TTreeCloner

Data Members

public:

enum ECloneMethod {	kDefault
	kSortBasketsByBranch
	kSortBasketsByOffset
	kSortBasketsByEntry
};

private:

UInt_t*	fBasketBranchNum	[fMaxBaskets] Index of the branch(es) of the basket.
Long64_t*	fBasketEntry	[fMaxBaskets] list of basket start entries.
UInt_t*	fBasketIndex	[fMaxBaskets] ordered list of basket indices to be written.
UInt_t*	fBasketNum	[fMaxBaskets] index of the basket within the branch.
Long64_t*	fBasketSeek	[fMaxBaskets] list of basket position to be read.
UInt_t	fCloneMethod	Indicates which cloning method was selected.
TObjArray	fFromBranches
TTree*	fFromTree
Bool_t	fIsValid
UInt_t	fMaxBaskets
Option_t*	fMethod
UShort_t	fPidOffset	Offset to be added to the copied key/basket.
TObjArray	fToBranches
Long64_t	fToStartEntries	Number of entries in the target tree before any addition.
TTree*	fToTree

Class Charts

Function documentation

TTreeCloner(TTree* from, TTree* to, Option_t* method)

 Constructor.  This object would transfer the data from
 'from' to 'to' using the method indicated in method.

 The value of the parameter 'method' determines in which
 order the branches' baskets are written to the output file.

 When a TTree is filled the data is stored in the individual
 branches' basket.  Each basket is written individually to
 the disk as soon as it is full.  In consequence the baskets
 of branches that contain 'large' data chunk are written to
 the disk more often.

 There is currently 3 supported sorting order:
    SortBasketsByOffset (the default)
    SortBasketsByBranch
    SortBasketsByEntry

 When using SortBasketsByOffset the baskets are written in
 the output file in the same order as in the original file
 (i.e. the basket are sorted on their offset in the original
 file; Usually this also means that the baskets are sorted
 on the index/number of the _last_ entry they contain)

 When using SortBasketsByBranch all the baskets of each
 individual branches are stored contiguously.  This tends to
 optimize reading speed when reading a small number (1->5) of
 branches, since all their baskets will be clustered together
 instead of being spread across the file.  However it might
 decrease the performance when reading more branches (or the full
 entry).

 When using SortBasketsByEntry the baskets with the lowest
 starting entry are written first.  (i.e. the baskets are
 sorted on the index/number of the first entry they contain).
 This means that on the file the baskets will be in the order
 in which they will be needed when reading the whole tree
 sequentially.

Bool_t Exec()

 Execute the cloning.

~TTreeCloner()

 TTreeCloner destructor

void CloseOutWriteBaskets()

 Before we can start adding new basket, we need to flush to
 disk the partially filled baskets (the WriteBasket)

UInt_t CollectBranches(TBranch *from, TBranch *to)

 Fill the array of branches, adding the branch 'from' and 'to',
 and matching the sub-branches of the 'from' and 'to' branches.
 Returns the total number of baskets in all the from branch and
 it sub-branches.

	TTreeCloner(TTree* from, TTree* to, Option_t* method)
virtual	~TTreeCloner()
static TClass*	Class()
void	CloseOutWriteBaskets()
void	CollectBaskets()
UInt_t	CollectBranches()
UInt_t	CollectBranches(TBranch* from, TBranch* to)
UInt_t	CollectBranches(TObjArray* from, TObjArray* to)
void	CopyMemoryBaskets()
void	CopyProcessIds()
void	CopyStreamerInfos()
Bool_t	Exec()
virtual TClass*	IsA() const
Bool_t	IsValid()
virtual void	ShowMembers(TMemberInspector& insp, char* parent)
void	SortBaskets()
virtual void	Streamer(TBuffer& b)
void	StreamerNVirtual(TBuffer& b)
void	WriteBaskets()

class TTreeCloner

Function Members (Methods)

Data Members

Class Charts

Function documentation