = Info2.Val1;
  if (Info1.Val1 == atStr || Info2.Val1 == atStr) {
    TExcept::Throw("Only numeric columns supported in arithmetic operations.");
  }
  if(Arg1Type == atInt && Arg2Type == atFlt && ResAttr == ""){
        TExcept::Throw("Trying to write float values to an existing int-typed column");
  }
  // source column indices
  TInt ColIdx1 = Info1.Val2;
  TInt ColIdx2 = Info2.Val2;

  // destination column index
  TInt ColIdx3 = AddToFirstTable ? ColIdx1 : ColIdx2;

  // Create empty result column in appropriate table with type that of first attribute
  if (ResAttr != "") {
    if (AddToFirstTable) {
      if (Arg1Type == atInt && Arg2Type == atInt) {
          AddIntCol(ResAttr);
      } else {
          AddFltCol(ResAttr);
      }
      ColIdx3 = GetColIdx(ResAttr);
    }
    else {
      if (Arg1Type == atInt && Arg2Type == atInt) {
          Table.AddIntCol(ResAttr);
      } else {
          Table.AddFltCol(ResAttr);
      }
      ColIdx3 = Table.GetColIdx(ResAttr);
    }
  }
  
  /*
  #ifdef USE_OPENMP
  if(GetMP()){
        ColGenericOpMP(Table, AddToFirstTable, ColIdx1, ColIdx2, Arg1Type, Arg2Type, ColIdx3, op);
        return;
  }
  #endif        //USE_OPENMP
  */
        
  TRowIterator RI1, RI2;
  RI1 = BegRI();
  RI2 = Table.BegRI();
  TAttrType ResType = atFlt;
  if(Arg1Type == atInt && Arg2Type == atInt){ ResType = atInt;}
  while (RI1 < EndRI() && RI2 < Table.EndRI()) {
    if (ResType == atInt) {
                TInt V1 = RI1.GetIntAttr(ColIdx1);
                TInt V2 = RI2.GetIntAttr(ColIdx2);
        if (AddToFirstTable) {
                if (op == aoAdd) { IntCols[ColIdx3][RI1.GetRowIdx()] = V1 + V2; }
                if (op == aoSub) { IntCols[ColIdx3][RI1.GetRowIdx()] = V1 - V2; }
                if (op == aoMul) { IntCols[ColIdx3][RI1.GetRowIdx()] = V1 * V2; }
                if (op == aoDiv) { IntCols[ColIdx3][RI1.GetRowIdx()] = V1 / V2; }
                if (op == aoMod) { IntCols[ColIdx3][RI1.GetRowIdx()] = V1 % V2; }
        }
        else {
                if (op == aoAdd) { Table.IntCols[ColIdx3][RI2.GetRowIdx()] = V1 + V2; }
                if (op == aoSub) { Table.IntCols[ColIdx3][RI2.GetRowIdx()] = V1 - V2; }
                if (op == aoMul) { Table.IntCols[ColIdx3][RI2.GetRowIdx()] = V1 * V2; }
                if (op == aoDiv) { Table.IntCols[ColIdx3][RI2.GetRowIdx()] = V1 / V2; }
                if (op == aoMod) { Table.IntCols[ColIdx3][RI2.GetRowIdx()] = V1 % V2; }
        }
    } else {
                TFlt V1 = (Arg1Type == atInt) ? (TFlt)RI1.GetIntAttr(ColIdx1) : RI2.GetFltAttr(ColIdx1);
                TFlt V2 = (Arg2Type == atInt) ? (TFlt)RI1.GetIntAttr(ColIdx2) : RI2.GetFltAttr(ColIdx2);
        if (AddToFirstTable) {
                if (op == aoAdd) { FltCols[ColIdx3][RI1.GetRowIdx()] = V1 + V2; }
                if (op == aoSub) { FltCols[ColIdx3][RI1.GetRowIdx()] = V1 - V2; }
                if (op == aoMul) { FltCols[ColIdx3][RI1.GetRowIdx()] = V1 * V2; }
                if (op == aoDiv) { FltCols[ColIdx3][RI1.GetRowIdx()] = V1 / V2; }
                if (op == aoMod) { TExcept::Throw("Cannot find modulo for float columns"); }
        } else {
                if (op == aoAdd) { Table.FltCols[ColIdx3][RI2.GetRowIdx()] = V1 + V2; }
                if (op == aoSub) { Table.FltCols[ColIdx3][RI2.GetRowIdx()] = V1 - V2; }
                if (op == aoMul) { Table.FltCols[ColIdx3][RI2.GetRowIdx()] = V1 * V2; }
                if (op == aoDiv) { Table.FltCols[ColIdx3][RI2.GetRowIdx()] = V1 / V2; }
                if (op == aoMod) { TExcept::Throw("Cannot find modulo for float columns"); }
        }
    }
    RI1++;
    RI2++;
  }

  if (RI1 != EndRI() || RI2 != Table.EndRI()) {
    TExcept::Throw("ColGenericOp: Iteration error");
  }
}

void TTable::ColGenericOp	(	const TStr &	Attr1,
		const TFlt &	Num,
		const TStr &	ResAttr,
		TArithOp	op,
		const TBool	floatCast
	)

Performs arithmetic op of column values and given Num.

Definition at line 4952 of file table.cpp.

                                                                                                                     {
  // check if attribute is valid
  if (!IsAttr(Attr1)) { TExcept::Throw("No attribute present: " + Attr1); }

  TPair<TAttrType, TInt> Info1 = GetColTypeMap(Attr1);
  TAttrType ArgType = Info1.Val1;
  if (ArgType == atStr) {
    TExcept::Throw("Only numeric columns supported in arithmetic operations.");
  }
  // source column index
  TInt ColIdx1 = Info1.Val2;
  // destination column index
  TInt ColIdx2 = ColIdx1;

  // Create empty result column with type that of first attribute
  TBool shouldCast = floatCast;
  if (ResAttr != "") {
      if ((ArgType == atInt) & !shouldCast) {
          AddIntCol(ResAttr);
      } else {
          AddFltCol(ResAttr);
      }
      ColIdx2 = GetColIdx(ResAttr);
  } else {
    // Cannot change type of existing attribute
    shouldCast = false;
  }
  
  #ifdef USE_OPENMP
  if(GetMP()){
        ColGenericOpMP(ColIdx1, ColIdx2, ArgType, Num, op, shouldCast);
        return;
  }
  #endif  //USE_OPENMP
        
  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    if ((ArgType == atInt) && !shouldCast) {
      TInt CurVal = RowI.GetIntAttr(ColIdx1);
      TInt Val = static_cast<int>(Num);
      if (op == aoAdd) { IntCols[ColIdx2][RowI.GetRowIdx()] = CurVal + Val; }
      if (op == aoSub) { IntCols[ColIdx2][RowI.GetRowIdx()] = CurVal - Val; }
      if (op == aoMul) { IntCols[ColIdx2][RowI.GetRowIdx()] = CurVal * Val; }
      if (op == aoDiv) { IntCols[ColIdx2][RowI.GetRowIdx()] = CurVal / Val; }
      if (op == aoMod) { IntCols[ColIdx2][RowI.GetRowIdx()] = CurVal % Val; }
    }
    else {
      TFlt CurVal = (ArgType == atFlt) ? RowI.GetFltAttr(ColIdx1) : (TFlt) RowI.GetIntAttr(ColIdx1);
      if (op == aoAdd) { FltCols[ColIdx2][RowI.GetRowIdx()] = CurVal + Num; }
      if (op == aoSub) { FltCols[ColIdx2][RowI.GetRowIdx()] = CurVal - Num; }
      if (op == aoMul) { FltCols[ColIdx2][RowI.GetRowIdx()] = CurVal * Num; }
      if (op == aoDiv) { FltCols[ColIdx2][RowI.GetRowIdx()] = CurVal / Num; }
      if (op == aoMod) { TExcept::Throw("Cannot find modulo for float columns"); }
    }
  }
}

void TTable::ColGenericOpMP	(	TInt	ArgColIdx1,
		TInt	ArgColIdx2,
		TAttrType	ArgType1,
		TAttrType	ArgType2,
		TInt	ResColIdx,
		TArithOp	op
	)

Definition at line 4685 of file table.cpp.

                                                                                                                                {
        TAttrType ResType = atFlt;
        if(ArgType1 == atInt && ArgType2 == atInt){ ResType = atInt;}
        TIntPrV Partitions;
        GetPartitionRanges(Partitions, omp_get_max_threads()*CHUNKS_PER_THREAD);
        TInt PartitionSize = Partitions[0].GetVal2()-Partitions[0].GetVal1()+1;
        #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD)
        for (int i = 0; i < Partitions.Len(); i++){
                TRowIterator RowI(Partitions[i].GetVal1(), this);
                TRowIterator EndI(Partitions[i].GetVal2(), this);
                while(RowI < EndI){
                        if(ResType == atInt){
                                TInt V1 = RowI.GetIntAttr(ArgColIdx1);
                                TInt V2 = RowI.GetIntAttr(ArgColIdx2);
                                if (op == aoAdd) { IntCols[ResColIdx][RowI.GetRowIdx()] = V1 + V2; }
                        if (op == aoSub) { IntCols[ResColIdx][RowI.GetRowIdx()] = V1 - V2; }
                        if (op == aoMul) { IntCols[ResColIdx][RowI.GetRowIdx()] = V1 * V2; }
                        if (op == aoDiv) { IntCols[ResColIdx][RowI.GetRowIdx()] = V1 / V2; }
                        if (op == aoMod) { IntCols[ResColIdx][RowI.GetRowIdx()] = V1 % V2; }
                        if (op == aoMin) { IntCols[ResColIdx][RowI.GetRowIdx()] = (V1 < V2) ? V1 : V2;}
                        if (op == aoMax) { IntCols[ResColIdx][RowI.GetRowIdx()] = (V1 > V2) ? V1 : V2;}
                        } else{
                            TFlt V1 = (ArgType1 == atInt) ? (TFlt)RowI.GetIntAttr(ArgColIdx1) : RowI.GetFltAttr(ArgColIdx1);
                            TFlt V2 = (ArgType2 == atInt) ? (TFlt)RowI.GetIntAttr(ArgColIdx2) : RowI.GetFltAttr(ArgColIdx2);
                                if (op == aoAdd) { FltCols[ResColIdx][RowI.GetRowIdx()] = V1 + V2; }
                        if (op == aoSub) { FltCols[ResColIdx][RowI.GetRowIdx()] = V1 - V2; }
                        if (op == aoMul) { FltCols[ResColIdx][RowI.GetRowIdx()] = V1 * V2; }
                        if (op == aoDiv) { FltCols[ResColIdx][RowI.GetRowIdx()] = V1 / V2; }
                        if (op == aoMod) { TExcept::Throw("Cannot find modulo for float columns");  }
                        if (op == aoMin) { FltCols[ResColIdx][RowI.GetRowIdx()] = (V1 < V2) ? V1 : V2;}
                        if (op == aoMax) { FltCols[ResColIdx][RowI.GetRowIdx()] = (V1 > V2) ? V1 : V2;}
                        }
                        RowI++;
                }
        }
}

void TTable::ColGenericOpMP	(	const TInt &	ColIdx1,
		const TInt &	ColIdx2,
		TAttrType	ArgType,
		const TFlt &	Num,
		TArithOp	op,
		TBool	ShouldCast
	)

Definition at line 5009 of file table.cpp.

                                                                                                                                      {
        TIntPrV Partitions;
        GetPartitionRanges(Partitions, omp_get_max_threads()*CHUNKS_PER_THREAD);
        TInt PartitionSize = Partitions[0].GetVal2()-Partitions[0].GetVal1()+1;
        #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD)
        for (int i = 0; i < Partitions.Len(); i++){
                TRowIterator RowI(Partitions[i].GetVal1(), this);
                TRowIterator EndI(Partitions[i].GetVal2(), this);
                while(RowI < EndI){
                        if ((ArgType == atInt) && !ShouldCast) {
                        TInt CurVal = RowI.GetIntAttr(ColIdx1);
                        TInt Val = static_cast<int>(Num);
                        if (op == aoAdd) { IntCols[ColIdx2][RowI.GetRowIdx()] = CurVal + Val; }
                        if (op == aoSub) { IntCols[ColIdx2][RowI.GetRowIdx()] = CurVal - Val; }
                        if (op == aoMul) { IntCols[ColIdx2][RowI.GetRowIdx()] = CurVal * Val; }
                        if (op == aoDiv) { IntCols[ColIdx2][RowI.GetRowIdx()] = CurVal / Val; }
                        if (op == aoMod) { IntCols[ColIdx2][RowI.GetRowIdx()] = CurVal % Val; }
                } else {
                        TFlt CurVal = (ArgType == atFlt) ? RowI.GetFltAttr(ColIdx1) : (TFlt) RowI.GetIntAttr(ColIdx1);
                        if (op == aoAdd) { FltCols[ColIdx2][RowI.GetRowIdx()] = CurVal + Num; }
                        if (op == aoSub) { FltCols[ColIdx2][RowI.GetRowIdx()] = CurVal - Num; }
                        if (op == aoMul) { FltCols[ColIdx2][RowI.GetRowIdx()] = CurVal * Num; }
                        if (op == aoDiv) { FltCols[ColIdx2][RowI.GetRowIdx()] = CurVal / Num; }
                        if (op == aoMod) { TExcept::Throw("Cannot find modulo for float columns"); }
                }
                RowI++;
                }
        }
}

void TTable::ColMax	(	const TStr &	Attr1,
		const TStr &	Attr2,
		const TStr &	ResultAttrName = ""
	)

Performs max of two columns. See TTable::ColGenericOp.

Definition at line 4817 of file table.cpp.

                                                                                    {
  ColGenericOp(Attr1, Attr2, ResultAttrName, aoMax);
}

void TTable::ColMin	(	const TStr &	Attr1,
		const TStr &	Attr2,
		const TStr &	ResultAttrName = ""
	)

Performs min of two columns. See TTable::ColGenericOp.

Definition at line 4813 of file table.cpp.

                                                                                    {
  ColGenericOp(Attr1, Attr2, ResultAttrName, aoMin);
}

void TTable::ColMod	(	const TStr &	Attr1,
		const TStr &	Attr2,
		const TStr &	ResultAttrName = ""
	)

Performs columnwise modulus. See TTable::ColGenericOp.

Definition at line 4809 of file table.cpp.

                                                                                    {
  ColGenericOp(Attr1, Attr2, ResultAttrName, aoMod);
}

void TTable::ColMod	(	const TStr &	Attr1,
		TTable &	Table,
		const TStr &	Attr2,
		const TStr &	ResAttr = "",
		TBool	AddToFirstTable = true
	)

Performs columnwise modulus with column of given table.

Definition at line 4946 of file table.cpp.

                                                    {
  ColGenericOp(Attr1, Table, Attr2, ResultAttrName, aoMod, AddToFirstTable);
}

void TTable::ColMod	(	const TStr &	Attr1,
		const TFlt &	Num,
		const TStr &	ResultAttrName = "",
		const TBool	floatCast = false
	)

Performs modulus of column values and given Num.

Definition at line 5056 of file table.cpp.

                                                                                                         {
  ColGenericOp(Attr1, Num, ResultAttrName, aoMod, floatCast);
}

void TTable::ColMul	(	const TStr &	Attr1,
		const TStr &	Attr2,
		const TStr &	ResultAttrName = ""
	)

Performs columnwise multiplication. See TTable::ColGenericOp.

Definition at line 4801 of file table.cpp.

                                                                                    {
  ColGenericOp(Attr1, Attr2, ResultAttrName, aoMul);
}

void TTable::ColMul	(	const TStr &	Attr1,
		TTable &	Table,
		const TStr &	Attr2,
		const TStr &	ResAttr = "",
		TBool	AddToFirstTable = true
	)

Performs columnwise multiplication with column of given table.

Definition at line 4936 of file table.cpp.

                                                    {
  ColGenericOp(Attr1, Table, Attr2, ResultAttrName, aoMul, AddToFirstTable);
}

void TTable::ColMul	(	const TStr &	Attr1,
		const TFlt &	Num,
		const TStr &	ResultAttrName = "",
		const TBool	floatCast = false
	)

Performs multiplication of column values and given Num.

Definition at line 5048 of file table.cpp.

                                                                                                         {
  ColGenericOp(Attr1, Num, ResultAttrName, aoMul, floatCast);
}

void TTable::ColSub	(	const TStr &	Attr1,
		const TStr &	Attr2,
		const TStr &	ResultAttrName = ""
	)

Performs columnwise subtraction. See TTable::ColGenericOp.

Definition at line 4797 of file table.cpp.

                                                                                    {
  ColGenericOp(Attr1, Attr2, ResultAttrName, aoSub);
}

void TTable::ColSub	(	const TStr &	Attr1,
		TTable &	Table,
		const TStr &	Attr2,
		const TStr &	ResAttr = "",
		TBool	AddToFirstTable = true
	)

Performs columnwise subtraction with column of given table.

Definition at line 4931 of file table.cpp.

                                                    {
  ColGenericOp(Attr1, Table, Attr2, ResultAttrName, aoSub, AddToFirstTable);
}

void TTable::ColSub	(	const TStr &	Attr1,
		const TFlt &	Num,
		const TStr &	ResultAttrName = "",
		const TBool	floatCast = false
	)

Performs subtraction of column values and given Num.

Definition at line 5044 of file table.cpp.

                                                                                                         {
  ColGenericOp(Attr1, Num, ResultAttrName, aoSub, floatCast);
}

TInt TTable::CompareKeyVal ( const TInt &  K1, const TInt &  V1, const TInt &  K2, const TInt &  V2  )

staticprotected

Definition at line 5274 of file table.cpp.

                                                                                         {
  // if (K1 == K2) { 
  //   if (V1 < V2) { return -1; }
  //   else if (V1 > V2) { return 1; }
  //   else return 0;
  // }
  // if (K1 < K2) { return -1; }
  // else { return 1; }

  if (K1 == K2) { return V1 - V2; }
  else { return K1 - K2; }
}

TInt TTable::CompareRows ( TInt  R1, TInt  R2, const TAttrType &  CompareByType, const TInt &  CompareByIndex, TBool  Asc = true  )

inlineprotected

Returns positive value if R1 is bigger, negative value if R2 is bigger, and 0 if they are equal (strcmp semantics).

Definition at line 3044 of file table.cpp.

                                                                                                                       {
  //printf("comparing rows %d %d by %s\n", R1.Val, R2.Val, CompareBy.CStr());
  switch (CompareByType) {
    case atInt:{
      if (IntCols[CompareByIndex][R1] > IntCols[CompareByIndex][R2]) { return (Asc ? 1 : -1); }
      if (IntCols[CompareByIndex][R1] < IntCols[CompareByIndex][R2]) { return (Asc ? -1 : 1); }
      return 0;
    }
    case atFlt:{
      if (FltCols[CompareByIndex][R1] > FltCols[CompareByIndex][R2]) { return (Asc ? 1 : -1); }
      if (FltCols[CompareByIndex][R1] < FltCols[CompareByIndex][R2]) { return (Asc ? -1 : 1); }
      return 0;
    }
    case atStr:{
      TStr S1 = GetStrVal(CompareByIndex, R1);
      TStr S2 = GetStrVal(CompareByIndex, R2);
      int CmpRes = strcmp(S1.CStr(), S2.CStr());
      return (Asc ? CmpRes : -CmpRes);
    }
  }
  // code should not come here, added to remove a compiler warning
  return 0;
}

TInt TTable::CompareRows ( TInt  R1, TInt  R2, const TVec< TAttrType > &  CompareByTypes, const TIntV &  CompareByIndices, TBool  Asc = true  )

inlineprotected

Returns positive value if R1 is bigger, negative value if R2 is bigger, and 0 if they are equal (strcmp semantics).

Definition at line 3068 of file table.cpp.

                                                                                                                                 {
  for (TInt i = 0; i < CompareByTypes.Len(); i++) {
    TInt res = CompareRows(R1, R2, CompareByTypes[i], CompareByIndices[i], Asc);
    if (res != 0) { return res; }
  }
  return 0;
}

void TTable::ConcatTable ( const PTable &  T )

inlineprotected

Appends all rows of T to this table, and recalculate indices.

Definition at line 693 of file table.h.

{AddTable(*T); Reindex(); }

void TTable::Count	(	const TStr &	CountColName,
		const TStr &	Col
	)

Counts number of unique elements.

Count the number of appearences of the different elements of column . Record results in column CountCol

Definition at line 1782 of file table.cpp.

                                                            {
  TStrV GroupByAttrs;
  GroupByAttrs.Add(CountColName);
  Aggregate(GroupByAttrs, aaCount, "", Col);
}

void TTable::Defrag

(

)

Releases memory of deleted rows, and defrags.

Also updates meta-data as row indices have changed Need some liveness analysis of columns

Definition at line 3291 of file table.cpp.

                    {
  TInt FreeIndex = 0;
  TIntV Mapping;  // Mapping[old_index] = new_index/invalid

  TInt IdColIdx = GetColIdx(IdColName);

  for (TInt i = 0; i < Next.Len(); i++) {
    if (Next[i] != TTable::Invalid) {  
      // "first row" properly set beforehand
      if (FreeIndex == 0) {
        Assert (i == FirstValidRow);
        FirstValidRow = 0;
      }
 
      if (Next[i] != Last) { 
        Next[FreeIndex] = FreeIndex + 1;
        Mapping.Add(FreeIndex);
      } else {
        Next[FreeIndex] = Last;
        LastValidRow = FreeIndex;
        Mapping.Add(Last);
      }

      RowIdMap.AddDat(IntCols[IdColIdx][i], FreeIndex);

      for (TInt j = 0; j < IntCols.Len(); j++) {
        IntCols[j][FreeIndex] = IntCols[j][i];
      }
      for (TInt j = 0; j < FltCols.Len(); j++) {
        FltCols[j][FreeIndex] = FltCols[j][i];
      }
      for (TInt j = 0; j < StrColMaps.Len(); j++) {
        StrColMaps[j][FreeIndex] = StrColMaps[j][i];
      }

      FreeIndex++;
    } else {
      NumRows--;
      Mapping.Add(TTable::Invalid);
    }
  }

  // should match, or bug somewhere
  Assert (NumValidRows == NumRows);
}

void TTable::DelColType ( const TStr &  ColName )

inlineprotected

Adds column with name ColName and type ColType to the ColTypeMap.

Definition at line 671 of file table.h.

                                       {
    TStr NColName = NormalizeColName(ColName);
    ColTypeMap.DelKey(NColName);
  }

TStr TTable::DenormalizeColName ( const TStr &  ColName ) const

protected

Removes suffix to column name if exists.

Definition at line 4625 of file table.cpp.

                                                         {
  TStr DColName = ColName;
  if (DColName.Len() == 0) { return DColName; }
  if (DColName.GetCh(0) == '_') { return DColName; }
  if (DColName.GetCh(DColName.Len()-2) == '-') { 
    DColName = DColName.GetSubStr(0,DColName.Len()-3); 
  }
  TInt Conflicts = 0;
  for (TInt i = 0; i < Sch.Len(); i++) {
    if (DColName == Sch[i].Val1.GetSubStr(0, Sch[i].Val1.Len()-3)) {
      Conflicts++;
    }
  }
  if (Conflicts > 1) { return ColName; } 
  else { return DColName; }
}

Schema TTable::DenormalizeSchema ( ) const

protected

Removes suffix to column names in the Schema.

Definition at line 4642 of file table.cpp.

                                       {
  Schema DSch;
  for (TInt i = 0; i < Sch.Len(); i++) {
    DSch.Add(TPair<TStr, TAttrType>(DenormalizeColName(Sch[i].Val1), Sch[i].Val2));
  }
  return DSch;
}

void TTable::Dump

(

FILE *

OutF = stdout

)

const

Prints table contents to a text file.

Definition at line 867 of file table.cpp.

                                  {
  TInt L = Sch.Len();
  Schema DSch = DenormalizeSchema();

  // LoadSS() will not throw away lines with #
  //fprintf(OutF, "# Table: rows: %d, columns: %d\n", GetNumValidRows(), GetNodes());
  // print title (schema), LoadSS() will take first line as (optional) schema
  fprintf(OutF, "# ");
  for (TInt i = 0; i < L-1; i++) {
    fprintf(OutF, "%s\t", DSch[i].Val1.CStr());
  }  
  fprintf(OutF, "%s\n", DSch[L-1].Val1.CStr());
  // print table contents
  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    for (TInt i = 0; i < L; i++) {
      char C = (i == L-1) ? '\n' : '\t';
      switch (GetSchemaColType(i)) {
        case atInt: {
          fprintf(OutF, "%d%c", RowI.GetIntAttr(GetSchemaColName(i)).Val, C);
          break;
        }
        case atFlt: {
          fprintf(OutF, "%f%c", RowI.GetFltAttr(GetSchemaColName(i)).Val, C);
          break;
        }
        case atStr: {
          fprintf(OutF, "%s%c", RowI.GetStrAttr(GetSchemaColName(i)).CStr(), C);
          break;
        }
      }
    }
  }
}

TRowIterator TTable::EndRI ( ) const

inline

Gets iterator to the last valid row of the table.

Definition at line 1234 of file table.h.

{ return TRowIterator(TTable::Last, this);}

TRowIteratorWithRemove TTable::EndRIWR ( )

inline

Gets iterator with reomve to the last valid row.

Definition at line 1238 of file table.h.

{ return TRowIteratorWithRemove(TTable::Last, this);}

void TTable::FillBucketsByInterval ( TStr  SplitAttr, TIntPrV  SplitIntervals  )

protected

Fills RowIdBuckets with sets of row ids.

Fill RowIdBuckets with sets of row ids, partitioned on the value of the column SplitAttr, according to the intervals specified by SplitIntervals. Called by ToVarGraphSequence and ToVarGraphSequenceIterator.

Definition at line 3577 of file table.cpp.

                                                                         {
  TInt SplitColId = GetColIdx(SplitAttr);
  int NumBuckets = SplitIntervals.Len();
  InitRowIdBuckets(NumBuckets);

  // populate RowIdSets by computing the range of buckets for each row
  for (TInt i = 0; i < Next.Len(); i++) {
    if (Next[i] == Invalid) { continue; }
    int SplitVal = IntCols[SplitColId][i];
    for (TInt j = 0; j < SplitIntervals.Len(); j++) { 
      if (SplitVal >= SplitIntervals[j].Val1 && SplitVal < SplitIntervals[j].Val2) {
        RowIdBuckets[j].Add(i);
      }
    }
  }
}

void TTable::FillBucketsByWindow ( TStr  SplitAttr, TInt  JumpSize, TInt  WindowSize, TInt  StartVal, TInt  EndVal  )

protected

Fills RowIdBuckets with sets of row ids.

Fill RowIdBuckets with sets of row ids partitioned on the value of the column SplitAttr, according to the windows specified by JumpSize and WindowSize. Called by ToGraphSequence and ToGraphSequenceIterator.

Definition at line 3527 of file table.cpp.

