a C# implementation of data structures known as succinct bit vectors.

a bit vector augmented with an "index", such that

• the size of the index is "insignificant" compared to that of the vector itself,

• the index is exploited to answer the following queries in sublinear time:

  1. rank: returns the number of 0s or 1s within a given range of the vector

    rank : BitVector -> Nat -> {0, 1} -> Nat
    rank vec k b = |{i | i <- [0, ..., k) vec[i] = b }|
  

  (N.B. The above claims the specified range is exclusive: i<- [0, ..., k), some formalisms instead employ inclusive definition for a range: i <- [0, ..., k]) 2. select: returns the position of the k-th occurrence of 0 or 1:

    select : BitVector -> Nat -> {0, 1} -> Nat
    select vec k b = i such that vec[i] = b && rank vec (i+1) b = k
  

two distinct representation of bit vectors:

• BitVector stores an uncompressed bit sequence with its index, and
• RRRBitVector stores a compressed bits with index, a la Raman, Raman and Rao, 2002.

For both representations, an instance is built from a plain-vanilla bit sequence stored in the helper class Bits:

1. Load your bulk data into a list of bytes:

IList<byte> bulk = /* ...my precious... */;

2. Instantiate a Bits and copy the bulk:

Bits bits = new Bits(bulk.Length / 8 /* is the initial capacity */);
bits.push(bulk);

3. Build a bit vector from bits:

(RRR)BitVector bv = new (RRR)BitVector(bits);

Supported operations are:

• rank()/select() as described above
• get() for random access to an arbitrary bit (N.B. random access over a compressed vector is slow)
• write()/read() for serialization/de-serialization, respectively

See the interface IBitVector for the complete list.

Two helper classes are supplied besides:

• Bits: is, as introduced above, a container for plain-vanilla bit sequences. A sequence can be of an arbitrary length (i.e. not necessarily be a multiple of 8, 16, etc.)
• EliasFanoSequence: is another instance of "succinct data structures" to encode a non-decreasing finite sequence of natural numbers in a "space efficient" way. It is used as a part of compressed vector representation, but might be applied for other purposes in your code.

Googling "succinct bit vector" or "succinct data structures" will give you a bunch of well-organized introductory articles.

Specific publications and softwares which my implementation based upon are listed in each source code.

MIT License.