void operator()( population& p ) const
{
const auto n = sizeof(integer_type) << 3;
vg::variate_generator<real_type> vg1(0, 1);
vg::variate_generator<integer_type> vg2(1, n-1);
for ( auto ch : p ) //for all chromosomes in the population
for ( auto g : ch ) //for all genes in the chromosome
if ( vg1() < r ) //case mutate
{
g.encode(); //real -> integer
auto c = gene_to_gray()(g.code()); // c -> gray
c ^= ( 1 << vg2() ); //reverse a random bit
g.code() = gray_to_gene()(c); // gray -> code
g.decode(); //integer -> real
}
}
int main(int argc, char** argv) {
if(argc == 1) {
// Print the help
help_main(argv);
return 1;
}
size_t kmerSize = 0;
size_t edgeMax = 0;
// Should we only merge on kmers and skip paths?
bool kmersOnly = false;
optind = 1; // Start at first real argument
bool optionsRemaining = true;
while(optionsRemaining) {
static struct option longOptions[] = {
{"kmer-size", required_argument, 0, 'k'},
{"edge-max", required_argument, 0, 'e'},
{"kmers-only", no_argument, 0, 'o'},
{"threads", required_argument, 0, 't'},
{"help", no_argument, 0, 'h'},
{0, 0, 0, 0}
};
int optionIndex = 0;
switch(getopt_long(argc, argv, "k:e:t:h", longOptions, &optionIndex)) {
// Option value is in global optarg
case -1:
optionsRemaining = false;
break;
case 'k': // Set the kmer size
kmerSize = atol(optarg);
break;
case 'e': // Set the edge max parameter for kmer enumeration
edgeMax = atol(optarg);
break;
case 'o': // Only merge on kmers
kmersOnly = true;
break;
case 't': // Set the openmp threads
omp_set_num_threads(atoi(optarg));
break;
case 'h': // When the user asks for help
case '?': // When we get options we can't parse
help_main(argv);
exit(1);
break;
default:
// TODO: keep track of the option
std::cerr << "Illegal option" << std::endl;
exit(1);
}
}
if(argc - optind < 2) {
// We don't have two positional arguments
// Print the help
help_main(argv);
return 1;
}
if(kmersOnly && kmerSize == 0) {
// We need a kmer size to use kmers
throw std::runtime_error("Can't merge only on kmers with no kmer size");
}
// Pull out the VG file names
std::string vgFile1 = argv[optind++];
std::string vgFile2 = argv[optind++];
// Guess index names (TODO: add options)
std::string indexDir1 = vgFile1 + ".index";
std::string indexDir2 = vgFile2 + ".index";
// Open the files
std::ifstream vgStream1(vgFile1);
if(!vgStream1.good()) {
std::cerr << "Could not read " << vgFile1 << std::endl;
exit(1);
}
std::ifstream vgStream2(vgFile2);
if(!vgStream2.good()) {
std::cerr << "Could not read " << vgFile2 << std::endl;
exit(1);
}
// We may have indexes. We need to use pointers because destructing an index
// that was never opened segfaults. TODO: fix vg
vg::Index* index1 = nullptr;
vg::Index* index2 = nullptr;
if(kmerSize) {
// Only go looking for indexes if we want to merge on kmers.
index1 = new vg::Index();
index1->open_read_only(indexDir1);
index2 = new vg::Index();
index2->open_read_only(indexDir2);
}
// Load up the first VG file
vg::VG vg1(vgStream1);
// And the second
vg::VG vg2(vgStream2);
// Make a way to track IDs
int64_t nextId = 1;
std::function<int64_t(void)> getId = [&]() {
return nextId++;
};
// Make a thread set
auto threadSet = stPinchThreadSet_construct();
// Make a place to keep track of the thread sequences.
// This will only contain sequences for threads that aren't staples.
// TODO: should this be by pointer instead?
std::map<int64_t, std::string> threadSequences;
// Add in each vg graph to the thread set
coregraph::EmbeddedGraph embedding1(vg1, threadSet, threadSequences, getId, vgFile1);
coregraph::EmbeddedGraph embedding2(vg2, threadSet, threadSequences, getId, vgFile2);
if(!kmersOnly) {
// We want to merge on shared paths in addition to kmers
// Complain if any of the graphs is not completely covered by paths
if(!embedding1.isCoveredByPaths()) {
std::cerr << "WARNING: " << embedding1.getName() << " contains nodes with no paths!" << std::endl;
}
if(!embedding2.isCoveredByPaths()) {
std::cerr << "WARNING: " << embedding2.getName() << " contains nodes with no paths!" << std::endl;
}
// Trace the paths and merge the embedded graphs.
std::cerr << "Pinching graphs on shared paths..." << std::endl;
embedding1.pinchWith(embedding2);
}
if(kmerSize > 0) {
// Merge on kmers that are unique in both graphs.
std::cerr << "Pinching graphs on shared " << kmerSize << "-mers..." << std::endl;
embedding1.pinchOnKmers(*index1, embedding2, *index2, kmerSize, edgeMax);
}
// Fix trivial joins so we don't produce more vg nodes than we really need to.
stPinchThreadSet_joinTrivialBoundaries(threadSet);
// Make another vg graph from the thread set
vg::VG core = pinchToVG(threadSet, threadSequences);
// Spit it out to standard output
core.serialize_to_ostream(std::cout);
// Tear everything down. TODO: can we somehow run this destruction function
// after all our other, potentially depending locals are destructed?
stPinchThreadSet_destruct(threadSet);
return 0;
}
