public static SizeBasedCompleteSignature CreateSignatureForLocalFile(string localFilePath)
{
var sig = new SizeBasedCompleteSignature();
var buffer = new byte[ConfigHelper.SignatureSize];
var sigDict = new Dictionary<int, List<BlockSignature>>();
using (var fs = new FileStream(localFilePath, FileMode.Open))
{
long offset = 0;
uint idCount = 0;
int bytesRead = 0;
while ((bytesRead = fs.Read(buffer, 0, ConfigHelper.SignatureSize)) > 0)
{
var blockSig = GenerateBlockSig(buffer, offset,bytesRead, idCount);
List<BlockSignature> sigList;
if (!sigDict.TryGetValue(bytesRead, out sigList))
{
sigList = new List<BlockSignature>();
sigDict[bytesRead] = sigList;
}
sigList.Add(blockSig);
offset += bytesRead;
idCount++;
}
}
var sizedBaseSignature = new SizeBasedCompleteSignature();
sizedBaseSignature.Signatures = new Dictionary<int, CompleteSignature>();
foreach (var key in sigDict.Keys)
{
var compSig = new CompleteSignature() {SignatureList = sigDict[key].ToArray()};
sizedBaseSignature.Signatures[key] = compSig;
}
return sizedBaseSignature;
}
// defrags a group of sigs... merges them together.
private void DefragSigGroup(SizeBasedCompleteSignature blobSig, DefragNode sig)
{
}
public void UploadSignatureForBlob(string blobName, string containerName, SizeBasedCompleteSignature sig)
{
var client = AzureHelper.GetCloudBlobClient();
var container = client.GetContainerReference(containerName);
// upload sig.
var sigBlobName = AzureHelper.SetSignatureName(containerName, blobName);
var sigBlob = container.GetBlockBlobReference(sigBlobName);
using (Stream s = new MemoryStream())
{
SerializationHelper.WriteBinarySizedBasedSignature(sig, s);
s.Seek(0, SeekOrigin.Begin);
sigBlob.UploadFromStream(s);
}
}
/// <summary>
/// Merge smaller blocks into something at least fragmentMergeSize bytes long.
/// Only upload at most maxUploadLimit (0 == no limit).
/// Should this be in CommonOps?
/// Lame... really? DEFRAG? Then again I suppose the term IS appropriate.
/// </summary>
/// <param name="containerName"></param>
/// <param name="blobName"></param>
public void DefragBlob(SizeBasedCompleteSignature blobSig, long maxUploadLimitMB = 2)
{
var allBlobSigs = blobSig.Signatures.Values.SelectMany(x => x.SignatureList).OrderBy(a => a.Offset).ToList();
var targetSigSize = ConfigHelper.SignatureSize;
// loop through sigs, merge what we can but dont exceed maxUploadLimit
long bytesToUpload = 0;
var byteRangesToUpload = new List<RemainingBytes>();
var defragNodeList = new List<DefragNode>();
for (var i = 0; i < allBlobSigs.Count; i++)
{
uint sigSize = 0;
var j = i;
while (j < allBlobSigs.Count)
{
var sig = allBlobSigs[j];
j++;
// break if we get too big.
if (sigSize + sig.Size > targetSigSize)
{
break;
}
sigSize += sig.Size;
}
defragNodeList.Add(new DefragNode { Offset = allBlobSigs[i].Offset, Size = sigSize, SigPos = i, NoSigs = j - i - 1 });
}
// defragNodeList is a list of sigs, and size. These ones will be merged.
var sortedList = defragNodeList.OrderByDescending(n => n.NoSigs).ToList();
// the entries in defragNodeList that has the max number of sigs in it (ie most fragmentation) will be the ones to get merged.
