Class WAH8DocIdSet


  • public final class WAH8DocIdSet
    extends DocIdSet
    DocIdSet implementation based on word-aligned hybrid encoding on words of 8 bits.

    This implementation doesn't support random-access but has a fast DocIdSetIterator which can advance in logarithmic time thanks to an index.

    The compression scheme is simplistic and should work well with sparse and very dense doc id sets while being only slightly larger than a FixedBitSet for incompressible sets (overhead<2% in the worst case) in spite of the index.

    Format: The format is byte-aligned. An 8-bits word is either clean, meaning composed only of zeros or ones, or dirty, meaning that it contains between 1 and 7 bits set. The idea is to encode sequences of clean words using run-length encoding and to leave sequences of dirty words as-is.

    TokenClean length+Dirty length+Dirty words
    1 byte0-n bytes0-n bytes0-n bytes
    • Token encodes whether clean means full of zeros or ones in the first bit, the number of clean words minus 2 on the next 3 bits and the number of dirty words on the last 4 bits. The higher-order bit is a continuation bit, meaning that the number is incomplete and needs additional bytes to be read.
    • Clean length+: If clean length has its higher-order bit set, you need to read a vint, shift it by 3 bits on the left side and add it to the 3 bits which have been read in the token.
    • Dirty length+ works the same way as Clean length+ but on 4 bits and for the length of dirty words.
    • Dirty words are the dirty words, there are Dirty length of them.

    This format cannot encode sequences of less than 2 clean words and 0 dirty word. The reason is that if you find a single clean word, you should rather encode it as a dirty word. This takes the same space as starting a new sequence (since you need one byte for the token) but will be lighter to decode. There is however an exception for the first sequence. Since the first sequence may start directly with a dirty word, the clean length is encoded directly, without subtracting 2.

    There is an additional restriction on the format: the sequence of dirty words is not allowed to contain two consecutive clean words. This restriction exists to make sure no space is wasted and to make sure iterators can read the next doc ID by reading at most 2 dirty words.

    • Field Detail

      • DEFAULT_INDEX_INTERVAL

        public static final int DEFAULT_INDEX_INTERVAL
        Default index interval.
        See Also:
        Constant Field Values
    • Method Detail

      • intersect

        public static WAH8DocIdSet intersect​(Collection<WAH8DocIdSet> docIdSets,
                                             int indexInterval)
        Compute the intersection of the provided sets. This method is much faster than computing the intersection manually since it operates directly at the byte level.
      • union

        public static WAH8DocIdSet union​(Collection<WAH8DocIdSet> docIdSets,
                                         int indexInterval)
        Compute the union of the provided sets. This method is much faster than computing the union manually since it operates directly at the byte level.
      • isCacheable

        public boolean isCacheable()
        Description copied from class: DocIdSet
        This method is a hint for CachingWrapperFilter, if this DocIdSet should be cached without copying it. The default is to return false. If you have an own DocIdSet implementation that does its iteration very effective and fast without doing disk I/O, override this method and return true.
        Overrides:
        isCacheable in class DocIdSet
      • iterator

        public org.apache.lucene.util.WAH8DocIdSet.Iterator iterator()
        Description copied from class: DocIdSet
        Provides a DocIdSetIterator to access the set. This implementation can return null if there are no docs that match.
        Specified by:
        iterator in class DocIdSet
      • cardinality

        public int cardinality()
        Return the number of documents in this DocIdSet in constant time.
      • ramBytesUsed

        public long ramBytesUsed()
        Return the memory usage of this class in bytes.