                                                                                                           {
  Assert (JumpSize <= WindowSize);
  int NumBuckets, MinBucket, MaxBucket;
  TInt SplitColId = GetColIdx(SplitAttr);

  if (StartVal == TInt::Mn || EndVal == TInt::Mx) {
    // calculate min and max value of the column 'SplitAttr'
    TInt MinValue = TInt::Mx;
    TInt MaxValue = TInt::Mn;
    for (TInt i = 0; i < Next.Len(); i++) {
      if (Next[i] != Invalid) { 
        if (MinValue > IntCols[SplitColId][i]) {
          MinValue = IntCols[SplitColId][i];
        }
        if (MaxValue < IntCols[SplitColId][i]) {
          MaxValue = IntCols[SplitColId][i];
        }
      }
    }

    if (StartVal == TInt::Mn) StartVal = MinValue;
    if (EndVal == TInt::Mx) EndVal = MaxValue;
  }

  // initialize buckets
  if (JumpSize == 0) { NumBuckets = (EndVal - StartVal)/JumpSize + 1; }
  else { NumBuckets = (EndVal - StartVal)/JumpSize + 1; }

  InitRowIdBuckets(NumBuckets);

  // populate RowIdSets by computing the range of buckets for each row
  for (TInt i = 0; i < Next.Len(); i++) {
    if (Next[i] == Invalid) { continue; }
    int SplitVal = IntCols[SplitColId][i];
    if (SplitVal < StartVal || SplitVal > EndVal) { continue; }
    int RowVal = SplitVal - StartVal;
    if (JumpSize == 0) { // expanding windows
      MinBucket = RowVal/WindowSize;
      MaxBucket = NumBuckets-1;
    } else if (JumpSize == WindowSize) { // disjoint windows
      MinBucket = MaxBucket = RowVal/JumpSize;  
    } else { // sliding windows
      if (RowVal < WindowSize) { MinBucket = 0; }
      else { MinBucket = (RowVal-WindowSize)/JumpSize + 1; }
      MaxBucket = RowVal/JumpSize;  
    }
    for (TInt j = MinBucket; j <= MaxBucket; j++) { RowIdBuckets[j].Add(i); }
  }
}

TInt TTable::GetColIdx ( const TStr &  ColName ) const

inline

Gets index of column ColName among columns of the same type in the schema.

Definition at line 1004 of file table.h.

                                            {
    TStr NColName = NormalizeColName(ColName);
    return ColTypeMap.IsKey(NColName) ? ColTypeMap.GetDat(NColName).Val2 : TInt(-1);
  }

void TTable::GetCollidingRows ( const TTable &  T, THashSet< TInt > &  Collisions  )

protected

Gets set of row ids of rows common with table T.

Definition at line 3991 of file table.cpp.

                                                                             {
  TIntV UniqueVec;
  THash<TGroupKey, TPair<TInt, TIntV> >Grouping;
  TStrV GroupBy;

  // indices of columns of each type
  TIntV IntGroupByCols;
  TIntV FltGroupByCols;
  TIntV StrGroupByCols;

  TInt IKLen, FKLen, SKLen;

  // check that schemas match
  for (TInt c = 0; c < Sch.Len(); c++) {
    if (Sch[c].Val1 == IdColName) {
      if (Table.Sch[c].Val1 != Table.GetIdColName()) {
        TExcept::Throw("GetCollidingRows: schemas do not match!");
      }
      continue;
    }
    if (Sch[c] != Table.Sch[c]) {
      printf("(%s,%d) != (%s,%d)\n", Sch[c].Val1.CStr(), Sch[c].Val2, Table.Sch[c].Val1.CStr(), Table.Sch[c].Val2); 
      TExcept::Throw("GetCollidingRows: schemas do not match!");
    }
    GroupBy.Add(NormalizeColName(Sch[c].Val1));
    TPair<TAttrType, TInt> ColType = Table.GetColTypeMap(Sch[c].Val1);
    switch (ColType.Val1) {
      case atInt:
        IntGroupByCols.Add(ColType.Val2);
        break;
      case atFlt:
        FltGroupByCols.Add(ColType.Val2);
        break;
      case atStr:
        StrGroupByCols.Add(ColType.Val2);
        break;
    }
  }

  IKLen = IntGroupByCols.Len();
  FKLen = FltGroupByCols.Len();
  SKLen = StrGroupByCols.Len();

  // group rows of first table
  GroupAux(GroupBy, Grouping, true, "", false, UniqueVec, true);

  // find colliding rows of second table
  for (TRowIterator it = Table.BegRI(); it < Table.EndRI(); it++) {
    // read keys from row
    TIntV IKey(IKLen + SKLen, 0);
    TFltV FKey(FKLen, 0);

    // find group key
    for (TInt c = 0; c < IKLen; c++) {
      IKey.Add(it.GetIntAttr(IntGroupByCols[c])); 
    }
    for (TInt c = 0; c < FKLen; c++) {
      FKey.Add(it.GetFltAttr(FltGroupByCols[c])); 
    }
    for (TInt c = 0; c < SKLen; c++) {
      IKey.Add(it.GetStrMapById(StrGroupByCols[c])); 
    }
    // look for group matching the key
    TGroupKey GroupKey = TGroupKey(IKey, FKey);

    TInt RowIdx = it.GetRowIdx();
    if (Grouping.IsKey(GroupKey)) {
      // row exists in first table
      Collisions.AddKey(RowIdx);
    }
  }
}

TAttrType TTable::GetColType ( const TStr &  ColName ) const

inline

Gets type of column ColName.

Definition at line 1218 of file table.h.

                                                        {
    TStr NColName = NormalizeColName(ColName);
    return ColTypeMap.GetDat(NColName).Val1;
  }

TPair<TAttrType, TInt> TTable::GetColTypeMap ( const TStr &  ColName ) const

inlineprotected

Gets column type and index of ColName.

Definition at line 676 of file table.h.

                                                                  {
    TStr NColName = NormalizeColName(ColName);
    return ColTypeMap.GetDat(NColName);
  }

TTableContext* TTable::GetContext ( )

inline

Returns the context.

Definition at line 996 of file table.h.

                              {
    return Context;
  }

const char* TTable::GetContextKey ( TInt  Val ) const

inlineprotected

Gets the Key of the Context StringVals pool. Used by ToGraph method in conv.cpp.

Definition at line 632 of file table.h.

                                            {
          return Context->StringVals.GetKey(Val);
  }

TSize TTable::GetContextMemUsedKB

(

)

Returns approximate memory used by table context in [KB].

Definition at line 3946 of file table.cpp.

                                 {
        TSize ApproxSize = 0;
        ApproxSize += Context->StringVals.GetMemUsed();
        return ApproxSize;
}

TStr TTable::GetDstCol ( ) const

inline

Gets the name of the column to be used as dst nodes in the graph.

Definition at line 1156 of file table.h.

{ return DstCol; }

TStrV TTable::GetDstNodeFltAttrV

(

)

const

Gets dst node float attribute name vector.

Definition at line 1029 of file table.cpp.

                                       {
  TStrV FltNA = TStrV(FltCols.Len(),0);
  for (TInt i = 0; i < DstNodeAttrV.Len(); i++) {
    TStr Attr = DstNodeAttrV[i];
    if (GetColType(Attr) == atFlt) {
      FltNA.Add(Attr);
    }
  }
  return FltNA;
}

TStrV TTable::GetDstNodeIntAttrV

(

)

const

Gets dst node int attribute name vector.

Definition at line 996 of file table.cpp.

                                       {
  TStrV IntNA = TStrV(IntCols.Len(),0);
  for (TInt i = 0; i < DstNodeAttrV.Len(); i++) {
    TStr Attr = DstNodeAttrV[i];
    if (GetColType(Attr) == atInt) {
      IntNA.Add(Attr);
    }
  }
  return IntNA;
}

TStrV TTable::GetDstNodeStrAttrV

(

)

const

Gets dst node str attribute name vector.

Definition at line 1062 of file table.cpp.

                                       {
  TStrV StrNA = TStrV(StrColMaps.Len(),0);
  for (TInt i = 0; i < DstNodeAttrV.Len(); i++) {
    TStr Attr = DstNodeAttrV[i];
    if (GetColType(Attr) == atStr) {
      StrNA.Add(Attr);
    }
  }
  return StrNA;
}

TStrV TTable::GetEdgeFltAttrV

(

)

const

Gets edge float attribute name vector.

Definition at line 1040 of file table.cpp.

                                    {
  TStrV FltEA = TStrV(FltCols.Len(),0);;
  for (TInt i = 0; i < EdgeAttrV.Len(); i++) {
    TStr Attr = EdgeAttrV[i];
    if (GetColType(Attr) == atFlt) {
      FltEA.Add(Attr);
    }
  }
  return FltEA;
}

TStrV TTable::GetEdgeIntAttrV

(

)

const

Gets edge int attribute name vector.

Definition at line 1007 of file table.cpp.

                                    {
  TStrV IntEA = TStrV(IntCols.Len(),0);
  for (TInt i = 0; i < EdgeAttrV.Len(); i++) {
    TStr Attr = EdgeAttrV[i];
    if (GetColType(Attr) == atInt) {
      IntEA.Add(Attr);
    }
  }
  return IntEA;
}

TStrV TTable::GetEdgeStrAttrV

(

)

const

Gets edge str attribute name vector.

Definition at line 1074 of file table.cpp.

                                    {
  TStrV StrEA = TStrV(StrColMaps.Len(),0);
  for (TInt i = 0; i < EdgeAttrV.Len(); i++) {
    TStr Attr = EdgeAttrV[i];
    if (GetColType(Attr) == atStr) {
      StrEA.Add(Attr);
    }
  }
  return StrEA;
}

PTable TTable::GetEdgeTable ( const PNEANet &  Network, TTableContext *  Context  )

static

Extracts edge TTable from PNEANet.

Definition at line 3719 of file table.cpp.

                                                                          {
  Schema SR;
  SR.Add(TPair<TStr,TAttrType>("edg_id",atInt));
  SR.Add(TPair<TStr,TAttrType>("src_id",atInt));
  SR.Add(TPair<TStr,TAttrType>("dst_id",atInt));

  TStrV IntAttrNames;
  TStrV FltAttrNames;
  TStrV StrAttrNames;

  TNEANet::TEdgeI EdgeI = Network->BegEI();
  EdgeI.GetIntAttrNames(IntAttrNames);
  EdgeI.GetFltAttrNames(FltAttrNames);
  EdgeI.GetStrAttrNames(StrAttrNames);
  for (TInt i = 0; i < IntAttrNames.Len(); i++) {
    SR.Add(TPair<TStr,TAttrType>(IntAttrNames[i],atInt));
  }
  for (TInt i = 0; i < FltAttrNames.Len(); i++) {
    SR.Add(TPair<TStr,TAttrType>(FltAttrNames[i],atFlt));
  }
  for (TInt i = 0; i < StrAttrNames.Len(); i++) {
    //printf("%s\n",StrAttrNames[i].CStr());
    SR.Add(TPair<TStr,TAttrType>(StrAttrNames[i],atStr));
  }

  PTable T = New(SR, Context);

  TInt Cnt = 0;
  // populate table columns
  while (EdgeI < Network->EndEI()) {
    T->IntCols[0].Add(EdgeI.GetId());
    T->IntCols[1].Add(EdgeI.GetSrcNId());
    T->IntCols[2].Add(EdgeI.GetDstNId());
    for (TInt i = 0; i < IntAttrNames.Len(); i++) {
      T->IntCols[i+3].Add(Network->GetIntAttrDatE(EdgeI,IntAttrNames[i]));
    }
    for (TInt i = 0; i < FltAttrNames.Len(); i++) {
      T->FltCols[i].Add(Network->GetFltAttrDatE(EdgeI,FltAttrNames[i]));
    }
    for (TInt i = 0; i < StrAttrNames.Len(); i++) {
      T->AddStrVal(i, Network->GetStrAttrDatE(EdgeI,StrAttrNames[i]));
    }
    Cnt++;
    EdgeI++;
  }
  // set number of rows and "Next" vector
  T->NumRows = Cnt;
  T->NumValidRows = T->NumRows;
  T->Next = TIntV(T->NumRows,0);
  for (TInt i = 0; i < T->NumRows-1; i++) {
    T->Next.Add(i+1);
  }
  T->LastValidRow = T->NumRows-1;
  T->Next.Add(Last);
  return T;
}

PTable TTable::GetEdgeTablePN ( const PNGraphMP &  Network, TTableContext *  Context  )

static

Extracts edge TTable from parallel graph PNGraphMP.

Definition at line 3777 of file table.cpp.

                                                                             {
  Schema SR;
  SR.Add(TPair<TStr,TAttrType>("src_id",atInt));
  SR.Add(TPair<TStr,TAttrType>("dst_id",atInt));

  TNGraphMP::TEdgeI FirstEI = Network->BegEI();
  PTable T = New(SR, Context);
  TInt NumEdges = Network->GetEdges();
  TInt NumPartitions = omp_get_max_threads()*CHUNKS_PER_THREAD;
  TInt PartitionSize = NumEdges/NumPartitions;
  if (PartitionSize*NumPartitions < NumEdges) { NumPartitions++;}

  typedef TPair<TNGraphMP::TEdgeI, TNGraphMP::TEdgeI> TEIPr;
  TVec<TEIPr> Partitions;
  TIntV PartitionSizes;
  TNGraphMP::TEdgeI currStart = FirstEI;
  TInt currCount = 0;
  while (FirstEI < Network->EndEI()){
    if (currCount == PartitionSize) {
      Partitions.Add(TEIPr(currStart, FirstEI));
      currStart = FirstEI;
      PartitionSizes.Add(currCount);
      //printf("added: %d\n", currCount.Val);
      currCount = 0;
    }
    //printf("%d\n", currCount.Val);
    FirstEI++;
    currCount++;
  }
  Partitions.Add(TEIPr(currStart, FirstEI));
  PartitionSizes.Add(currCount);

  T->ResizeTable(NumEdges);
  #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD)
  for (int p = 0; p < Partitions.Len(); p++) {
    TNGraphMP::TEdgeI EdgeI = Partitions[p].GetVal1();
    TNGraphMP::TEdgeI EndI = Partitions[p].GetVal2();
    //printf("Thread = %d, p = %d, size = %d\n", omp_get_thread_num(), p, PartitionSizes[p].Val);
    int start = T->GetEmptyRowsStart(PartitionSizes[p]);
    while (EdgeI < EndI) {
      T->IntCols[0][start] = EdgeI.GetSrcNId();
      T->IntCols[1][start] = EdgeI.GetDstNId();
      EdgeI++;
      if (EdgeI < EndI) { T->Next[start] = start+1;}
      start++;
    }
  }

  Assert(T->NumRows == NumEdges);
  return T;
}

int TTable::GetEmptyRowsStart ( int  NewRows )

protected

Gets the start index to a chunk of empty rows of size NewRows.

Definition at line 4353 of file table.cpp.

                                         {
  int start = -1;
#ifdef USE_OPENMP
  #pragma omp critical
  {
#endif
    start = NumRows;
    NumRows += NewRows;
    NumValidRows += NewRows;
    // To make this function thread-safe, the following call must be done before the 
    // code enters parallel region.
    // ResizeTable(NumRows);
    Assert(NumRows <= Next.Len());
    if (LastValidRow >= 0) {Next[LastValidRow] = start;}
    LastValidRow = start+NewRows-1;
    Next[LastValidRow] = Last;
#ifdef USE_OPENMP
  }
#endif
  Assert (start >= 0);
  return start;
}

PNEANet TTable::GetFirstGraphFromSequence ( TAttrAggr  AggrPolicy )

protected

Returns the first graph of the sequence.

Return the first graph of the sequence corresponding to the sets of row ids in RowIdBuckets. This is used by the ToGraph*Iterator functions.

Definition at line 3606 of file table.cpp.

                                                              {
  CurrBucket = -1;
  this->AggrPolicy = AggrPolicy;
  return GetNextGraphFromSequence();
}

PTable TTable::GetFltNodePropertyTable ( const PNEANet &  Network, const TIntFltH &  Property, const TStr &  NodeAttrName, const TAttrType &  NodeAttrType, const TStr &  PropertyAttrName, TTableContext *  Context  )

static

Extracts node and edge property TTables from THash.

Definition at line 3830 of file table.cpp.

                         {
  Schema SR;
  // Determine type of node id
  SR.Add(TPair<TStr,TAttrType>(NodeAttrName,NodeAttrType));
  SR.Add(TPair<TStr,TAttrType>(PropertyAttrName,atFlt));
  PTable T = New(SR, Context);
  TInt NodeColIdx = T->GetColIdx(NodeAttrName);
  TInt Cnt = 0;
  // populate table columns
  for (TNEANet::TNodeI NodeI = Network->BegNI(); NodeI < Network->EndNI(); NodeI++) {
    switch (NodeAttrType) {
      case atInt:
        T->IntCols[NodeColIdx].Add(Network->GetIntAttrDatN(NodeI,NodeAttrName));
        break;
      case atFlt:
        T->FltCols[NodeColIdx].Add(Network->GetFltAttrDatN(NodeI,NodeAttrName));
        break;
      case atStr:
        T->AddStrVal(TInt(0), Network->GetStrAttrDatN(NodeI,NodeAttrName));
        break;
    }
    T->FltCols[0].Add(Property.GetDat(NodeI.GetId()));
    Cnt++;
  }
  // set number of rows and "Next" vector
  T->NumRows = Cnt;
  T->NumValidRows = T->NumRows;
  T->Next = TIntV(T->NumRows,0);
  for (TInt i = 0; i < T->NumRows-1; i++) {
    T->Next.Add(i+1);
  }
  T->LastValidRow = T->NumRows-1;
  T->Next.Add(Last);
  return T;
}

TIntV TTable::GetFltRowIdxByVal	(	const TStr &	ColName,
		const TFlt &	Val
	)		const

Gets the rows containing Val in flt column ColName.

Returns the RowIdxs in the float column given by ColName which have value Val, as a Vector. (If no such value is found, returns an empty vector.) Uses an index created by RequestIndex method if it exists, else loops over the entire table (which can be slow, so it is recommended to request an index if multiple queries must be made).

Definition at line 5430 of file table.cpp.

                                                                          {

  if (FltColIndexes.IsKey(ColName)) {
    THash<TFlt, TIntV> ColIndex = FltColIndexes.GetDat(ColName);
    if (ColIndex.IsKey(Val)) {
      return ColIndex.GetDat(Val);
    }
    else {
      TIntV Empty;
      return Empty;
    }
  }

  TIntV ToReturn;
  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    TFlt ValAtRow = RowI.GetFltAttr(ColName);
    if ( Val == ValAtRow) {
      ToReturn.Add(RowI.GetRowIdx());
    }
  }
  return ToReturn;
}

TFlt TTable::GetFltVal ( const TStr &  ColName, const TInt &  RowIdx  )

inline

Gets the value of float attribute ColName at row RowIdx.

Definition at line 1015 of file table.h.

                                                          {
    return FltCols[GetColIdx(ColName)][RowIdx];
  }

TFlt TTable::GetFltValAtRowIdx ( const TInt &  ColIdx, const TInt &  RowIdx  )

inline

Get the float value at column ColIdx and row RowIdx.

Definition at line 1111 of file table.h.

                                                                 {
    return FltCols[ColIdx][RowIdx];
  }

TVec< PNEANet > TTable::GetGraphsFromSequence ( TAttrAggr  AggrPolicy )

protected

Returns a sequence of graphs.

Return a sequence of graphs, each constructed from the set of row ids corresponding to a particular bucket in RowIdBuckets.

Definition at line 3594 of file table.cpp.

                                                                {
  //call BuildGraph on each row id set - parallelizable!
  TVec<PNEANet> GraphSequence;
  for (TInt i = 0; i < RowIdBuckets.Len(); i++) {
    if (RowIdBuckets[i].Len() == 0) { continue; }
    PNEANet PNet = BuildGraph(RowIdBuckets[i], AggrPolicy);
    GraphSequence.Add(PNet);
  }

  return GraphSequence;
}

TStr TTable::GetIdColName ( ) const

inlineprotected

Gets name of the id column of this table.

Definition at line 646 of file table.h.

{ return IdColName; }

TIntV TTable::GetIntRowIdxByVal	(	const TStr &	ColName,
		const TInt &	Val
	)		const

Gets the rows containing Val in int column ColName.

Returns the RowIdxs in the integer column given by ColName which have value Val, as a Vector. (If no such value is found, returns an empty vector.) Uses an index created by RequestIndex method if it exists, else loops over the entire table (which can be slow, so it is recommended to request an index if multiple queries must be made).

Definition at line 5387 of file table.cpp.

                                                                          {

  if (IntColIndexes.IsKey(ColName)) {
    THash<TInt, TIntV> ColIndex = IntColIndexes.GetDat(ColName);
    if (ColIndex.IsKey(Val)) {
      return ColIndex.GetDat(Val);
    }
    else {
      TIntV Empty;
      return Empty;
    }
  }
  TIntV ToReturn;
  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    TInt ValAtRow = RowI.GetIntAttr(ColName);
    if ( Val == ValAtRow) {
      ToReturn.Add(RowI.GetRowIdx());
    }
  }
  return ToReturn;
}

TInt TTable::GetIntVal ( const TStr &  ColName, const TInt &  RowIdx  )

inline

Gets the value of integer attribute ColName at row RowIdx.

Definition at line 1011 of file table.h.

                                                          {
    return IntCols[GetColIdx(ColName)][RowIdx];
  }

TInt TTable::GetIntValAtRowIdx ( const TInt &  ColIdx, const TInt &  RowIdx  )

inline

Get the integer value at column ColIdx and row RowIdx.

Definition at line 1107 of file table.h.

                                                                 {
    return IntCols[ColIdx][RowIdx];
  }

TInt TTable::GetLastValidRowIdx ( )

protected

Gets the id of the last valid row of the table.

static TTableIterator TTable::GetMapHitsIterator ( const TVec< PNEANet > &  GraphSeq, TTableContext *  Context, const int &  MaxIter = 20  )

inlinestatic

Gets sequence of Hits tables from given GraphSeq.

Definition at line 1518 of file table.h.

                                                    {
    TVec<PTable> TableSeq(GraphSeq.Len());
    TSnap::MapHits(GraphSeq, TableSeq, Context, MaxIter);
    return TTableIterator(TableSeq);
  }

static TTableIterator TTable::GetMapPageRank ( const TVec< PNEANet > &  GraphSeq, TTableContext *  Context, const double &  C = 0.85, const double &  Eps = 1e-4, const int &  MaxIter = 100  )

inlinestatic

Gets sequence of PageRank tables from given GraphSeq.

Definition at line 1510 of file table.h.

                                                                               {
    TVec<PTable> TableSeq(GraphSeq.Len());
    TSnap::MapPageRank(GraphSeq, TableSeq, Context, C, Eps, MaxIter);
    return TTableIterator(TableSeq);
  }

TSize TTable::GetMemUsedKB

(

)

Returns approximate memory used by table in [KB].

Definition at line 3918 of file table.cpp.

                           {
  TSize ApproxSize = 0;
  ApproxSize += Next.GetMemUsed()/1000;  // Next vector
  for(int i = 0; i < IntCols.Len(); i++){
        ApproxSize += IntCols[i].GetMemUsed()/1000;
  }
  for(int i = 0; i < FltCols.Len(); i++){
        ApproxSize += FltCols[i].GetMemUsed()/1000;
  }
  for(int i = 0; i < StrColMaps.Len(); i++){
        ApproxSize += StrColMaps[i].GetMemUsed()/1000;
  }
  ApproxSize += RowIdMap.GetMemUsed()/1000;
  ApproxSize += GroupIDMapping.GetMemUsed()/1000;
  ApproxSize += GroupMapping.GetMemUsed()/1000;
  ApproxSize += RowIdBuckets.GetMemUsed() / 1000;
  return ApproxSize;
}

static TInt TTable::GetMP ( )

inlinestatic

Definition at line 537 of file table.h.

{ return UseMP; }

PNEANet TTable::GetNextGraphFromSequence ( )

protected

Returns the next graph in sequence corresponding to RowIdBuckets.

Returns the next graph in sequence corresponding to RowIdBuckets. This is used to iterate over the graph sequence by constructing one graph at a time. Called by NextGraphIterator().

Definition at line 3612 of file table.cpp.

                                         {
  CurrBucket++;
  while (CurrBucket < RowIdBuckets.Len() && RowIdBuckets[CurrBucket].Len() == 0) {
    CurrBucket++;
  }
  if (CurrBucket >= RowIdBuckets.Len()) { return NULL; }
  return BuildGraph(RowIdBuckets[CurrBucket], AggrPolicy);
}

PTable TTable::GetNodeTable ( const PNEANet &  Network, TTableContext *  Context  )

static

Extracts node TTable from PNEANet.

Definition at line 3667 of file table.cpp.

                                                                          {
  Schema SR;
  SR.Add(TPair<TStr,TAttrType>("node_id",atInt));

  TStrV IntAttrNames;
  TStrV FltAttrNames;
  TStrV StrAttrNames;

  TNEANet::TNodeI NodeI = Network->BegNI();
  NodeI.GetIntAttrNames(IntAttrNames);
  NodeI.GetFltAttrNames(FltAttrNames);
  NodeI.GetStrAttrNames(StrAttrNames);
  for (TInt i = 0; i < IntAttrNames.Len(); i++) {
    SR.Add(TPair<TStr,TAttrType>(IntAttrNames[i],atInt));
  }
  for (TInt i = 0; i < FltAttrNames.Len(); i++) {
    SR.Add(TPair<TStr,TAttrType>(FltAttrNames[i],atFlt));
  }
  for (TInt i = 0; i < StrAttrNames.Len(); i++) {
    SR.Add(TPair<TStr,TAttrType>(StrAttrNames[i],atStr));
  }

  PTable T = New(SR, Context);

  TInt Cnt = 0;
  // populate table columns
  while (NodeI < Network->EndNI()) {
    T->IntCols[0].Add(NodeI.GetId());
    for (TInt i = 0; i < IntAttrNames.Len(); i++) {
      T->IntCols[i+1].Add(Network->GetIntAttrDatN(NodeI,IntAttrNames[i]));
    }
    for (TInt i = 0; i < FltAttrNames.Len(); i++) {
      T->FltCols[i].Add(Network->GetFltAttrDatN(NodeI,FltAttrNames[i]));
    }
    for (TInt i = 0; i < StrAttrNames.Len(); i++) {
      T->AddStrVal(i, Network->GetStrAttrDatN(NodeI,StrAttrNames[i]));
    }
    Cnt++;
    NodeI++;
  }
  // set number of rows and "Next" vector
  T->NumRows = Cnt;
  T->NumValidRows = T->NumRows;
  T->Next = TIntV(T->NumRows,0);
  for (TInt i = 0; i < T->NumRows-1; i++) {
    T->Next.Add(i+1);
  }
  T->LastValidRow = T->NumRows-1;
  T->Next.Add(Last);
  return T;
}

TInt TTable::GetNumRows ( ) const

inline

Gets total number of rows in this table.

Definition at line 1223 of file table.h.

{ return NumRows;}

TInt TTable::GetNumValidRows ( ) const

inline

Gets number of valid, i.e. not deleted, rows in this table.

Definition at line 1225 of file table.h.

{ return NumValidRows;}

void TTable::GetPartitionRanges	(	TIntPrV &	Partitions,
		TInt	NumPartitions
	)		const

Partitions the table into NumPartitions and populate Partitions with the ranges.

