private void ProcessFile(FileInfo file)
{
if (!file.Exists)
{
return;
}
if (!extensions.Contains(file.Extension))
{
return;
}
// \u2022 is the unicode for a bullet symbol.
var separators = new[]
{
' ', '\u2022', '’', '\"', '“', '!', '\'', '\\', '/', '_', '(', ')', '-', ',', ':', '?', ';', '.',
'\r', '\n', '|'
};
try
{
//use toxy to extract string from files.
//parser = ParserFactory.CreateText(new ParserContext(file.FullName));
//checks if file has an html or xml extension.
string document;
ITextParser parser;
if (file.Extension == ".html" || file.Extension == ".htm" || file.Extension == ".xml")
{
parser = ParserFactory.CreateText(new ParserContext(file.FullName));
string textWithTags = parser.Parse();
document = RemoveAllTags(textWithTags);
}
else if (file.Extension == ".pptx")
{
document = ExtractPptxText(file);
}
else
{
parser = ParserFactory.CreateText(new ParserContext(file.FullName));
document = parser.Parse();
}
// Split with separators and ignore empty spaces.
foreach (var word in document.ToLower().Split(separators, StringSplitOptions.RemoveEmptyEntries))
{
// Remove stop words and numeric data.
if (stopwords.Contains(word) || Regex.IsMatch(word, "\\d+"))
{
continue;
}
//stems word before adding it to the inverted index.
InvertedIndex.GetInstance()
.Add(stemmer.StemWord(word.Trim()), new InvertedIndex.Tuple(docId, wordPosition++));
}
}
catch (Exception e) when(e is IOException || e is NullReferenceException || e is ZipException)
{
MessageBox.Show(@"Please close all programs using the files you want to search.");
}
catch (Exception e) when(e is InvalidDataException)
{
MessageBox.Show(@"Invalid file format.");
}
FileMatch.GetInstance().Add(docId, file);
docId++;
}