foreach( var sig in sortedList)
{
DefragSigGroup(blobSig, sig);
bytesToUpload += sig.Size;
if (bytesToUpload > maxUploadLimitMB)
{
break;
}
}
}
/// <summary>
/// Existing blocks + sigs are in searchResults
/// new
/// </summary>
/// <param name="bytesUploaded"></param>
/// <returns></returns>
internal static SizeBasedCompleteSignature CreateSignatureFromNewAndReusedBlocks(List<UploadedBlock> allBlocks)
{
var sigDict = new Dictionary<int, List<BlockSignature>>();
List<BlockSignature> sigList;
// new blocks
foreach (var newBlock in allBlocks )
{
if (!sigDict.TryGetValue((int)newBlock.Sig.Size, out sigList))
{
sigList = new List<BlockSignature>();
sigDict[(int) newBlock.Sig.Size] = sigList;
}
// add sig to the list.
sigList.Add( newBlock.Sig);
}
var sizedBaseSignature = new SizeBasedCompleteSignature();
sizedBaseSignature.Signatures = new Dictionary<int, CompleteSignature>();
foreach (var key in sigDict.Keys)
{
var compSig = new CompleteSignature() {SignatureList = sigDict[key].ToArray()};
sizedBaseSignature.Signatures[key] = compSig;
}
return sizedBaseSignature;
}
/// <summary>
/// Gets SizeBasedCompleteSignature.
/// Format is: first 4 bytes are number of CompleteSig's there are.
/// For each complete sig, the format is 4 bytes, number of entries.
/// </summary>
/// <param name="s"></param>
/// <returns></returns>
public static SizeBasedCompleteSignature ReadSizeBasedBinarySignature(Stream s)
{
var sig = new SizeBasedCompleteSignature();
sig.Signatures = new Dictionary<int, CompleteSignature>();
// always go to beginning of stream.
s.Seek(0, SeekOrigin.Begin);
var reader = new BinaryReader(s);
int numberOfCompleteSignatures = reader.ReadInt32();
for (var i = 0; i < numberOfCompleteSignatures; i++)
{
int keySize = reader.ReadInt32();
var completeSig = ReadBinaryCompleteSignature(s);
sig.Signatures[keySize] = completeSig;
}
return sig;
}
public static SignatureSearchResult SearchLocalFileForSignatures(string localFilePath, SizeBasedCompleteSignature sig)
{
var result = new SignatureSearchResult();
// length of file.
var tempFile = File.Open(localFilePath, FileMode.Open);
var fileLength = tempFile.Length;
tempFile.Close();
var offset = 0;
var windowSize = ConfigHelper.SignatureSize;
var windowBuffer = new byte[windowSize];
// signatures we can reuse.
var signaturesToReuse = new List<BlockSignature>();
// get sizes of signatures (block sizes) from existing sig.
// then loop through all sizes looking for matches in local file.
// important to search from largest to smallest.
var signatureSizes = sig.Signatures.Keys.ToList();
signatureSizes.Sort();
signatureSizes.Reverse();
// byte ranges that have not been matched to existing blocks yet.
var remainingByteList = new List<RemainingBytes>();
remainingByteList.Add(new RemainingBytes {BeginOffset = 0, EndOffset = fileLength - 1});
// Create the memory-mapped file.
using (var mmf = MemoryMappedFile.CreateFromFile(localFilePath, FileMode.Open))
{
using (var accessor = mmf.CreateViewAccessor())
{
// Any sigs smaller than 100 bytes? skip?
// Valid?
// Really want to avoid searching for single bytes everywhere.