Definition at line 1157 of file table.cpp.

                                                                             {
  TInt PartitionSize = NumValidRows / (NumPartitions);
  if (NumValidRows % NumPartitions != 0) PartitionSize++;
  if (PartitionSize < 10) { 
    PartitionSize = 10;
    NumPartitions = NumValidRows / PartitionSize; 
  }
  Partitions.Reserve(NumPartitions+1);

  TInt currRow = FirstValidRow;
  TInt currStart = currRow;
  if (IsNextDirty) {
    TInt currCount = PartitionSize;
    while (currRow != TTable::Last) {
      if (currCount == 0) {
        Partitions.Add(TIntPr(currStart, currRow));
        currStart = currRow;
        currCount = PartitionSize;
      }
      currRow = Next[currRow];
      currCount--;
    }
    Partitions.Add(TIntPr(currStart, currRow));
  } else {
    // Optimize for the case when rows are logically in sequence.
    currRow += PartitionSize;
    while (currRow != TTable::Last && currRow < Next.Len()) {
      if (Next[currRow] == TTable::Invalid) { currRow++; continue; }
      Partitions.Add(TIntPr(currStart, currRow));
      currStart = currRow;
      currRow += PartitionSize;
    }
    Partitions.Add(TIntPr(currStart, TTable::Last));
  }
  //printf("Num partitions: %d\n", Partitions.Len());
}

TInt TTable::GetPivot ( TIntV &  V, TInt  StartIdx, TInt  EndIdx, const TVec< TAttrType > &  SortByTypes, const TIntV &  SortByIndices, TBool  Asc  )

protected

Gets pivot element for QSort.

Definition at line 3090 of file table.cpp.

                                                                                                                                     {
  TInt L = EndIdx - StartIdx + 1;
  const TInt Idx1 = StartIdx + TInt::GetRnd(L);
  const TInt Idx2 = StartIdx + TInt::GetRnd(L);
  const TInt Idx3 = StartIdx + TInt::GetRnd(L);
  if (CompareRows(V[Idx1], V[Idx2], SortByTypes, SortByIndices, Asc) < 0) {
    if (CompareRows(V[Idx2], V[Idx3], SortByTypes, SortByIndices, Asc) < 0) { return Idx2; }
    if (CompareRows(V[Idx1], V[Idx3], SortByTypes, SortByIndices, Asc) < 0) { return Idx3; }
    return Idx1;
  } else {
    if (CompareRows(V[Idx3], V[Idx2], SortByTypes, SortByIndices, Asc) < 0) { return Idx2; }
    if (CompareRows(V[Idx3], V[Idx1], SortByTypes, SortByIndices, Asc) < 0) { return Idx3; }
    return Idx1;
  }
}

TInt TTable::GetPivotKeyVal ( TIntV &  Key, TIntV &  Val, TInt  Start, TInt  End  )

staticprotected

Definition at line 5315 of file table.cpp.

                                                                        {
  TInt L = End - Start + 1;
  const TInt Idx1 = Start + TInt::GetRnd(L);
  const TInt Idx2 = Start + TInt::GetRnd(L);
  const TInt Idx3 = Start + TInt::GetRnd(L);
  if (CompareKeyVal(Key[Idx1], Val[Idx1], Key[Idx2], Val[Idx2]) < 0) {
    if (CompareKeyVal(Key[Idx2], Val[Idx2], Key[Idx3], Val[Idx3]) < 0) { return Idx2; }
    if (CompareKeyVal(Key[Idx1], Val[Idx1], Key[Idx3], Val[Idx3]) < 0) { return Idx3; }
    return Idx1;
  } else {
    if (CompareKeyVal(Key[Idx3], Val[Idx3], Key[Idx2], Val[Idx2]) < 0) { return Idx2; }
    if (CompareKeyVal(Key[Idx3], Val[Idx3], Key[Idx1], Val[Idx1]) < 0) { return Idx3; }
    return Idx1;
  }
}

THash<TInt, TInt> TTable::GetRowIdMap ( ) const

inline

Gets a map of logical to physical row ids.

Definition at line 1228 of file table.h.

{ return RowIdMap;}

void TTable::GetSchema ( const TStr &  InFNm, Schema &  S, const char &  Separator = '\t'  )

static

Returns pointer to a new table created from given Table, with name set to TableName.

Automatically detects the Schema of a input file (data is assumed to be in tsv format)

Definition at line 435 of file table.cpp.

                                                                          {
  // Determine Attr Type
  // Assume that the data is tab separated
  TSsParser Ss(InFNm, '\t', false, false, false);
  TInt rowsToPeek = 1000;
  TInt currRow = 0;
  TInt lastComment = 0;
  while (Ss.Next()) {
    if (Ss.IsCmt()) {
      lastComment += 1;
    }
    else break;
  }
  if (Ss.Eof()) {TExcept::Throw("No Data to determine attribute types!");}
  TInt numCols = Ss.GetFlds();
  TVec<TAttrType> colAttrV(numCols);
  colAttrV.PutAll(atInt);
  while (true) {
    for (TInt i = 0; i < numCols; i++) {
      if (Ss.IsInt(i)) {
      }
      else if (Ss.IsFlt(i)) {
        colAttrV[i] = atFlt;
      }
      else {
        colAttrV[i] = atStr;
      }
    }
    currRow++;
    if (currRow > rowsToPeek || Ss.Eof()) break;
    Ss.Next();
  }
  // Default Separator is tab
  TSsParser SsNames(InFNm, Separator, false, false, false);
  for (int i = 0; i < lastComment; i++) { SsNames.Next();}
  TVec<TStr> attrV;
  TStr first(SsNames[0]);
  int begin = 0;
  TStr comment('#');
  if (first != comment) {
    for (int i = 1; i < first.Len(); i++){
      if (first[i] != ' ') { begin = i; break;}
    }
    attrV.Add(first.GetSubStr(begin));
  }
  for (int i = 1; i < SsNames.GetFlds(); i++) {attrV.Add(SsNames[i]);}
  for (TInt i = 0; i < numCols; i++) {
    S.Add(TPair<TStr,TAttrType>(attrV[i],colAttrV[i]));
  } 
}

Schema TTable::GetSchema ( )

inline

Gets the schema of this table.

Definition at line 1116 of file table.h.

{ return DenormalizeSchema(); }

TStr TTable::GetSchemaColName ( TInt  Idx ) const

inlineprotected

Gets name of the column with index Idx in the schema.

Definition at line 648 of file table.h.

{ return Sch[Idx].Val1; }

TAttrType TTable::GetSchemaColType ( TInt  Idx ) const

inlineprotected

Gets type of the column with index Idx in the schema.

Definition at line 650 of file table.h.

{ return Sch[Idx].Val2; }

TStr TTable::GetSrcCol ( ) const

inline

Gets the name of the column to be used as src nodes in the graph.

Definition at line 1149 of file table.h.

{ return SrcCol; }

TStrV TTable::GetSrcNodeFltAttrV

(

)

const

Gets src node float attribute name vector.

Definition at line 1018 of file table.cpp.

                                       {
  TStrV FltNA = TStrV(FltCols.Len(),0);
  for (TInt i = 0; i < SrcNodeAttrV.Len(); i++) {
    TStr Attr = SrcNodeAttrV[i];
    if (GetColType(Attr) == atFlt) {
      FltNA.Add(Attr);
    }
  }
  return FltNA;
}

TStrV TTable::GetSrcNodeIntAttrV

(

)

const

Gets src node int attribute name vector.

Definition at line 985 of file table.cpp.

                                       {
  TStrV IntNA = TStrV(IntCols.Len(),0);
  for (TInt i = 0; i < SrcNodeAttrV.Len(); i++) {
    TStr Attr = SrcNodeAttrV[i];
    if (GetColType(Attr) == atInt) {
      IntNA.Add(Attr);
    }
  }
  return IntNA;
}

TStrV TTable::GetSrcNodeStrAttrV

(

)

const

Gets src node str attribute name vector.

Definition at line 1051 of file table.cpp.

                                       {
  TStrV StrNA = TStrV(StrColMaps.Len(),0);
  for (TInt i = 0; i < SrcNodeAttrV.Len(); i++) {
    TStr Attr = SrcNodeAttrV[i];
    if (GetColType(Attr) == atStr) {
      StrNA.Add(Attr);
    }
  }
  return StrNA;
}

TStr TTable::GetStr ( const TInt &  KeyId ) const

inline

Gets the string with KeyId.

Definition at line 1100 of file table.h.

                                       {
    return Context->StringVals.GetKey(KeyId);
  }

TInt TTable::GetStrMapById ( TInt  ColIdx, TInt  RowIdx  ) const

inline

Gets the integer mapping of the string at column ColIdx at row RowIdx.

Definition at line 1024 of file table.h.

                                                     {
    return StrColMaps[ColIdx][RowIdx];
  }

TInt TTable::GetStrMapByName ( const TStr &  ColName, TInt  RowIdx  ) const

inline

Gets the integer mapping of the string at column ColName at row RowIdx.

Definition at line 1029 of file table.h.

                                                               {
    return StrColMaps[GetColIdx(ColName)][RowIdx];
  }

TIntV TTable::GetStrRowIdxByMap	(	const TStr &	ColName,
		const TInt &	Map
	)		const

Gets the rows containing int mapping Map in str column ColName.

Returns the RowIdxs in the string column given by ColName which have the string with integer mapping Map, as a Vector. (If no such value is found, returns an empty vector.) Uses an index created by RequestIndex method if it exists, else loops over the entire table (which can be slow, so it is recommended to request an index if multiple queries must be made).

Definition at line 5408 of file table.cpp.

                                                                          {

  if (StrMapColIndexes.IsKey(ColName)) {
    THash<TInt, TIntV> ColIndex = StrMapColIndexes.GetDat(ColName);
    if (ColIndex.IsKey(Map)) {
      return ColIndex.GetDat(Map);
    }
    else {
      TIntV Empty;
      return Empty;
    }
  }
  TIntV ToReturn;
  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    TInt MapAtRow = RowI.GetStrMapByName(ColName);
    if ( Map == MapAtRow) {
      ToReturn.Add(RowI.GetRowIdx());
    }
  }
  return ToReturn;
}

TStr TTable::GetStrVal ( TInt  ColIdx, TInt  RowIdx  ) const

inlineprotected

Gets the value in column with id ColIdx at row RowIdx.

Definition at line 636 of file table.h.

                                                 {
    return TStr(Context->StringVals.GetKey(StrColMaps[ColIdx][RowIdx]));
  }

TStr TTable::GetStrVal ( const TStr &  ColName, const TInt &  RowIdx  ) const

inline

Gets the value of string attribute ColName at row RowIdx.

Definition at line 1019 of file table.h.

                                                                {
    return GetStrVal(GetColIdx(ColName), RowIdx);
  }

TStr TTable::GetStrValById ( TInt  ColIdx, TInt  RowIdx  ) const

inline

Gets the value of the string attribute at column ColIdx at row RowIdx.

Definition at line 1034 of file table.h.

                                                     {
    return GetStrVal(ColIdx, RowIdx);
  }

TStr TTable::GetStrValByName ( const TStr &  ColName, const TInt &  RowIdx  ) const

inline

Gets the value of the string attribute at column ColName at row RowIdx.

Definition at line 1039 of file table.h.

                                                                      {
    return GetStrVal(ColName, RowIdx);
  }

void TTable::Group	(	const TStrV &	GroupBy,
		const TStr &	GroupColName,
		TBool	Ordered = true,
		TBool	UsePhysicalIds = true
	)

Groups rows depending on values of GroupBy columns.

Specify columns to group by, name of column in new table, whether to treat columns as ordered If name of column is an empty string, no column is created

Definition at line 1549 of file table.cpp.

                                                                                                      {
  TStrV NGroupBy = NormalizeColNameV(GroupBy);
  TStr NGroupColName = NormalizeColName(GroupColName);
  TIntV UniqueVec;
  THash<TGroupKey, TPair<TInt, TIntV> > Grouping;
  GroupAux(NGroupBy, Grouping, Ordered, NGroupColName, false, UniqueVec, UsePhysicalIds);
}

void TTable::GroupAux ( const TStrV &  GroupBy, THash< TGroupKey, TPair< TInt, TIntV > > &  Grouping, TBool  Ordered, const TStr &  GroupColName, TBool  KeepUnique, TIntV &  UniqueVec, TBool  UsePhysicalIds = true  )

protected

Helper function for grouping.

If KeepUnique is true, UniqueVec will be modified to contain a row from each group If KeepUnique is false, then normal grouping is done and a new column is added depending on whether GroupColName is empty

Definition at line 1302 of file table.cpp.

                                                                                                    {
  TInt IdColIdx = GetColIdx(IdColName);
  if(!UsePhysicalIds && IdColIdx < 0){
        TExcept::Throw("Grouping: Either use physical row ids, or have an id column");
  }
  TIntV IntGroupByCols;
  TIntV FltGroupByCols;
  TIntV StrGroupByCols;
  // get indices for each column type
  for (TInt c = 0; c < GroupBy.Len(); c++) {
        //printf("GroupBy col %d: %s\n", c.Val, GroupBy[c].CStr());
    if (!IsColName(GroupBy[c])) { 
      TExcept::Throw("no such column " + GroupBy[c]); 
    }

    TPair<TAttrType, TInt> ColType = GetColTypeMap(GroupBy[c]);
    switch (ColType.Val1) {
      case atInt:
        IntGroupByCols.Add(ColType.Val2);
        break;
      case atFlt:
        FltGroupByCols.Add(ColType.Val2);
        break;
      case atStr:
        StrGroupByCols.Add(ColType.Val2);
        break;
    }
  }
  
  TInt IKLen = IntGroupByCols.Len();
  TInt FKLen = FltGroupByCols.Len();
  TInt SKLen = StrGroupByCols.Len();

  TInt GroupNum = 0;
  TVec<TPair<TInt, TInt> > GroupAndRowIds;
  //printf("done GroupAux initialization\n");

  // iterate over rows
  for (TRowIterator it = BegRI(); it < EndRI(); it++) {
    TIntV IKey(IKLen + SKLen, 0);
    TFltV FKey(FKLen, 0);
    TIntV SKey(SKLen, 0);

    // find group key
    for (TInt c = 0; c < IKLen; c++) {
      IKey.Add(it.GetIntAttr(IntGroupByCols[c])); 
    }
    for (TInt c = 0; c < FKLen; c++) {
      FKey.Add(it.GetFltAttr(FltGroupByCols[c])); 
    }
    for (TInt c = 0; c < SKLen; c++) {
      SKey.Add(it.GetStrMapById(StrGroupByCols[c])); 
    }
    if (!Ordered) {
      if (IKLen > 0) { IKey.ISort(0, IKey.Len()-1, true); }
      if (FKLen > 0) { FKey.ISort(0, FKey.Len()-1, true); }
      if (SKLen > 0) { SKey.ISort(0, SKey.Len()-1, true); }
    }
    for (TInt c = 0; c < SKLen; c++) {
      IKey.Add(SKey[c]);
    }
    
    // look for group matching the key
    TGroupKey GroupKey = TGroupKey(IKey, FKey);

    TInt RowIdx = it.GetRowIdx();
    TInt idx = UsePhysicalIds ? it.GetRowIdx() : IntCols[IdColIdx][it.GetRowIdx()];
    if (!Grouping.IsKey(GroupKey)) {
      // Grouping key hasn't been seen before, create a new group
      TPair<TInt, TIntV> NewGroup;
      NewGroup.Val1 = GroupNum;
      NewGroup.Val2.Add(idx);
      Grouping.AddDat(GroupKey, NewGroup);
      if (GroupColName != "") {
        GroupAndRowIds.Add(TPair<TInt, TInt>(GroupNum, RowIdx));
      }
      if (KeepUnique) { 
        UniqueVec.Add(idx);
      }
      GroupNum++;
    } else {
      // Grouping key has been seen before, update corresponding group
      if (!KeepUnique) {
        TPair<TInt, TIntV>& NewGroup = Grouping.GetDat(GroupKey);
        NewGroup.Val2.Add(idx);
        if (GroupColName != "") {
          GroupAndRowIds.Add(TPair<TInt, TInt>(NewGroup.Val1, RowIdx));
        }
      }
    }
  }
  // printf("KeepUnique: %d\n", KeepUnique.Val);
  // update group mapping
  if (!KeepUnique) {
    GroupStmt Stmt(NormalizeColNameV(GroupBy), Ordered, UsePhysicalIds);
    GroupStmtNames.AddDat(GroupColName, Stmt);
    GroupIDMapping.AddKey(Stmt);
    GroupMapping.AddKey(Stmt);
    //printf("Adding statement: ");
    //Stmt.Print();
    for (THash<TGroupKey, TPair<TInt, TIntV> >::TIter it = Grouping.BegI(); it < Grouping.EndI(); it++) {
      TGroupKey key = it.GetKey();
      TPair<TInt, TIntV> group = it.GetDat();
      GroupIDMapping.GetDat(Stmt).AddDat(group.Val1, TGroupKey(key));
      GroupMapping.GetDat(Stmt).AddDat(TGroupKey(key), TIntV(group.Val2));
    }
  }

  // add a column to the table
  if (GroupColName != "") {
    StoreGroupCol(GroupColName, GroupAndRowIds);
    AddSchemaCol(GroupColName, atInt);  // update schema
  }
}

template<class T >

void TTable::GroupByFltCol ( const TStr &  GroupBy, T &  Grouping, const TIntV &  IndexSet, TBool  All, TBool  UsePhysicalIds = true  ) const

protected

Groups/hashes by a single column with float values. Returns hash table with grouping.

Definition at line 1633 of file table.h.

                                                               {
  TInt IdColIdx = GetColIdx(IdColName);
  if(!UsePhysicalIds && IdColIdx < 0){
        TExcept::Throw("Grouping: Either use physical row ids, or have an id column");
  }
  GroupingSanityCheck(GroupBy, atFlt);
  if (All) {
     // Optimize for the common and most expensive case - iterate over only valid rows.
    for (TRowIterator it = BegRI(); it < EndRI(); it++) {
      TInt idx = UsePhysicalIds ? it.GetRowIdx() : it.GetIntAttr(IdColIdx);
      UpdateGrouping<TFlt>(Grouping, it.GetFltAttr(GroupBy), idx);
    }
  } else {
    // Consider only rows in IndexSet.
    for (TInt i = 0; i < IndexSet.Len(); i++) {
      if (IsRowValid(IndexSet[i])) {
        TInt RowIdx = IndexSet[i];
        const TFltV& Col = FltCols[GetColIdx(GroupBy)];   
        TInt idx = UsePhysicalIds ? RowIdx : IntCols[IdColIdx][RowIdx];     
        UpdateGrouping<TFlt>(Grouping, Col[RowIdx], idx);
      }
    }
  }
}

template<class T >

void TTable::GroupByIntCol ( const TStr &  GroupBy, T &  Grouping, const TIntV &  IndexSet, TBool  All, TBool  UsePhysicalIds = true  ) const

protected

Groups/hashes by a single column with integer values.

Group/hash by a single column with integer values. Returns hash table with grouping. IndexSet tells what rows to consider (vector of physical row ids). It is used only if All == true. Note that the IndexSet option is currently not used anywhere.

Definition at line 1605 of file table.h.

                                                               {
  TInt IdColIdx = GetColIdx(IdColName);
  if(!UsePhysicalIds && IdColIdx < 0){
        TExcept::Throw("Grouping: Either use physical row ids, or have an id column");
  }
 // TO do: add a check if grouping already exists and is valid
  GroupingSanityCheck(GroupBy, atInt);
  if (All) {
     // Optimize for the common and most expensive case - iterate over only valid rows.
    for (TRowIterator it = BegRI(); it < EndRI(); it++) {
      TInt idx = UsePhysicalIds ? it.GetRowIdx() : it.GetIntAttr(IdColIdx);
      UpdateGrouping<TInt>(Grouping, it.GetIntAttr(GroupBy), idx);
    }
  } else {
    // Consider only rows in IndexSet.
    for (TInt i = 0; i < IndexSet.Len(); i++) {
      if (IsRowValid(IndexSet[i])) {
        TInt RowIdx = IndexSet[i];
        const TIntV& Col = IntCols[GetColIdx(GroupBy)];
        TInt idx = UsePhysicalIds ? RowIdx : IntCols[IdColIdx][RowIdx];       
        UpdateGrouping<TInt>(Grouping, Col[RowIdx], idx);
      }
    }
  }
}

void TTable::GroupByIntColMP	(	const TStr &	GroupBy,
		THashMP< TInt, TIntV > &	Grouping,
		TBool	UsePhysicalIds = true
	)		const

Groups/hashes by a single column with integer values, using OpenMP multi-threading.

Definition at line 1205 of file table.cpp.

                                                                                                            {
  timeval timer0;
  gettimeofday(&timer0, NULL);
  double t1 = timer0.tv_sec + (timer0.tv_usec/1000000.0);
  //printf("X\n");
  TInt IdColIdx = GetColIdx(IdColName);
  TInt GroupByColIdx = GetColIdx(GroupBy);
  if(!UsePhysicalIds && IdColIdx < 0){
        TExcept::Throw("Grouping: Either use physical row ids, or have an id column");
  }
  //double startFn = omp_get_wtime();
  GroupingSanityCheck(GroupBy, atInt);
  TIntPrV Partitions;
  GetPartitionRanges(Partitions, 8*CHUNKS_PER_THREAD);
  TInt PartitionSize = Partitions[0].GetVal2()-Partitions[0].GetVal1()+1;
  //double endPart = omp_get_wtime();
  //printf("Partition time = %f\n", endPart-startFn);

  Grouping.Gen(NumValidRows);
  //double endGen = omp_get_wtime();
  //printf("Gen time = %f\n", endGen-endPart);
  //printf("S\n");
  #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD) //num_threads(1)
  for (int i = 0; i < Partitions.Len(); i++){
    TRowIterator RowI(Partitions[i].GetVal1(), this);
    TRowIterator EndI(Partitions[i].GetVal2(), this);
    while (RowI < EndI) {
      TInt idx = UsePhysicalIds ? RowI.GetRowIdx() : RowI.GetIntAttr(IdColIdx);
      // printf("updating grouping with key = %d, row_id = %d\n", RowI.GetIntAttr(GroupBy).Val, idx.Val);
      UpdateGrouping<TInt>(Grouping, RowI.GetIntAttr(GroupByColIdx), idx);
      RowI++;
    }
  }
  gettimeofday(&timer0, NULL);
  double t2 = timer0.tv_sec + (timer0.tv_usec/1000000.0);
  printf("Grouping time: %f\n", t2 - t1);
  //double endAdd = omp_get_wtime();
  //printf("Add time = %f\n", endAdd-endGen);
}

template<class T >

void TTable::GroupByStrCol ( const TStr &  GroupBy, T &  Grouping, const TIntV &  IndexSet, TBool  All, TBool  UsePhysicalIds = true  ) const

protected

Groups/hashes by a single column with string values. Returns hash table with grouping.

Definition at line 1660 of file table.h.

                                                               {
  TInt IdColIdx = GetColIdx(IdColName);
  if(!UsePhysicalIds && IdColIdx < 0){
        TExcept::Throw("Grouping: Either use physical row ids, or have an id column");
  }
  GroupingSanityCheck(GroupBy, atStr);
  if (All) {
    // Optimize for the common and most expensive case - iterate over all valid rows.
    for (TRowIterator it = BegRI(); it < EndRI(); it++) {
      TInt idx = UsePhysicalIds ? it.GetRowIdx() : it.GetIntAttr(IdColIdx);
      UpdateGrouping<TInt>(Grouping, it.GetStrMapByName(GroupBy), idx);
    }
  } else {
    // Consider only rows in IndexSet.
    for (TInt i = 0; i < IndexSet.Len(); i++) {
      if (IsRowValid(IndexSet[i])) {
        TInt RowIdx = IndexSet[i];
        TInt ColIdx = GetColIdx(GroupBy);
        TInt idx = UsePhysicalIds ? RowIdx : IntCols[IdColIdx][RowIdx];  
        UpdateGrouping<TInt>(Grouping, StrColMaps[ColIdx][RowIdx], idx);
      }
    }
  }
}

void TTable::GroupingSanityCheck ( const TStr &  GroupBy, const TAttrType &  AttrType  ) const

protected

Checks if grouping key exists and matches given attr type.

Definition at line 1195 of file table.cpp.

                                                                                     {
  if (!IsColName(GroupBy)) {
    TExcept::Throw("no such column " + GroupBy);
  }
  if (GetColType(GroupBy) != AttrType) {
    TExcept::Throw(GroupBy + " values are not of expected type");
  }
}

void TTable::IncrementNext ( )

protected

Increments the next vector and set last, NumRows and NumValidRows.

Definition at line 2235 of file table.cpp.

{
        // Advance the Next vector
        NumRows++;
        NumValidRows++;
        if (!Next.Empty()) {
                Next[Next.Len()-1] = NumValidRows-1;
                LastValidRow = NumValidRows-1;
        }
        Next.Add(Last);
}

PTable TTable::InitializeJointTable ( const TTable &  Table )

protected

Initializes an empty table for the join of this table with the given table.

Definition at line 1896 of file table.cpp.

                                                        {
  PTable JointTable = New(Context);
  JointTable->IntCols = TVec<TIntV>(IntCols.Len() + Table.IntCols.Len() + 1);
  JointTable->FltCols = TVec<TFltV>(FltCols.Len() + Table.FltCols.Len());
  JointTable->StrColMaps = TVec<TIntV>(StrColMaps.Len() + Table.StrColMaps.Len());
  for (TInt i = 0; i < Sch.Len(); i++) {
    TStr ColName = GetSchemaColName(i);
    TAttrType ColType = GetSchemaColType(i);
    TStr CName = JointTable->RenumberColName(ColName);
    TPair<TAttrType, TInt> TypeMap = GetColTypeMap(ColName);
    JointTable->AddColType(CName, TypeMap);
    //JointTable->AddLabel(CName, ColName);
    JointTable->AddSchemaCol(CName, ColType);
  }
  for (TInt i = 0; i < Table.Sch.Len(); i++) {
    TStr ColName = Table.GetSchemaColName(i);
    TAttrType ColType = Table.GetSchemaColType(i);
    TStr CName = JointTable->RenumberColName(ColName);
    TPair<TAttrType, TInt> NewDat = Table.GetColTypeMap(ColName);
    Assert(ColType == NewDat.Val1);
    // add offsets
    switch (NewDat.Val1) {
      case atInt:
        NewDat.Val2 += IntCols.Len();
        break;
      case atFlt:
        NewDat.Val2 += FltCols.Len();
        break;
      case atStr:
        NewDat.Val2 += StrColMaps.Len();
        break;
    }
    JointTable->AddColType(CName, NewDat);
    JointTable->AddSchemaCol(CName, ColType);
  }
  TStr IdColName = "_id";
  JointTable->AddColType(IdColName, atInt, IntCols.Len() + Table.IntCols.Len());
  JointTable->AddSchemaCol(IdColName, atInt);
  return JointTable;
}

void TTable::InitIds

(

)

Adds explicit row ids, initialize hash set mapping ids to physical rows.

