public static void OutputIndex(Index <RawShapeletFormat> index, String parent)
{
int splitDepth = index.SplitDepth;
foreach (IndexEntry i in index.GetIndexEntries())
{
if (i is TermEntry)
{
TermEntry term = ((TermEntry)i);
String saxWord = term.SaxWord;
String iSaxWord = term.iSaxWord;
String fileName = term.FileName;
int numNodes = term.NumNodes;
int numTimeSeries = term.NumTimeSeries;
System.Console.WriteLine("SAX word: {0}, iSax word: {1}, file name: {2}, parent: {3}, split depth: {4}, num nodes: {5}, num TS: {6}", saxWord, iSaxWord, fileName, parent, splitDepth, numNodes, numTimeSeries);
String[] line = { saxWord, iSaxWord, "TermEntry", "", fileName, parent, splitDepth.ToString(), "", numNodes.ToString(), numTimeSeries.ToString() };
output = output + String.Join(";", line) + "\n";
}
else
{
SplitEntry <RawShapeletFormat> split = (SplitEntry <RawShapeletFormat>)i;
String saxWord = split.SaxWord;
String iSaxWord = split.iSaxWord;
String baseDir = split.Options.BaseDir;
String maskValue = split.Options.maskValue();
int numNodes = split.NumNodes;
int numTimeSeries = split.NumTimeSeries;
System.Console.WriteLine("SAX word: {0}, iSax word: {1}, dir: {2}, mask: {3}, split depth: {4}, num nodes: {5}, num TS: {6}", saxWord, iSaxWord, baseDir, maskValue, splitDepth, numNodes, numTimeSeries);
String[] line = { saxWord, iSaxWord, "SplitEntry", baseDir, "", parent, splitDepth.ToString(), maskValue, numNodes.ToString(), numTimeSeries.ToString() };
output = output + String.Join(";", line) + "\n";
Index <RawShapeletFormat> splitIndex = split.GetIndex();
OutputIndex(splitIndex, saxWord);
}
}
}
public CostCounter ExactSearch(double[] ts, out IndexFileDist bsf)
{
CostCounter meas = new CostCounter(0, 0);
IntervalHeap <IndexEntryDist> pq = new IntervalHeap <IndexEntryDist>(NumIndexEntries);
// approx search
TermEntry approx = ApproximateSearch(ts);
bsf = Index <DATAFORMAT> .MinFileEucDist(ts, approx.FileName);
meas.IO++;
meas.distance += approx.NumTimeSeries;
// initalize pq with IndexEntries at root node
foreach (IndexEntry e in index.Values)
{
pq.Add(new IndexEntryDist(e, Sax.MinDistPAAToiSAX(
Sax.SaxStrToSaxVals(e.SaxWord), options.SaxOpts, ts)));
}
while (!pq.IsEmpty)
{
IndexEntryDist minInfo = pq.DeleteMin();
IndexEntry minEntry = minInfo.entry;
if (minInfo.dist >= bsf.distance)
{
break;
}
if (minEntry is TermEntry)
{
IndexFileDist posMin = Index <DATAFORMAT> .MinFileEucDist(ts, ((TermEntry)minEntry).FileName);
meas.IO++;
meas.distance += minEntry.NumTimeSeries;
// update bsf
if (posMin.distance < bsf.distance)
{
bsf = posMin;
}
}
else if (minEntry is SplitEntry <DATAFORMAT> )
{
SplitEntry <DATAFORMAT> sEntry = minEntry as SplitEntry <DATAFORMAT>;
foreach (IndexEntry e in sEntry.GetIndexEntries())
{
pq.Add(new IndexEntryDist(e, Sax.MinDistPAAToiSAX(
Sax.SaxStrToSaxVals(e.SaxWord), sEntry.Options.SaxOpts, ts)));
}
}
}
return(meas);
}
public void Insert(SaxData input)
{
string saxString = Sax.SaxDataRepToSaxStr(input, options.SaxOpts);
if (splitDepth == 0 && flush == false)
{
if (!buffer.ContainsKey(saxString))
{
buffer.Add(saxString, new List <SaxData>());
}
buffer[saxString].Add(input);
}
else
{
if (index.ContainsKey(saxString))
{
IndexEntry entry = index[saxString];
if (entry is TermEntry)// if terminal, then search path terminates here
{
TermEntry tentry = (TermEntry)entry;
string oldFileName = tentry.FileName;
if (SplitEntry(tentry) == false) // check bucket requires a split
{
tentry.InsertToBuffer(input);
}
else
{
List <SaxData> B = tentry.getbuffer();
if (B == null)
{
B = new List <SaxData>();
}
DiskCost.increasesavedcost(B.Count);
ushort[] newMask = this.options.MaskCopy;
ushort[] newSaxString = Sax.SaxStrToSaxVals(saxString);
string newName = "";
for (int i = 0; i < newMask.Length; i++)
{
newName = newName + newSaxString[i].ToString() + "." + newMask[i].ToString() + "_";
}
newName = newName.Substring(0, newName.Length - 1);
string[] files = Directory.GetFiles(WorkingFolder,
string.Concat(newName, "*.txt"));
//string[] files = Directory.GetFiles(tentry.FileName);
if (tentry.OnDisk == true)
{
Assert.AreEqual(files.Length, 1);
}
else
{
Assert.AreEqual(files.Length, 0);
}
byte[] temp;
int pos = 0;
long length = -1;
int bytesToRead = SaxData.ByteLength(typeof(DATAFORMAT));
foreach (string f in files)
{
using (BinaryReader br = new BinaryReader(new FileStream(f, FileMode.Open, FileAccess.Read)))
{
length = br.BaseStream.Length;
if (length != 0)
{
DiskCost.increaserandomcost();
}
if (Math.IEEERemainder(length, bytesToRead) != 0)
{
throw new ApplicationException("Math.IEEERemainder(br.BaseStream.Length, bytesToRead) != 0");
}
while (pos < length)
{
temp = br.ReadBytes(bytesToRead);
if (temp.Length != bytesToRead)
{
throw new ApplicationException("temp.Length != bytesToRead");
}
B.Add(SaxData.Parse <DATAFORMAT>(temp));
DiskCost.increasereadcost();
pos += bytesToRead;
}
}
File.Delete(f);
}
SplitEntry <DATAFORMAT> newSplit;
if (Globals.NewSplitPolicy)
{
newSplit = new SplitEntry <DATAFORMAT>(saxString, UpdateOptions(B), (byte)(1 + splitDepth));
}
else
{
newSplit = new SplitEntry <DATAFORMAT>(saxString, UpdateOptions(null), (byte)(1 + splitDepth));
}
newSplit.Insert(input);
foreach (SaxData S in B)
{
newSplit.Insert(S);
}
// update index entry from terminal to split
index[saxString] = newSplit;
}
}
else if (entry is SplitEntry <DATAFORMAT> ) // internal node
{
((SplitEntry <DATAFORMAT>)entry).Insert(input);
}
}
else // saxString has not been seen before, create new file and entry
{
ushort[] newMask = this.options.MaskCopy;
ushort[] newSaxString = Sax.SaxStrToSaxVals(saxString);
string newName = "";
for (int i = 0; i < newMask.Length; i++)
{
newName = newName + newSaxString[i].ToString() + "." + newMask[i].ToString() + "_";
}
newName = newName.Substring(0, newName.Length - 1);
string newfile = Path.Combine(WorkingFolder, string.Concat(newName, ".0.txt"));
TermEntry newEntry = new TermEntry(saxString, newfile);
newEntry.InsertToBuffer(input);
index.Add(saxString, newEntry);
}
}
}