foreach (var sigSize in signatureSizes)
{
var sigs = sig.Signatures[sigSize];
var newRemainingByteList = SearchLocalFileForSignaturesBasedOnSize(sigs, accessor, remainingByteList, sigSize, fileLength, signaturesToReuse);
remainingByteList = newRemainingByteList;
}
}
}
result.ByteRangesToUpload = remainingByteList;
result.SignaturesToReuse = signaturesToReuse;
return result;
}
private void PopulateSignatureTreeByOffset(TreeView sigTV, SizeBasedCompleteSignature sig, Dictionary<TreeNode, List<BlockSignature>> sigDict)
{
var sigList = new List<BlockSignature>();
if (sig.Signatures != null)
{
foreach (var size in sig.Signatures.Keys)
{
foreach (var sSig in sig.Signatures[size].SignatureList)
{
sigList.Add(sSig);
}
}
var sortedSigList = (from s in sigList orderby s.Offset select s).ToList<BlockSignature>();
PopulateRootNodes(sigTV, sortedSigList, sigDict);
}
}
private void ProcessDoubleClick(TreeNode selectedNode, SizeBasedCompleteSignature sig)
{
if (selectedNode != null && selectedNode.Parent != null)
{
var sp = selectedNode.Parent.Text.Split();
var sigSize = Convert.ToInt32(sp[0]);
var offset = Convert.ToInt64(selectedNode.Text.Split()[0]);
var specificSig = (from s in sig.Signatures[sigSize].SignatureList where s.Offset == offset select s).First<BlockSignature>();
var md5String = ByteArrayToString(specificSig.MD5Signature);
var rollingSig = string.Format("{0}:{1}", specificSig.RollingSig.Sig1, specificSig.RollingSig.Sig2);
var msg = string.Format("Offset: {0}\nSize: {1}\nRollingSig: {2}\nMD5: {3}", specificSig.Offset.ToString(),
specificSig.Size.ToString(), rollingSig, md5String);
var dialog = MessageBox.Show(msg);
}
}
/// <summary>
/// Loads the sig file.
/// </summary>
/// <param name="filename"></param>
private void LoadSigFile(string filename, ref SizeBasedCompleteSignature sig, Dictionary<TreeNode, List<BlockSignature>> sigDict, TreeView sigTV)
{
using (var fs = new FileStream(filename, FileMode.Open))
{
sig = SerializationHelper.ReadSizeBasedBinarySignature( fs );
VerifySignature(sig);
if (sigTV.Name == "sigTreeView")
{
sig1MD5Dict = GenerateMD5DictFromSig(sig);
file1Size = CalculateFileSize(sig);
}
else
{
sig2MD5Dict = GenerateMD5DictFromSig(sig);
file2Size = CalculateFileSize(sig);
}
}
bothFilesShared = 0;
PopulateSignatureTree(sig, sigDict, sigTV);
}
private void PopulateSignatureTree(SizeBasedCompleteSignature sig, Dictionary<TreeNode, List<BlockSignature>> sigDict, TreeView sigTV)
{
sigTV.Nodes.Clear();
bool isLeftTree;
if (sigTV.Name == "sigTreeView")
{
isLeftTree = true;
}
else
{
isLeftTree = false;
}
bothFilesShared = 0;
PopulateSignatureTreeByOffset(sigTV, sig, sigDict);
sharedSize.Text = bothFilesShared.ToString("N0");
newSize.Text = (file2Size - bothFilesShared).ToString("N0");
if (isLeftTree)
{
file1TotalSize.Text = file1Size.ToString("N0");
sigTV.Update();
}
else
{
file2TotalSize.Text = file2Size.ToString("N0");
sigTV.Update();
}
}
private List<Tuple<byte[], long>> GenerateMD5ListFromSig(SizeBasedCompleteSignature? sig)
{
var md5List = new List< Tuple<byte[], long>>();
if (sig.HasValue && sig.Value.Signatures != null)
{
foreach (var size in sig.Value.Signatures.Keys)
{
foreach (var sSig in sig.Value.Signatures[size].SignatureList)
{
var tuple = new Tuple<byte[], long>(sSig.MD5Signature, sSig.Offset);
md5List.Add(tuple);
}
}
}
return md5List;
}
private Dictionary<string, long> GenerateMD5DictFromSig(SizeBasedCompleteSignature? sig)
{
var list = GenerateMD5ListFromSig(sig);
var dict = GenerateMD5Dict(list);
return dict;
}
private long CalculateFileSize(SizeBasedCompleteSignature sig)
{
long fileSize = 0;
foreach( var sigSize in sig.Signatures.Keys)
{
fileSize += (long) sigSize * (long) sig.Signatures[sigSize].SignatureList.Count();
}
return fileSize;
}
private List<RemainingBytes> GenerateByteRangesOfBlobToDownload(List<BlockSignature> sigsToReuseList, SizeBasedCompleteSignature cloudBlobSig, string containerName, string blobName)
{
var blobSize = AzureHelper.GetBlobSize(containerName, blobName);
var remainingBytesList = new List<RemainingBytes>();
var allBlobSigs = cloudBlobSig.Signatures.Values.SelectMany(x => x.SignatureList).OrderBy(a => a.Offset).ToList();
var sortedSigs = (from sig in sigsToReuseList orderby sig.Offset ascending select sig).ToList();
long startOffsetToCopy = 0;
// loop through all cloudBlobSigs.
