Notes.txt

CMakeLists.txt
    In order to add custom library path to CMake (like zlib),
    set environment variable CMAKE_PREFIX_PATH to the folder,
    where you have installed the libraries

*.mta
	meta info for the db
	each record is represented as (seq_name, offset, seq_length)
	seq_name is serialized as mmstring data structure with length first, then
	contents

*.cat
	convert Ns to randomly selected ACGT as seqs (seq string)
	output seqs
	output reverse complementary of seqs
	Note: use another random step in reverse complementary step might lead to
	better randomness, (need to consider and prove)
	put an '$' in the end.

*.mfi
	my fm-index aux file, combination of cc, oo, bwt, need to think and implement
	Current format:
	CC [256 x uint64]
	OO [o_ratio int, o_len uint64, OO o_len x uint64]
	bwt [len uint64, bwt len x char]

*.sa5
	sa using 5 bytes data structure for serialization. Created by pSAscan.
	The ui40_t data structure in sa_use may not be always compatible with .sa5, need further investigate
	pSAscan cannot load gzfile directly, could be further improved by our own library

fmidx/fmidx.c:_bwt_from_sa5
	could be in 2 cases: ram_use < 2n or ram_use >= 2n.
	First case is the same with current approach (use a buffer)
	Second case could be improved by loading the file directly into memory.
	n here means the length of the prefix file


lchash/fmindex bug
	currently the C table is built to distinguish 'a' and 'A', however, this might lead to inconsistent behaviours. Need to fix later. Currently only support all small case in db. Controlled with the _seq_from_num function

Current problem:
	Only first iteration has output
	Need to find a good threshold for uninformative seeds
	Some of the reads are locating at 0 (all uninformative seeds)

Known issues:
	Nov 12:
		using khash instead of a vector
		the key stored in histo is the original location (bucketing)
		need to improve sensitivity