FM-index is a static, compact, and fast index for full-text search.
The index type, FMIndex{w,T}
, is able to index an arbitrary byte sequence.
w
is the number of bits required to encode the alphabet of the sequence and T
is the type of positions of the sequence.
julia> using FMIndexes
julia> fmindex = FMIndex("abracadabra");
julia> count("abra", fmindex) # count the number of occurrences of a query
2
julia> for loc in locate("ra", fmindex) # return the iterator of positions of a query
println(loc)
end
10
3
julia> locateall("ra", fmindex) # return the all positions of a query
2-element Array{Int64,1}:
10
3
julia> String(restore(fmindex)) # restore a byte sequence from the index
"abracadabra"
The following is a general constructor:
FMIndex(seq, σ=256; r=32, program=:SuffixArrays, mmap::Bool=false, opts...)
seq
is expected be a byte sequence; seq[i]
should return a value of UInt8
.
If the alphabet of a sequence can be encoded less that 8 bits, IntArrays.jl would be helpful to save the space.
σ
is the size of the alphabet; for example, if the sequence is a DNA sequence, setting σ
to 4 (four nucleotides) is the best choice in terms of efficiency.
Setting larger σ
value than necessary is just a waste of query time and index space.
The positions of the sequence are sampled every r
elements. There is a trade-off between query time and index space about this value: the smaller r
is, the faster it is to locate positions but the larger the index is.
program
is used to construct the suffix array of the sequence. The SuffixArrays.jl package is used by default, but if you want to create the index for a very long sequence it is recommended to use the pSAscan program.
Also, the mmap
flag determines wheather the suffix array is stored in a memory-mapped array or not. This flag would be necessary for a long sequence because the temporary suffix array often consumes larger memory space than the index itself (for instance, the suffix array of a sequence of 2^32 length consumes 16GiB RAM).