// ------------------------------------------------------------------------------------------------------------
// Name: getXmlStats
// Goal: Get all the text from the <s><t> nodes and compute a hash code from it
// Also compute the number of words and sentences
// History:
// 04-02-2016 ERK Created
// ------------------------------------------------------------------------------------------------------------
public bool getXmlStats(String sFileIn, ref String sSimHash, List <String> lStat,
ref int iWords, ref int iSents)
{
String sMethod = "simhash";
try {
// Initialise
iWords = 0; iSents = 0;
bool bUseNext = false;
// We need to have our own XmlDocument ready
XmlDocument pdxLocal = new XmlDocument();
// Prepare a string reader to read all we need
StringBuilder sbThis = new StringBuilder();
// Alsoo prepare a string reader to store potential StatusInfo
StringBuilder sbStat = new StringBuilder();
// Create an XmlReader to get to the <s><t> nodes...
using (StreamReader rdFileTmp = new StreamReader(sFileIn))
using (XmlReader rdFolia = XmlReader.Create(rdFileTmp)) {
// (1) Walk through the bare folia input file
while (!rdFolia.EOF && rdFolia.Read())
{
// (2) Check the input element
if (rdFolia.IsStartElement("t"))
{
// It needs to have an attribute [class]
if (rdFolia.HasAttributes)
{
// Get the @class attribute
String sClass = rdFolia.GetAttribute("class");
// Check the value
if (sClass == "nld" || sClass == "nl")
{
// Correct attribute: read the node
String sContent = rdFolia.ReadInnerXml();
String sLine = sContent + "\n";
sbThis.Append(sLine);
// Check for StatusInfo
if (bUseNext)
{
sbStat.Append(" // " + sContent);
// Add to the list of statusinfo evidence
lStat.Add(sbStat.ToString());
sbStat.Clear();
bUseNext = false;
}
else if (General.DoLike(sLine.ToLower(),
"*vertaald*|*vertaling*|*ondertiteling*|*bewerkt*|*ripped*|*download*|*copyright*"))
{
// Is this the first one?
if (sbStat.Length > 0)
{
sbStat.Append("\n");
}
sbStat.Append(sContent);
bUseNext = true;
}
}
}
}
else if (rdFolia.IsStartElement("w"))
{
// Get the @class attribute
String sClass = rdFolia.GetAttribute("class");
if (sClass == "Vern")
{
iWords += 1;
}
}
else if (rdFolia.IsStartElement("s"))
{
// Keep track of the number of sentences
iSents += 1;
}
}
}
// Create one string from the whole
String sTotal = sbThis.ToString();
// sStat = sbStat.ToString();
// =============== DEBUG ===============
// Store the string into a text file
File.WriteAllText(sFileIn + ".txt", sTotal, System.Text.Encoding.UTF8);
// =====================================
switch (sMethod)
{
case "md5":
// Method #1: compute the hash from this string
var md5 = MD5.Create();
MemoryStream mStrm = new MemoryStream(Encoding.UTF8.GetBytes(sTotal));
byte[] hashBytes = md5.ComputeHash(mStrm);
// Convert the byte array to a hash string
sSimHash = ByteArrayToString(hashBytes);
break;
case "simhash":
// Method #2: compute the simhash from this string
SimHashAnalyzer oAna = new SimHashAnalyzer();
// errHandle.Status("input = [" + sTotal + "]");
UInt64 iSimHash = oAna.DoCalculateSimHash(sTotal);
// Convert integer to string
sSimHash = Convert.ToString(iSimHash);
break;
}
return(true);
} catch (Exception ex) {
// Warn the user
errHandle.DoError("xmlTools/getXmlStats", ex);
return(false);
}
}
/* -------------------------------------------------------------------------------------
* Name: findDuplicates
* Goal: Go through the list and find any duplicates
* Add this duplicate information to the .cmdi file
* History:
* 10/feb/2016 ERK Created
------------------------------------------------------------------------------------- */
public bool findDuplicates(ref List<SubInstance> lSubInst, int iGoodHd, ref omdbapi objOmdb)
{
try {
// Need to have a hash analyzer object
util.SimHashAnalyzer oSim = new util.SimHashAnalyzer();
// Walk through the list of instances
for (int i=0;i<lSubInst.Count;i++) {
// Walk through all other instances that could match up with me
for (int j=0;j< i;j++) {
// Check if the idmove or the imdbid is similar between <i,j>
if (lSubInst[i].sIdMovie == lSubInst[j].sIdMovie ||
lSubInst[i].sImdbId == lSubInst[j].sImdbId) {
// Get the hamming distance between items <i,j>
int iHdist = oSim.GetHammingDistance(lSubInst[i].simhash, lSubInst[j].simhash);
// Is this on or below the threshold?
