Class for performing a radix sort (fast comparison-less sort based on byte value) on various standard STL containers. More...

Public Types
typedef TContainer::iterator	ContainerIter

Public Member Functions
	RadixSort ()

	~RadixSort ()

template<class TFunction >
void	sort (TContainer &container, TFunction func)
	Main sort function.

Protected Types
typedef std::vector< SortEntry, STLAllocator< SortEntry, GeneralAllocPolicy > >	SortVector
	Temp sort storage.

Protected Member Functions
void	finalPass (int byteIndex, float val)

void	finalPass (int byteIndex, int val)

template<typename T >
void	finalPass (int byteIndex, T val)

unsigned char	getByte (int byteIndex, TCompValueType val)

void	sortPass (int byteIndex)

Protected Attributes
int	mCounters [4][256]
	Alpha-pass counters of values (histogram) 4 of them so we can radix sort a maximum of a 32bit value.

SortVector *	mDest

int	mNumPasses
	Number of passes for this type.

int	mOffsets [256]
	Beta-pass offsets.

SortVector	mSortArea1

SortVector	mSortArea2

int	mSortSize
	Sort area size.

SortVector *	mSrc

TContainer	mTmpContainer

Detailed Description

template<class TContainer, class TContainerValueType, typename TCompValueType>
class Ogre::RadixSort< TContainer, TContainerValueType, TCompValueType >

Class for performing a radix sort (fast comparison-less sort based on byte value) on various standard STL containers.

Remarks: A radix sort is a very fast sort algorithm. It doesn't use comparisons and thus is able to break the theoretical minimum O(N*logN) complexity. Radix sort is complexity O(k*N), where k is a constant. Note that radix sorting is not in-place, it requires additional storage, so it trades memory for speed. The overhead of copying means that it is only faster for fairly large datasets, so you are advised to only use it for collections of at least a few hundred items.

: This is a template class to allow it to deal with a variety of containers, and a variety of value types to sort on. In addition to providing the container and value type on construction, you also need to supply a functor object which will retrieve the value to compare on for each item in the list. For example, if you had an std::vector of by-value instances of an object of class 'Bibble', and you wanted to sort on Bibble::getDoobrie(), you'd have to firstly create a functor like this:
struct BibbleSortFunctor

{

float operator()(const Bibble& val) const

{

return val.getDoobrie();

}

}

Ogre::SharedPtr
Reference-counted shared pointer, used for objects where implicit destruction is required.
Definition OgreSharedPtr.h:118

Then, you need to declare a RadixSort class which names the container type, the value type in the container, and the type of the value you want to sort by. You can then call the sort function. E.g.
RadixSort<BibbleList, Bibble, float> radixSorter;

BibbleSortFunctor functor;

radixSorter.sort(myBibbleList, functor);

You should try to reuse RadixSort instances, since repeated allocation of the internal storage is then avoided.

Note: Radix sorting is often associated with just unsigned integer values. Our implementation can handle both unsigned and signed integers, as well as floats (which are often not supported by other radix sorters). doubles are not supported; you will need to implement your functor object to convert to float if you wish to use this sort routine.

Definition at line 88 of file OgreRadixSort.h.