Definition at line 1863 of file table.cpp.

                     {
  IdColName = "_id";
  //Assert(NumRows == NumValidRows);
  AddIdColumn(IdColName);
}

void TTable::InitRowIdBuckets ( int  NumBuckets )

protected

Initializes the RowIdBuckets vector which will be used for the graph sequence creation.

Definition at line 3515 of file table.cpp.

                                            {
  for (TInt i = 0; i < RowIdBuckets.Len(); i++) {
    RowIdBuckets[i].Clr();
  }
  RowIdBuckets.Clr();

  RowIdBuckets.Gen(NumBuckets);
  for (TInt i = 0; i < NumBuckets; i++) {
    RowIdBuckets[i].Gen(10, 0);
  }
}

PTable TTable::Intersection

(

const TTable &

Table

)

Returns intersection of this table with given Table.

Definition at line 4544 of file table.cpp.

                                               {
  Schema NewSchema;
  THashSet<TInt> Collisions;

  for (TInt c = 0; c < Sch.Len(); c++) {
    if (Sch[c].Val1 != GetIdColName()) {
      NewSchema.Add(TPair<TStr,TAttrType>(Sch[c].Val1, Sch[c].Val2));
    }
  }
  PTable result = TTable::New(NewSchema, Context);

  GetCollidingRows(Table, Collisions);

  // this part should be made faster by adding all the rows in one go
  for (TRowIterator it = Table.BegRI(); it < Table.EndRI(); it++) {
    if (Collisions.IsKey(it.GetRowIdx())) {
      result->AddRow(it);
    }
  }
  result->InitIds();
  return result;
}

PTable TTable::Intersection ( const PTable &  Table )

inline

Definition at line 1413 of file table.h.

{ return Intersection(*Table); };

void TTable::InvalidateAffectedGroupings ( const TStr &  Attr )

protected

Definition at line 1561 of file table.cpp.

                                                        {
        //TODO
}

void TTable::InvalidatePhysicalGroupings ( )

protected

A mapping between grouping statement (group-by attribute names and 'Ordered' flag) to a hash map between group-by keys to the ids of records that share the group-by key. Can be used as hash-index for the table.

Definition at line 1557 of file table.cpp.

                                        {
        //TODO
}

TBool TTable::IsAttr ( const TStr &  Attr )

protected

Checks if Attr is an attribute of this table schema.

Definition at line 4605 of file table.cpp.

                                     {
  return IsColName(Attr);
}

TBool TTable::IsColName ( const TStr &  ColName ) const

inlineprotected

Definition at line 656 of file table.h.

                                             {
    TStr NColName = NormalizeColName(ColName);
    return ColTypeMap.IsKey(NColName);
  }

TBool TTable::IsLastGraphOfSequence

(

)

Checks if the end of the graph sequence is reached.

Definition at line 3663 of file table.cpp.

                                    {
  return CurrBucket >= RowIdBuckets.Len() - 1;
}

PTable TTable::IsNextK	(	const TStr &	OrderCol,
		TInt	K,
		const TStr &	GroupBy,
		const TStr &	RankColName = ""
	)

Distance based filter.

Creates a table T' where the rows are joint rows (T[r1],T[r2]) such that r2 is one of the successive rows to r1 when this table is ordered by OrderCol, and both r1 and r2 have the same value of GroupBy column

Definition at line 3869 of file table.cpp.

                                                                                                 {
  TStrV OrderBy;
  if (GroupBy.Empty()) {
    OrderBy.Add(OrderCol);
  } else {
    OrderBy.Add(GroupBy);
    OrderBy.Add(OrderCol);
  }
  if (RankColName.Empty()) {
    Order(OrderBy);
  } else {
    Order(OrderBy, RankColName, true);
  }
  TAttrType GroupByAttrType = GetColType(GroupBy);
  PTable T = InitializeJointTable(*this);
  for (TRowIterator RI = BegRI(); RI < EndRI(); RI++) {
    TInt Succ = RI.GetRowIdx();
    TBool OutOfGroup = false;
    for (TInt i = 0; i < K; i++) {
      Succ = Next[Succ];
      if (Succ == Last) { break; }
      switch (GroupByAttrType) {
        case atInt:
          if (GetIntVal(GroupBy, Succ) != RI.GetIntAttr(GroupBy)) { OutOfGroup = true; }
          break;
        case atFlt:
          if (GetFltVal(GroupBy, Succ) != RI.GetFltAttr(GroupBy)) { OutOfGroup = true; }
          break;
        case atStr:
          if (GetStrVal(GroupBy, Succ) != RI.GetStrAttr(GroupBy)) { OutOfGroup = true; }
          break;
      }
      if (OutOfGroup) { break; }  // break out of inner for loop
      T->AddJointRow(*this, *this, RI.GetRowIdx(), Succ);
    }
  }
  return T;
}

void TTable::ISort ( TIntV &  V, TInt  StartIdx, TInt  EndIdx, const TVec< TAttrType > &  SortByTypes, const TIntV &  SortByIndices, TBool  Asc = true  )

protected

Performs insertion sort on given vector V.

Definition at line 3076 of file table.cpp.

                                                                                                                                  {
  if (StartIdx < EndIdx) {
    for (TInt i = StartIdx+1; i <= EndIdx; i++) {
      TInt Val = V[i];
      TInt j = i;
      while ((StartIdx < j) && (CompareRows(V[j-1], Val, SortByTypes, SortByIndices, Asc) > 0)) {
        V[j] = V[j-1];
        j--;
      }
      V[j] = Val;
    }
  }
}

void TTable::ISortKeyVal ( TIntV &  Key, TIntV &  Val, TInt  Start, TInt  End  )

staticprotected

Definition at line 5298 of file table.cpp.

                                                                     {
  if (Start < End) {
    for (TInt i = Start+1; i <= End; i++) {
      TInt K = Key[i];
      TInt V = Val[i];
      TInt j = i;
      while ((Start < j) && (CompareKeyVal(Key[j-1], Val[j-1], K, V) > 0)) {
        Key[j] = Key[j-1];
        Val[j] = Val[j-1];
        j--;
      }
      Key[j] = K;
      Val[j] = V;
    }
  }
}

bool TTable::IsRowValid ( TInt  RowIdx ) const

inlineprotected

Checks if RowIdx corresponds to a valid (i.e. not deleted) row.

Definition at line 811 of file table.h.

{ return Next[RowIdx] != Invalid;}

PTable TTable::Join	(	const TStr &	Col1,
		const TTable &	Table,
		const TStr &	Col2
	)

Performs equijoin.

Perform equi-join with given columns - i.e. keep tuple pairs where this->Col1 == Table->Col2 Implementation: Hash-Join - build a hash out of the smaller table hash the larger table and check for collisions

Definition at line 2252 of file table.cpp.

                                                                           {
  // double startFn = omp_get_wtime();
  if (!IsColName(Col1)) {
    TExcept::Throw("no such column " + Col1);
    printf("no such column %s\n", Col1.CStr());
  }
  if (!Table.IsColName(Col2)) {
    TExcept::Throw("no such column " + Col2);
    printf("no such column %s\n", Col2.CStr());
  }
  if (GetColType(Col1) != Table.GetColType(Col2)) {
    TExcept::Throw("Trying to Join on columns of different type");
    printf("Trying to Join on columns of different type\n");
  }
  //printf("passed initial checks\n");
  // initialize result table
  PTable JointTable = InitializeJointTable(Table);
  //printf("initialized joint table\n");
  // hash smaller table (group by column)
  TAttrType ColType = GetColType(Col1);
  TBool ThisIsSmaller = (NumValidRows <= Table.NumValidRows);
  const TTable& TS = ThisIsSmaller ? *this : Table;
  const TTable& TB = ThisIsSmaller ?  Table : *this;
  TStr ColS = ThisIsSmaller ? Col1 : Col2;
  TStr ColB = ThisIsSmaller ? Col2 : Col1;
  TInt ColBId = ThisIsSmaller ? Table.GetColIdx(ColB) : GetColIdx(ColB);
  // double endInit = omp_get_wtime();
  // printf("Init time = %f\n", endInit-startFn);
  // iterate over the rows of the bigger table and check for "collisions" 
  // with the group keys for the small table.
#ifdef GCC_ATOMIC
  if (GetMP()) {
    switch(ColType){
      case atInt:{
        THashMP<TInt, TIntV> T(TS.GetNumValidRows());
        TS.GroupByIntColMP(ColS, T, true);
        // double endGroup = omp_get_wtime();
        // printf("Group time = %f\n", endGroup-endInit);
        
        TIntPrV Partitions;
        TB.GetPartitionRanges(Partitions, omp_get_max_threads()*CHUNKS_PER_THREAD);
        TInt PartitionSize = Partitions[0].GetVal2()-Partitions[0].GetVal1()+1;
        TVec<TIntPrV> JointRowIDSet(Partitions.Len());
        // double endPart = omp_get_wtime();
        // printf("Partition time = %f\n", endPart-endGroup);

        #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD) 
        for (int i = 0; i < Partitions.Len(); i++){
          //double start = omp_get_wtime();
          JointRowIDSet[i].Reserve(PartitionSize);
          TRowIterator RowI(Partitions[i].GetVal1(), &TB);
          TRowIterator EndI(Partitions[i].GetVal2(), &TB);
          while (RowI < EndI) {
            TInt K = RowI.GetIntAttr(ColBId);
            if(T.IsKey(K)){
              TIntV& Group = T.GetDat(K);
              for(TInt j = 0; j < Group.Len(); j++){
                if(ThisIsSmaller){
                  JointRowIDSet[i].Add(TIntPr(Group[j], RowI.GetRowIdx()));
                } else{
                  JointRowIDSet[i].Add(TIntPr(RowI.GetRowIdx(), Group[j]));
                }
              }
            }
            RowI++;
          }
          //double end = omp_get_wtime();
          //printf("END: Thread %d: i = %d, start = %d, end = %d, num = %d, time = %f\n", omp_get_thread_num(), i,
          //    Partitions[i].GetVal1().Val, Partitions[i].GetVal2().Val, JointRowIDSet[i].Len(), end-start);
        }
        // double endJoin = omp_get_wtime();
        // printf("Iterate time = %f\n", endJoin-endPart);
        JointTable->AddNJointRowsMP(*this, Table, JointRowIDSet);      
        // double endAdd = omp_get_wtime();
        // printf("Add time = %f\n", endAdd-endJoin);
        break;
      }
      case atFlt:{
        THashMP<TFlt, TIntV> T(TS.GetNumValidRows());
        TS.GroupByFltCol(ColS, T, TIntV(), true);

        TIntPrV Partitions;
        TB.GetPartitionRanges(Partitions, omp_get_max_threads()*CHUNKS_PER_THREAD);
        TInt PartitionSize = Partitions[0].GetVal2()-Partitions[0].GetVal1()+1;
        TVec<TIntPrV> JointRowIDSet(Partitions.Len());

        #pragma omp parallel for schedule(dynamic) 
        for (int i = 0; i < Partitions.Len(); i++){
          JointRowIDSet[i].Reserve(PartitionSize);
          TRowIterator RowI(Partitions[i].GetVal1(), &TB);
          TRowIterator EndI(Partitions[i].GetVal2(), &TB);
          while (RowI < EndI) {
            TFlt K = RowI.GetFltAttr(ColBId);
            if(T.IsKey(K)){
              TIntV& Group = T.GetDat(K);
              for(TInt j = 0; j < Group.Len(); j++){
                if(ThisIsSmaller){
                  JointRowIDSet[i].Add(TIntPr(Group[j], RowI.GetRowIdx()));
                } else{
                  JointRowIDSet[i].Add(TIntPr(RowI.GetRowIdx(), Group[j]));
                }
              }
            }
            RowI++;
          }
        }
        JointTable->AddNJointRowsMP(*this, Table, JointRowIDSet);
        break;
      }
      case atStr:{
        THashMP<TInt, TIntV> T(TS.GetNumValidRows());
        TS.GroupByStrCol(ColS, T, TIntV(), true);

        TIntPrV Partitions;
        TB.GetPartitionRanges(Partitions, omp_get_max_threads()*CHUNKS_PER_THREAD);
        TInt PartitionSize = Partitions[0].GetVal2()-Partitions[0].GetVal1()+1;
        TVec<TIntPrV> JointRowIDSet(Partitions.Len());

        #pragma omp parallel for schedule(dynamic) 
        for (int i = 0; i < Partitions.Len(); i++){
          JointRowIDSet[i].Reserve(PartitionSize);
          TRowIterator RowI(Partitions[i].GetVal1(), &TB);
          TRowIterator EndI(Partitions[i].GetVal2(), &TB);
          while (RowI < EndI) {
            TInt K = RowI.GetStrMapById(ColBId);
            if(T.IsKey(K)){
              TIntV& Group = T.GetDat(K);
              for(TInt j = 0; j < Group.Len(); j++){
                if(ThisIsSmaller){
                  JointRowIDSet[i].Add(TIntPr(Group[j], RowI.GetRowIdx()));
                } else{
                  JointRowIDSet[i].Add(TIntPr(RowI.GetRowIdx(), Group[j]));
                }
              }
            }
            RowI++;
          }
        }
        JointTable->AddNJointRowsMP(*this, Table, JointRowIDSet);
      }
      break;
    }
  } else {
#endif // GCC_ATOMIC
    switch (ColType) {
      case atInt:{
        TIntIntVH T;
        TS.GroupByIntCol(ColS, T, TIntV(), true);
        for (TRowIterator RowI = TB.BegRI(); RowI < TB.EndRI(); RowI++) {
          TInt K = RowI.GetIntAttr(ColBId);
          if (T.IsKey(K)) {
            TIntV& Group = T.GetDat(K);
            for (TInt i = 0; i < Group.Len(); i++) {
              if (ThisIsSmaller) {
                JointTable->AddJointRow(*this, Table, Group[i], RowI.GetRowIdx());
              } else {
                JointTable->AddJointRow(*this, Table, RowI.GetRowIdx(), Group[i]);
              }
            }
          }
        }
        break;
      }
      case atFlt:{
        THash<TFlt, TIntV> T;
        TS.GroupByFltCol(ColS, T, TIntV(), true);
        for (TRowIterator RowI = TB.BegRI(); RowI < TB.EndRI(); RowI++) {
          TFlt K = RowI.GetFltAttr(ColBId);
          if (T.IsKey(K)) {
            TIntV& Group = T.GetDat(K);
            for (TInt i = 0; i < Group.Len(); i++) {
              if (ThisIsSmaller) {
                JointTable->AddJointRow(*this, Table, Group[i], RowI.GetRowIdx());
              } else {
                JointTable->AddJointRow(*this, Table, RowI.GetRowIdx(), Group[i]);
              }
            }
          }
        }
        break;
      }
      case atStr:{
        TIntIntVH T;
        TS.GroupByStrCol(ColS, T, TIntV(), true);
        for (TRowIterator RowI = TB.BegRI(); RowI < TB.EndRI(); RowI++) {
          TInt K = RowI.GetStrMapById(ColBId);
          if (T.IsKey(K)) {
            TIntV& Group = T.GetDat(K);
            for (TInt i = 0; i < Group.Len(); i++) {
              if (ThisIsSmaller) {
                JointTable->AddJointRow(*this, Table, Group[i], RowI.GetRowIdx());
              } else {
                JointTable->AddJointRow(*this, Table, RowI.GetRowIdx(), Group[i]);
              }
            }
          }
        }
      }
      break;
    }
#ifdef GCC_ATOMIC
  }
#endif
  return JointTable; 
}

PTable TTable::Join ( const TStr &  Col1, const PTable &  Table, const TStr &  Col2  )

inline

Definition at line 1351 of file table.h.

                                                                       {
    return Join(Col1, *Table, Col2);
  }

void TTable::KeepSortedRows ( const TIntV &  KeepV )

protected

Removes all rows that are not mentioned in the SORTED vector KeepV.

Definition at line 1132 of file table.cpp.

                                              {
  TIntIntH KeepH(KeepV.Len());
  for (TInt i = 0; i < KeepV.Len(); i++) {
    KeepH.AddKey(KeepV[i]);
  }

  TRowIteratorWithRemove RowI = BegRIWR();
  TInt KeepSize = 0;
  while (RowI.GetNextRowIdx() != Last) {
    if (KeepSize < KeepV.Len()) {
      if (KeepH.IsKey(RowI.GetNextRowIdx())) {
        KeepSize++;
        RowI++;
      } else {
        RowI.RemoveNext();
      }
    } else {
      // Covered all of KeepV. Remove the rest of the rows.
      // Current RowI.CurrRowIdx is the last element of KeepV.
      RowI.RemoveNext();
    }
  }
  LastValidRow = KeepV[KeepV.Len()-1];
}

static PTable TTable::Load ( TSIn &  SIn, TTableContext *  Context  )

inlinestatic

Loads table from a binary format.

TTableContext Context must be provided as a parameter and loaded separately from a table load as it can be shared among multiple tables. Context can be loaded either before and after the table load, but must be available for operations that require string values (as opposed to string references).

Definition at line 970 of file table.h.

{ return new TTable(SIn, Context);}

PTable TTable::LoadSS ( const Schema &  S, const TStr &  InFNm, TTableContext *  Context, const char &  Separator = '\t', TBool  HasTitleLine = false  )

static

Loads table from spread sheet (TSV, CSV, etc). Note: HasTitleLine = true is not supported. Please comment title lines instead.

Definition at line 775 of file table.cpp.

                                            {
  return LoadSS(S, InFNm, Context, TIntV(), Separator, HasTitleLine);
}

PTable TTable::LoadSS ( const Schema &  S, const TStr &  InFNm, TTableContext *  Context, const TIntV &  RelevantCols, const char &  Separator = '\t', TBool  HasTitleLine = false  )

static

Loads table from spread sheet - but only load the columns specified by RelevantCols. Note: HasTitleLine = true is not supported. Please comment title lines instead.

Definition at line 737 of file table.cpp.

                                                                       {
  TVec<uint64> IntGroupByCols;
  bool NoStringCols = true;

  // find the schema for the new table which contains only relevant columns
  Schema SR;
  if (RelevantCols.Len() == 0) {
    SR = S;
  } else {
    for (int i = 0; i < RelevantCols.Len(); i++) {
      SR.Add(S[RelevantCols[i]]);
    }
  }
  PTable T = New(SR, Context);

  // find col types and check for string cols
  for (int i = 0; i < SR.Len(); i++) {
    if (T->GetSchemaColType(i) == atStr) {
      NoStringCols = false;
      break;
    }
  }

  if (GetMP() && NoStringCols) {
    // Right now, can load in parallel only in Linux (for mmap) and if
    // there are no string columns
#ifdef GLib_LINUX
    LoadSSPar(T, S, InFNm, RelevantCols, Separator, HasTitleLine);
#else
    LoadSSSeq(T, S, InFNm, RelevantCols, Separator, HasTitleLine);
#endif
  } else {
    LoadSSSeq(T, S, InFNm, RelevantCols, Separator, HasTitleLine);
  }
  return T;
}

void TTable::LoadSSPar ( PTable &  NewTable, const Schema &  S, const TStr &  InFNm, const TIntV &  RelevantCols, const char &  Separator, TBool  HasTitleLine  )

staticprotected

Parallelly loads data from input file at InFNm into NewTable. Only work when NewTable has no string columns.

Definition at line 487 of file table.cpp.

                                                                   {
  // preloaded necessary variables
  TInt RowLen = T->Sch.Len();
  TVec<TAttrType> ColTypes = TVec<TAttrType>(RowLen);
  for (TInt i = 0; i < RowLen; i++) {
    ColTypes[i] = T->GetSchemaColType(i);
  }

  TSsParserMP Ss(InFNm, Separator);
  Ss.SkipCommentLines();

  // if title line (i.e. names of the columns) is included as first row in the
  // input file - use it to validate schema
  if (HasTitleLine) {
    Ss.Next();
    if (S.Len() != Ss.GetFlds()) {
      printf("%s\n", Ss[0]); TExcept::Throw("Table Schema Mismatch!");
    }
    for (TInt i = 0; i < Ss.GetFlds(); i++) {
      // remove carriage return char
      TInt L = strlen(Ss[i]);
      if (Ss[i][L-1] < ' ') { Ss[i][L-1] = 0; }
      if (NormalizeColName(S[i].Val1) != NormalizeColName(Ss[i])) { TExcept::Throw("Table Schema Mismatch!"); }
    }
  }

  // Divide remaining part of stream into equal sized chunks
  // Find starting position in stream for each thread
  int64 Cnt = 0;
  uint64 Pos = Ss.GetStreamPos();
  uint64 Len = Ss.GetStreamLen();
  uint64 Rem = Len - Pos;
  int NumThreads = omp_get_max_threads();

  uint64 Delta = Rem / NumThreads;
  if (Delta < 1) Delta = 1;

  TVec<uint64> StartIntV(NumThreads);
  TVec<uint64> LineCountV(NumThreads);
  TVec<uint64> PrefixSumV(NumThreads);

  StartIntV[0] = Pos;
  for (int i = 1; i < NumThreads; i++) {
    StartIntV[i] = StartIntV[i-1] + Delta;
  }
  StartIntV.Add(Len);

  // Find number of lines handled by each thread
  omp_set_num_threads(NumThreads);
  #pragma omp parallel for schedule(dynamic) reduction(+:Cnt)
  for (int i = 0; i < NumThreads; i++) {
    LineCountV[i] = Ss.CountNewLinesInRange(StartIntV[i], StartIntV[i+1]);
    Cnt += LineCountV[i];
  }

  // Calculate row index offsets for each thread
  PrefixSumV[0] = 0;
  for (int i = 1; i < NumThreads; i++) {
    PrefixSumV[i] = PrefixSumV[i-1] + LineCountV[i-1];
  }
  Ss.SetStreamPos(Pos);

  // allocate memory for columns
  TInt IntColIdx = 0;
  TInt FltColIdx = 0;
  for (TInt i = 0; i < RowLen; i++) {
    switch (ColTypes[i]) {
      case atInt:
        T->IntCols[IntColIdx].Gen(Cnt);
        IntColIdx++;
        break;
      case atFlt:
        T->FltCols[FltColIdx].Gen(Cnt);
        FltColIdx++;
        break;
      case atStr:
        break;
    }
  }

  Cnt = 0;
  omp_set_num_threads(NumThreads);
  #pragma omp parallel for schedule(dynamic) reduction(+:Cnt)
  for (int i = 0; i < NumThreads; i++) {
    // calculate beginning of each line handled by thread
    TVec<uint64> LineStartPosV = Ss.GetStartPosV(StartIntV[i], StartIntV[i+1]);

    // parse line and fill rows
    for (uint64 k = 0; k < (uint64) LineStartPosV.Len(); k++) {
      TVec<char*> FieldsV;
      Ss.NextFromIndex(LineStartPosV[k], FieldsV);
      if (FieldsV.Len() != S.Len()) {
        TExcept::Throw("Error reading tsv file");
      }
      TInt IntColIdx = 0;
      TInt FltColIdx = 0;
      TInt RowIdx = PrefixSumV[i] + k;

      for (TInt j = 0; j < RowLen; j++) {
        switch (ColTypes[j]) {
          case atInt:
            if (RelevantCols.Len() == 0) {
              T->IntCols[IntColIdx][RowIdx] = \
                (Ss.GetIntFromFldV(FieldsV, j));
            } else {
              T->IntCols[IntColIdx][RowIdx] = \
                (Ss.GetIntFromFldV(FieldsV, RelevantCols[j]));
            }
            IntColIdx++;
            break;
          case atFlt:
            if (RelevantCols.Len() == 0) {
              T->FltCols[FltColIdx][RowIdx] = \
                (Ss.GetFltFromFldV(FieldsV, j));
            } else {
              T->FltCols[FltColIdx][RowIdx] = \
                (Ss.GetFltFromFldV(FieldsV, RelevantCols[j]));
            }
            FltColIdx++;
            break;
          case atStr:
            TExcept::Throw("TTable::LoadSS:: Str Col found\n");
            break;
        }
      }
      Cnt++;
    }
  }

  // set number of rows and "Next" vector
  T->NumRows = Cnt;
  T->NumValidRows = T->NumRows;

  T->Next.Clr();
  T->Next.Gen(Cnt);

  omp_set_num_threads(NumThreads);
  #pragma omp parallel for schedule(dynamic, 10000)
  for (int64 i = 0; i < Cnt-1; i++) {
    T->Next[i] = i+1;
  }
  T->IsNextDirty = 0;
  T->Next[Cnt-1] = Last;
  T->LastValidRow = T->NumRows - 1;

  T->IdColName = "_id";
  TInt IdCol = T->IntCols.Add();
  T->IntCols[IdCol].Gen(Cnt);

  // initialize ID column
  omp_set_num_threads(NumThreads);
  #pragma omp parallel for schedule(dynamic, 10000)
  for (int64 i = 0; i < Cnt; i++) {
    T->IntCols[IdCol][i] = i;
  }

  T->AddSchemaCol(T->IdColName, atInt);
  T->AddColType(T->IdColName, atInt, T->IntCols.Len()-1);
}

void TTable::LoadSSSeq ( PTable &  NewTable, const Schema &  S, const TStr &  InFNm, const TIntV &  RelevantCols, const char &  Separator, TBool  HasTitleLine  )

staticprotected

Sequentially loads data from input file at InFNm into NewTable.

Definition at line 649 of file table.cpp.

                                            {
  // preloaded necessary variables
  int RowLen = T->Sch.Len();
  TVec<TAttrType> ColTypes = TVec<TAttrType>(RowLen);
  for (int i = 0; i < RowLen; i++) {
    ColTypes[i] = T->GetSchemaColType(i);
  }

  // Sequential load
  TSsParser Ss(InFNm, Separator);
  // if title line (i.e. names of the columns) is included as first row in the
  // input file - use it to validate schema
  if (HasTitleLine) {
    Ss.Next();
    if (S.Len() != Ss.GetFlds()) {
      printf("%s\n", Ss[0]); TExcept::Throw("Table Schema Mismatch!");
    }
    for (int i = 0; i < Ss.GetFlds(); i++) {
      // remove carriage return char
      int L = strlen(Ss[i]);
      if (Ss[i][L-1] < ' ') { Ss[i][L-1] = 0; }
      if (NormalizeColName(S[i].Val1) != NormalizeColName(Ss[i])) { TExcept::Throw("Table Schema Mismatch!"); }
    }
  }

  // populate table columns
  //printf("starting to populate table\n");
  uint64 Cnt = 0;
  while (Ss.Next()) {
    int IntColIdx = 0;
    int FltColIdx = 0;
    int StrColIdx = 0;
    Assert(Ss.GetFlds() == S.Len()); // compiled only in debug
    if (Ss.GetFlds() != S.Len()) {
      printf("%s\n", Ss[S.Len()]); TExcept::Throw("Error reading tsv file");
    }
    for (int i = 0; i < RowLen; i++) {
      switch (ColTypes[i]) {
        case atInt:
          if (RelevantCols.Len() == 0) {
            T->IntCols[IntColIdx].Add(Ss.GetInt(i));
          } else {
            T->IntCols[IntColIdx].Add(Ss.GetInt(RelevantCols[i]));
          }
          IntColIdx++;
          break;
        case atFlt:
          if (RelevantCols.Len() == 0) {
            T->FltCols[FltColIdx].Add(Ss.GetFlt(i));
          } else {
            T->FltCols[FltColIdx].Add(Ss.GetFlt(RelevantCols[i]));
          }
          FltColIdx++;
          break;
        case atStr:
          int ColIdx;
          if (RelevantCols.Len() == 0) {
            ColIdx = i;
          } else {
            ColIdx = RelevantCols[i];
          }
          TStr Sval = TStr(Ss[ColIdx]);
          T->AddStrVal(StrColIdx, Sval);
          StrColIdx++;
          break;
      }
    }
    Cnt += 1;
  }
  //printf("finished populating table\n");
  // set number of rows and "Next" vector
  T->NumRows = static_cast<int>(Cnt);
  T->NumValidRows = T->NumRows;

  T->Next.Clr();
  T->Next.Gen(static_cast<int>(Cnt));
  for (uint64 i = 0; i < Cnt-1; i++) {
    T->Next[static_cast<int>(i)] = static_cast<int>(i+1);
  }
  T->IsNextDirty = 0;
  T->Next[static_cast<int>(Cnt-1)] = Last;
  T->LastValidRow = T->NumRows - 1;

  T->InitIds();
}

void TTable::Merge ( TIntV &  V, TInt  Idx1, TInt  Idx2, TInt  Idx3, const TVec< TAttrType > &  SortByTypes, const TIntV &  SortByIndices, TBool  Asc = true  )

protected

Helper function for parallel QSort.