if (iHdist <= iGoodHd) {
// So this is probably a duplicate of me -- add it
lSubInst[i].addDuplicate(j);
// DO NOT add me to the list of the other one --> EXTINCT
// lSubInst[j].addDuplicate(i);
// DO add me to the list of the other one (otherwise order of occurrance plays a role)
lSubInst[j].addDuplicate(i);
}
}
}
}
// Walk through all the instances again
for (int i = 0; i < lSubInst.Count; i++) {
String sTargetDir = sDirRoot; // Directory where we will store the result
// Get this instance
SubInstance oOrg = lSubInst[i];
// Initialisations
String sLink = "";
List<String> lSimilar = new List<string>();
List<String> lEvid = new List<string>();
// Does this one have duplicates?
if (oOrg.lDup.Count>0) {
// Find the first longest text both in words and sentences
int iWords = oOrg.words;
int iSents = oOrg.sents;
int iLongest = -1;
bool bEqual = true; // Assume copies are NOT equal...
for (int j=0;j<oOrg.lDup.Count;j++) {
SubInstance oThis = lSubInst[oOrg.lDup[j]];
if (oThis.words> iWords && oThis.sents >= iSents) {
// Adapt the new maximum
iWords = oThis.words;
iSents = oThis.sents;
iLongest = j;
}
// Check for inequality
if (oThis.words != iWords || oThis.sents != iSents) bEqual = false;
}
// Check what the longest is; that's the most 'original'
if (iLongest<0) {
// Are they equal?
if (bEqual) {
// The [oOrg] is the longest
oOrg.license = "equal";
sLink = "list";
} else {
// The [oOrg] is the longest
oOrg.license = "largest";
sLink = "this";
}
// Create a list of similar ones
for (int j = 0; j < oOrg.lDup.Count; j++) {
// aSimilar.Add(lSubInst[oOrg.lDup[j]].name);
lSimilar.Add(lSubInst[oOrg.lDup[j]].name);
}
} else {
// Another one is the longest
oOrg.license = "copy";
sLink = lSubInst[iLongest].name;
}
} else {
// This is a unique subtitle file
oOrg.license = "unique";
sLink = "none";
}
// Adapt the .cmdi.xml file for this item
String sFileCmdi = oOrg.file.Replace(".folia.xml", ".cmdi.xml");
// Do we need to continue?
