Class IntArrayFrontCodedBigList

All Implemented Interfaces:
BigList<int[]>, ObjectBigList<int[]>, ObjectCollection<int[]>, ObjectIterable<int[]>, Size64, Stack<int[]>, Serializable, Cloneable, Comparable<BigList<? extends int[]>>, Iterable<int[]>, Collection<int[]>, RandomAccess

public class IntArrayFrontCodedBigList extends AbstractObjectBigList<int[]> implements Serializable, Cloneable, RandomAccess
Compact storage of big lists of arrays using front-coding (also known as prefix-omission) compression.

This class stores immutably a big list of arrays in a single big array using front coding (of course, the compression will be reasonable only if the list is sorted lexicographically—see below). It implements an immutable type-specific list that returns the i-th array when calling get(i). The returned array may be freely modified.

Front-coding (also known as prefix-omission) compression is based on the idea that if the i-th and the (i+1)-th array have a common prefix, we might store the length of the common prefix, and then the rest of the second array.

This approach, of course, requires that once in a while an array is stored entirely. The ratio of a front-coded list defines how often this happens (once every ratio() arrays). A higher ratio means more compression, but means also a longer access time, as more arrays have to be probed to build the result. Note that we must build an array every time get(long) is called, but this class provides also methods that extract one of the stored arrays in a given array, reducing garbage collection. See the documentation of the family of get() methods.

By setting the ratio to 1 we actually disable front coding: however, we still have a data structure storing large list of arrays with a reduced overhead (just one integer per array, plus the space required for lengths).

Note that the typical usage of front-coded lists is under the form of serialized objects; usually, the data that has to be compacted is processed offline, and the resulting structure is stored permanently. Since the pointer array is not stored, the serialized format is very small.

Implementation Details

All arrays are stored in a big array. A separate array of pointers indexes arrays whose position is a multiple of the ratio: thus, a higher ratio means also less pointers.

More in detail, an array whose position is a multiple of the ratio is stored as the array length, followed by the elements of the array. The array length is coded by a simple variable-length list of k-1 bit blocks, where k is the number of bits of the underlying primitive type. All other arrays are stored as follows: let common the length of the maximum common prefix between the array and its predecessor. Then we store the array length decremented by common, followed by common, followed by the array elements whose index is greater than or equal to common. For instance, if we store foo, foobar, football and fool in a front-coded character-array list with ratio 3, the character array will contain

 3 f o o 3 3 b a r 5 3 t b a l l 4 f o o l
 
See Also:
  • Constructor Details

    • IntArrayFrontCodedBigList

      public IntArrayFrontCodedBigList(Iterator<int[]> arrays, int ratio)
      Creates a new front-coded list containing the arrays returned by the given iterator.
      Parameters:
      arrays - an iterator returning arrays.
      ratio - the desired ratio.
    • IntArrayFrontCodedBigList

      public IntArrayFrontCodedBigList(Collection<int[]> c, int ratio)
      Creates a new front-coded list containing the arrays in the given collection.
      Parameters:
      c - a collection containing arrays.
      ratio - the desired ratio.
  • Method Details

    • ratio

      public int ratio()
    • arrayLength

      public int arrayLength(long index)
      Computes the length of the array at the given index.
      Parameters:
      index - an index.
      Returns:
      the length of the index-th array.
    • get

      public int[] get(long index)
      Returns the element at the specified position.
      Specified by:
      get in interface BigList<int[]>
      Parameters:
      index - a position in the big list.
      Returns:
      the element at the specified position.
      See Also:
      Implementation Specification:
      This implementation delegates to getArray(long).
    • getArray

      public int[] getArray(long index)
      Returns an array stored in this front-coded list.
      Parameters:
      index - an index.
      Returns:
      the corresponding array stored in this front-coded list.
    • get

      public int get(long index, int[] a, int offset, int length)
      Stores in the given array elements from an array stored in this front-coded list.
      Parameters:
      index - an index.
      a - the array that will store the result.
      offset - an offset into a where elements will be store.
      length - a maximum number of elements to store in a.
      Returns:
      if a can hold the extracted elements, the number of extracted elements; otherwise, the number of remaining elements with the sign changed.
    • get

      public int get(long index, int[] a)
      Stores in the given array an array stored in this front-coded list.
      Parameters:
      index - an index.
      a - the array that will store the content of the result (we assume that it can hold the result).
      Returns:
      if a can hold the extracted elements, the number of extracted elements; otherwise, the number of remaining elements with the sign changed.
    • size64

      public long size64()
      Description copied from interface: Size64
      Returns the size of this data structure as a long.
      Specified by:
      size64 in interface Size64
      Returns:
      the size of this data structure.
    • listIterator

      public ObjectBigListIterator<int[]> listIterator(long start)
      Description copied from class: AbstractObjectBigList
      Returns a type-specific list iterator on this type-specific big list starting at a given index.
      Specified by:
      listIterator in interface BigList<int[]>
      Specified by:
      listIterator in interface ObjectBigList<int[]>
      Overrides:
      listIterator in class AbstractObjectBigList<int[]>
      Parameters:
      start - index of first element to be returned from the big-list iterator.
      Returns:
      a big-list iterator of the elements in this big list, starting at the specified position in this big list.
      See Also:
    • clone

      public IntArrayFrontCodedBigList clone()
      Returns a copy of this list.
      Returns:
      a copy of this list.
    • toString

      public String toString()
      Overrides:
      toString in class AbstractObjectBigList<int[]>
    • dump

      public void dump(DataOutputStream array, DataOutputStream pointers) throws IOException
      Throws:
      IOException