//TODO: translate this void haplo_d::initialize_skeleton(thread_t& t, pair<int,int> interval, cross_section& prevAs, xg::XG& graph) { rectangle rect; int new_height; int last_height = prevAs.S[0].J; bool add_rectangle; bool add_A; //TODO: fix this int width = 0; for(int i = interval.first; i <= interval.second; i++) { // Count the number of base pairs since the last entry or exit node width += graph.node_length(t[i-1].node_id); new_height = graph.node_height(t[i]); if(cs.back().S.size() != 0) { if(i == interval.first) { prevAs.S[0]; } else { rect = cs.back().S[0]; } rect.J = rect.get_next_J(t[i],graph); // step this strip forward // Did any threads leave? if(last_height > rect.J) { add_A = 1; } // Are there any threads here which didn't come from the previous node? if(rect.J < new_height) { add_rectangle = 1; add_A = 1; } // This is an entry or exit node, add a cross-section to the vector of // cross-sections (which corresponds to the "A" set in the theory doc) if(add_A) { cs.back().width = width; width = 0; cs.push_back(cross_section(new_height,i,t[i])); } else { // This isn't a node where anything leaves or joins, let's skip over it cs.back().bridge.push_back(t[i]); for (size_t a = 0; a < cs.back().S.size(); a++) { cs.back().S[a].extend(t[i],graph); } } // This is an entry node; we also need a new rectangle corresponding to the // new strip. We need to do this *before* we populate since cross_sections // arrange rectangles newest -> oldest // NB that add_rectangle implies add_A if(add_rectangle) { rectangle new_rect; new_rect.extend(t[i],graph); new_rect.J = new_height; cs.back().height = new_rect.J; cs.back().S.push_back(new_rect); cs.back().S.back().I = new_rect.J - rect.J; } if(add_A) { int b = cs.size()-1; if(rect.J > 0) { cs[b].S.push_back(rect); cs[b].S.back().prev = 0; cs[b-1].S[0].next = cs[b].S.size()-1; } } last_height = new_height; add_A = 0; add_rectangle = 0; } else { cs.back().width = width; width = 0; cs.push_back(cross_section(new_height,i,t[i])); if(new_height > 0) { rectangle new_rect; new_rect.extend(t[i],graph); new_rect.J = new_height; cs.back().height = new_rect.J; cs.back().S.push_back(new_rect); cs.back().S.back().I = new_rect.J - rect.J; } } } if(cs.size() == 1) { cs.back().width = width; } cs.back().width += graph.node_length(t.back().node_id) - 1; for(int i = 0; i < cs.size(); i++) { tot_width += cs[i].width; } }
PathIndex::PathIndex(const Path& path, const xg::XG& index) { // Trace the given path in the given XG graph, collecting sequence // We're going to build the sequence string std::stringstream seq_stream; // What base are we at in the path? size_t path_base = 0; // What was the last rank? Ranks must always go up. int64_t last_rank = -1; for (size_t i = 0; i < path.mapping_size(); i++) { auto& mapping = path.mapping(i); if (!by_id.count(mapping.position().node_id())) { // This is the first time we have visited this node in the path. // Add in a mapping. by_id[mapping.position().node_id()] = std::make_pair(path_base, mapping.position().is_reverse()); #ifdef debug #pragma omp critical (cerr) std::cerr << "Node " << mapping.position().node_id() << " rank " << mapping.rank() << " starts at base " << path_base << " with " << index.node_sequence(mapping.position().node_id()) << std::endl; #endif // Make sure ranks are monotonically increasing along the path, or // unset. assert(mapping.rank() > last_rank || (mapping.rank() == 0 && last_rank == 0)); last_rank = mapping.rank(); } // Say that this node appears here along the reference in this // orientation. by_start[path_base] = NodeSide(mapping.position().node_id(), mapping.position().is_reverse()); // Remember that occurrence by node ID. node_occurrences[mapping.position().node_id()].push_back(by_start.find(path_base)); // Find the node's sequence std::string node_sequence = index.node_sequence(mapping.position().node_id()); while(path_base == 0 && node_sequence.size() > 0 && (node_sequence[0] != 'A' && node_sequence[0] != 'T' && node_sequence[0] != 'C' && node_sequence[0] != 'G' && node_sequence[0] != 'N')) { // If the path leads with invalid characters (like "X"), throw them // out when computing path positions. // TODO: this is a hack to deal with the debruijn-brca1-k63 graph, // which leads with an X. #pragma omp critical (cerr) std::cerr << "Warning: dropping invalid leading character " << node_sequence[0] << " from node " << mapping.position().node_id() << std::endl; node_sequence.erase(node_sequence.begin()); } if (mapping.position().is_reverse()) { // Put the reverse sequence in the path seq_stream << reverse_complement(node_sequence); } else { // Put the forward sequence in the path seq_stream << node_sequence; } // Whether we found the right place for this node in the reference or // not, we still need to advance along the reference path. We assume the // whole node (except any leading bogus characters) is included in the // path (since it sort of has to be, syntactically, unless it's the // first or last node). path_base += node_sequence.size(); // TODO: handle leading bogus characters in calls on the first node. } // Record the length of the last mapping's node, since there's no next mapping to work it out from last_node_length = path.mapping_size() > 0 ? index.node_length(path.mapping(path.mapping_size() - 1).position().node_id()) : 0; // Create the actual reference sequence we will use sequence = seq_stream.str(); #ifdef debug // Announce progress. #pragma omp critical (cerr) std::cerr << "Traced " << path_base << " bp path." << std::endl; if (sequence.size() < 100) { #pragma omp critical (cerr) std::cerr << "Sequence: " << sequence << std::endl; } #endif }