if (!File.Exists(sFileCmdi)) {
// If the CMDI file does not exist, we cannot adapt it
// It is not really an error, but we should give a warning
errHandle.Status("findDuplicates: skipping non-existent [" + sFileCmdi + "]");
return true;
}
// Read the CMDI
XmlDocument pdxCmdi = new XmlDocument();
pdxCmdi.Load(sFileCmdi);
oTools.SetXmlDocument(pdxCmdi, CLARIN_CMDI);
// Get correct namespace manager
XmlNamespaceManager nsFolia = new XmlNamespaceManager(pdxCmdi.NameTable);
nsFolia.AddNamespace("f", pdxCmdi.DocumentElement.NamespaceURI);
// Zoek het <Subtitle> element
XmlNode ndxSubtitle = pdxCmdi.SelectSingleNode("./descendant::f:Subtitle", nsFolia);
if (ndxSubtitle != null) {
// Get the movie id
XmlNode ndxMovieId = pdxCmdi.SelectSingleNode("./descendant::f:MovieId", nsFolia);
String sIdMovie = ndxMovieId.InnerText;
// Get the list of languages
String sSubLangs = "";
if (!getSubtitleLanguages(sIdMovie, ref sSubLangs)) return false;
// Remove any available list
XmlNode ndxAvailableList = ndxSubtitle.SelectSingleNode("./child::f:AvailableList", nsFolia);
if (ndxAvailableList != null) {
ndxAvailableList.RemoveAll();
ndxSubtitle.RemoveChild(ndxAvailableList);
XmlNode ndxTmp = ndxSubtitle.SelectSingleNode("./child::f:languageAvailable", nsFolia);
if (ndxTmp != null) {
ndxTmp.RemoveAll();
ndxSubtitle.RemoveChild(ndxTmp);
}
}
// Process the languages available list:
XmlNode ndxLngAvail = ndxSubtitle.SelectSingleNode("./child::f:languageAvailable", nsFolia);
if (ndxLngAvail == null) {
ndxLngAvail = oTools.AddXmlChild(ndxSubtitle, "languageAvailable", "", sSubLangs, "text");
} else {
// Check if this string is already in there
if (!ndxLngAvail.InnerText.Contains(sSubLangs)) {
// Add the string
ndxLngAvail.InnerText = ndxLngAvail.InnerText + " " + sSubLangs;
}
}
// Check if there is a statusinfo child
XmlNode ndxStatusInfo = ndxSubtitle.SelectSingleNode("./child::f:StatusInfo", nsFolia);
if (ndxStatusInfo == null) {
// Create such a child
ndxStatusInfo = oTools.AddXmlChild(ndxSubtitle, "StatusInfo",
"status", "", "attribute",
"link", "", "attribute");
} else {
// Get any status info evidence there is
XmlNode ndxEvid = ndxStatusInfo.SelectSingleNode("./child::f:Evidence", nsFolia);
while (ndxEvid != null) {
lEvid.Add(ndxEvid.InnerText);
ndxEvid = ndxEvid.SelectSingleNode("./following-sibling::f:Evidence", nsFolia);
}
}
// Add the information into the status info node
ndxStatusInfo.Attributes["status"].Value = oOrg.license;
ndxStatusInfo.Attributes["link"].Value = sLink;
// Remove any previous links that might be still in here
XmlNode ndxWork = ndxStatusInfo.SelectSingleNode("./child::f:Similar", nsFolia);
while (ndxWork != null) {
XmlNode ndxRemove = ndxWork;
ndxWork = ndxWork.SelectSingleNode("./following-sibling::f:Similar", nsFolia);
// Remove worknode contents
ndxRemove.RemoveAll();
// Remove the worknode itself
ndxStatusInfo.RemoveChild(ndxRemove);
}
// Add any links
for (int j=0;j<lSimilar.Count;j++) {
XmlNode ndxSimi = oTools.AddXmlChild(ndxStatusInfo, "Similar",
"SimilarId", Convert.ToString(j+1), "attribute");
ndxSimi.InnerText = Path.GetFileNameWithoutExtension( lSimilar[j]);
}
// Try to find out more information, depending on the status we have found
switch (oOrg.license) {
case "copy":
case "largest":
case "equal":
case "unique":
bool bCopyright = false;
bool bTranslated = false;
bool bDownload = false;
String sDetails = "";
// Walk through all evidence
for (int j=0; j<lEvid.Count;j++) {
// Get this evidence
String sEvid = lEvid[j].ToLower();
// Check for RIP information
// "*vertaald*|*vertaling*|*ondertiteling*|*bewerkt*|*ripped*|*download*|*copyright*"
if (util.General.DoLike(sEvid, "*ripped*|*copyright*")) {
// Add the copyright information as evidence if it is necessary
if (bTranslated || bDownload || !bCopyright) sDetails = sEvid;
bCopyright = true;
