Global design of gatb::core package

Introduction

As its name suggests, the gatb::core package holds all the fundamental packages needed for writting assembly algorithms.

It holds some generic services, like:

operating system abstraction,
collections management
design patterns
genomic banks read/write
kmers management
De Bruijn graph management

Global view of the gatb::core package

Package gatb::core::system

The gatb::core::system package holds operations related to operating system.

Making such an abstraction is interesting when considering the following points:

client code should rely on interfaces defined here and not on specific implementations (for linux, macos, ...)
client code should not use compilation flags for one specific system (like #ifdef LINUX)
improved operating system operations should be available to clients without modification in their code

The idea is so to list all necessary operating system operations that are likely to be used, to make several interfaces that gather some operations set and to provide specific implementations for several system (linux, macos, windows).

Global view of the gatb::core::system package

Package gatb::core::tools

The gatb::core::tools package holds generic operations used throughout the project. They are not specific to genome nor assembly concerns and are defined here in a separate package.

Roughly speaking, we may find the following operations:

design patterns tools (iterator, observer, smart pointer, etc...)
collections (container, bag, iterable, etc...)
misc

Global view of the gatb::core::tools package

Package gatb::core::bank

The gatb::core::bank package holds operations related to genomic databases management.

In particular, we define a gatb::core::bank::IBank interface that is an iterable over gatb::core::bank::Sequence instances. This interface has a lot of implementations, including FASTA format.

Package gatb::core::kmer

The gatb::core::kmer package holds operations related to kmer management.

A typical case is to create a IBank instance and then an iterator on it; such an iterator can feed the kmer model for iterating all the kmer from the sequences.

Package gatb::core::debruijn

The gatb::core::debruijn package holds operations related to De Bruijn graph.

The main API is the gatb::core::debruijn::impl::Graph class. It holds several features:

create graph
load graph
iterate nodes of the graph
get neighbors of a node

The implementation of the class relies on a probabilistic data structure named Bloom filter. Its usage allows to have a very low memory footprint, even for big graphs.