public override void Start()
{
if (!IsRunning)
{
mThread = new Thread(new ThreadStart(
delegate() {
while (!mStopped && !mMessenger.isMessagingFinished())
{
string message = mMessenger.getMessage();
if (message != null)
{
try
{
DocumentCorpus dc = new DocumentCorpus();
XmlReader reader = new XmlTextReader(new StringReader(message));
dc.ReadXml(reader);
reader.Close();
DispatchData(dc);
}
catch (Exception e)
{
mLogger.Error("ZeroMqReceiverComponent", e);
//File.WriteAllText(@"C:\Users\Administrator\Desktop\err\" + Guid.NewGuid().ToString("N") + ".xml", message, Encoding.UTF8);
}
}
Thread.Sleep(1);
}
}
));
mStopped = false;
mThread.Start();
}
}
public static List<String> convertFIRSTXMLtoGATE(String file)
{
List<String> documents = new List<string>();
// conversion from FIRST XML to GATE XML
// load FIRST XML corpus
DocumentCorpus corpus = new DocumentCorpus();
corpus.ReadXml(new XmlTextReader(new StreamReader(file)));
// save documents as GATE XML
XmlWriterSettings xmlSettings = new XmlWriterSettings();
xmlSettings.Indent = true;
xmlSettings.NewLineOnAttributes = true;
xmlSettings.CheckCharacters = false;
xmlSettings.Encoding = Encoding.UTF8;
//int i = 0;
foreach (Document doc in corpus.Documents)
{
StringBuilder docXML = new StringBuilder();
XmlWriter writer = XmlWriter.Create(docXML, xmlSettings);
doc.WriteGateXml(writer, /*writeTopElement=*/true, /*removeBoilerplate=*/true);
//doc.WriteXml(writer, /*writeTopElement=*/true);
String docstr = docXML.ToString();
documents.Add(docstr);
writer.Close();
}
return documents;
}
static void Main(string[] args)
{
// conversion from FIRST XML to GATE XML
// load FIRST XML corpus
DocumentCorpus corpus = new DocumentCorpus();
corpus.ReadXml(new XmlTextReader(new StreamReader(@"D:\streamer\files\23_55_08_450f24c0969d49d2883fc17a6f4e2af0.xml")));
// save documents as GATE XML
XmlWriterSettings xmlSettings = new XmlWriterSettings();
xmlSettings.Indent = true;
xmlSettings.NewLineOnAttributes = true;
xmlSettings.CheckCharacters = false;
xmlSettings.Encoding = Encoding.UTF8;
int i = 0;
foreach (Document doc in corpus.Documents)
{
StreamWriter streamWriter = new StreamWriter(string.Format(@"D:\streamer\output\{0}.xml", ++i));
XmlWriter writer = XmlWriter.Create(streamWriter, xmlSettings);
doc.WriteGateXml(writer, /*writeTopElement=*/true, /*removeBoilerplate=*/true);
//doc.WriteXml(writer, /*writeTopElement=*/true);
String docstr = writer.ToString();
writer.Close();
streamWriter.Close();
}
Console.ReadKey();
}
public string GetDoc(string corpusId, string docId, string format, bool rmvRaw, bool changesOnly, string corpusTime)
{
string dataPath = Utils.GetConfigValue("DataPath", ".");
if (corpusId == null || corpusId.Replace("-", "").Length != 32) { return "*** Invalid corpus ID."; }
corpusId = corpusId.Replace("-", "");
if (docId == null || docId.Replace("-", "").Length != 32) { return "*** Invalid document ID."; }
docId = docId.Replace("-", "");
string[] fileNames = null;
if (!string.IsNullOrEmpty(corpusTime))
{
try
{
DateTime dt = DateTime.Parse(corpusTime);
string prefix = dt.ToString("HH_mm_ss_");
string path = "\\" + dt.Year + "\\" + dt.Month + "\\" + dt.Day + "\\";
string fileName = dataPath.TrimEnd('\\') + path + prefix + corpusId + ".xml";
if (!Utils.VerifyFileNameOpen(fileName)) { return "*** Corpus not found."; }
fileNames = new string[] { fileName };
}
catch { return "*** Unable to parse time."; }
}
if (fileNames == null) { fileNames = Directory.GetFiles(dataPath, "*" + corpusId + ".xml", SearchOption.AllDirectories); }
if (fileNames.Length == 0) { return "*** Corpus not found."; }
DocumentCorpus corpus = new DocumentCorpus();
StreamReader reader = new StreamReader(fileNames[0]);
XmlTextReader xmlReader = new XmlTextReader(reader);
corpus.ReadXml(xmlReader);
xmlReader.Close();
reader.Close();
Document document = null;
foreach (Document doc in corpus.Documents)
{
if (new Guid(doc.Features.GetFeatureValue("guid")).ToString("N") == docId) { document = doc; break; }
}
if (document == null) { return "*** Document not found."; }
if (rmvRaw) { document.Features.RemoveFeature("raw"); }
string response;
if (format == "html")
{
StringWriter writer = new StringWriter();
document.MakeHtmlPage(writer, /*inlineCss=*/true);
string html = new Regex(@"<!--back_button-->.*?<!--/back_button-->").Replace(writer.ToString(), "");
response = html;
}
else if (format == "txt")
{
StringBuilder txt = new StringBuilder();
string selector = "TextBlock/Content";
if (changesOnly && document.Features.GetFeatureValue("rev") != "1") { selector = "TextBlock/Content/Unseen"; }
foreach (TextBlock block in document.GetAnnotatedBlocks(selector))
{
txt.AppendLine(block.Text);
}
response = document.Name + "\r\n\r\n" + txt.ToString();
}
else
{
StringWriter writer = new StringWriter();
XmlWriterSettings xmlSettings = new XmlWriterSettings();
xmlSettings.Indent = true;
xmlSettings.NewLineOnAttributes = true;
xmlSettings.CheckCharacters = false;
XmlWriter xmlWriter = XmlWriter.Create(writer, xmlSettings);
if (format == "gate_xml")
{
document.WriteGateXml(xmlWriter, /*writeTopElement=*/true, /*removeBoilerplate=*/true);
xmlWriter.Flush();
response = writer.ToString();
}
else // xml
{
document.WriteXml(xmlWriter, /*writeTopElement=*/true);
xmlWriter.Flush();
response = writer.ToString().Replace("<?xml version=\"1.0\" encoding=\"utf-16\"?>",
"<?xml version=\"1.0\" encoding=\"utf-8\"?>");
}
xmlWriter.Close();
}
return response;
}
protected override object ProduceData()
{
// are we done?