} else if (util.General.DoLike(sEvid, "*vertaald*|*vertaling*|*ondertiteling*|*bewerkt*")) {
// Double check the text of the evidence
if (util.General.DoLike(sEvid, "*broadcast text*|*bti *")) {
// This is 'stolen' from BTI or its predecessor
bCopyright = true;
sEvid = "BTI: " + sEvid;
} else {
bTranslated = true;
}
if (sDetails == "" || bDownload) sDetails = sEvid;
} else if (util.General.DoLike(sEvid, "*download*")) {
bDownload = true;
if (sDetails == "") sDetails = sEvid;
}
}
// Additional check
XmlNode ndxUserClass = ndxSubtitle.SelectSingleNode("./descendant::f:Author/child::f:UserClass", nsFolia);
if (ndxUserClass != null) {
String sUserClass = ndxUserClass.InnerText.ToLower();
if (sUserClass == "subtranslator") {
bTranslated = true;
sDetails = "userclass=SubTranslator";
}
}
// We should be able to determine the license information
String sLicense = "";
if (bCopyright)
sLicense = "copyright";
else if (bTranslated)
sLicense = "translation";
else if (bDownload)
sLicense = "download";
else
sLicense = "unknown";
// Find the location where we are going to put this information
XmlNode ndxLicenseType = ndxSubtitle.SelectSingleNode("./descendant::f:LicenseType", nsFolia);
ndxLicenseType.InnerText = sLicense;
XmlNode ndxLicenseDetails = ndxSubtitle.SelectSingleNode("./descendant::f:LicenseDetails", nsFolia);
ndxLicenseDetails.InnerText = sDetails;
// Adapt the target directory
XmlNode ndxYear = pdxCmdi.SelectSingleNode("./descendant::f:Year", nsFolia);
XmlNode ndxImdbId = pdxCmdi.SelectSingleNode("./descendant::f:ImdbId", nsFolia);
String sYear = ndxYear.InnerText;
String sImdbId = ndxImdbId.InnerText;
// Determine the target directory...
// WAS: sTargetDir += oOrg.license + "/" + sLicense + "/";
if (sYear != "")
sTargetDir += sYear + "/";
else
sTargetDir += "unknown/";
if (sImdbId != "") sTargetDir += sImdbId + "/";
break;
default:
// No further license determination is needed, since this is a copy
break;
}
}
// Zoek het <Movie> element
XmlNode ndxMovie = pdxCmdi.SelectSingleNode("./descendant::f:Movie", nsFolia);
if (ndxMovie != null) {
// Movie information needs to be gathered *ALWAYS*
XmlNode ndxImdbId = ndxMovie.SelectSingleNode("./child::f:ImdbId", nsFolia);
String sImdbId = ndxImdbId.InnerText;
// Get the movie information
MovieInfo oInfo = objOmdb.getInfo(sImdbId);
if (oInfo == null) {
// Not sure what to do now
int iError = 1;
errHandle.Status("findDuplicates: could not get information for imdb="+sImdbId);
} else {
// (1) Add the runtime information
if (!addOneInfo(ndxMovie, nsFolia, "Runtime", oInfo.runtime)) return false;
// (2) Add the COUNTRY information
if (!addMultiInfo(ndxMovie, nsFolia, "Country", oInfo.country)) return false;
// (3) Add the GENRE information
if (!addMultiInfo(ndxMovie, nsFolia, "Genre", oInfo.genre)) return false;
// (4) Add the LANGUAGE information
if (!addMultiInfo(ndxMovie, nsFolia, "Language", oInfo.language)) return false;
// (5) Add the DIRECTOR information
if (!addMultiInfo(ndxMovie, nsFolia, "Director", oInfo.director)) return false;
// (6) Add the WRITER information
if (!addMultiInfo(ndxMovie, nsFolia, "Writer", oInfo.writer)) return false;
// (7) Add the ACTOR information
if (!addMultiInfo(ndxMovie, nsFolia, "Actor", oInfo.actors)) return false;
// (8) Add other information: rated, released, plot, awards, imdbRating, imdbVotes
if (!addOneInfo(ndxMovie, nsFolia, "Rated", oInfo.rated)) return false;
if (!addOneInfo(ndxMovie, nsFolia, "Released", oInfo.released)) return false;
if (!addOneInfo(ndxMovie, nsFolia, "Plot", oInfo.plot)) return false;
if (!addOneInfo(ndxMovie, nsFolia, "Awards", oInfo.awards)) return false;
if (!addOneInfo(ndxMovie, nsFolia, "imdbRating", oInfo.imdbRating)) return false;
if (!addOneInfo(ndxMovie, nsFolia, "imdbVotes", oInfo.imdbVotes.Replace(",", ""))) return false;
// (9) Look for the <Series>...