Definition at line 3158 of file table.cpp.

                                                                                                                                       {
  TInt i = Idx1, j = Idx2;
  TIntV SortedV;
  while  (i < Idx2 && j < Idx3) {
    if (CompareRows(V[i], V[j], SortByTypes, SortByIndices, Asc) <= 0) {
      SortedV.Add(V[i]);
      i++;
    }
    else {
      SortedV.Add(V[j]);
      j++;
    }
  }
  while (i < Idx2) {
    SortedV.Add(V[i]);
    i++;
  }
  while (j < Idx3) {
    SortedV.Add(V[j]);
    j++;
  }

  for (TInt sz = 0; sz < Idx3 - Idx1; sz++) {
    V[Idx1 + sz] = SortedV[sz];
  }
}

PTable TTable::Minus

(

TTable &

Table

)

Returns table with rows that are present in this table but not in given Table.

Definition at line 4569 of file table.cpp.

                                  {
  Schema NewSchema;
  THashSet<TInt> Collisions;

  for (TInt c = 0; c < Sch.Len(); c++) {
    if (Sch[c].Val1 != GetIdColName()) {
      NewSchema.Add(TPair<TStr,TAttrType>(Sch[c].Val1, Sch[c].Val2));
    }
  }
  PTable result = TTable::New(NewSchema, Context);

  Table.GetCollidingRows(*this, Collisions);

  // this part should be made faster by adding all the rows in one go
  for (TRowIterator it = BegRI(); it < EndRI(); it++) {
    if (!Collisions.IsKey(it.GetRowIdx())) {
      result->AddRow(it);
    }
  }
  result->InitIds();
  return result;
}

PTable TTable::Minus ( const PTable &  Table )

inline

Definition at line 1416 of file table.h.

{ return Minus(*Table); };

static PTable TTable::New ( )

inlinestatic

Definition at line 931 of file table.h.

{ return new TTable(); }

static PTable TTable::New ( TTableContext *  Context )

inlinestatic

Definition at line 932 of file table.h.

{ return new TTable(Context); }

static PTable TTable::New ( const Schema &  S, TTableContext *  Context  )

inlinestatic

Definition at line 933 of file table.h.

                                                             {
    return new TTable(S, Context);
  }

static PTable TTable::New ( const THash< TInt, TInt > &  H, const TStr &  Col1, const TStr &  Col2, TTableContext *  Context, const TBool  IsStrKeys = false  )

inlinestatic

Returns pointer to a table constructed from given int->int hash.

Definition at line 937 of file table.h.

                                                                            {
    return new TTable(H, Col1, Col2, Context, IsStrKeys);
  }

static PTable TTable::New ( const THash< TInt, TFlt > &  H, const TStr &  Col1, const TStr &  Col2, TTableContext *  Context, const TBool  IsStrKeys = false  )

inlinestatic

Returns pointer to a table constructed from given int->float hash.

Definition at line 942 of file table.h.

                                                                            {
    return new TTable(H, Col1, Col2, Context, IsStrKeys);
  }

static PTable TTable::New ( const PTable  Table )

inlinestatic

Returns pointer to a new table created from given Table.

Definition at line 947 of file table.h.

{ return new TTable(*Table); }

PNEANet TTable::NextGraphIterator

(

)

Calls to this must be preceded by a call to one of the above ToGraph*Iterator functions.

Definition at line 3659 of file table.cpp.

                                  {
  return GetNextGraphFromSequence();
}

static TStr TTable::NormalizeColName ( const TStr &  ColName )

inlinestatic

Adds suffix to column name if it doesn't exist.

Definition at line 540 of file table.h.

                                                    {
    TStr Result = ColName;
    int RLen = Result.Len();
    if (RLen == 0) { return Result; }
    if (Result.GetCh(0) == '_') { return Result; }
    if (RLen >= 2  &&  Result.GetCh(RLen-2) == '-') { return Result; }
    return Result + "-1";
  }

static TStrV TTable::NormalizeColNameV ( const TStrV &  Cols )

inlinestatic

Adds suffix to column name if it doesn't exist.

Definition at line 549 of file table.h.

                                                    {
    TStrV NCols;
    for (TInt i = 0; i < Cols.Len(); i++) { NCols.Add(NormalizeColName(Cols[i])); }
    return NCols;
  }

void TTable::Order	(	const TStrV &	OrderBy,
		TStr	OrderColName = "",
		TBool	ResetRankByMSC = false,
		TBool	Asc = true
	)

Orders the rows according to the values in columns of OrderBy (in descending lexicographic order).

Definition at line 3220 of file table.cpp.

                                                                                           {
  // get a vector of all valid row indices
  TIntV ValidRows = TIntV(NumValidRows);
  if (NumRows == NumValidRows) {
    for (TInt i = 0; i < NumValidRows; i++) {
      ValidRows[i] = i;
    }
  } else {
    TInt i = 0;
    for (TRowIterator RI = BegRI(); RI < EndRI(); RI++) {
      ValidRows[i] = RI.GetRowIdx();
      i++;
    }
  }
  TVec<TAttrType> OrderByTypes(OrderBy.Len());
  TIntV OrderByIndices(OrderBy.Len());
  for (TInt i = 0; i < OrderBy.Len(); i++) {
    OrderByTypes[i] = GetColType(OrderBy[i]);
    OrderByIndices[i] = GetColIdx(OrderBy[i]);
  }

  // sort that vector according to the attributes given in "OrderBy" in lexicographic order
#ifdef USE_OPENMP
  if (GetMP()) {
    QSortPar(ValidRows, OrderByTypes, OrderByIndices, Asc);
  } else {
#endif
    QSort(ValidRows, 0, NumValidRows-1, OrderByTypes, OrderByIndices, Asc);
#ifdef USE_OPENMP
  }
#endif

  // rewire Next vector
  IsNextDirty = 1;
  if (NumValidRows > 0) {
    FirstValidRow = ValidRows[0];
  } else {
    FirstValidRow = Last;
  }
  for (TInt i = 0; i < NumValidRows-1; i++) {
    Next[ValidRows[i]] = ValidRows[i+1];
  }
  if (NumValidRows > 0) {
    Next[ValidRows[NumValidRows-1]] = Last;
    LastValidRow = ValidRows[NumValidRows-1];
  } else {
    LastValidRow = Last;
  }

  // add rank column
  if (!OrderColName.Empty()) {
    TIntV RankCol = TIntV(NumRows);
    for (TInt i = 0; i < NumValidRows; i++) {
      RankCol[ValidRows[i]] = i;
    }
    if (ResetRankByMSC) {
      for (TInt i = 1; i < NumValidRows; i++) {
        TStr GroupName = OrderBy[0];
        if (GetStrVal(GroupName, ValidRows[i]) != GetStrVal(GroupName, ValidRows[i-1])) { 
          RankCol[ValidRows[i]] = 0;
        } else {
          RankCol[ValidRows[i]] = RankCol[ValidRows[i-1]] + 1;
        }
      }
    }
    IntCols.Add(RankCol);
    AddSchemaCol(OrderColName, atInt);
    AddColType(OrderColName, atInt, IntCols.Len()-1);
  }
}

TInt TTable::Partition ( TIntV &  V, TInt  StartIdx, TInt  EndIdx, const TVec< TAttrType > &  SortByTypes, const TIntV &  SortByIndices, TBool  Asc  )

protected

Partitions vector for QSort.

Definition at line 3106 of file table.cpp.

                                                                                                                                      {

  // test if the elements are already sorted
  TInt j;
  for (j = StartIdx; j < EndIdx; j++) {
    if (CompareRows(V[j], V[j+1], SortByTypes, SortByIndices, Asc) > 0) {
      break;
    }
  }
  if (j >= EndIdx) {
    return EndIdx+1;
  }

  TInt PivotIdx = GetPivot(V, StartIdx, EndIdx, SortByTypes, SortByIndices, Asc);
  TInt Pivot = V[PivotIdx];
  V.Swap(PivotIdx, EndIdx);
  TInt StoreIdx = StartIdx;
  for (TInt i = StartIdx; i < EndIdx; i++) {
    if (CompareRows(V[i], Pivot, SortByTypes, SortByIndices, Asc) <= 0) {
      V.Swap(i, StoreIdx);
      StoreIdx++;
    }
  }
  // move pivot value to its place
  V.Swap(StoreIdx, EndIdx);
  return StoreIdx;
}

TInt TTable::PartitionKeyVal ( TIntV &  Key, TIntV &  Val, TInt  Start, TInt  End  )

staticprotected

Definition at line 5332 of file table.cpp.

                                                                         {
  TInt Pivot = GetPivotKeyVal(Key, Val, Start, End);
  //printf("Pivot=%d\n", Pivot.Val);
  TInt PivotKey = Key[Pivot];
  TInt PivotVal = Val[Pivot];
  Key.Swap(Pivot, End);
  Val.Swap(Pivot, End);
  TInt StoreIdx = Start;
  for (TInt i = Start; i < End; i++) {
    //printf("%d %d %d %d\n", Key[i].Val, Val[i].Val, PivotKey.Val, PivotVal.Val);
    if (CompareKeyVal(Key[i], Val[i], PivotKey, PivotVal) <= 0) {
      Key.Swap(i, StoreIdx);
      Val.Swap(i, StoreIdx);
      StoreIdx++;
    }
  }
  //printf("StoreIdx=%d\n", StoreIdx.Val);
  // move pivot value to its place
  Key.Swap(StoreIdx, End);
  Val.Swap(StoreIdx, End);
  return StoreIdx;
}

void TTable::PrintContextSize

(

)

Definition at line 3937 of file table.cpp.

                             {
        printf("Number of strings in pool: ");
        printf("%d\n", Context->StringVals.Len());
        printf("Number of entries in hash table: ");
        printf("%d\n", Context->StringVals.Reserved());
        TSize MemUsed = GetContextMemUsedKB();
        printf("Approximate memory used for Context: %lu KB\n", MemUsed);
}

void TTable::PrintGrouping ( const THash< TGroupKey, TIntV > &  Grouping ) const

protected

Definition at line 1768 of file table.cpp.

                                                                      {
        for(THash<TGroupKey, TIntV>::TIter it = Mapping.BegI(); it < Mapping.EndI(); it++){
                TGroupKey gk = it.GetKey();
                TIntV ik = gk.Val1;
                TFltV fk = gk.Val2;
                for(int i = 0; i < ik.Len(); i++){ printf("%d ",ik[i].Val);} 
                for(int i = 0; i < fk.Len(); i++){ printf("%f ",fk[i].Val);} 
                printf("-->");
                TIntV v = it.GetDat();
                for(int i = 0; i < v.Len(); i++){ printf("%d ",v[i].Val);} 
                printf("\n");
        }
}

void TTable::PrintSize

(

)

Definition at line 3908 of file table.cpp.

                      {
        printf("Total number of rows: %d\n", NumRows.Val);
        printf("Number of valid rows: %d\n", NumValidRows.Val);
        printf("Number of Int columns: %d\n", IntCols.Len());
        printf("Number of Flt columns: %d\n", FltCols.Len());
        printf("Number of Str columns: %d\n", StrColMaps.Len());
        TSize MemUsed = GetMemUsedKB();
        printf("Approximated size is %lu KB\n", MemUsed);
}

PTable TTable::Project

(

const TStrV &

ProjectCols

)

Returns table with only the columns in ProjectCols.

Definition at line 4592 of file table.cpp.

                                               {
  Schema NewSchema;
  for (TInt c = 0; c < ProjectCols.Len(); c++) {
    if (!IsColName(ProjectCols[c])) { TExcept::Throw("no such column " + ProjectCols[c]); }
    NewSchema.Add(TPair<TStr,TAttrType>(ProjectCols[c], GetColType(ProjectCols[c])));
  }

  PTable result = TTable::New(NewSchema, Context);
  result->AddTable(*this);
  result->InitIds();
  return result;
}

void TTable::ProjectInPlace

(

const TStrV &

ProjectCols

)

Keeps only the columns specified in ProjectCols.

Definition at line 5216 of file table.cpp.

                                                    {
  TStrV NProjectCols = NormalizeColNameV(ProjectCols);
  for (TInt c = 0; c < NProjectCols.Len(); c++) {
    if (!IsColName(NProjectCols[c])) { TExcept::Throw("no such column " + NProjectCols[c]); }
  }
  THashSet<TStr> ProjectColsSet = THashSet<TStr>(NProjectCols);
  // Delete the column vectors
  for (TInt i = Sch.Len() - 1; i >= 0; i--) {
    TStr ColName = GetSchemaColName(i);
    if (ProjectColsSet.IsKey(ColName) || ColName == IdColName) { continue; }
    TAttrType ColType = GetSchemaColType(i);
    TInt ColId = GetColIdx(ColName);
    switch (ColType) {
      case atInt:
        IntCols.Del(ColId);
        break;
      case atFlt:
        FltCols.Del(ColId);
        break;
      case atStr:
        StrColMaps.Del(ColId);
        break;
    }
  }

  // Rebuild the ColTypeMap with new indexes of the column vectors
  TInt IntColCnt = 0;
  TInt FltColCnt = 0;
  TInt StrColCnt = 0;
  ColTypeMap.Clr();
  for (TInt i = 0; i < Sch.Len(); i++) {
    TStr ColName = GetSchemaColName(i);
    if (!ProjectColsSet.IsKey(ColName) && ColName != IdColName) { continue; }
    TAttrType ColType = GetSchemaColType(i);
    switch (ColType) {
      case atInt:
        AddColType(ColName, atInt, IntColCnt);
        IntColCnt++;
        break;
      case atFlt:
        AddColType(ColName, atFlt, FltColCnt);
        FltColCnt++;
        break;
      case atStr:
        AddColType(ColName, atStr, StrColCnt);
        StrColCnt++;
        break;
    }
  }

  // Update schema
  for (TInt i = Sch.Len() - 1; i >= 0; i--) {
    TStr ColName = GetSchemaColName(i);
    if (ProjectColsSet.IsKey(ColName) || ColName == IdColName) { continue; }
    Sch.Del(i);
  }
}

void TTable::QSort ( TIntV &  V, TInt  StartIdx, TInt  EndIdx, const TVec< TAttrType > &  SortByTypes, const TIntV &  SortByIndices, TBool  Asc = true  )

protected

Performs QSort on given vector V.

Definition at line 3134 of file table.cpp.

                                                                                                                                  {
  if (StartIdx < EndIdx) {
    if (EndIdx - StartIdx < 20) {
      ISort(V, StartIdx, EndIdx, SortByTypes, SortByIndices, Asc);
    } else {
      TInt Pivot = Partition(V, StartIdx, EndIdx, SortByTypes, SortByIndices, Asc);
      if (Pivot > EndIdx) {
        return;
      }
      // Everything <= Pivot will be in StartIdx, Pivot-1. Shrink this
      // range to ignore elements equal to the pivot in the first
      // recursive call, to optimize for the case when a lot of
      // rows are equal.
      int Ub = Pivot - 1;
      while (Ub >= StartIdx && CompareRows(
        V[Ub], V[Pivot], SortByTypes, SortByIndices, Asc) == 0) {
        Ub -= 1;
      }
      QSort(V, StartIdx, Ub, SortByTypes, SortByIndices, Asc);
      QSort(V, Pivot+1, EndIdx, SortByTypes, SortByIndices, Asc);
    }
  }
}

void TTable::QSortKeyVal ( TIntV &  Key, TIntV &  Val, TInt  Start, TInt  End  )

staticprotected

Definition at line 5355 of file table.cpp.

                                                                     {
  //printf("Thread=%d, Start=%d, End=%d\n", omp_get_thread_num(), Start.Val, End.Val);
  TInt L = End-Start;
  if (L <= 0) { return; }
  if (CheckSortedKeyVal(Key, Val, Start, End) == 0) { return; }
  
  if (L <= 20) { ISortKeyVal(Key, Val, Start, End); }
  else {
    TInt Pivot = PartitionKeyVal(Key, Val, Start, End);
    
    if (Pivot > End) { return; }
    if (L <= 500000) {
      QSortKeyVal(Key, Val, Start, Pivot-1);
      QSortKeyVal(Key, Val, Pivot+1, End);
    } else {
#ifdef USE_OPENMP
#ifndef GLib_WIN32
      #pragma omp task untied shared(Key, Val)
#endif
#endif
      { QSortKeyVal(Key, Val, Start, Pivot-1); }
        
#ifdef USE_OPENMP
#ifndef GLib_WIN32
      #pragma omp task untied shared(Key, Val)
#endif
#endif
      { QSortKeyVal(Key, Val, Pivot+1, End); }
    }
  }
}

void TTable::QSortPar ( TIntV &  V, const TVec< TAttrType > &  SortByTypes, const TIntV &  SortByIndices, TBool  Asc = true  )

protected

Performs QSort in parallel on given vector V.

Definition at line 3186 of file table.cpp.

                                                                                                         {
  TInt NumThreads = 8; // Setting this to 8 because that results in the fastest sorting on Madmax.
  TInt Sz = V.Len();
  TIntV IndV, NextV;
  for (TInt i = 0; i < NumThreads; i++) {
    IndV.Add(i * (Sz / NumThreads));
  }
  IndV.Add(Sz);

  omp_set_num_threads(NumThreads);
  #pragma omp parallel for
  for (int i = 0; i < NumThreads; i++) {
    QSort(V, IndV[i], IndV[i+1] - 1, SortByTypes, SortByIndices, Asc);
  }

  while (NumThreads > 1) {
    omp_set_num_threads(NumThreads / 2);
    #pragma omp parallel for
    for (int i = 0; i < NumThreads; i += 2) {
      Merge(V, IndV[i], IndV[i+1], IndV[i+2], SortByTypes, SortByIndices, Asc);
    }

    NextV.Clr();
    for (TInt i = 0; i < NumThreads; i+=2) {
      NextV.Add(IndV[i]);
    }
    NextV.Add(Sz);
    IndV = NextV;

    NumThreads = NumThreads / 2;
  }
}

void TTable::ReadFltCol	(	const TStr &	ColName,
		TFltV &	Result
	)		const

Reads values of entire float column into Result.

Definition at line 5198 of file table.cpp.

                                                               {
  if (!IsColName(ColName)) { TExcept::Throw("no such column " + ColName); }
  if (GetColType(ColName) != atFlt) { TExcept::Throw("not a floating point column " + ColName); }
  TInt ColId = GetColIdx(ColName);
  for (TRowIterator it = BegRI(); it < EndRI(); it++) {
    Result.Add(it.GetFltAttr(ColId));
  }
}

void TTable::ReadIntCol	(	const TStr &	ColName,
		TIntV &	Result
	)		const

Reads values of entire int column into Result.

Definition at line 5189 of file table.cpp.

                                                               {
  if (!IsColName(ColName)) { TExcept::Throw("no such column " + ColName); }
  if (GetColType(ColName) != atInt) { TExcept::Throw("not an integer column " + ColName); }
  TInt ColId = GetColIdx(ColName);
  for (TRowIterator it = BegRI(); it < EndRI(); it++) {
    Result.Add(it.GetIntAttr(ColId));
  }
}

void TTable::ReadStrCol	(	const TStr &	ColName,
		TStrV &	Result
	)		const

Reads values of entire string column into Result.

Definition at line 5207 of file table.cpp.

                                                               {
  if (!IsColName(ColName)) { TExcept::Throw("no such column " + ColName); }
  if (GetColType(ColName) != atStr) { TExcept::Throw("not a string column " + ColName); }
  TInt ColId = GetColIdx(ColName);
  for (TRowIterator it = BegRI(); it < EndRI(); it++) {
    Result.Add(it.GetStrAttr(ColId));
  }
}

void TTable::Reindex ( )

protected

Reinitializes row ids.

Register (cache) result of a grouping statement by a single group-by attribute T is a hash table mapping a key x to rows keyed by x => DISABLED FOR NOW

Definition at line 1869 of file table.cpp.

                     {
  RowIdMap.Clr();
  TInt IdColIdx = GetColIdx(IdColName);
  TInt IdCnt = 0;
  for (TRowIterator RI = BegRI(); RI < EndRI(); RI++) {
    IntCols[IdColIdx][RI.GetRowIdx()] = IdCnt;
    RowIdMap.AddDat(RI.GetRowIdx(), IdCnt);
    IdCnt++;
  }
}

void TTable::RemoveFirstRow ( )

protected

Removes first valid row of the table.

Definition at line 1102 of file table.cpp.

                            {
  if (FirstValidRow == LastValidRow) {
    LastValidRow = -1;
  }

  TInt Old = FirstValidRow;
  FirstValidRow = Next[FirstValidRow];
  Next[Old] = TTable::Invalid;
  NumValidRows--;
  TInt IdColIdx = GetColIdx(GetIdColName());
  RowIdMap.AddDat(IntCols[IdColIdx][Old], Invalid);
}

void TTable::RemoveRow ( TInt  RowIdx, TInt  PrevRowIdx  )

protected

Removes row with id RowIdx.

Definition at line 1115 of file table.cpp.

                                                   {
  if (RowIdx == FirstValidRow) {
    RemoveFirstRow();
    return;
  }
  Assert(RowIdx != TTable::Invalid);
  if (RowIdx == TTable::Last) { return; }
  Next[PrevRowIdx] = Next[RowIdx];
  if (LastValidRow == RowIdx) {
    LastValidRow = RowIdx;
  }
  Next[RowIdx] = TTable::Invalid;
  NumValidRows--;
  TInt IdColIdx = GetColIdx(GetIdColName());
  RowIdMap.AddDat(IntCols[IdColIdx][RowIdx], Invalid);
}

void TTable::Rename	(	const TStr &	Column,
		const TStr &	NewLabel
	)

Renames a column.

Definition at line 1085 of file table.cpp.

                                                            {
  // This function is necessary, for example to take the union of two tables 
  // where the attribute names don't match.
  if (!IsColName(column)) { TExcept::Throw("no such column " + column); }
  TPair<TAttrType,TInt> ColVal = GetColTypeMap(column);
  DelColType(column);
  AddColType(NewLabel, ColVal);
  TStr NColName = NormalizeColName(column);
  TStr NLabel = NormalizeColName(NewLabel);
  for (TInt c = 0; c < Sch.Len(); c++) {
    if (Sch[c].Val1 == NColName) {
      Sch.SetVal(c, TPair<TStr, TAttrType>(NLabel, Sch[c].Val2));
      break;
    }
  }
}

TStr TTable::RenumberColName ( const TStr &  ColName ) const

protected

Returns a re-numbered column name based on number of existing columns with conflicting names.

Definition at line 4609 of file table.cpp.

                                                      {
  TStr NColName = ColName;
  if (NColName.GetCh(NColName.Len()-2) == '-') { 
    NColName = NColName.GetSubStr(0,NColName.Len()-3); 
  }
  TInt Conflicts = 0;
  for (TInt i = 0; i < Sch.Len(); i++) {
    if (NColName == Sch[i].Val1.GetSubStr(0, Sch[i].Val1.Len()-3)) {
      Conflicts++;
    }
  }
  Conflicts++;
  NColName = NColName + "-" + Conflicts.GetStr();
  return NColName;
}

TInt TTable::RequestIndexFlt

(

const TStr &

ColName

)

Creates Index for Flt Column ColName.

Creates an Index on float column ColName. The index is hash-based, going from the column value to a vector of RowIdxs in the table that correspond to the value. If it exists, the index is used by the Get*RowIdxByVal functions; else, those functions will loop over the entire table. The index is NOT updated automatically when the table is modified; it is the user's responsibility to call RequestIndex after modifying the table if the index is necessary.

Definition at line 5472 of file table.cpp.

                                                {

  THash<TFlt, TIntV> NewIndex;
  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    TFlt ValAtRow = RowI.GetFltAttr(ColName);
    TInt RowIdx = RowI.GetRowIdx();
    if (NewIndex.IsKey(ValAtRow)) {
       TIntV Curr_V = NewIndex.GetDat(ValAtRow);
       Curr_V.Add(RowIdx);
    }
    else {
      TIntV New_V;
      New_V.Add(RowIdx);
      NewIndex.AddDat(ValAtRow, New_V);
    }
  }
  FltColIndexes.AddDat(ColName, NewIndex); 
  return 0;
}

TInt TTable::RequestIndexInt

(

const TStr &

ColName

)

Creates Index for Int Column ColName.

Creates an Index on integer column ColName. The index is hash-based, going from the column value to a vector of RowIdxs in the table that correspond to the value. If it exists, the index is used by the Get*RowIdxByVal functions; else, those functions will loop over the entire table. The index is NOT updated automatically when the table is modified; it is the user's responsibility to call RequestIndex after modifying the table if the index is necessary.

Definition at line 5453 of file table.cpp.

                                                {

  THash<TInt, TIntV> NewIndex;
  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    TInt ValAtRow = RowI.GetIntAttr(ColName);
    TInt RowIdx = RowI.GetRowIdx();
    if (NewIndex.IsKey(ValAtRow)) {
       TIntV Curr_V = NewIndex.GetDat(ValAtRow);
       Curr_V.Add(RowIdx);
    }
    else {
      TIntV New_V;
      New_V.Add(RowIdx);
      NewIndex.AddDat(ValAtRow, New_V);
    }
  }
  IntColIndexes.AddDat(ColName, NewIndex); 
  return 0;
}

TInt TTable::RequestIndexStrMap

(

const TStr &

ColName

)

Creates Index for Str Column ColName.

Creates an Index on string column given by ColName. The index is hash-based, going from the column value (that is, the integer mapping of the string value) to a vector of RowIdxs in the table that correspond to the value. If it exists, the index is used by the Get*RowIdxByVal functions; else, those functions will loop over the entire table. The index is NOT updated automatically when the table is modified; it is the user's responsibility to call RequestIndex after modifying the table if the index is necessary.

