/**
* Updates the tree starting from inputNode and by adding stringPart.
*
* Returns a reference (NodeA<T>, string) Tuple for the string that has been added so far.
* This means:
* - the NodeA<T> will be the NodeA<T> that can be reached by the longest path string (S1)
* that can be obtained by concatenating consecutive edges in the tree and
* that is a substring of the string added so far to the tree.
* - the string will be the remainder that must be added to S1 to get the string
* added so far.
*
* @param inputNode the NodeA<T> to start from
* @param stringPart the string to add to the tree
* @param rest the rest of the string
* @param value the value to add to the index
*/
private Tuple<Node<T>, List<int>> Update(Node<T> inputNode, List<int> stringPart, List<int> rest, T value)
{
var s = inputNode;
var tempstr = stringPart;
var newChar = stringPart[stringPart.Count - 1];
// Console.WriteLine("stringPart: " + stringPart.Count + ", Rest count = " + rest.Count);
// line 1
var oldroot = _root;
// line 1b
var ret = TestAndSplit(s, tempstr.Take(tempstr.Count - 1).ToList(), newChar, rest, value);
var r = ret.Item2;
var endpoint = ret.Item1;
// line 2
while (!endpoint)
{
// line 3
var tempEdge = r.GetEdge(newChar);
Node<T> leaf;
if (null != tempEdge)
{
// such a NodeA<T> is already present. This is one of the main differences from Ukkonen's case:
// the tree can contain deeper nodes at this stage because different strings were added by previous iterations.
leaf = tempEdge.Target;
}
else
{
// must build a new leaf
leaf = new Node<T>();
leaf.AddRef(value);
var newedge = new Edge<T>(rest, leaf);
r.AddEdge(newChar, newedge);
}
// update suffix link for newly created leaf
if (_activeLeaf != _root)
{
_activeLeaf.Suffix = leaf;
}
_activeLeaf = leaf;
// line 4
if (oldroot != _root)
{
oldroot.Suffix = r;
}
// line 5
oldroot = r;
// line 6
if (null == s.Suffix)
{
// root NodeA<T>
//TODO Check why assert
//assert (root == s);
// this is a special case to handle what is referred to as NodeA<T> _|_ on the paper
tempstr = tempstr.Skip(1).ToList();
}
else
{
var canret = Canonize(s.Suffix, SafeCutLastChar(tempstr));
s = canret.Item1;
// use intern to ensure that tempstr is a reference from the string pool
var lastChar = tempstr[tempstr.Count - 1];
tempstr = new List<int>();
tempstr.AddRange(canret.Item2);
tempstr.Add(lastChar); // tempstr = canret.Item2 + tempstr[-1]
}
// line 7
ret = TestAndSplit(s, SafeCutLastChar(tempstr), newChar, rest, value);
r = ret.Item2;
endpoint = ret.Item1;
}
// line 8
if (oldroot != _root)
{
oldroot.Suffix = r;
}
return new Tuple<Node<T>, List<int>>(s, tempstr);
}
// refactored
private static Tuple<bool, Node<T>> TestAndSplit(Node<T> inputs, List<int> stringPart, int t, List<int> remainder, T value)
{
// descend the tree as far as possible
var ret = Canonize(inputs, stringPart);
var s = ret.Item1;
var str = ret.Item2;
// if (!(string.Empty.Equals(str)))
if (str.Count != 0)
{
var g = s.GetEdge(str[0]);
var label = g.Label;
// must see whether "str" is substring of the label of an EdgeA<T>
if (label.Count > str.Count && label[str.Count] == t)
{
return new Tuple<bool, Node<T>>(true, s);
}
// need to split the EdgeA<T>
var newlabel = label.Skip(str.Count).ToList();
//assert (label.startsWith(str));
// build a new NodeA<T>
var r = new Node<T>();
// build a new EdgeA<T>
var newedge = new Edge<T>(str, r);
g.Label = newlabel;
// link s -> r
r.AddEdge(newlabel[0], g);
s.AddEdge(str[0], newedge);
return new Tuple<bool, Node<T>>(false, r);
}
var e = s.GetEdge(t);
if (null == e)
{
// if there is no t-transtion from s
return new Tuple<bool, Node<T>>(false, s);
}
// if (remainder.Equals(e.Label))
if (Equals(remainder, e.Label))
{
// update payload of destination NodeA<T>
e.Target.AddRef(value);
return new Tuple<bool, Node<T>>(true, s);
}
// if (remainder.StartsWith(e.Label))
if (StartsWith(remainder, e.Label))
{
return new Tuple<bool, Node<T>>(true, s);
}
if (!StartsWith(e.Label, remainder))
{
return new Tuple<bool, Node<T>>(true, s);
}
// need to split as above
var newNode = new Node<T>();
newNode.AddRef(value);
var newEdge = new Edge<T>(remainder, newNode);
e.Label = e.Label.Skip(remainder.Count).ToList();
newNode.AddEdge(e.Label[0], e);
s.AddEdge(t, newEdge);
return new Tuple<bool, Node<T>>(false, s);
// they are different words. No prefix. but they may still share some common substr
}