if (mCurrentDirIdx >= mDataDirs.Length)
{
Stop();
return(null);
}
// do we need to get more files?
if (mFiles == null)
{
mFiles = Directory.GetFiles(mDataDirs[mCurrentDirIdx], "*.xml");
Array.Sort(mFiles);
}
// did we process all currently available files?
if (mCurrentFileIdx >= mFiles.Length)
{
mFiles = null;
mCurrentFileIdx = 0;
mCurrentDirIdx++;
return(null);
}
try
{
// read next file
mLogger.Info("ProduceData", "Reading " + mFiles[mCurrentFileIdx] + " ...");
DocumentCorpus corpus = new DocumentCorpus();
StreamReader reader = new StreamReader(mFiles[mCurrentFileIdx]);
corpus.ReadXml(new XmlTextReader(reader));
//string fileName = new FileInfo(mFiles[mCurrentFileIdx]).Name;
//string corpusId = new Guid(fileName.Split('_', '.')[3]).ToString();
//corpus.Features.SetFeatureValue("guid", corpusId);
reader.Close();
// refresh corpus ID (to avoid conflicts)
corpus.Features.SetFeatureValue("guid", Guid.NewGuid().ToString());
// remove underscores in feature names
string[] tmp = new string[corpus.Features.Names.Count];
corpus.Features.Names.CopyTo(tmp, /*index=*/ 0);
foreach (string featureName in tmp)
{
if (featureName.StartsWith("_"))
{
corpus.Features.SetFeatureValue(featureName.TrimStart('_'), corpus.Features.GetFeatureValue(featureName));
corpus.Features.RemoveFeature(featureName);
}
}
foreach (Document doc in corpus.Documents)
{
// remove annotations
doc.ClearAnnotations();
// remove underscores in feature names
tmp = new string[doc.Features.Names.Count];
doc.Features.Names.CopyTo(tmp, /*index=*/ 0);
foreach (string featureName in tmp)
{
if (featureName.StartsWith("_"))
{
doc.Features.SetFeatureValue(featureName.TrimStart('_'), doc.Features.GetFeatureValue(featureName));
doc.Features.RemoveFeature(featureName);
}
}
// remove processing-specific features
foreach (string featureName in new string[] {
"detectedLanguage",
"detectedCharRange",
"bprBoilerplateCharCount",
"bprContentCharCount",
"domainName",
"urlKey",
"rev",
"blacklisted"
})
{
doc.Features.RemoveFeature(featureName);
}
// if there's raw data available, reset the content
string raw = doc.Features.GetFeatureValue("raw");
if (raw != null)
{
doc.Features.SetFeatureValue("contentType", "Html");
doc.Text = GetEncoding(doc.Features.GetFeatureValue("charSet")).GetString(Convert.FromBase64String(raw));
}
}
mCurrentFileIdx++;
while (WorkflowUtils.GetBranchLoadMax(this) > 10) // I'm giving it all she's got, Captain!
{
Thread.Sleep(1000);
}
return(corpus);
}
catch (Exception e)
{
mCurrentFileIdx++;
throw e;
}
}
public override void Start()
{
if (!IsRunning)
{
mThread = new Thread(new ThreadStart(
delegate() {
while (!mStopped && !mMessenger.isMessagingFinished())
{
string message = mMessenger.getMessage();
if (message != null)
{
try
{
DocumentCorpus dc = new DocumentCorpus();
XmlReader reader = new XmlTextReader(new StringReader(message));
dc.ReadXml(reader);
reader.Close();
DispatchData(dc);
}
catch (Exception e)
{
mLogger.Error("ZeroMqReceiverComponent", e);
//File.WriteAllText(@"C:\Users\Administrator\Desktop\err\" + Guid.NewGuid().ToString("N") + ".xml", message, Encoding.UTF8);
}
}
Thread.Sleep(1);
}
}
));
mStopped = false;
mThread.Start();
}
}