Definition at line 5491 of file table.cpp.

                                                   {
  THash<TInt, TIntV> NewIndex;
  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    TInt MapAtRow = RowI.GetStrMapByName(ColName);
    TInt RowIdx = RowI.GetRowIdx();
    if (NewIndex.IsKey(MapAtRow)) {
       TIntV Curr_V = NewIndex.GetDat(MapAtRow);
       Curr_V.Add(RowIdx);
    }
    else {
      TIntV New_V;
      New_V.Add(RowIdx);
      NewIndex.AddDat(MapAtRow, New_V);
    }
  }
  StrMapColIndexes.AddDat(ColName, NewIndex); 
  return 0;
}

void TTable::ResizeTable ( int  RowCount )

protected

Resizes the table to hold RowCount rows.

Definition at line 4307 of file table.cpp.

                                     {
  if (RowCount == 0) {
    // initialize empty table
    NumValidRows = 0;
    FirstValidRow = TTable::Invalid;
    LastValidRow = TTable::Invalid;
  }
  if (Next.Len() < RowCount) {
    TInt FltOffset = IntCols.Len();
    TInt StrOffset = FltOffset + FltCols.Len();
    TInt TotalCols = StrOffset + StrColMaps.Len();
#ifdef USE_OPENMP
    #pragma omp parallel for schedule(static)
#endif
    for (int i = 0; i < TotalCols+1; i++) {
      if (i < FltOffset) {
        IntCols[i].Reserve(RowCount, RowCount); 
      } else if (i < StrOffset) {
        FltCols[i-FltOffset].Reserve(RowCount, RowCount);
      } else if (i < TotalCols) {
        StrColMaps[i-StrOffset].Reserve(RowCount, RowCount);  
      } else {
        Next.Reserve(RowCount, RowCount);    
      }
    }
  } else if (Next.Len() > RowCount) {
    TInt FltOffset = IntCols.Len();
    TInt StrOffset = FltOffset + FltCols.Len();
    TInt TotalCols = StrOffset + StrColMaps.Len();
#ifdef USE_OPENMP
    #pragma omp parallel for schedule(static)
#endif
    for (int i = 0; i < TotalCols+1; i++) {
      if (i < FltOffset) {
        IntCols[i].Trunc(RowCount); 
      } else if (i < StrOffset) {
        FltCols[i-FltOffset].Trunc(RowCount); 
      } else if (i < TotalCols) {
        StrColMaps[i-StrOffset].Trunc(RowCount); 
      } else {
        Next.Trunc(RowCount); 
      }
    }
  }
}

void TTable::Save

(

TSOut &

SOut

)

Saves table schema and content to a binary format.

Note that TTableContext must be saved separately as it can be shared among multiple tables.

Definition at line 834 of file table.cpp.

                             {
  NumRows.Save(SOut);
  NumValidRows.Save(SOut);
  FirstValidRow.Save(SOut);
  LastValidRow.Save(SOut);
  Next.Save(SOut);
  IntCols.Save(SOut);
  FltCols.Save(SOut);
  StrColMaps.Save(SOut);

  THash<TStr,TPair<TInt,TInt> > ColTypeIntMap;
  TInt atIntVal = TInt(0);
  TInt atFltVal = TInt(1);
  TInt atStrVal = TInt(2);
  for (THash<TStr,TPair<TAttrType,TInt> >::TIter it = ColTypeMap.BegI(); it < ColTypeMap.EndI(); it++) {
    TPair<TAttrType,TInt> dat = it.GetDat();
    TStr DColName = DenormalizeColName(it.GetKey());
    switch (dat.GetVal1()) {
      case atInt:
        ColTypeIntMap.AddDat(DColName, TPair<TInt,TInt>(atIntVal, dat.GetVal2()));
        break;
      case atFlt:
        ColTypeIntMap.AddDat(DColName, TPair<TInt,TInt>(atFltVal, dat.GetVal2()));
        break;
      case atStr:
        ColTypeIntMap.AddDat(DColName, TPair<TInt,TInt>(atStrVal, dat.GetVal2()));
        break;
    }
  }
  ColTypeIntMap.Save(SOut);
  SOut.Flush();
}

void TTable::SaveBin

(

const TStr &

OutFNm

)

Saves table schema and content to a binary file.

Definition at line 829 of file table.cpp.

                                       {
  TFOut SOut(OutFNm);
  Save(SOut);
}

void TTable::SaveSS

(

const TStr &

OutFNm

)

Saves table schema and content to a TSV file.

Definition at line 780 of file table.cpp.

                                      {
  if (NumValidRows == 0) {
    printf("Table is empty");
    return;
  }
  FILE* F = fopen(OutFNm.CStr(), "w");
  // debug
  if (F == NULL) {
    printf("failed to open file %s\n", OutFNm.CStr());
    perror("fail ");
    return;
  }

  Dump(F);

#if 0
  Schema DSch = DenormalizeSchema();

  TInt L = Sch.Len();
  // print title (schema)
  fprintf(F, "# ");
  for (TInt i = 0; i < L-1; i++) {
    fprintf(F, "%s\t", DSch[i].Val1.CStr());
  }  
  fprintf(F, "%s\n", DSch[L-1].Val1.CStr());
  // print table contents
  for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
    for (TInt i = 0; i < L; i++) {
      char C = (i == L-1) ? '\n' : '\t';
      switch (GetSchemaColType(i)) {
        case atInt: {
          fprintf(F, "%d%c", RowI.GetIntAttr(GetSchemaColName(i)).Val, C);
          break;
        }
        case atFlt: {
          fprintf(F, "%f%c", RowI.GetFltAttr(GetSchemaColName(i)).Val, C);
          break;
        }
        case atStr: {
          fprintf(F, "%s%c", RowI.GetStrAttr(GetSchemaColName(i)).CStr(), C);
          break;
        }
      }
    }
  }
#endif
  fclose(F);
}

void TTable::Select	(	TPredicate &	Predicate,
		TIntV &	SelectedRows,
		TBool	Remove = true
	)

Selects rows that satisfy given Predicate.

Select. Has two modes of operation:

1. If Remove == true then (logically) remove the rows for which the predicate doesn't hold
2. If Remove == false then add the physical indices of the rows for which the predicate holds to the vactor SelectedRows

Definition at line 2730 of file table.cpp.

                                                                            {
  TIntV Selected;
  TStrV RelevantCols;
  Predicate.GetVariables(RelevantCols);
  TInt NumRelevantCols = RelevantCols.Len();
  TVec<TAttrType> ColTypes = TVec<TAttrType>(NumRelevantCols);
  TIntV ColIndices = TIntV(NumRelevantCols);
  for (TInt i = 0; i < NumRelevantCols; i++) {
    ColTypes[i] = GetColType(RelevantCols[i]);
    ColIndices[i] = GetColIdx(RelevantCols[i]);
  } 
  
  if (Remove) {
    TRowIteratorWithRemove RowI = BegRIWR();
    while (RowI.GetNextRowIdx() != Last) {
      // prepare arguments for predicate evaluation
      for (TInt i = 0; i < NumRelevantCols; i++) {
        switch (ColTypes[i]) {
        case atInt:
          Predicate.SetIntVal(RelevantCols[i], RowI.GetNextIntAttr(ColIndices[i]));
          break;
        case atFlt:
          Predicate.SetFltVal(RelevantCols[i], RowI.GetNextFltAttr(ColIndices[i]));
          break;
        case atStr:
          Predicate.SetStrVal(RelevantCols[i], RowI.GetNextStrAttr(ColIndices[i]));
          break;
        }
      }
      if (!Predicate.Eval()) { 
        RowI.RemoveNext();
      } else {
        RowI++;
      }
    }
  } else {
    for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
      for (TInt i = 0; i < NumRelevantCols; i++) {
        switch (ColTypes[i]) {
        case atInt:
          Predicate.SetIntVal(RelevantCols[i], RowI.GetIntAttr(RelevantCols[i]));
          break;
        case atFlt:
          Predicate.SetFltVal(RelevantCols[i], RowI.GetFltAttr(RelevantCols[i]));
          break;
        case atStr:
          Predicate.SetStrVal(RelevantCols[i], RowI.GetStrAttr(RelevantCols[i]));
          break;
        }
      }
      if (Predicate.Eval()) { SelectedRows.Add(RowI.GetRowIdx()); }
    }
  }
}

void TTable::Select ( TPredicate &  Predicate )

inline

Definition at line 1257 of file table.h.

                                     {
    TIntV SelectedRows;
    Select(Predicate, SelectedRows, true);
  }

void TTable::SelectAtomic	(	const TStr &	Col1,
		const TStr &	Col2,
		TPredComp	Cmp,
		TIntV &	SelectedRows,
		TBool	Remove = true
	)

Selects rows using atomic compare operation.

Select atomic - optimized cases of select with predicate of an atomic form: compare attribute to attribute or compare attribute to a constant

Definition at line 2793 of file table.cpp.

                                                                                                              {
  const TAttrType Ty1 = GetColType(Col1);
  const TAttrType Ty2 = GetColType(Col2);
  const TInt ColIdx1 = GetColIdx(Col1);
  const TInt ColIdx2 = GetColIdx(Col2);
  if (Ty1 != Ty2) {
    TExcept::Throw("SelectAtomic: diff types");
  }
  if (Cmp == SUBSTR || Cmp == SUPERSTR) { Assert(Ty1 == atStr); }

  if (Remove) {
    TRowIteratorWithRemove RowI = BegRIWR();
    while (RowI.GetNextRowIdx() != Last) {

      TBool Result;
      switch (Ty1) {
        case atInt:
          Result = TPredicate::EvalAtom(RowI.GetNextIntAttr(ColIdx1), RowI.GetNextIntAttr(ColIdx2), Cmp);
          break;
        case atFlt:
          Result = TPredicate::EvalAtom(RowI.GetNextFltAttr(ColIdx1), RowI.GetNextFltAttr(ColIdx2), Cmp);
          break;
        case atStr:
          Result = TPredicate::EvalStrAtom(RowI.GetNextStrAttr(ColIdx1), RowI.GetNextStrAttr(ColIdx2), Cmp);
          break;
      }
                        
      if (!Result) { 
        RowI.RemoveNext();
      } else {
        RowI++;
      }

    }
  } else {
    for (TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++) {
      TBool Result;
      switch (Ty1) {
        case atInt:
          Result = TPredicate::EvalAtom(RowI.GetIntAttr(Col1), RowI.GetIntAttr(Col2), Cmp);
          break;
        case atFlt:
          Result = TPredicate::EvalAtom(RowI.GetFltAttr(Col1), RowI.GetFltAttr(Col2), Cmp);
          break;
        case atStr:
          Result = TPredicate::EvalStrAtom(RowI.GetStrAttr(Col1), RowI.GetStrAttr(Col2), Cmp);
          break;
      }
      if (Result) { SelectedRows.Add(RowI.GetRowIdx()); }
    }
  }
}

void TTable::SelectAtomic ( const TStr &  Col1, const TStr &  Col2, TPredComp  Cmp  )

inline

Definition at line 1269 of file table.h.

                                                                       {
    TIntV SelectedRows;
    SelectAtomic(Col1, Col2, Cmp, SelectedRows, true);
  }

void TTable::SelectAtomicConst	(	const TStr &	Col,
		const TPrimitive &	Val,
		TPredComp	Cmp,
		TIntV &	SelectedRows,
		PTable &	SelectedTable,
		TBool	Remove = true,
		TBool	Table = true
	)

Selects rows where the value of Col matches given primitive Val.

Definition at line 2853 of file table.cpp.

                                                                         {
  //double startFn = omp_get_wtime();
  TStr ValTStr(Val.GetStr());
  TAttrType Type = GetColType(Col);
  TInt ColIdx = GetColIdx(Col);

  if (Type != Val.GetType()) { 
    TExcept::Throw("SelectAtomicConst: coltype does not match const type"); 
  }

  if(Remove){
#ifdef USE_OPENMP
    if (GetMP()) {
      //double endInit = omp_get_wtime();
      //printf("Init time = %f\n", endInit-startFn);
      TIntPrV Partitions;
      GetPartitionRanges(Partitions, omp_get_max_threads()*CHUNKS_PER_THREAD);
      TInt PartitionSize = Partitions[0].GetVal2()-Partitions[0].GetVal1()+1;
      int RemoveCount = 0;
      //double endPart = omp_get_wtime();
      //printf("Partition time = %f\n", endPart-endInit);

      TIntPrV Bounds(Partitions.Len());
    
      // #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD) reduction(+:RemoveCount) shared(Val)
      #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD) reduction(+:RemoveCount)
      for (int i = 0; i < Partitions.Len(); i++){
        //TPrimitive ThreadLocalVal(Val);
        TRowIterator RowI(Partitions[i].GetVal1(), this);
        TRowIterator EndI(Partitions[i].GetVal2(), this);
        TInt FirstRowIdx = TTable::Invalid;
        TInt LastRowIdx = TTable::Invalid;
        TBool First = true;
        while (RowI < EndI) {
          TInt CurrRowIdx = RowI.GetRowIdx();
          TBool Result;
          if (Type != atStr) {
            Result = RowI.CompareAtomicConst(ColIdx, Val, Cmp);
          } else {
            Result = RowI.CompareAtomicConstTStr(ColIdx, ValTStr, Cmp);
          }
          RowI++;
          if(!Result) {
            Next[CurrRowIdx] = TTable::Invalid;
            RemoveCount++;
          } else { 
            if (First) { FirstRowIdx = CurrRowIdx; First = false; } 
            else { Next[LastRowIdx] = CurrRowIdx; }
            LastRowIdx = CurrRowIdx; 
          }
        }
        Bounds[i] = TIntPr(FirstRowIdx, LastRowIdx);
        //printf("Thread %d: i = %d, start = %d, end = %d\n", omp_get_thread_num(), i,
        //  Partitions[i].GetVal1().Val, Partitions[i].GetVal2().Val);
      }
      //double endIter = omp_get_wtime();
      //printf("Iter time = %f\n", endIter-endPart);
      
      // repair the next vector
      TInt CurrBound = 0;
      while (CurrBound < Bounds.Len() && Bounds[CurrBound].Val1 == TTable::Invalid) {
        CurrBound++;
      }
      if (CurrBound == Bounds.Len()) {
        // selected table is empty
        Assert(NumValidRows == RemoveCount);
        NumValidRows = 0;
        FirstValidRow = TTable::Invalid;
        LastValidRow = TTable::Invalid;
      } else {
        NumValidRows -= RemoveCount;
        FirstValidRow = Bounds[CurrBound].Val1;
        LastValidRow = Bounds[CurrBound].Val2;
        TInt PrevBound = CurrBound;
        CurrBound++;
        while (CurrBound < Bounds.Len()) {
          if (Bounds[CurrBound].Val1 == TTable::Invalid) { CurrBound++; continue; }
          Next[Bounds[PrevBound].Val2] = Bounds[CurrBound].Val1;
          LastValidRow = Bounds[CurrBound].Val2;
          PrevBound = CurrBound;
          CurrBound++;
        }
        Next[Bounds[PrevBound].Val2] = TTable::Last;
      }
      IsNextDirty = 1;
      //double endRepair = omp_get_wtime();
      //printf("Repair time = %f\n", endRepair-endIter);
    } else {
#endif
      TRowIteratorWithRemove RowI = BegRIWR();
      while(RowI.GetNextRowIdx() != Last){
        if (!RowI.CompareAtomicConst(ColIdx, Val, Cmp)) {
          RowI.RemoveNext();
        } else {
          RowI++;
        }
      }
      IsNextDirty = 1;
#ifdef USE_OPENMP
    }
#endif
  } else if (Table) {
#ifdef USE_OPENMP
    if (GetMP()) {
      //double endInit = omp_get_wtime();
      //printf("Init time = %f\n", endInit-startFn);
      TIntPrV Partitions;
      GetPartitionRanges(Partitions, omp_get_max_threads()*CHUNKS_PER_THREAD);
      TInt PartitionSize = Partitions[0].GetVal2()-Partitions[0].GetVal1()+1;
      //double endPart = omp_get_wtime();
      //printf("Partition time = %f\n", endPart-endInit);

      int TotalSelectedRows = 0;
      #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD) reduction(+:TotalSelectedRows)
      for (int i = 0; i < Partitions.Len(); i++){
        TRowIterator RowI(Partitions[i].GetVal1(), this);
        TRowIterator EndI(Partitions[i].GetVal2(), this);
        while (RowI < EndI) {
          if (Type != atStr) {
            if (RowI.CompareAtomicConst(ColIdx, Val, Cmp)) { 
              TotalSelectedRows++;
            }
          } else {
            if (RowI.CompareAtomicConstTStr(ColIdx, ValTStr, Cmp)) { 
              TotalSelectedRows++;
            }
          }
          RowI++;
        }
      }
      //double endCount = omp_get_wtime();
      //printf("Count time = %f\n", endCount-endPart);

      SelectedTable->ResizeTable(TotalSelectedRows);
      //double endResize = omp_get_wtime();
      //printf("Resize time = %f\n", endResize-endCount);

      if (TotalSelectedRows == 0) {
        // printf("Select: Empty output!\n");
        return;
      }
    
      #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD)
      for (int i = 0; i < Partitions.Len(); i++){
        TIntV LocalSelectedRows;
        LocalSelectedRows.Reserve(PartitionSize);
        TRowIterator RowI(Partitions[i].GetVal1(), this);
        TRowIterator EndI(Partitions[i].GetVal2(), this);
        while (RowI < EndI) {
          if (Type != atStr) {
            if (RowI.CompareAtomicConst(ColIdx, Val, Cmp)) { 
              LocalSelectedRows.Add(RowI.GetRowIdx());
            }
          } else {
            if (RowI.CompareAtomicConstTStr(ColIdx, ValTStr, Cmp)) { 
              LocalSelectedRows.Add(RowI.GetRowIdx());
            }
          }
          RowI++;
        }
        SelectedTable->AddSelectedRows(*this, LocalSelectedRows);
        //printf("Thread %d: i = %d, start = %d, end = %d\n", omp_get_thread_num(), i,
        //  Partitions[i].GetVal1().Val, Partitions[i].GetVal2().Val);
      }
      //double endIter = omp_get_wtime();
      //printf("Iter time = %f\n", endIter-endResize);

      //SelectedTable->ResizeTable(SelectedTable->GetNumValidRows());
      //double endResize2 = omp_get_wtime();
      //printf("Resize2 time = %f\n", endResize2-endIter);      
      SelectedTable->SetFirstValidRow();
    } else {
#endif
      for(TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++){
        if (RowI.CompareAtomicConst(ColIdx, Val, Cmp)) { 
          SelectedTable->AddRow(RowI);
        }
      }
#ifdef USE_OPENMP
    }
#endif
  } else {
    for(TRowIterator RowI = BegRI(); RowI < EndRI(); RowI++){
      if (RowI.CompareAtomicConst(ColIdx, Val, Cmp)) { 
        SelectedRows.Add(RowI.GetRowIdx());
      }
    }
  }
}

template<class T >

void TTable::SelectAtomicConst ( const TStr &  Col, const T &  Val, TPredComp  Cmp  )

inline

Definition at line 1281 of file table.h.

                                                                       {
    TIntV SelectedRows;
    PTable SelectedTable;
    SelectAtomicConst(Col, TPrimitive(Val), Cmp, SelectedRows, SelectedTable, true, false);
  }

template<class T >

void TTable::SelectAtomicConst ( const TStr &  Col, const T &  Val, TPredComp  Cmp, PTable &  SelectedTable  )

inline

Definition at line 1287 of file table.h.

                                                                                              {
    TIntV SelectedRows;
    SelectAtomicConst(Col, TPrimitive(Val), Cmp, SelectedRows, SelectedTable, false, true);
  }

void TTable::SelectAtomicFltConst ( const TStr &  Col, const TFlt &  Val, TPredComp  Cmp  )

inline

Definition at line 1314 of file table.h.

                                                                             {
    SelectAtomicConst(Col, Val, Cmp);
  }

void TTable::SelectAtomicFltConst ( const TStr &  Col, const TFlt &  Val, TPredComp  Cmp, PTable &  SelectedTable  )

inline

Definition at line 1317 of file table.h.

                                                                                                    {
    SelectAtomicConst(Col, Val, Cmp, SelectedTable);
  }

void TTable::SelectAtomicIntConst ( const TStr &  Col, const TInt &  Val, TPredComp  Cmp  )

inline

Definition at line 1300 of file table.h.

                                                                             {
    SelectAtomicConst(Col, Val, Cmp);
  }

void TTable::SelectAtomicIntConst ( const TStr &  Col, const TInt &  Val, TPredComp  Cmp, PTable &  SelectedTable  )

inline

Definition at line 1303 of file table.h.

                                                                                                    {
    SelectAtomicConst(Col, Val, Cmp, SelectedTable);
  }

void TTable::SelectAtomicStrConst ( const TStr &  Col, const TStr &  Val, TPredComp  Cmp  )

inline

Definition at line 1307 of file table.h.

                                                                             {
    SelectAtomicConst(Col, Val, Cmp);
  }

void TTable::SelectAtomicStrConst ( const TStr &  Col, const TStr &  Val, TPredComp  Cmp, PTable &  SelectedTable  )

inline

Definition at line 1310 of file table.h.

                                                                                                    {
    SelectAtomicConst(Col, Val, Cmp, SelectedTable);
  }

void TTable::SelectFirstNRows

(

const TInt &

N

)

Selects first N rows from the table.

Definition at line 3337 of file table.cpp.

                                           {
  if (N == 0) {
    LastValidRow = -1;
    return;
  }
  TRowIterator RowI = BegRI();
  TInt count = 1;
  while (count < N) {
    if (!(RowI < EndRI())) {
      return; // The table contains less than N rows
    }
    RowI++;
    count++;
  }
  NumValidRows = N;
  TInt LastId = RowI.GetRowIdx();
  if (Next[LastId] == Last) {
    return; // The table contains exactly N rows
  }
  // The table contains more than N rows
  TInt CurrId = LastId;
  while (Next[CurrId] != Last) {
    Assert(Next[CurrId] != Invalid);
    TInt NextId = Next[CurrId];
    Next[CurrId] = Invalid;
    CurrId = NextId;
  }
  Next[LastId] = Last;
  LastValidRow = LastId;
}

PTable TTable::SelfJoin ( const TStr &  Col )

inline

Joins table with itself, on values of Col.

Definition at line 1357 of file table.h.

{ return Join(Col, *this, Col); }

PTable TTable::SelfSimJoin ( const TStrV &  Cols, const TStr &  DistanceColName, const TSimType &  SimType, const TFlt &  Threshold  )

inline

Definition at line 1358 of file table.h.

{ return SimJoin(Cols, *this, Cols, DistanceColName, SimType, Threshold); }

PTable TTable::SelfSimJoinPerGroup	(	const TStr &	GroupAttr,
		const TStr &	SimCol,
		const TStr &	DistanceColName,
		const TSimType &	SimType,
		const TFlt &	Threshold
	)

Performs join if the distance between two rows is less than the specified threshold.

Returns table with schema (GroupId1, GroupId2, Similarity).

Definition at line 2074 of file table.cpp.

{
        if(!IsColName(SimCol) || !IsColName(GroupAttr)){
                TExcept::Throw("No such column found in table");
        }

  PTable JointTable = New(Context);
        // Initialize the joint table - (GroupId1, GroupId2, Similarity)
        JointTable->IntCols = TVec<TIntV>(2);
        JointTable->FltCols = TVec<TFltV>(1);

        for(TInt i=0;i<2;i++){
    TInt Suffix = i+1;
    TStr CName = "GroupId_" + Suffix.GetStr();
    TPair<TAttrType, TInt> Group(atInt, (int)i);
    JointTable->AddColType(CName, Group);
    JointTable->AddSchemaCol(CName, atInt);
  }

        TPair<TAttrType, TInt> Group(atFlt, 0);
        JointTable->AddColType(DistanceColName, Group);
        JointTable->AddSchemaCol(DistanceColName, atFlt);

        THash<TInt, THash<TInt, TInt> > TIntHH;

        TAttrType attrType = GetColType(SimCol);
        TInt GroupColIdx = GetColIdx(GroupAttr);
        TInt SimColIdx = GetColIdx(SimCol);

        for (TRowIterator RowI = this->BegRI(); RowI < this->EndRI(); RowI++) {
                TInt GroupId = IntCols[GroupColIdx][RowI.GetRowIdx()];
        
                if(attrType==atInt || attrType==atStr)
                {
                        if(!TIntHH.IsKey(GroupId)){
                                THash<TInt, TInt> TIntH;
                                TIntHH.AddDat(GroupId, TIntH);
                        }

                        THash<TInt, TInt>& TIntH = TIntHH.GetDat(GroupId);
                        TInt SimAttrVal = (attrType==atInt ? IntCols[SimColIdx][RowI.GetRowIdx()] : StrColMaps[SimColIdx][RowI.GetRowIdx()]);
                        TIntH.AddDat(SimAttrVal, 0);
                }
                else
                {
                        TExcept::Throw("Attribute type not supported.");
                }
        }

        // Iterate through every pair of groups and calculate the distance
        for (THash<TInt, THash<TInt, TInt> >::TIter it1 = TIntHH.BegI(); it1 < TIntHH.EndI(); it1++) {
                THash<TInt, TInt> Vals1H = it1.GetDat();
                TInt GroupId1 = it1.GetKey();

                for (THash<TInt, THash<TInt, TInt> >::TIter it2 = TIntHH.BegI(); it2 < TIntHH.EndI(); it2++) {
                                int intersectionCount = 0;
                                TInt GroupId2 = it2.GetKey();
                                THash<TInt, TInt> Vals2H = it2.GetDat();

                                for(THash<TInt, TInt>::TIter it = Vals1H.BegI(); it < Vals1H.EndI(); it++)
                                {
                                        TInt Val = it.GetKey();
                                        if(Vals2H.IsKey(Val)){
                                                intersectionCount+=1;
                                        }
                                }

                                int unionCount = Vals1H.Len() + Vals2H.Len() - intersectionCount;
                                float distance = 1.0f - (float)intersectionCount/unionCount;

                                // Add a new row to the JointTable
                                if(distance<=Threshold){
                                                JointTable->IntCols[0].Add(GroupId1);
                                                JointTable->IntCols[1].Add(GroupId2);
                                                JointTable->FltCols[0].Add(distance);
                                                JointTable->IncrementNext();
                        }
                }
        }

  JointTable->InitIds();
        return JointTable;
}

PTable TTable::SelfSimJoinPerGroup	(	const TStrV &	GroupBy,
		const TStr &	SimCol,
		const TStr &	DistanceColName,
		const TSimType &	SimType,
		const TFlt &	Threshold
	)

Performs join if the distance between two rows is less than the specified threshold.

SimJoinPerGroup performs SimJoin based on a set of attributes. Performs the grouping internally and returns a projection of the columns on which groupby was performed along with the similarity.

Definition at line 2160 of file table.cpp.

