// Method to parse the main directory
// dataSetType: Train or Test
//
public void Parse(String rootDirectory, String dataSetType, String parsingMode, String inputFilesFormat)
{
// Not need to write to output file if maxIDRun
if (!maxIDRun)
{
// Start the file writer
outFileWriter = new OutputFileWriter(configManager, logger, rootDirectory, dataSetType);
// Start the file
outFileWriter.WriteOutputFile(null, OutFileMode.START);
}// end if (!maxIDRun)
switch (parsingMode)
{
case "AnyFile":
ParseAnyFile(rootDirectory, dataSetType, inputFilesFormat);
break;
case "FolderStructure":
ParseFolderStructure(rootDirectory, dataSetType, inputFilesFormat);
break;
}// end switch
// Not need to write to output file if maxIDRun
if (!maxIDRun)
{
// Finalize the output file
outFileWriter.WriteOutputFile(null, OutFileMode.FINISH);
}
}// end Parse()
// Method to log the parsing info of a parse process in the trace file
protected void LogParsingInfo(String[] categoryFolders, uint numFiles, OutputFileWriter outFileWriter)
{
logger.LogTrace("Finished parsing");
logger.LogTrace("Total number of categories: " + categoryFolders.Length.ToString());
logger.LogTrace("Total parsed Files: " + numFiles.ToString());
logger.LogTrace("Total number of words: " + words.Length.ToString());
if (outFileWriter != null)
{
logger.LogTrace("Total number of words actually written to file: " + outFileWriter.numExamplesInOutFile.ToString());
}
logger.LogTrace("Max ID of mrfType is " + maxIDs.mrfType + " needs " + GetNumBits(maxIDs.mrfType).ToString() + " bits");
logger.LogTrace("Max ID of prefix is " + maxIDs.p + " needs " + GetNumBits(maxIDs.p).ToString() + " bits");
logger.LogTrace("Max ID of root is " + maxIDs.r + " needs " + GetNumBits(maxIDs.r).ToString() + " bits");
logger.LogTrace("Max ID of form is " + maxIDs.f + " needs " + GetNumBits(maxIDs.f).ToString() + " bits");
logger.LogTrace("Max ID of suffix is " + maxIDs.s + " needs " + GetNumBits(maxIDs.s).ToString() + " bits");
logger.LogTrace("Max ID of POS is " + maxIDs.POS_IDs[0] + " needs " + GetNumBits(maxIDs.POS_IDs[0]).ToString() + " bits");
logger.LogTrace("Total number of needed bits for binary representation (POS is bit-field): " + (GetNumBits(maxIDs.mrfType) + GetNumBits(maxIDs.p) + GetNumBits(maxIDs.r) + GetNumBits(maxIDs.f) + GetNumBits(maxIDs.s) + maxIDs.POS_IDs[0] + 1).ToString());
// The +1 is added because maxID value means we could have positions from 0 to this maxID, so total of maxID + 1 positions
logger.LogTrace("Total number of needed bits for bit-field representation: " + ((maxIDs.mrfType + 1) + (maxIDs.p + 1) + (maxIDs.r + 1) + (maxIDs.f + 1) + (maxIDs.s + 1) + (maxIDs.POS_IDs[0] + 1)).ToString());
logger.LogInfo();
}
// Method to parse the main directory
public void Parse(String rootDirectory, String mode)
{
// Traverse the root directory
String[] categoryFolders = Directory.GetDirectories(rootDirectory);
// Temp string to build the current mrf folder name in it
String currentMrfFolderName;
// Temp words List list over all files to accomodate words. To be converted to words [] when full length is known.
