/// <summary>
/// This should be the only function setting any internal variables other than parent.
/// </summary>
private void Split(List <BookData> books, List <string> splittingWords, List <int> splittingBookIndices, int splitIndex, int maxBooksPerLeaf)
{
_splittingWord = splittingWords[splitIndex];
books.OrderBy(book => book.WordCounts[SplittingWord]);
int median = books[books.Count / 2].WordCounts[SplittingWord];
if (books.Count % 2 == 0)
{
median = (median + books[(books.Count - 1) / 2].WordCounts[SplittingWord]) / 2;
}
_splittingValue = median;
var leftIndices = new List <int>();
var rightIndices = new List <int>();
foreach (int i in splittingBookIndices)
{
if (books[i].WordCounts[SplittingWord] <= SplittingValue)
{
leftIndices.Add(i);
}
else
{
rightIndices.Add(i);
}
}
_leftChild = new KDTreeNode(this, books, splittingWords, leftIndices, splitIndex, maxBooksPerLeaf);
_rightChild = new KDTreeNode(this, books, splittingWords, rightIndices, splitIndex, maxBooksPerLeaf);
}
private KDTreeNode(KDTreeNode parent, List <BookData> books, List <string> splittingWords, List <int> splittingBookIndices, int lastSplitIndex, int maxBooksPerLeaf)
{
_parent = parent;
if (splittingBookIndices.Count > maxBooksPerLeaf)
{
// Last split index + 1 to get current index.
Split(books, splittingWords, splittingBookIndices, lastSplitIndex + 1, maxBooksPerLeaf);
}
else
{
// Each indes remaining is a leaf.
_containedBookIndices = splittingBookIndices;
}
}
/// <summary>
/// Creates a KDTree to speed up the K Nearest Neighbor algorithm.
/// </summary>
/// <param name="books">Books contained in the tree.</param>
/// <param name="wordsToSplitOn">Words (most common words) used in the algorithm.</param>
public KDTree(List <BookData> books, List <string> wordsToSplitOn, int maxBooksPerLeaf = 4)
{
_books = books;
_wordsToSplitOn = wordsToSplitOn;
_root = new KDTreeNode(books, wordsToSplitOn, maxBooksPerLeaf);
}
