public static async Task <PlagiarismInWeb> Find(string simplifiedText)
{
var words = TextManager.WordsFromText(simplifiedText).ToArray();
var wordCount = words.Length;
string path = @"D:\urlToWordsIndexes.txt";
//var shingles = Shingle.ShinglesFromWords(words);
//var urlToWordsIndexes = WebManager.WebSitesFromWordsParallel(shingles);
//File.WriteAllText(path, JsonConvert.SerializeObject(urlToWordsIndexes));
var urlToWordsIndexes = JsonConvert.DeserializeObject <Dictionary <string, List <int> > >(File.ReadAllText(path));
var orderedUrlToWordsIndexes = urlToWordsIndexes.OrderByDescending(kvp => kvp.Value.Count).ToDictionary(pair => pair.Key, pair => pair.Value);
var orderedUrls = orderedUrlToWordsIndexes.Keys.ToList();
var urlsCountCap = (int)Math.Ceiling(wordCount * 0.1);
var webPagesSimplifiedTexts = await WebManager.UrlsToSimplifiedTextsAsync(urlsCountCap, orderedUrls);
var urlToCommonTextParts = TextComparer.CommonTextParts(orderedUrls, orderedUrlToWordsIndexes, words, webPagesSimplifiedTexts);
double originalTextCharactersCount = simplifiedText.RemoveWhiteSpaces().Length;
var webResults = Logic.GetCheckResults(orderedUrls, urlsCountCap, urlToCommonTextParts, originalTextCharactersCount);
var orderedWebResults = webResults.OrderByDescending(res => res.CharactersPercentage).ToList();
return(new PlagiarismInWeb
{
Words = words,
WordCount = wordCount,
OrderedUrlToWordsIndexes = orderedUrlToWordsIndexes,
OrderedUrls = orderedUrls,
OrderedWebResults = orderedWebResults
});
}
public static async Task <Plagiarism <string> > FindAsync(string[] initialWords, Dictionary <int, string> initialDocIndexToSimplifiedWord, int[] initialDocIndexes, string[] simplifiedWords, string serverMapPath)
{
string debugLog = "";
var plagiarismWeb = new Plagiarism <string>();
Stopwatch stopwatch = new Stopwatch();
//stopwatch.Start();
//Dictionary<string, SortedSet<int>> urlToInitialDocWordsIndexesSet = GetUrlToInitialDocWordsIndexesFromGoogleAPIParallel(initialDocIndexes, simplifiedWords);
//stopwatch.Stop();
//Debug.WriteLine("GSEARCH TIME " + stopwatch.ElapsedMilliseconds);
var fileName = "serializedGS.txt";
var path = Path.Combine(serverMapPath, fileName);
//var serializedGS = JsonConvert.SerializeObject(urlToInitialDocWordsIndexesSet);
//File.WriteAllText(path, serializedGS);
var urlToInitialDocWordsIndexesSet = JsonConvert.DeserializeObject <Dictionary <string, SortedSet <int> > >(File.ReadAllText(path));
var urlToInitialDocWordsIndexesList = urlToInitialDocWordsIndexesSet.ToDictionary(pair => pair.Key, pair => pair.Value.ToList());
var mostPopularUrls = GetMostPopularUrls(urlToInitialDocWordsIndexesSet);
stopwatch.Restart();
var urlsSimplifiedTexts = await WebManager.TextsAsync(mostPopularUrls);
stopwatch.Stop();
debugLog += "WebManager.TextsAsync TIME " + stopwatch.ElapsedMilliseconds + " ";
stopwatch.Restart();
var tasks = new Task <List <Plagiarism <string> > > [mostPopularUrls.Count];
for (int i = 0; i < mostPopularUrls.Count; i++)
{
int savei = i;
tasks[i] = Task <List <Plagiarism <string> > > .Factory.StartNew(() =>
{
TextManager.PrepareText(urlsSimplifiedTexts[savei], out _, out _, out int[] urlInitialDocIndexes, out string[] urlSimplifiedWords, out _);
var indexedUrlText = Indexer.Indexing(urlInitialDocIndexes, urlSimplifiedWords);
var initialDocIndexesFoundOnUrl = urlToInitialDocWordsIndexesList[mostPopularUrls[savei]];
var plagiarismWebs = new List <Plagiarism <string> >();
for (int j = 0; j <= initialDocIndexesFoundOnUrl.Count - Shingle.Lenght; j++)
{
List <string> wordsList = new List <string>();
for (int k = 0; k < Shingle.Lenght; k++)
{
wordsList.Add(initialDocIndexToSimplifiedWord[initialDocIndexesFoundOnUrl[j + k]]);
}
var urlTextWordsPositionsForShingle = TextComparer.FindWordsInIndexedText(wordsList, indexedUrlText);
if (urlTextWordsPositionsForShingle.Count == 0)
{
continue;
}
var initialDocIndexesForShingle = new List <int>();
for (int k = 0; k < Shingle.Lenght; k++)
{
initialDocIndexesForShingle.Add(initialDocIndexesFoundOnUrl[j + k]);
}
var urlToWebPageWordsPositionsForShingle = new Dictionary <string, List <List <int> > >
{
{ mostPopularUrls[savei], urlTextWordsPositionsForShingle }
};
var plagiarismWebForShingle = Plagiarism <string> .FindPlagiarism(urlToWebPageWordsPositionsForShingle, initialDocIndexesForShingle);
plagiarismWebs.Add(plagiarismWebForShingle);
}
return(plagiarismWebs);
});
}
Task.WaitAll(tasks);
for (int i = 0; i < mostPopularUrls.Count; i++)
{
var plagiarismWebs = tasks[i].Result;
for (int j = 0; j < plagiarismWebs.Count; j++)
{
var plagiarismWebForShingle = plagiarismWebs[j];
plagiarismWeb.Add(plagiarismWebForShingle);
}
}
stopwatch.Stop();
foreach (var kvp in plagiarismWeb.SourceIdToInitialWordsIndexes)
{
plagiarismWeb.SourceIdToInitialDocumentHtml.Add(kvp.Key, TextManager.ComposeHtmlText(initialWords, kvp.Value));
}
plagiarismWeb.DebugLogs = debugLog;
return(plagiarismWeb);
}
