private static void BuildIndexForDirectory(string folder, MainWindow window)
{
// The inverted index.
var index = new PositionalInvertedIndex();
//Get the number of documents
_numberOfDocuments = Directory.EnumerateFiles(folder, "*.txt").Count();
//Initiate the progress Bar
window.InitiateprogressBar(_numberOfDocuments);
// Index the directory using a naive index
IndexFiles(folder, index, window);
foreach (var subfolder in Directory.EnumerateDirectories(folder))
{
IndexFiles(subfolder, index, window);
}
//Hide the progress bar to allow the user to start searching for terms
window.HideProgressBar();
index.ComputeStatistics();
index.StatToDisk(folder);
// at this point, "index" contains the in-memory inverted index
// now we save the index to disk, building three files: the postings index,
// the vocabulary list, and the vocabulary table.
// the array of terms
string[] dictionary = index.GetDictionary();
// an array of positions in the vocabulary file
long[] vocabPositions = new long[dictionary.Length];
BuildVocabFile(folder, dictionary, vocabPositions);
BuildPostingsFile(folder, index, dictionary, vocabPositions);
}
public void BuildIndex(MainWindow mainWindow)
{
BuildIndexForDirectory(_mPath, mainWindow);
}
private static void IndexFiles(string folder, PositionalInvertedIndex index, MainWindow window)
{
var documentId = 0;
FileStream writer = new FileStream(Path.Combine(folder, "docWeights.bin"), FileMode.Create);
foreach (string fileName in Directory.EnumerateFiles(Path.Combine(Environment.CurrentDirectory,
folder)))
{
if (fileName.EndsWith(".txt"))
{
var termToOccurence = IndexFile(fileName, index, documentId);
// Calculate document weight.
// Compute all wdts.
var wdts = new List<double>();
foreach (var pair in termToOccurence)
{
var wdt = 1.0 + Math.Log(pair.Value);
QueryReformulation.AddWeightToMatrix(pair.Key, pair.Value, documentId, _numberOfDocuments);
wdts.Add(wdt);
}
// Calculate ld for this document.
double sumTemp = 0.0;
foreach (var wdt in wdts)
sumTemp += wdt*wdt;
double ld = Math.Sqrt(sumTemp);
// Write ld in docWeights.bin.
var buffer = BitConverter.GetBytes(ld);
if (BitConverter.IsLittleEndian)
Array.Reverse(buffer);
writer.Write(buffer, 0, buffer.Length);
documentId++;
}
window.IncrementProgressBar();
}
// Compute the coocurence Matrix and put it to disk.
QueryReformulation.CreateMatrix(_numberOfDocuments);
QueryReformulation.ToDisk(folder);
writer.Close();
}
