/**
* Would factoring out this suffix result in elimating the reference source vertex?
*
* @param graph the graph
* @param commonSuffix the common suffix of all toSplits
* @param toSplits the list of vertices we're are trying to split
* @return true if toSplit contains the reference source and this ref source has all and only the
* bases of commonSuffix
*/
private boolean wouldEliminateRefSource(
final SeqGraph graph, final SeqVertex commonSuffix, final Collection<SeqVertex> toSplits) {
for (final SeqVertex toSplit : toSplits) {
if (graph.isRefSource(toSplit)) return toSplit.length() == commonSuffix.length();
}
return false;
}
/**
* Would all vertices that we'd split just result in the common suffix?
*
* <p>That is, suppose we have prefix nodes ABC and ABC. After splitting all of the vertices would
* just be ABC again, and we'd enter into an infinite loop.
*
* @param commonSuffix the common suffix of all vertices in toSplits
* @param toSplits the collection of vertices we want to split
* @return true if all of the vertices are equal to the common suffix
*/
private boolean allVerticesAreTheCommonSuffix(
final SeqVertex commonSuffix, final Collection<SeqVertex> toSplits) {
for (final SeqVertex toSplit : toSplits) {
if (toSplit.length() != commonSuffix.length()) return false;
}
return true;
}
/**
* Simple single-function interface to split and then update a graph
*
* @param graph the graph containing the vertices in toMerge
* @param v The bottom node whose incoming vertices we'd like to split
* @return true if some useful splitting was done, false otherwise
*/
public boolean split(final SeqGraph graph, final SeqVertex v) {
if (graph == null) throw new IllegalArgumentException("graph cannot be null");
if (v == null) throw new IllegalArgumentException("v cannot be null");
if (!graph.vertexSet().contains(v))
throw new IllegalArgumentException("graph doesn't contain vertex v " + v);
final Collection<SeqVertex> toSplit = graph.incomingVerticesOf(v);
if (toSplit.size() < 2)
// Can only split at least 2 vertices
return false;
else if (!safeToSplit(graph, v, toSplit)) {
return false;
} else {
final SeqVertex suffixVTemplate = commonSuffix(toSplit);
if (suffixVTemplate.isEmpty()) {
return false;
} else if (wouldEliminateRefSource(graph, suffixVTemplate, toSplit)) {
return false;
} else if (allVerticesAreTheCommonSuffix(suffixVTemplate, toSplit)) {
return false;
} else {
final List<BaseEdge> edgesToRemove = new LinkedList<BaseEdge>();
// graph.printGraph(new File("split.pre_" + v.getSequenceString() + "." +
// counter + ".dot"), 0);
for (final SeqVertex mid : toSplit) {
// create my own copy of the suffix
final SeqVertex suffixV = new SeqVertex(suffixVTemplate.getSequence());
graph.addVertex(suffixV);
final SeqVertex prefixV = mid.withoutSuffix(suffixV.getSequence());
final BaseEdge out = graph.outgoingEdgeOf(mid);
final SeqVertex incomingTarget;
if (prefixV == null) {
// this node is entirely explained by suffix
incomingTarget = suffixV;
} else {
incomingTarget = prefixV;
graph.addVertex(prefixV);
graph.addEdge(prefixV, suffixV, new BaseEdge(out.isRef(), 1));
edgesToRemove.add(out);
}
graph.addEdge(suffixV, graph.getEdgeTarget(out), out.copy());
for (final BaseEdge in : graph.incomingEdgesOf(mid)) {
graph.addEdge(graph.getEdgeSource(in), incomingTarget, in.copy());
edgesToRemove.add(in);
}
}
graph.removeAllVertices(toSplit);
graph.removeAllEdges(edgesToRemove);
// graph.printGraph(new File("split.post_" + v.getSequenceString() + "." +
// counter++ + ".dot"), 0);
return true;
}
}
}