public IList <string> TestNuveTokenizerReturnDelimsFalse(string text)
{
var tokenizer = new NuveTokenizer();
IList <string> tokens = tokenizer.Tokenize(text);
return(tokens);
}
public IList <string> TestNuveTokenizerReturnDelimsTrue(string text)
{
var tokenizer = new NuveTokenizer(true);
IList <string> tokens = tokenizer.Tokenize(text);
int length = string.Join("", tokens).Length;
Assert.AreEqual(length, text.Length);
return(tokens);
}
static void Main(string[] args)
{
MPI.Environment.Run(ref args, communicator =>
{
if (communicator.Rank == 0) //Master process
{
Console.WriteLine("Process with rank " + communicator.Rank + " successfully running. Total rank " + communicator.Size);
//Timer
System.Diagnostics.Stopwatch watch = System.Diagnostics.Stopwatch.StartNew();
long totalLineNumber = 0;
long intervalForEachProcess = 0;
FileStream readInput = File.OpenRead("news.txt");
totalLineNumber = StreamExtensions.CountLines(readInput, default);
Console.WriteLine("Total line number in the input text: " + totalLineNumber);
intervalForEachProcess = totalLineNumber / (communicator.Size - 1);
long startIndexOfEachProcess = 0;
//Reading number of lines in the input, and distributing them equally to all machines in the cluster
for (int rank = 1; rank < communicator.Size; rank++)
{
communicator.Send(startIndexOfEachProcess, rank, 0);
if (rank != communicator.Size - 1)
{
communicator.Send(startIndexOfEachProcess + intervalForEachProcess, rank, 0);
}
else
{
//Simply add the remaining part to last process
communicator.Send(totalLineNumber, rank, 0);
}
startIndexOfEachProcess += intervalForEachProcess;
}
//BiGram - count
Dictionary <String, int> hashMap = new Dictionary <String, int>();
int terminator = 0;
while (terminator != communicator.Size - 1)
{
String recievedBiGramString = communicator.Receive <String>(MPI.Communicator.anySource, 0);
//Console.WriteLine(recievedBiGramString);
if (recievedBiGramString == "###-TERMINATION-###")
{
terminator++;
}
else if (!hashMap.ContainsKey(recievedBiGramString))
{
hashMap.Add(recievedBiGramString, 1);
}
else
{
hashMap[recievedBiGramString]++;
}
}
Console.WriteLine("Loop terminated.");
//Handles with biGrams which are obtained from last word of an interval and first word of next interval
for (int i = 1; i < communicator.Size - 1; i++)
{
String firstWordOfInterval = communicator.Receive <String>(i, i);
String lastWordOfLine = communicator.Receive <String>(i + 1, i + 1);
String recievedBiGramString = lastWordOfLine + " " + firstWordOfInterval;
Console.WriteLine(recievedBiGramString + " " + i + " " + (i + 1));
if (!hashMap.ContainsKey(recievedBiGramString))
{
hashMap.Add(recievedBiGramString, 1);
}
else
{
hashMap[recievedBiGramString]++;
}
}
Console.WriteLine("Separation points done.");
watch.Stop();
long elapsedMs = watch.ElapsedMilliseconds;
long elapsedSecond = elapsedMs / 1000;
Console.WriteLine("biGram analysis time: " + elapsedSecond / 60 + " min" + elapsedSecond % 60 + " sec");
watch.Start();
KeyValuePair <String, int>[] output = countSort(hashMap);
watch.Stop();
long overall = watch.ElapsedMilliseconds;
long overallSec = overall / 1000;
Console.WriteLine("Sort time: " + (overallSec - elapsedSecond) / 60 + " min" + (overallSec - elapsedSecond) % 60 + " sec");
Console.WriteLine("Overall time: " + overallSec / 60 + " min" + overallSec % 60 + " sec");
StreamWriter writer = File.CreateText("out.txt");
for (int i = 0; i < output.Length; i++)
{
writer.WriteLine(output[i]);
}
writer.Close();
Console.WriteLine("Output.txt has been written successfully.");
}
else // Processes that makes bigram analysis
{
Console.WriteLine("Process with rank " + communicator.Rank + " successfully running.");
long intervalStart = communicator.Receive <long>(0, 0);
long intervalEnd = communicator.Receive <long>(0, 0);
Console.WriteLine(intervalStart + " " + intervalEnd);
StreamReader readInput = new StreamReader("news.txt");
//Carries readStream's current index to the given line number by master process.
for (long i = 0; i < intervalStart; i++)
{
readInput.ReadLine();
}
NuveTokenizer tokenize = new NuveTokenizer(false);
IList <String> tokenizedInput;
String firstWordOfInterval = " ";
bool isFirstWord = true;
// Last word of each line is temporarily held in this variable to be matched with first word of next line
// Otherwise, 1 biGram will be lost by every line
// Also used for holding last word of each interval too
String lastWordOfLine = " ";
//The main loop which makes bigram analysis
for (long i = intervalStart; i < intervalEnd; i++)
{
tokenizedInput = tokenize.Tokenize(readInput.ReadLine().ToString());
//Instead of natural language processing library Nuve, simply following line could have been used.
//tokenizedInput = readInput.ReadLine().ToString().Split();
if (isFirstWord && tokenizedInput.IsNotNullOrEmpty())
{
firstWordOfInterval = tokenizedInput[0];
isFirstWord = false;
}
if (tokenizedInput.IsNotNullOrEmpty() && lastWordOfLine != " ")
{
String endLineStartingLineMergedBiGram = tokenizedInput[0] + " " + lastWordOfLine;
communicator.Send(endLineStartingLineMergedBiGram, 0, 0);
}
for (int k = 0; k < tokenizedInput.Count - 1; k++)
{
String biGram = tokenizedInput[k] + " " + tokenizedInput[k + 1];
communicator.Send(biGram, 0, 0);
lastWordOfLine = tokenizedInput[k + 1];
}
}
communicator.Send("###-TERMINATION-###", 0, 0);
//Console.WriteLine("###-TERMINATION-### sent from: " + communicator.Rank);
//Handles with biGrams which are obtained from last word of an interval and first word of next interval
if (communicator.Rank == 1)
{
Console.WriteLine("Trying to send from " + communicator.Rank);
communicator.Send(lastWordOfLine, 0, communicator.Rank);
}
else if (communicator.Rank == communicator.Size - 1)
{
Console.WriteLine("Trying to send from " + communicator.Rank);
communicator.Send(firstWordOfInterval, 0, communicator.Rank);
}
else
{
Console.WriteLine("Trying to send from " + communicator.Rank);
communicator.Send(firstWordOfInterval, 0, communicator.Rank);
communicator.Send(lastWordOfLine, 0, communicator.Rank);
}
}
});
}
