Update: This work here is made obsolete by Ben Ward's superior NTupleKmers, which as of 2020-09-25 is planned to become the default kmer type in BioJulia

This is a proof-of-concept package. It is intended show a working example of a more general, very "hackable", yet still fast, kmer type for use in BioSequences.

Extra features compared to existing kmer types:

• Supports arbitrary alphabets
• Variable-length encoding from 8- to 1024 bits.

The type is parameterized by three parameters: Kmer{K, A, T}:

• K is the value of k
• A is the alphabet type, a subtype of Alphabet
• T is the storage type, a subtype of Unsigned

Arbitrary alphabets

julia> m = Kmer{4, CharAlphabet, UInt128}("读写汉字")
4-mer of CharAlphabet:
读写汉字

julia> reverse(m)
4-mer of CharAlphabet:
字汉写读

Supports large kmers through BitIntegers.jl

julia> Kmer(RNAAlphabet{2}, randrnaseq(500))
500-mer of RNAAlphabet{2}:
CAUAUGAUGGAUGGGUUUGGUGCGCAGACUUUAGGACUA…GGCAGUAGAUAAAUAUUCAACGGAGUGUCUAUAGCUGUG

julia> kmer"GWYFPPNML"aa
9-mer of AminoAcidAlphabet:
GWYFPPNML

Simple, efficient kmer iterator of any alphabet

julia>  it = SimpleKmerIterator{AAKmer{5}}(randaaseq(10^6))
Main.t.SimpleKmerIterator{Main.t.Kmer{5,AminoAcidAlphabet,UInt64},LongSequence{AminoAcidAlphabet}}(DDHMGFAHPCAYFVNMFPTCVFSAVSSTHNVKLYFTQTY…MGRIPSPPWCWHIIKDQGDIFWNHWLDSCCCRKDYTIRL)

julia> @btime sum(i.data for i in it)
  875.581 μs (2 allocations: 32 bytes)
0x0091802f3c41a9b9

Uses the high-level BioSequences v2 API internally, with sane fallbacks This allows very generic code, with following features being example of emergent properties:

• Creating kmers from kmers

julia> T1 = t.Kmer{3, RNAAlphabet{4}, UInt32};

julia> mer = t.kmer"TAGTCGCGAGAA"
12-mer of DNAAlphabet{2}:
TAGTCGCGAGAA

julia> it = t.StandardKmerIterator{T1, typeof(mer)}(mer);

julia> [canonical(i) for i in it]
10-element Array{Main.t.Kmer{3,RNAAlphabet{4},UInt32},1}:
 CUA
 ACU
 GAC
 CGA
 CGC
 CGC
 CGA
 CUC
 AGA
 GAA

• Many basic properties like getindex just uses the BioSequence fallback