XmlNode ndxSeries = pdxCmdi.SelectSingleNode("./descendant::f:Series", nsFolia);
if (ndxSeries != null) {
// Get the nodes we are interested in
XmlNode ndxSeason = ndxSeries.SelectSingleNode("./child::f:Season", nsFolia);
XmlNode ndxEpisode = ndxSeries.SelectSingleNode("./child::f:Episode", nsFolia);
XmlNode ndxParent = ndxSeries.SelectSingleNode("./child::f:ParentImdbId", nsFolia);
MovieInfo oParent = null;
if (ndxParent != null && ndxParent.InnerText != "") {
String sParentImdbId = ndxParent.InnerText;
oParent = objOmdb.getInfo(sParentImdbId);
}
// Add the season/episode information
if (ndxSeason != null && ndxEpisode != null) {
if (oParent== null) {
oTools.AddAttribute(ndxSeason, "Name", "");
oTools.AddAttribute(ndxEpisode, "Name", "");
} else {
oTools.AddAttribute(ndxSeason, "Name", "");
oTools.AddAttribute(ndxEpisode, "Name", "");
}
}
}
}
}
// Save the adapted CMDI
pdxCmdi.Save(sFileCmdi);
// Create the target directory if it does not exist yet
if (!Directory.Exists(sTargetDir)) {
Directory.CreateDirectory(sTargetDir);
}
// Get the file name
String sName = Path.GetFileNameWithoutExtension(sFileCmdi).Replace(".cmdi", "");
String sSrc = Path.GetDirectoryName(sFileCmdi);
if (!sSrc.EndsWith("/") && !sSrc.EndsWith("\\")) sSrc += "/";
// Copy the CMDI
File.Copy(sFileCmdi, sTargetDir + sName + ".cmdi.xml", true);
// Copy the folia
sName = sName + ".folia.xml.gz";
File.Copy(sSrc + sName, sTargetDir + sName, true);
// Show where we are
errHandle.Status("copying:\t" + oOrg.name + "\t" + oOrg.license + "\t" + sLink + "\t" + sTargetDir);
}
// Return positively
return true;
} catch (Exception ex) {
errHandle.DoError("oprConv/findDuplicates", ex);
return false;
}
}
/* -------------------------------------------------------------------------------------
* Name: getDistanceOview
* Goal: Get an overview of the distances between files
* History:
* 8/feb/2016 ERK Created
------------------------------------------------------------------------------------- */
public String getDistanceOview()
{
String sBack = "";
StringBuilder sbThis = new StringBuilder();
try {
util.SimHashAnalyzer oSim = new util.SimHashAnalyzer();
// Compare all similarity hashes with one another
for (int i=0;i<lstSimHash.Count;i++) {
for (int j=0;j< i;j++) {
// Compare items <i,j>
float fDist = oSim.GetLikenessValue(lstSimHash[i].iSimHash, lstSimHash[j].iSimHash);
int iHdist = oSim.GetHammingDistance(lstSimHash[i].iSimHash, lstSimHash[j].iSimHash);
sbThis.AppendLine( fDist +
"\t" + iHdist +
"\t" + lstSimHash[i].sFile +
"\t" + lstSimHash[j].sFile +
"\tfc " + lstSimHash[i].sFile + " " + lstSimHash[j].sFile);
}
}
// Combine
sBack = sbThis.ToString();
// Return the result
return sBack;
} catch (Exception ex) {
errHandle.DoError("oprConv/getDistanceOview", ex);
return "";
}
}
// ------------------------------------------------------------------------------------------------------------
// Name: getXmlStats
// Goal: Get all the text from the <s><t> nodes and compute a hash code from it
// Also compute the number of words and sentences