ArrayList wordsList = new ArrayList();
// The features formatter
FeaturesFormatter featuresFormatter;
// Start the file writer
OutputFileWriter outFileWriter = new OutputFileWriter(configManager, logger, rootDirectory, mode);
// Start the file
outFileWriter.WriteOutputFile(null, OutFileMode.START);
// Counter of parsed files
uint numFiles = 0;
// Counter of truly written words to out file
// int numExamplesInOutFile = 0;
// Parse files of each category
foreach (String category in categoryFolders)
{
logger.LogTrace("Parsing files of category: " + category + "...");
// Form the string of mrf folder in the current category
currentMrfFolderName = category + configManager.directorySeparator + configManager.mrfFolderName;
// Parse files of mrf folder
foreach (String file in Directory.GetFiles(currentMrfFolderName))
{
// Increment number of files
numFiles++;
logger.LogTrace("Parsing file: " + numFiles.ToString() + "- " + file + "...");
if (numFiles == 44)
{
int x;
}
// Temp array list to hold the words parsed from the file
ArrayList fileWordsList = new ArrayList();
// Temp array to hold the words parsed from the file
Word[] fileWords;
// Parse words in file into its structure
fileWordsList = FileParse(file);
//if (fileWordsList[)
// Add the word to the global words list
// Copy to be done by AddRange--Working
wordsList.AddRange(fileWordsList);
if (fileWordsList.Count != 0)
{
// Set the words array to the words list parsed by FileParse
fileWordsList.TrimToSize();
fileWords = (Word[])fileWordsList.ToArray(fileWordsList[0].GetType());
// Decide which type of formatter to use
switch (configManager.featuresFormat)
{
case "Normal":
featuresFormatter = new NormalFeaturesFormatter(configManager, logger, fileWords);
break;
case "Binary":
featuresFormatter = new BinaryFeaturesFormatter(configManager, logger, fileWords);
break;
case "Bitfield":
featuresFormatter = new BitFieldFeaturesFormatter(configManager, logger, fileWords);
break;
case "Raw":
featuresFormatter = new RawFeaturesFormatter(configManager, logger, fileWords);
break;
default:
Console.WriteLine("Incorrect features format configuration. {0} is invalid configuration. Valid configurations are: Normal, Binary and Bitfield", configManager.featuresFormat);
throw (new IndexOutOfRangeException());
}// end switch
// Format the words features of the file
try
{
featuresFormatter.FormatFeatures();
}
catch (OutOfMemoryException)
{
Console.WriteLine("Ooops! Out of memory");
}
// Start the context extractor
ContextExtractor contextExtractor = new ContextExtractor(featuresFormatter.wordsFeatures, logger, configManager);
// Extract the context extraction
contextExtractor.ContextExtract();
// Write (append) to output file
outFileWriter.WriteOutputFile(contextExtractor.contextFeatures, OutFileMode.APPEND);
// Accumulate numExamplesInOutFile
//numExamplesInOutFile += outFileWriter.numExamplesInOutFile;
/*if (numExamplesInOutFile == 460)
* {
* int x = 1;
* }*/
logger.LogTrace("Parsing done successfully for the file");
// Free the file words list
fileWordsList = null;
// Free the features formatter for this file
featuresFormatter = null;
// Free the context extractor
contextExtractor = null;
// Force memory freeing
GC.Collect();
}// end if(fileWordsList.Count != 0)
else
{
logger.LogTrace("Empty File");
}
}// end foreach mrf directory parse
logger.LogTrace("Finished parsing of category " + category);
} // end forach categories traversing
// Copy words list to words array
// First limit the size
wordsList.TrimToSize();
logger.LogTrace("POS:");
for (int j = 0; j < posNames.Length; j++)
{
logger.LogTrace(posNames[j]);
}
// Copy
this.words = (Word[])wordsList.ToArray(wordsList[0].GetType());
logger.LogTrace("Finished parsing");
logger.LogTrace("Total number of categories: " + categoryFolders.Length.ToString());
logger.LogTrace("Total parsed Files: " + numFiles.ToString());
logger.LogTrace("Total number of words: " + words.Length.ToString());
logger.LogTrace("Total number of words actually written to file: " + outFileWriter.numExamplesInOutFile.ToString());
logger.LogTrace("Max ID of mrfType is " + maxIDs.mrfType + " needs " + GetNumBits(maxIDs.mrfType).ToString() + " bits");
logger.LogTrace("Max ID of prefix is " + maxIDs.p + " needs " + GetNumBits(maxIDs.p).ToString() + " bits");
logger.LogTrace("Max ID of root is " + maxIDs.r + " needs " + GetNumBits(maxIDs.r).ToString() + " bits");
logger.LogTrace("Max ID of form is " + maxIDs.f + " needs " + GetNumBits(maxIDs.f).ToString() + " bits");
logger.LogTrace("Max ID of suffix is " + maxIDs.s + " needs " + GetNumBits(maxIDs.s).ToString() + " bits");
logger.LogTrace("Max ID of POS is " + maxIDs.POS_IDs[0] + " needs " + GetNumBits(maxIDs.POS_IDs[0]).ToString() + " bits");
logger.LogTrace("Total number of needed bits for binary representation (POS is bit-field): " + (GetNumBits(maxIDs.mrfType) + GetNumBits(maxIDs.p) + GetNumBits(maxIDs.r) + GetNumBits(maxIDs.f) + GetNumBits(maxIDs.s) + maxIDs.POS_IDs[0] + 1).ToString());
// The +1 is added because maxID value means we could have positions from 0 to this maxID, so total of maxID + 1 positions
logger.LogTrace("Total number of needed bits for bit-field representation: " + ((maxIDs.mrfType + 1) + (maxIDs.p + 1) + (maxIDs.r + 1) + (maxIDs.f + 1) + (maxIDs.s + 1) + (maxIDs.POS_IDs[0] + 1)).ToString());
logger.LogInfo();
// Finalize the file
outFileWriter.WriteOutputFile(null, OutFileMode.FINISH);
}// end Parse()