                                                                              {
  TStrV NGroupBy = NormalizeColNameV(GroupBy);
        TStrV ProjectionV;
        
        // Only keep the GroupBy cols and the SimCol
        for(TInt i=0; i<GroupBy.Len(); i++)
        {
                ProjectionV.Add(GroupBy[i]);
        }

        ProjectionV.Add(SimCol);
        ProjectInPlace(ProjectionV);

        TStr CName = "Group";
  TIntV UniqueVec;
  THash<TGroupKey, TPair<TInt, TIntV> > Grouping;
  GroupAux(NGroupBy, Grouping, false, CName, false, UniqueVec);
        PTable GroupJointTable = SelfSimJoinPerGroup(CName, SimCol, DistanceColName, SimType, Threshold);
        PTable JointTable = InitializeJointTable(*this);

        // Hash of groupid to any arbitrary row of that group. Arbitrary because the GroupBy 
        // columns within that group are the same, so we can choose any one. 
        THash<TInt, TInt> GroupIdH;

        for(THash<TGroupKey, TPair<TInt, TIntV> >::TIter it=Grouping.BegI(); it<Grouping.EndI(); it++)
        {
                TPair<TInt, TIntV> group = it.GetDat();
                TInt GroupNum = group.Val1;
                TIntV RowIds = group.Val2;

                if(!GroupIdH.IsKey(GroupNum))
                {
                        TInt RandomRowId = RowIds[0];  // Arbitrarily select the 1st row. 
                        GroupIdH.AddDat(GroupNum, RandomRowId);
                }
        }

        for(TRowIterator RowI = GroupJointTable->BegRI(); RowI < GroupJointTable->EndRI(); RowI++)
        {
                // The GroupJoinTable has a well defined structure - columns 0 and 1 are GroupIds
                TInt GroupId1 = GroupJointTable->IntCols[0][RowI.GetRowIdx()];
                TInt GroupId2 = GroupJointTable->IntCols[1][RowI.GetRowIdx()];

                // Get the rows for groupid1 and groupid and arbitrary select one row
                TInt RowId1 = GroupIdH.GetDat(GroupId1);
                TInt RowId2 = GroupIdH.GetDat(GroupId2);
                JointTable->AddJointRow(*this, *this, RowId1, RowId2);
        } 

        // Add the simiarlity column from the GroupJointTable - GroupJointTable has a 
        // well defined structure - The first float column is the similarity;
        JointTable->StoreFltCol(DistanceColName, GroupJointTable->FltCols[0]);
        ProjectionV.Clr();
        ProjectionV.Add(DistanceColName);

        // Find the GroupBy columns in the JointTable by matching the Suffix of the Schema
        // columns with the original GroupBy columns - Note that Join renames columns. 
        for(TInt i=0; i<GroupBy.Len(); i++){
                for(TInt j=0; j<JointTable->Sch.Len(); j++)
                {
                        TStr ColName = JointTable->Sch[j].Val1;
                        if(ColName.IsStrIn(GroupBy[i]))
                        {
                                ProjectionV.Add(ColName);
                        }
                }
        }

        JointTable->ProjectInPlace(ProjectionV);
        JointTable->InitIds();
        return JointTable;
}

void TTable::SetCommonNodeAttrs ( const TStr &  SrcAttr, const TStr &  DstAttr, const TStr &  CommonAttrName  )

inline

Sets the columns to be used as both src and dst node attributes.

Definition at line 1179 of file table.h.

                                                                                               {
    CommonNodeAttrs.Add(TStrTr(NormalizeColName(SrcAttr), NormalizeColName(DstAttr), NormalizeColName(CommonAttrName)));
  }

void TTable::SetDstCol ( const TStr &  Dst )

inline

Sets the name of the column to be used as dst nodes in the graph.

Definition at line 1158 of file table.h.

                                  {
    if (!IsColName(Dst)) { TExcept::Throw(Dst + ": no such column"); }
    DstCol = NormalizeColName(Dst);
  }

void TTable::SetFirstValidRow ( )

inlineprotected

Sets the first valid row of the TTable.

Definition at line 821 of file table.h.

                          {
    for (int i = 0; i < Next.Len(); i++) {
      if(Next[i] != TTable::Invalid) { FirstValidRow = i; return;}
    }
    TExcept::Throw("SetFirstValidRow: Table is empty");
  }

void TTable::SetFltColToConstMP	(	TInt	UpdateColIdx,
		TFlt	DefaultFltVal
	)

Definition at line 4129 of file table.cpp.

                                                                    {
    if(!GetMP()){ TExcept::Throw("Not Using MP!");}
        TIntPrV Partitions;
        GetPartitionRanges(Partitions, omp_get_max_threads()*CHUNKS_PER_THREAD);
        TInt PartitionSize = Partitions[0].GetVal2()-Partitions[0].GetVal1()+1;
        #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD)
        for (int i = 0; i < Partitions.Len(); i++){
                TRowIterator RowI(Partitions[i].GetVal1(), this);
                TRowIterator EndI(Partitions[i].GetVal2(), this);
                while(RowI < EndI){
                        FltCols[UpdateColIdx][RowI.GetRowIdx()] = DefaultFltVal;
                        RowI++;
                }
        }
}

static void TTable::SetMP ( TInt  Value )

inlinestatic

Definition at line 536 of file table.h.

{ UseMP = Value; }

void TTable::SetSrcCol ( const TStr &  Src )

inline

Sets the name of the column to be used as src nodes in the graph.

Definition at line 1151 of file table.h.

                                  {
    if (!IsColName(Src)) { TExcept::Throw(Src + ": no such column"); }
    SrcCol = NormalizeColName(Src);
  }

PTable TTable::SimJoin	(	const TStrV &	Cols1,
		const TTable &	Table,
		const TStrV &	Cols2,
		const TStr &	DistanceColName,
		const TSimType &	SimType,
		const TFlt &	Threshold
	)

Performs join if the distance between two rows is less than the specified threshold.

Returns Similarity based join of two tables based on a given distance metric and a given threshold. Records (r1, r2) that are returned satisfy the criterion: d(r1, r2) <= Threshold

Definition at line 1974 of file table.cpp.

{
        Assert(Cols1.Len() == Cols2.Len());

        if(Cols1.Len()!=Cols2.Len()){
                TExcept::Throw("Column vectors must match in type and length");
        }

        for (TInt i = 0; i < Cols1.Len(); i++) {
                if(!IsColName(Cols1[i]) || !Table.IsColName(Cols2[i])){
                        TExcept::Throw("Column not found in Table");
                }

                TAttrType Type1 = GetColType(Cols1[i]);
                TAttrType Type2 = GetColType(Cols2[i]);

                if(Type1!=Type2){
                        TExcept::Throw("Column types on the two tables must match.");
                }

                // When supporting more distance metrics, check if the types are supported for given metric. 
                if((Type1!=atInt && Type1!=atFlt) || (Type2!=atInt && Type2!=atFlt)){
                        TExcept::Throw("Column type not supported. Only Flt and Int column types are supported.");
                }
  }

        // Initialize Join table and add the similarity column
  PTable JointTable = InitializeJointTable(Table);
        TFltV DistanceV;

        // O(n^2): Parallelize 
        for(TRowIterator RowI = this->BegRI(); RowI < this->EndRI(); RowI++) {
                for(TRowIterator RowI2 = Table.BegRI(); RowI2 < Table.EndRI(); RowI2++) {
                        float distance = 0;

                        switch(SimType)
                        {
                                // Calculate the distance metric
                                case L2Norm:
                                        for(TInt i = 0; i < Cols1.Len(); i++) {
                                                float attrVal1, attrVal2;
                                                attrVal1 = GetColType(Cols1[i])==atInt ? (float)RowI.GetIntAttr(Cols1[i]) : (float)RowI.GetFltAttr(Cols1[i]);
                                                attrVal2 = Table.GetColType(Cols2[i])==atInt ? (float)RowI2.GetIntAttr(Cols2[i]) : (float)RowI2.GetFltAttr(Cols2[i]);
                                                distance += pow(attrVal1 - attrVal2, 2);
                                        }
                                
                                        distance = sqrt(distance);

                                        if(distance<=Threshold){
                                                JointTable->AddJointRow(*this, Table, RowI.GetRowIdx(), RowI2.GetRowIdx());
                                                DistanceV.Add(distance);
                                        }

                                        // Add row to the joint table if distance <= Threshold
                                        break;
                                // Haversine distance to calculate the distance between two points on Earth from latitude/longitude
                                case Haversine:
                                        {
                                                if(Cols1.Len()!=2){
                                                        TExcept::Throw("Haversine disance expects exactly two attributes - latitude and longitude - in that order.");
                                                }
                                                
                                                // Block to prevent cross-initialization error from compiler
                                                TFlt Radius = 6373; // km
                                                float Latitude1  = GetColType(Cols1[0])==atInt ? (float)RowI.GetIntAttr(Cols1[0]) : (float)RowI.GetFltAttr(Cols1[0]);
                                                float Latitude2 = Table.GetColType(Cols2[0])==atInt ? (float)RowI2.GetIntAttr(Cols2[0]) : (float)RowI2.GetFltAttr(Cols2[0]);

                                                float Longitude1  = GetColType(Cols1[1])==atInt ? (float)RowI.GetIntAttr(Cols1[1]) : (float)RowI.GetFltAttr(Cols1[1]);
                                                float Longitude2  = Table.GetColType(Cols2[1])==atInt ? (float)RowI2.GetIntAttr(Cols2[1]) : (float)RowI2.GetFltAttr(Cols2[1]);

                                                Latitude1 *= static_cast<float>(M_PI/180.0);
                                                Latitude2 *= static_cast<float>(M_PI/180.0);
                                                Longitude1 *= static_cast<float>(M_PI/180.0);
                                                Longitude2 *= static_cast<float>(M_PI/180.0);

                                                float dlon = Longitude2 - Longitude1;
                                                float dlat = Latitude2 - Latitude1;
                                                float a = pow(sin(dlat/2), 2) + cos(Latitude1)*cos(Latitude2)*pow(sin(dlon/2), 2);
                                                float c = 2*atan2(sqrt(a), sqrt(1-a));
                                                distance = (static_cast<float>(Radius.Val))*c;

                                                if(distance<=Threshold){
                                                        JointTable->AddJointRow(*this, Table, RowI.GetRowIdx(), RowI2.GetRowIdx());
                                                        DistanceV.Add(distance);
                                                }
                                        }
                                        break;
                                case L1Norm:
                                case Jaccard:
                                        TExcept::Throw("This distance metric is not supported");
                        }
                }       
        }

        // Add the value for the similarity column 
        JointTable->StoreFltCol(DistanceColName, DistanceV);
        JointTable->InitIds();
        return JointTable;
}

TVec< PTable > TTable::SpliceByGroup	(	const TStrV &	GroupByAttrs,
		TBool	Ordered = true
	)

Splices table into subtables according to a grouping statement.

Definition at line 1788 of file table.cpp.

                                                                      {
  TStrV NGroupBy = NormalizeColNameV(GroupBy);
  TIntV UniqueVec;
  THash<TGroupKey, TPair<TInt, TIntV> >Grouping;
  TVec<PTable> Result;

  Schema NewSchema;
  for (TInt c = 0; c < Sch.Len(); c++) {
    if (Sch[c].Val1 != GetIdColName()) {
      NewSchema.Add(Sch[c]);
    }
  }

  GroupAux(NGroupBy, Grouping, Ordered, "", false, UniqueVec);

  TInt cnt = 0;
  // iterate over groups
  for (THash<TGroupKey, TPair<TInt, TIntV> >::TIter it = Grouping.BegI(); it != Grouping.EndI(); it++) {
    PTable GroupTable = TTable::New(NewSchema, Context);

    TVec<TPair<TAttrType, TInt> > ColInfo;
    TIntV V;
    for (TInt i = 0; i < Sch.Len(); i++) {
      ColInfo.Add(GroupTable->GetColTypeMap(Sch[i].Val1));
      if (Sch[i].Val1 == IdColName()) {
        ColInfo[i].Val2 = -1;
      }
      V.Add(GetColIdx(Sch[i].Val1));
    }

    TIntV& Rows = it.GetDat().Val2;

    // iterate over rows in group
    for (TInt i = 0; i < Rows.Len(); i++) {
      // convert from permanent ID to row ID
      TInt RowIdx = RowIdMap.GetDat(Rows[i]);

      // iterate over schema
      for (TInt c = 0; c < Sch.Len(); c++) {
        TPair<TAttrType, TInt> Info = ColInfo[c];
        TInt ColIdx = Info.Val2;

        if (ColIdx == -1) { continue; }

        // add row to new group
        switch (Info.Val1) {
          case atInt:
            GroupTable->IntCols[ColIdx].Add(IntCols[V[c]][RowIdx]);
            break;
          case atFlt:
            GroupTable->FltCols[ColIdx].Add(FltCols[V[c]][RowIdx]);
            break;
          case atStr:
            GroupTable->StrColMaps[ColIdx].Add(StrColMaps[V[c]][RowIdx]);
            break;
        }

      }
      if (GroupTable->LastValidRow >= 0) {
        GroupTable->Next[GroupTable->LastValidRow] = GroupTable->NumRows;
      }
      GroupTable->Next.Add(GroupTable->Last);
      GroupTable->LastValidRow = GroupTable->NumRows;

      GroupTable->NumRows++;
      GroupTable->NumValidRows++;
    }
    GroupTable->InitIds();
    Result.Add(GroupTable);

    cnt += 1;
  }
  return Result;
}

void TTable::StoreFltCol	(	const TStr &	ColName,
		const TFltV &	ColVals
	)

Adds entire flt column to table.

Definition at line 4081 of file table.cpp.

                                                                  {
  if (ColVals.Len() != NumRows) {
    printf("new column dimension must agree with number of rows\n");
    return;
  }
  AddSchemaCol(ColName, atFlt);
  FltCols.Add(TFltV(NumRows));
  TInt ColIdx = FltCols.Len()-1;
  TInt i = 0;
  for (TRowIterator RI = BegRI(); RI < EndRI(); RI++) {
    FltCols[ColIdx][RI.GetRowIdx()] = ColVals[i];
    i++;
  }
  TInt L = FltCols.Len();
  AddColType(ColName, atFlt, L-1);
}

void TTable::StoreGroupCol ( const TStr &  GroupColName, const TVec< TPair< TInt, TInt > > &  GroupAndRowIds  )

protected

Parallel helper function for grouping. - we currently don't support such parallel grouping by complex keys.

Stores column for a group. Physical row ids have to be passed.

Definition at line 1290 of file table.cpp.

                                                                                                   {
  // Add a column where the value of the i'th row is the group id of row i.
  IntCols.Add(TIntV(NumRows));
  TInt L = IntCols.Len();
  AddColType(GroupColName, atInt, L-1);
  // Store group id for each row.
  for (TInt i = 0; i < GroupAndRowIds.Len(); i++) {
    IntCols[L-1][GroupAndRowIds[i].Val2] = GroupAndRowIds[i].Val1;
  }
}

void TTable::StoreIntCol	(	const TStr &	ColName,
		const TIntV &	ColVals
	)

Adds entire int column to table.

Definition at line 4064 of file table.cpp.

                                                                  {
  if (ColVals.Len() != NumRows) {
    printf("new column dimension must agree with number of rows\n");
    return;
  }
  AddSchemaCol(ColName, atInt);
  IntCols.Add(TIntV(NumRows));
  TInt ColIdx = IntCols.Len()-1;
  TInt i = 0;
  for (TRowIterator RI = BegRI(); RI < EndRI(); RI++) {
    IntCols[ColIdx][RI.GetRowIdx()] = ColVals[i];
    i++;
  }
  TInt L = IntCols.Len();
  AddColType(ColName, atInt, L-1);
}

void TTable::StoreStrCol	(	const TStr &	ColName,
		const TStrV &	ColVals
	)

Adds entire str column to table.

Definition at line 4098 of file table.cpp.

                                                                  {
  if (ColVals.Len() != NumRows) {
    printf("new column dimension must agree with number of rows\n");
    return;
  }
  AddSchemaCol(ColName, atStr);
  StrColMaps.Add(TIntV(NumRows,0));
  TInt ColIdx = FltCols.Len()-1;
  TInt i = 0;
  for (TRowIterator RI = BegRI(); RI < EndRI(); RI++) {
    TInt Key = Context->StringVals.GetKeyId(ColVals[i]);
    if (Key == -1) { Context->StringVals.AddKey(ColVals[i]); }
    StrColMaps[ColIdx][RI.GetRowIdx()] = Key;
    i++;
  }
  TInt L = StrColMaps.Len();
  AddColType(ColName, atStr, L-1);
}

static PTable TTable::TableFromHashMap ( const THash< TInt, TInt > &  H, const TStr &  Col1, const TStr &  Col2, TTableContext *  Context, const TBool  IsStrKeys = false  )

inlinestatic

Builds table from hash table of int->int.

Definition at line 979 of file table.h.

                                                          {
    PTable T = New(H, Col1, Col2, Context, IsStrKeys);
    T->InitIds();
    return T;
  }

static PTable TTable::TableFromHashMap ( const THash< TInt, TFlt > &  H, const TStr &  Col1, const TStr &  Col2, TTableContext *  Context, const TBool  IsStrKeys = false  )

inlinestatic

Builds table from hash table of int->float.

Definition at line 986 of file table.h.

                                                          {
    PTable T = New(H, Col1, Col2, Context, IsStrKeys);
    T->InitIds();
    return T;
  }

PTable TTable::ThresholdJoin	(	const TStr &	KeyCol1,
		const TStr &	JoinCol1,
		const TTable &	Table,
		const TStr &	KeyCol2,
		const TStr &	JoinCol2,
		TInt	Threshold,
		TBool	PerJoinKey = false
	)

Definition at line 2624 of file table.cpp.

                                                                              {
  // test input correctness
  ThresholdJoinInputCorrectness(KeyCol1, JoinCol1, Table, KeyCol2, JoinCol2);
  //printf("verified input correctness\n");
  // type of column on which we join (currently support only int)
  TAttrType JoinColType = GetColType(JoinCol1);
  // type of key column (currently support only int)
  TAttrType KeyType = GetColType(KeyCol1);
  // Determine which table is smaller
  TBool ThisIsSmaller = (NumValidRows <= Table.NumValidRows);
  const TTable& TS = ThisIsSmaller ? *this : Table;
  const TTable& TB = ThisIsSmaller ?  Table : *this;
  TStr JoinColS = JoinCol1;
  TInt JoinColIdxB = GetColIdx(JoinCol2);
  TInt KeyColIdxS = GetColIdx(KeyCol1);
  TInt KeyColIdxB = GetColIdx(KeyCol2);
  if(!ThisIsSmaller){
        JoinColS = JoinCol2;
    JoinColIdxB = GetColIdx(JoinCol1);
        KeyColIdxS = GetColIdx(KeyCol2);
    KeyColIdxB = GetColIdx(KeyCol1);
  }
  
  // debug print
  //printf("JoinColS = %d, JoinColIdxB = %d, KeyColIdxS = %d, KeyColIdxB = %d\n", 
        //GetColIdx(JoinColS).Val, JoinColIdxB.Val, KeyColIdxS.Val, KeyColIdxB.Val);
  //printf("starting switch-case\n");
  
  if(KeyType != atInt && KeyType != atStr){
    printf("ThresholdJoin only supports integer or string key attributes\n");
    TExcept::Throw("ThresholdJoin only supports integer or string key attributes");
  }
  if(JoinColType != atInt && JoinColType != atStr){
    printf("ThresholdJoin only supports integer or string join attributes\n");
    TExcept::Throw("ThresholdJoin only supports integer or string join attributes");
  }
  //printf("starting the real stuff!\n");
  // hash the smaller table T: join col value --> physical row ids of rows with that value
  TIntIntVH T;
  if(JoinColType == atInt){
    TS.GroupByIntCol(JoinColS, T, TIntV(), true);
  } else if(JoinColType == atStr){
    TS.GroupByStrCol(JoinColS, T, TIntV(), true);
  } else{
    TExcept::Throw("ThresholdJoin only supports integer or string join attributes");
  } 
  
 /*
  for(THash<TInt,TIntV>::TIter it = T.BegI(); it < T.EndI(); it++){
    if(JoinColType == atStr){
      printf("%s -->", Context.StringVals.GetKey(it.GetKey().Val));
    } else{
      printf("%d -->", it.GetKey().Val);
    }
    const TIntV& V = it.GetDat();
    for(int sr = 0; sr < V.Len(); sr++){
      printf(" %d", V[sr].Val);
    }
    printf("\n");
  }
  */
  
  // Counters: (K1,K2) --> (RowIdx1,RowIdx2, count) where K1 is a key from KeyCol1, 
  // K2 is a key from Table's KeyCol2; RowIdx1 and RowIdx2 are physical row ids
  // that participates in a joint tuple that satisfies (1).
  // count is the count of joint records that satisfy (1).
  // In case of string attributes - the integer mappings of the key attribute values are used.
  if(PerJoinKey){
    //printf("PerJoinKey\n");
    THash<TIntTr,TIntTr> Counters;
    ThresholdJoinCountPerJoinKeyCollisions(TB, TS, T, JoinColIdxB, KeyColIdxB, KeyColIdxS, Counters, ThisIsSmaller, JoinColType, KeyType);
    /*
    for(THash<TIntTr,TIntTr>::TIter it = Counters.BegI(); it < Counters.EndI(); it++){
      const TIntTr& K = it.GetKey();
      const TIntTr& V = it.GetDat();
      if(KeyType == atStr){
        printf("%s %s --> %d %d %d\n", Context->StringVals.GetKey(K.Val1), Context->StringVals.GetKey(K.Val2), V.Val1.Val, V.Val2.Val, V.Val3.Val);
      } else{
        printf("%d %d --> %d %d %d\n", K.Val1.Val, K.Val2.Val, V.Val1.Val, V.Val2.Val, V.Val3.Val); 
      }
    }
    */
    //printf("found collisions\n");
    return ThresholdJoinPerJoinKeyOutputTable(Counters, Threshold, Table);
  } else{
    //printf("not PerJoinKey\n");
    THash<TIntPr,TIntTr> Counters;
    ThresholdJoinCountCollisions(TB, TS, T, JoinColIdxB, KeyColIdxB, KeyColIdxS, Counters, ThisIsSmaller, JoinColType, KeyType);
    /*
    for(THash<TIntPr,TIntTr>::TIter it = Counters.BegI(); it < Counters.EndI(); it++){
      const TIntPr& K = it.GetKey();
      const TIntTr& V = it.GetDat();
      if(KeyType == atStr){
        printf("%s %s --> %d %d %d\n", Context->StringVals.GetKey(K.Val1), Context->StringVals.GetKey(K.Val2), V.Val1.Val, V.Val2.Val, V.Val3.Val);
      } else{
        printf("%d %d --> %d %d %d\n", K.Val1.Val, K.Val2.Val, V.Val1.Val, V.Val2.Val, V.Val3.Val); 
      }
    }
    */
    //printf("found collisions\n");
    return ThresholdJoinOutputTable(Counters, Threshold, Table);
  }
}

void TTable::ThresholdJoinCountCollisions ( const TTable &  TB, const TTable &  TS, const TIntIntVH &  T, TInt  JoinColIdxB, TInt  KeyColIdxB, TInt  KeyColIdxS, THash< TIntPr, TIntTr > &  Counters, TBool  ThisIsSmaller, TAttrType  JoinColType, TAttrType  KeyType  )

protected

Definition at line 2486 of file table.cpp.

                                                                                                {
    // iterate over big table and count / record joint tuples
    for (TRowIterator RowI = TB.BegRI(); RowI < TB.EndRI(); RowI++) {
    // value to join on from big table
      TInt JVal = 0;
      if(JoinColType == atStr){
        JVal = RowI.GetStrMapById(JoinColIdxB);
      } else{
        JVal = RowI.GetIntAttr(JoinColIdxB);
      }
      //printf("JVal: %d\n", JVal.Val);
      if(T.IsKey(JVal)){
        // read key attribute of big table row
        TInt KeyB = 0;
        if(KeyType == atStr){
          KeyB = RowI.GetStrMapById(KeyColIdxB);
        } else{
          KeyB = RowI.GetIntAttr(KeyColIdxB);
        } 
        // read row ids from small table with join attribute value of JVal
        const TIntV& RelevantRows = T.GetDat(JVal);
        for(int i = 0; i < RelevantRows.Len(); i++){
          // read key attribute of relevant row from small table
          TInt KeyS = 0;
          if(KeyType == atStr){
            KeyS = TS.StrColMaps[KeyColIdxS][RelevantRows[i]];
          } else{
            KeyS = TS.IntCols[KeyColIdxS][RelevantRows[i]];
          }
          // create a pair of keys - serves as a key in Counters
          TIntPr Keys = ThisIsSmaller ? TIntPr(KeyS, KeyB) : TIntPr(KeyB, KeyS);
          if(Counters.IsKey(Keys)){
          // if the key pair has been seen before - increment its counter by 1
            TIntTr& V = Counters.GetDat(Keys);
            V.Val3 = V.Val3 + 1;
          } else{
            // if the key pair hasn't been seen before - add it with value of 
            // row indices that create a joint record with this key pair
            if(ThisIsSmaller){
              Counters.AddDat(Keys, TIntTr(RelevantRows[i], RowI.GetRowIdx(),1));
            } else{
              Counters.AddDat(Keys, TIntTr(RowI.GetRowIdx(), RelevantRows[i],1));
            }
          }
        }       // end of for loop
      } // end of if statement
    } // end of for loop
}

void TTable::ThresholdJoinCountPerJoinKeyCollisions ( const TTable &  TB, const TTable &  TS, const TIntIntVH &  T, TInt  JoinColIdxB, TInt  KeyColIdxB, TInt  KeyColIdxS, THash< TIntTr, TIntTr > &  Counters, TBool  ThisIsSmaller, TAttrType  JoinColType, TAttrType  KeyType  )

protected

Definition at line 2537 of file table.cpp.

                                                                                                {
    for (TRowIterator RowI = TB.BegRI(); RowI < TB.EndRI(); RowI++) {
      // value to join on from big table
      TInt JVal = 0;
      if(JoinColType == atStr){
        JVal = RowI.GetStrMapById(JoinColIdxB);
       } else{
        JVal = RowI.GetIntAttr(JoinColIdxB);
       }
      //printf("JVal: %d\n", JVal.Val);
      if(T.IsKey(JVal)){
        // read key attribute of big table row
        TInt KeyB = 0;
        if(KeyType == atStr){
          KeyB = RowI.GetStrMapById(KeyColIdxB);
        } else{
          KeyB = RowI.GetIntAttr(KeyColIdxB);
        } 
        // read row ids from small table with join attribute value of JVal
        const TIntV& RelevantRows = T.GetDat(JVal);
        for(int i = 0; i < RelevantRows.Len(); i++){
          // read key attribute of relevant row from small table
          TInt KeyS = 0;
          if(KeyType == atStr){
            KeyS = TS.StrColMaps[KeyColIdxS][RelevantRows[i]];
          } else{
            KeyS = TS.IntCols[KeyColIdxS][RelevantRows[i]];
          }
                // create a pair of keys - serves as a key in Counters
                TIntPr Keys = ThisIsSmaller ? TIntPr(KeyS, KeyB) : TIntPr(KeyB, KeyS);
                TIntTr K(Keys.Val1,Keys.Val2,JVal);
          if(Counters.IsKey(K)){
            // if the key pair has been seen before - increment its counter by 1
            TIntTr& V = Counters.GetDat(K);
            V.Val3 = V.Val3 + 1;
          } else{
            // if the key pair hasn't been seen before - add it with value of 
            // row indices that create a joint record with this key pair
            if(ThisIsSmaller){
              Counters.AddDat(K, TIntTr(RelevantRows[i], RowI.GetRowIdx(),1));
            } else{
              Counters.AddDat(K, TIntTr(RowI.GetRowIdx(), RelevantRows[i],1));
            }
          }
        }       // end of for loop
      } // end of if statement
    } // end of for loop
  }

void TTable::ThresholdJoinInputCorrectness ( const TStr &  KeyCol1, const TStr &  JoinCol1, const TTable &  Table, const TStr &  KeyCol2, const TStr &  JoinCol2  )

protected

Definition at line 2458 of file table.cpp.