// History:
// 04-02-2016 ERK Created
// ------------------------------------------------------------------------------------------------------------
public bool getXmlStats(String sFileIn, ref String sSimHash, List<String> lStat,
ref int iWords, ref int iSents)
{
String sMethod = "simhash";
try {
// Initialise
iWords = 0; iSents = 0;
bool bUseNext = false;
// We need to have our own XmlDocument ready
XmlDocument pdxLocal = new XmlDocument();
// Prepare a string reader to read all we need
StringBuilder sbThis = new StringBuilder();
// Alsoo prepare a string reader to store potential StatusInfo
StringBuilder sbStat = new StringBuilder();
// Create an XmlReader to get to the <s><t> nodes...
using (StreamReader rdFileTmp = new StreamReader(sFileIn))
using (XmlReader rdFolia = XmlReader.Create(rdFileTmp)) {
// (1) Walk through the bare folia input file
while (!rdFolia.EOF && rdFolia.Read()) {
// (2) Check the input element
if (rdFolia.IsStartElement("t")) {
// It needs to have an attribute [class]
if (rdFolia.HasAttributes) {
// Get the @class attribute
String sClass = rdFolia.GetAttribute("class");
// Check the value
if (sClass == "nld" || sClass == "nl") {
// Correct attribute: read the node
String sContent = rdFolia.ReadInnerXml();
String sLine = sContent + "\n";
sbThis.Append(sLine);
// Check for StatusInfo
if (bUseNext) {
sbStat.Append(" // "+ sContent );
// Add to the list of statusinfo evidence
lStat.Add(sbStat.ToString());
sbStat.Clear();
bUseNext = false;
} else if (General.DoLike(sLine.ToLower(),
"*vertaald*|*vertaling*|*ondertiteling*|*bewerkt*|*ripped*|*download*|*copyright*")) {
// Is this the first one?
if (sbStat.Length > 0) sbStat.Append("\n");
sbStat.Append(sContent);
bUseNext = true;
}
}
}
} else if (rdFolia.IsStartElement("w")) {
// Get the @class attribute
String sClass = rdFolia.GetAttribute("class");
if (sClass == "Vern") iWords += 1;
} else if (rdFolia.IsStartElement("s")) {
// Keep track of the number of sentences
iSents += 1;
}
}
}
// Create one string from the whole
String sTotal = sbThis.ToString();
// sStat = sbStat.ToString();
// =============== DEBUG ===============
// Store the string into a text file
File.WriteAllText(sFileIn + ".txt", sTotal, System.Text.Encoding.UTF8);
// =====================================
switch (sMethod) {
case "md5":
// Method #1: compute the hash from this string
var md5 = MD5.Create();
MemoryStream mStrm = new MemoryStream(Encoding.UTF8.GetBytes(sTotal));
byte[] hashBytes = md5.ComputeHash(mStrm);
// Convert the byte array to a hash string
sSimHash = ByteArrayToString(hashBytes);
break;
case "simhash":
// Method #2: compute the simhash from this string
SimHashAnalyzer oAna = new SimHashAnalyzer();
// errHandle.Status("input = [" + sTotal + "]");
UInt64 iSimHash = oAna.DoCalculateSimHash(sTotal);
// Convert integer to string
sSimHash = Convert.ToString(iSimHash);
break;
}
return true;
} catch (Exception ex) {
// Warn the user
errHandle.DoError("xmlTools/getXmlStats", ex);
return false;
}
}