                                            {
  if (!IsColName(KeyCol1)) {
    printf("no such column %s\n", KeyCol1.CStr());
    TExcept::Throw("no such column " + KeyCol1);
  }
  if (!Table.IsColName(KeyCol2)) {
    printf("no such column %s\n", KeyCol2.CStr());
    TExcept::Throw("no such column " + KeyCol2);
  }
  if (!IsColName(JoinCol1)) {
    printf("no such column %s\n", JoinCol1.CStr());
    TExcept::Throw("no such column " + JoinCol1);
  }
  if (!Table.IsColName(JoinCol2)) {
    printf("no such column %s\n", JoinCol2.CStr());
    TExcept::Throw("no such column " + JoinCol2);
  }
  if (GetColType(JoinCol1) != Table.GetColType(JoinCol2)) {
    printf("Trying to Join on columns of different type\n");
    TExcept::Throw("Trying to Join on columns of different type");
  }
  if (GetColType(KeyCol1) != Table.GetColType(KeyCol2)) {
    printf("Key type mismatch\n");
    TExcept::Throw("Key type mismatch");
  }
}

PTable TTable::ThresholdJoinOutputTable ( const THash< TIntPr, TIntTr > &  Counters, TInt  Threshold, const TTable &  Table  )

protected

Definition at line 2588 of file table.cpp.

                                                                                                                {
  // initialize result table
  PTable JointTable = InitializeJointTable(Table);
  for(THash<TIntPr,TIntTr>::TIter iter = Counters.BegI(); iter < Counters.EndI(); iter++){
    TIntTr& Counter = iter.GetDat();
    //printf("keys: %d, %d\n", iter.GetKey().Val1.Val, iter.GetKey().Val2.Val);
    //printf("selected rows: %d,%d, counter: %d\n", Counter.Val1.Val, Counter.Val2.Val, Counter.Val3.Val);
    if(Counter.Val3 >= Threshold){
      JointTable->AddJointRow(*this, Table, Counter.Val1, Counter.Val2);
    }
  }
  return JointTable;
}

PTable TTable::ThresholdJoinPerJoinKeyOutputTable ( const THash< TIntTr, TIntTr > &  Counters, TInt  Threshold, const TTable &  Table  )

protected

Definition at line 2602 of file table.cpp.

                                                                                                                          {
  PTable JointTable = InitializeJointTable(Table);
  for(THash<TIntTr,TIntTr>::TIter iter = Counters.BegI(); iter < Counters.EndI(); iter++){
    const TIntTr& Counter = iter.GetDat();
    const TIntTr& Keys = iter.GetKey();
    THashSet<TIntPr> Pairs;
    if(Counter.Val3 >= Threshold){
      TIntPr K(Keys.Val1,Keys.Val2);
      if(!Pairs.IsKey(K)){
        Pairs.AddKey(K);
        JointTable->AddJointRow(*this, Table, Counter.Val1, Counter.Val2);
      }
    }
  }
  return JointTable;
}

TVec< PNEANet > TTable::ToGraphPerGroup	(	TStr	GroupAttr,
		TAttrAggr	AggrPolicy
	)

Creates a sequence of graphs based on grouping specified by GroupAttr.

Definition at line 3640 of file table.cpp.

                                                                          {
  return ToGraphSequence(GroupAttr, AggrPolicy, TInt(1), TInt(1), TInt::Mn, TInt::Mx);
}

PNEANet TTable::ToGraphPerGroupIterator	(	TStr	GroupAttr,
		TAttrAggr	AggrPolicy
	)

Creates the graph sequence one at a time.

Create the graph sequence one at a time, to allow efficient use of memory. A call to this function must be followed by subsequent calls to NextGraphIterator().

Definition at line 3654 of file table.cpp.

                                                                            {
  return ToGraphSequenceIterator(GroupAttr, AggrPolicy, TInt(1), TInt(1), TInt::Mn, TInt::Mx);
}

TVec< PNEANet > TTable::ToGraphSequence	(	TStr	SplitAttr,
		TAttrAggr	AggrPolicy,
		TInt	WindowSize,
		TInt	JumpSize,
		TInt	StartVal = TInt::Mn,
		TInt	EndVal = TInt::Mx
	)

Creates a sequence of graphs based on values of column SplitAttr and windows specified by JumpSize and WindowSize.

Definition at line 3629 of file table.cpp.

                                                                                                                                      {
  FillBucketsByWindow(SplitAttr, JumpSize, WindowSize, StartVal, EndVal);
  printf("buckets filled\n");
  return GetGraphsFromSequence(AggrPolicy);  
}

PNEANet TTable::ToGraphSequenceIterator	(	TStr	SplitAttr,
		TAttrAggr	AggrPolicy,
		TInt	WindowSize,
		TInt	JumpSize,
		TInt	StartVal = TInt::Mn,
		TInt	EndVal = TInt::Mx
	)

Creates the graph sequence one at a time.

Create the graph sequence one at a time, to allow efficient use of memory. A call to this function must be followed by subsequent calls to NextGraphIterator().

Definition at line 3644 of file table.cpp.

                                                                                                                                        {
  FillBucketsByWindow(SplitAttr, JumpSize, WindowSize, StartVal, EndVal);
  return GetFirstGraphFromSequence(AggrPolicy);  
}

TVec< PNEANet > TTable::ToVarGraphSequence	(	TStr	SplitAttr,
		TAttrAggr	AggrPolicy,
		TIntPrV	SplitIntervals
	)

Creates a sequence of graphs based on values of column SplitAttr and intervals specified by SplitIntervals.

Definition at line 3635 of file table.cpp.

                                                                                                     {
  FillBucketsByInterval(SplitAttr, SplitIntervals);
  return GetGraphsFromSequence(AggrPolicy);
}

PNEANet TTable::ToVarGraphSequenceIterator	(	TStr	SplitAttr,
		TAttrAggr	AggrPolicy,
		TIntPrV	SplitIntervals
	)

Creates the graph sequence one at a time.

Create the graph sequence one at a time, to allow efficient use of memory. A call to this function must be followed by subsequent calls to NextGraphIterator().

Definition at line 3649 of file table.cpp.

                                                                                                       {
  FillBucketsByInterval(SplitAttr, SplitIntervals);
  return GetFirstGraphFromSequence(AggrPolicy);
}

PTable TTable::Union

(

const TTable &

Table

)

Returns union of this table with given Table.

Definition at line 4508 of file table.cpp.

                                        {
  Schema NewSchema;
  THashSet<TInt> Collisions;
  TStrV ColNames;

  for (TInt c = 0; c < Sch.Len(); c++) {
    if (Sch[c].Val1 != GetIdColName()) {
      NewSchema.Add(TPair<TStr,TAttrType>(Sch[c].Val1, Sch[c].Val2));
      ColNames.Add(Sch[c].Val1);
    }
  }
  PTable result = TTable::New(NewSchema, Context);
  
  GetCollidingRows(Table, Collisions); 
  
  result->AddTable(*this);

  result->Unique(ColNames);

  // this part should be made faster by adding all the rows in one go
  for (TRowIterator it = Table.BegRI(); it < Table.EndRI(); it++) {
    if (!Collisions.IsKey(it.GetRowIdx())) {
      result->AddRow(it);
    }
  }

  // printf("this: %d %d, table: %d %d, result: %d %d\n", 
  //   this->GetNumRows().Val, this->GetNumValidRows().Val,
  //   Table.GetNumRows().Val, Table.GetNumValidRows().Val, 
  //   result->GetNumRows().Val, result->GetNumValidRows().Val);

  result->InitIds();
  return result;
}

PTable TTable::Union ( const PTable &  Table )

inline

Definition at line 1404 of file table.h.

{ return Union(*Table); };

PTable TTable::UnionAll

(

const TTable &

Table

)

Returns union of this table with given Table, preserving duplicates.

Definition at line 4488 of file table.cpp.

                                           {
  Schema NewSchema;
  for (TInt c = 0; c < Sch.Len(); c++) {
    if (Sch[c].Val1 != GetIdColName()) {
      NewSchema.Add(TPair<TStr,TAttrType>(Sch[c].Val1, Sch[c].Val2));
    }
  }
  PTable result = TTable::New(NewSchema, Context);
  result->AddTable(*this);
  result->UnionAllInPlace(Table);
  return result;
}

PTable TTable::UnionAll ( const PTable &  Table )

inline

Definition at line 1407 of file table.h.

{ return UnionAll(*Table); };

void TTable::UnionAllInPlace

(

const TTable &

Table

)

Same as TTable::ConcatTable.

Definition at line 4501 of file table.cpp.

                                                {
  AddTable(Table);
  // TODO: For the moment, IDs are not initialized (to avoid having too many ID columns)
  //result->InitIds();
}

void TTable::UnionAllInPlace ( const PTable &  Table )

inline

Definition at line 1410 of file table.h.

{ return UnionAllInPlace(*Table); };

void TTable::Unique

(

const TStr &

Col

)

Removes rows with duplicate values in given column.

Definition at line 1246 of file table.cpp.

                                   {
  TIntV RemainingRows;
  TStr NCol = NormalizeColName(Col);
  switch (GetColType(NCol)) {
    case atInt: {
      TIntIntVH Grouping;
      GroupByIntCol(NCol, Grouping, TIntV(), true, true);
      for (TIntIntVH::TIter it = Grouping.BegI(); it < Grouping.EndI(); it++) {
        RemainingRows.Add(it->Dat[0]);
      }
      break;
    }
    case atFlt: {
      THash<TFlt,TIntV> Grouping;
      GroupByFltCol(NCol, Grouping, TIntV(), true, true);
      for (THash<TFlt,TIntV>::TIter it = Grouping.BegI(); it < Grouping.EndI(); it++) {
        RemainingRows.Add(it->Dat[0]);
      }
      break;
    } 
    case atStr: {
      TIntIntVH Grouping;
      GroupByStrCol(NCol, Grouping, TIntV(), true, true);
      for (TIntIntVH::TIter it = Grouping.BegI(); it < Grouping.EndI(); it++) {
        RemainingRows.Add(it->Dat[0]);
      }
      break;
    }
  }
  KeepSortedRows(RemainingRows);
}

void TTable::Unique	(	const TStrV &	Cols,
		TBool	Ordered = true
	)

Removes rows with duplicate values in given columns.

Definition at line 1278 of file table.cpp.

                                                    {
  if(Cols.Len() == 1){ 
        Unique(Cols[0]);
        return;
  }
  TStrV NCols = NormalizeColNameV(Cols);
  THash<TGroupKey, TPair<TInt, TIntV> > Grouping;
  TIntV UniqueVec;
  GroupAux(NCols, Grouping, Ordered, "", true, UniqueVec, true);
  KeepSortedRows(UniqueVec);
}

void TTable::UpdateFltFromTable	(	const TStr &	KeyAttr,
		const TStr &	UpdateAttr,
		const TTable &	Table,
		const TStr &	FKeyAttr,
		const TStr &	ReadAttr,
		TFlt	DefaultFltVal = 0.0
	)

Definition at line 4219 of file table.cpp.

                                                                 {
  if(!IsColName(KeyAttr)){ TExcept::Throw("Bad KeyAttr parameter");}
  if(!IsColName(UpdateAttr)){ TExcept::Throw("Bad UpdateAttr parameter");}
  if(!Table.IsColName(FKeyAttr)){ TExcept::Throw("Bad FKeyAttr parameter");}
  if(!Table.IsColName(ReadAttr)){ TExcept::Throw("Bad ReadAttr parameter");}
  
#ifdef GCC_ATOMIC
  if(GetMP()){
    UpdateFltFromTableMP(KeyAttr, UpdateAttr,Table, FKeyAttr, ReadAttr, DefaultFltVal);
    return;
  }
#endif  // GCC_ATOMIC
        
  TAttrType KeyType = GetColType(KeyAttr);
  TAttrType FKeyType = Table.GetColType(FKeyAttr);
  if(KeyType != FKeyType){TExcept::Throw("Key Type Mismatch");}
  if(GetColType(UpdateAttr) != atFlt || Table.GetColType(ReadAttr) != atFlt){
    TExcept::Throw("Expecting Float values");
  }
  TStr NKeyAttr = NormalizeColName(KeyAttr);
  TStr NUpdateAttr = NormalizeColName(UpdateAttr);
  TStr NFKeyAttr = Table.NormalizeColName(FKeyAttr);
  TStr NReadAttr = Table.NormalizeColName(ReadAttr);
  TInt UpdateColIdx = GetColIdx(UpdateAttr);
        
  for(TRowIterator iter = BegRI(); iter < EndRI(); iter++){
    FltCols[UpdateColIdx][iter.GetRowIdx()] = DefaultFltVal;
  }
        
  switch(KeyType) {
    // TODO: add support for other cases of KeyType
    case atInt: {
        TIntIntVH Grouping;
        GroupByIntCol(NKeyAttr, Grouping, TIntV(), true, true);
        for (TRowIterator RI = Table.BegRI(); RI < Table.EndRI(); RI++) {
          TInt K = RI.GetIntAttr(NFKeyAttr);
          if (Grouping.IsKey(K)) {
            TIntV& UpdateRows = Grouping.GetDat(K);
            for (int i = 0; i < UpdateRows.Len(); i++) {
              FltCols[UpdateColIdx][UpdateRows[i]] = RI.GetFltAttr(NReadAttr);
            } // end of for loop
          } // end of if statement
        } // end of for loop
      } // end of case atInt
      break;
    default:
      break;
  } // end of outer switch statement
}

void TTable::UpdateFltFromTableMP	(	const TStr &	KeyAttr,
		const TStr &	UpdateAttr,
		const TTable &	Table,
		const TStr &	FKeyAttr,
		const TStr &	ReadAttr,
		TFlt	DefaultFltVal = 0.0
	)

Definition at line 4151 of file table.cpp.

                        {
        if (!GetMP()) {
    TExcept::Throw("Not Using MP!");
  }

  TAttrType KeyType = GetColType(KeyAttr);
  TAttrType FKeyType = Table.GetColType(FKeyAttr);
  if(KeyType != FKeyType){TExcept::Throw("Key Type Mismatch");}
  if(GetColType(UpdateAttr) != atFlt || Table.GetColType(ReadAttr) != atFlt){
    TExcept::Throw("Expecting Float values");
  }
  TStr NKeyAttr = NormalizeColName(KeyAttr);
  //TStr NUpdateAttr = NormalizeColName(UpdateAttr);
  //TStr NFKeyAttr = Table.NormalizeColName(FKeyAttr);
  //TStr NReadAttr = Table.NormalizeColName(ReadAttr);
  TInt UpdateColIdx = GetColIdx(UpdateAttr);
  TInt FKeyColIdx = GetColIdx(FKeyAttr);
  TInt ReadColIdx = GetColIdx(ReadAttr);

  // TODO: this should be a generic vector operation
  SetFltColToConstMP(UpdateColIdx, DefaultFltVal);
                
        TIntPrV Partitions;
        Table.GetPartitionRanges(Partitions, omp_get_max_threads()*CHUNKS_PER_THREAD);
        TInt PartitionSize = Partitions[0].GetVal2()-Partitions[0].GetVal1()+1;
        TIntV Locks(NumRows);
        Locks.PutAll(0);        // need to parallelize this...

  switch (KeyType) {
    // TODO: add support for other cases of KeyType
    case atInt: {
        THashMP<TInt,TIntV> Grouping;
        // must use physical row ids
        GroupByIntColMP(NKeyAttr, Grouping, true);
        #pragma omp parallel for schedule(dynamic, CHUNKS_PER_THREAD) // num_threads(1)
                          for (int i = 0; i < Partitions.Len(); i++) {
                                  TRowIterator RowI(Partitions[i].GetVal1(), &Table);
                                  TRowIterator EndI(Partitions[i].GetVal2(), &Table);
                                  while (RowI < EndI) {
            TInt K = RowI.GetIntAttr(FKeyColIdx);
            if (Grouping.IsKey(K)) {
              TIntV& UpdateRows = Grouping.GetDat(K);
              for (int j = 0; j < UpdateRows.Len(); j++) {
                int* lock = &Locks[UpdateRows[j]].Val;
                // OP RS 2016/06/30: needed to define a wrapper function
                //   for the code to compile on Mac OS X gcc 4.2.1
                //if (!__sync_bool_compare_and_swap(lock, 0, 1)) {
                if (!sync_bool_compare_and_swap(lock)) {
                  continue;
                }
                //printf("key = %d, row = %d, old_score = %f\n", K.Val, j, UpdateRows[j].Val, FltCols[UpdateColIdx][UpdateRows[j]].Val);
                                                          FltCols[UpdateColIdx][UpdateRows[j]] = RowI.GetFltAttr(ReadColIdx);
                                                          //printf("key = %d, new_score = %f\n", K.Val, j, FltCols[UpdateColIdx][UpdateRows[j]].Val);
              } // end of for loop
            } // end of if statement
            RowI++;
          } // end of while loop
        }       // end of for loop
      } // end of case atInt
      break;
    default:
      break;
  } // end of outer switch statement
}

template<class T >

void TTable::UpdateGrouping ( THash< T, TIntV > &  Grouping, T  Key, TInt  Val  ) const

protected

Template for utility function to update a grouping hash map.

Definition at line 1687 of file table.h.

                                                                          {
  if (Grouping.IsKey(Key)) {
    Grouping.GetDat(Key).Add(Val);
  } else {
    TIntV NewGroup;
    NewGroup.Add(Val);
    Grouping.AddDat(Key, NewGroup);
  }
}

template<class T >

void TTable::UpdateGrouping ( THashMP< T, TIntV > &  Grouping, T  Key, TInt  Val  ) const

protected

Template for utility function to update a parallel grouping hash map.

Definition at line 1699 of file table.h.

                                                                            {
  if (Grouping.IsKey(Key)) {
        //printf("y\n");
    Grouping.GetDat(Key).Add(Val);
  } else {
        //printf("n\n");
    TIntV NewGroup;
    NewGroup.Add(Val);
    Grouping.AddDat(Key, NewGroup);
  }
}

void TTable::UpdateTableForNewRow ( )

protected

Updates table state after adding one or more rows.

Definition at line 4117 of file table.cpp.

                                  {
  if (LastValidRow >= 0) {
    Next[LastValidRow] = NumRows;
  }
  Next.Add(Last);
  LastValidRow = NumRows;

  NumRows++;
  NumValidRows++;
}

Friends And Related Function Documentation

friend class TPt< TTable >

friend

Definition at line 1532 of file table.h.

friend class TRowIterator

friend

Definition at line 1533 of file table.h.

friend class TRowIteratorWithRemove

friend

Definition at line 1534 of file table.h.

int TSnap::LoadCrossNet ( TCrossNet &  Graph, PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TStrV &  EdgeAttrV  )

friend

int TSnap::LoadMode ( TModeNet &  Graph, PTable  Table, const TStr &  NCol, TStrV &  NodeAttrV  )

friend

template<class PGraph >

PGraph TSnap::ToGraph ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TAttrAggr  AggrPolicy  )

friend

template<class PGraphMP >

PGraphMP TSnap::ToGraphMP ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol  )

friend

template<class PGraphMP >

PGraphMP TSnap::ToGraphMP3 ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol  )

friend

template<class PGraph >

PGraph TSnap::ToNetwork ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TStrV &  SrcAttrs, TStrV &  DstAttrs, TStrV &  EdgeAttrs, TAttrAggr  AggrPolicy  )

friend

template<class PGraph >

PGraph TSnap::ToNetwork ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TStrV &  EdgeAttrV, TAttrAggr  AggrPolicy  )

friend

template<class PGraph >

PGraph TSnap::ToNetwork ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TAttrAggr  AggrPolicy  )

friend

template<class PGraph >

PGraph TSnap::ToNetwork ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TStrV &  EdgeAttrV, PTable  NodeTable, const TStr &  NodeCol, TStrV &  NodeAttrV, TAttrAggr  AggrPolicy  )

friend

template<class PGraphMP >

PGraphMP TSnap::ToNetworkMP ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TStrV &  SrcAttrs, TStrV &  DstAttrs, TStrV &  EdgeAttrs, TAttrAggr  AggrPolicy  )

friend

template<class PGraphMP >

PGraphMP TSnap::ToNetworkMP ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TStrV &  EdgeAttrV, TAttrAggr  AggrPolicy  )

friend

template<class PGraphMP >

PGraphMP TSnap::ToNetworkMP ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TAttrAggr  AggrPolicy  )

friend

template<class PGraphMP >

PGraphMP TSnap::ToNetworkMP ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TStrV &  EdgeAttrV, PTable  NodeTable, const TStr &  NodeCol, TStrV &  NodeAttrV, TAttrAggr  AggrPolicy  )

friend

template<class PGraphMP >

PGraphMP TSnap::ToNetworkMP2 ( PTable  Table, const TStr &  SrcCol, const TStr &  DstCol, TStrV &  SrcAttrs, TStrV &  DstAttrs, TStrV &  EdgeAttrs, TAttrAggr  AggrPolicy  )

friend

Member Data Documentation

TAttrAggr TTable::AggrPolicy

protected

Aggregation policy used for solving conflicts between different values of an attribute of the same node.

Definition at line 611 of file table.h.

THash<TStr,TPair<TAttrType,TInt> > TTable::ColTypeMap

protected

String columns are implemented using a string pool to fight memory fragmentation. The value of string column c in row r is Context.StringVals.GetKey(StrColMaps[c][r])

Definition at line 574 of file table.h.

TStrTrV TTable::CommonNodeAttrs

protected

List of attribute pairs with values common to source and destination and their common given name.

Definition at line 604 of file table.h.

TTableContext* TTable::Context

protected

Execution Context.

Definition at line 555 of file table.h.

TCRef TTable::CRef

protected

Definition at line 560 of file table.h.

TInt TTable::CurrBucket

protected

Current row id bucket - used when generating a sequence of graphs using an iterator.

Definition at line 610 of file table.h.

TStr TTable::DstCol

protected

Column (attribute) to serve as dst nodes when constructing the graph.

Definition at line 600 of file table.h.

TStrV TTable::DstNodeAttrV

protected

List of columns (attributes) to serve as destination node attributes.

Definition at line 603 of file table.h.

TStrV TTable::EdgeAttrV

protected

List of columns (attributes) to serve as edge attributes.

Definition at line 601 of file table.h.

TInt TTable::FirstValidRow

protected

Physical index of first valid row.

Definition at line 563 of file table.h.

THash<TStr, THash<TFlt, TIntV> > TTable::FltColIndexes

protected

Indexes for Float Columns.

Definition at line 580 of file table.h.

TVec<TFltV> TTable::FltCols

protected

Data columns of floating point attributes.

Definition at line 569 of file table.h.

THash<GroupStmt, THash<TInt, TGroupKey> > TTable::GroupIDMapping

protected

Maps grouping statements to their (group id –> group-by key) mapping.

A mapping between the newly-added group id column name of a grouping statement to a vector of the group-by attribute names and a flag specifying whether those attributes are ordered or not.

Definition at line 587 of file table.h.

THash<GroupStmt, THash<TGroupKey, TIntV> > TTable::GroupMapping

protected

Maps grouping statements to their (group-by key –> group id) mapping.

A mapping between grouping statement (group-by attribute names and 'Ordered' flag) to a hash map between given group ids to their corresponding group-by key.

Definition at line 591 of file table.h.

THash<TStr, GroupStmt > TTable::GroupStmtNames

protected

Maps user-given grouping statement names to their group-by attributes.

Definition at line 583 of file table.h.

TStr TTable::IdColName

protected

A mapping from column name to column type and column index among columns of the same type.

Name of column associated with (optional) permanent row identifiers.

Definition at line 575 of file table.h.

THash<TStr, THash<TInt, TIntV> > TTable::IntColIndexes

protected

Indexes for Int Columns.

Definition at line 578 of file table.h.

TVec<TIntV> TTable::IntCols

protected

Next[i] is the successor of row i. Table iterators follow the order dictated by Next

Data columns of integer attributes.

Definition at line 568 of file table.h.

TInt const TTable::Invalid = -2

staticprotected

Special value for Next vector entry - logically removed row.

Definition at line 498 of file table.h.

TInt TTable::IsNextDirty

protected

Flag to signify whether the rows are stored in logical sequence or reordered. Used for optimizing GetPartitionRanges.

Definition at line 613 of file table.h.

TInt const TTable::Last = -1

staticprotected

Special value for Next vector entry - last row in table.

Definition at line 497 of file table.h.

TInt TTable::LastValidRow

protected

Physical index of last valid row.

Definition at line 564 of file table.h.

TIntV TTable::Next

protected

A vector describing the logical order of the rows.

Definition at line 565 of file table.h.

TInt TTable::NumRows

protected

Number of rows in the table (valid and invalid).

Definition at line 561 of file table.h.

TInt TTable::NumValidRows

protected

Number of valid rows in the table (i.e. rows that were not logically removed).

Definition at line 562 of file table.h.

TVec<TIntV> TTable::RowIdBuckets

protected

Partitioning of row ids into buckets corresponding to different graph objects when generating a sequence of graphs.

Example: <T_1.age,T_2.age, age> - T_1.age is a src node attribute, T_2.age is a dst node attribute. However, since all nodes refer to the same universe of entities (users) we just do one assignment of age per node, and call that attribute 'age'. This list should be very small.

Definition at line 609 of file table.h.

TIntIntH TTable::RowIdMap

protected

Mapping of permanent row ids to physical id.

Definition at line 576 of file table.h.

Schema TTable::Sch

protected

Table Schema.

Execution context includes a global string pool for all string values of tables in current session. Access to the pool is done via Context.StringVals.

Definition at line 559 of file table.h.

TStr TTable::SrcCol

protected

Column (attribute) to serve as src nodes when constructing the graph.

Definition at line 599 of file table.h.

TStrV TTable::SrcNodeAttrV

protected

List of columns (attributes) to serve as source node attributes.

Definition at line 602 of file table.h.

TVec<TIntV> TTable::StrColMaps

protected

Data columns of integer mappings of string attributes.

Definition at line 570 of file table.h.

THash<TStr, THash<TInt, TIntV> > TTable::StrMapColIndexes

protected

Indexes for String Columns.

Definition at line 579 of file table.h.

TInt TTable::UseMP = 1

staticprotected

Global switch for choosing multi-threaded versions of TTable functions.

Definition at line 500 of file table.h.

---

The documentation for this class was generated from the following files:

• snap-core/table.h
• snap-core/table.cpp