// If have a match in sigsToReuse, skip it.
// otherwise, take note of offset and size to download.
foreach( var sig in allBlobSigs)
{
var haveMatchingSig = sigsToReuseList.Any(s => s.MD5Signature.SequenceEqual(sig.MD5Signature));
if (!haveMatchingSig)
{
// if no match then we need to copy everything from startOffsetToCopy to sig.Offset + sig.Size
remainingBytesList.Add(new RemainingBytes()
{
BeginOffset = startOffsetToCopy,
EndOffset = sig.Offset + sig.Size -1
});
startOffsetToCopy = sig.Offset + sig.Size;
}
else
{
// we have a match therefore dont need to copy the data.
// change startOffsetToCopy to just after current sig.
startOffsetToCopy = sig.Offset + sig.Size;
}
}
return remainingBytesList;
}
/// <summary>
/// Quick check to make sure we dont have any dupe md5's
/// </summary>
/// <param name="sig"></param>
/// <returns></returns>
private Boolean VerifySignature(SizeBasedCompleteSignature sig)
{
var myDict = new Dictionary<string, int>();
var valid = true;
var count = 0;
foreach (var size in sig.Signatures.Keys)
{
foreach (var sSig in sig.Signatures[size].SignatureList)
{
var md5 = sSig.MD5Signature;
var md5Str = ByteArrayToString(md5);
if (myDict.ContainsKey( md5Str) )
{
valid = false;
count++;
}
else
{
myDict[md5Str] = 1;
}
}
}
return valid;
}
// regenerate blob locally.
// we need to either download byte ranges from Azure.
// OR
// need to copy from local file.
private void RegenerateBlob(string containerName, string blobName, List<RemainingBytes> byteRangesToDownload, string localFilePath, List<BlockSignature> reusableBlockSignatures, SizeBasedCompleteSignature blobSig, int parallelFactor = 2)
{
// removing size from the equation.
var allBlobSigs =
blobSig.Signatures.Values.SelectMany(x => x.SignatureList).OrderBy(a => a.Offset).ToList();
// LUT to see if block is to be reused or not.
var reusableBlockDict = CommonOps.GenerateBlockDict(reusableBlockSignatures.ToArray());
var offset = 0L;
using (var localStream = new FileStream( localFilePath, FileMode.Open))
using (var newStream = new FileStream( localFilePath+".new", FileMode.Create))
{
// go through all sigs in offset order.... determine if can reuse or need to download.
foreach (var sig in allBlobSigs)
{
var haveMatch = false;
if (reusableBlockDict.ContainsKey(sig.RollingSig))
{
// have a match... so will reuse local file.
var localSig = reusableBlockDict[sig.RollingSig];
var matchingLocalSigs =
localSig.Where(s => s.MD5Signature.SequenceEqual(sig.MD5Signature))
.Select(n => n)
.ToList();
if (matchingLocalSigs.Any())
{
// have a match.
var matchingLocalSig = matchingLocalSigs[0];
// huge amount of wasted allocations... maybe move this.
var buffer = new byte[matchingLocalSig.Size];
localStream.Seek(matchingLocalSig.Offset, SeekOrigin.Begin);
localStream.Read(buffer, 0, (int) matchingLocalSig.Size);
newStream.Seek(sig.Offset, SeekOrigin.Begin);
newStream.Write( buffer, 0, (int) matchingLocalSig.Size);
haveMatch = true;
offset += matchingLocalSig.Size;
}
}
if (!haveMatch)
{
// check if we have byte ranges starting at offset.
var byteRange =
(from b in byteRangesToDownload where b.BeginOffset == offset select b).FirstOrDefault();
if (byteRange != null)
{
// download bytes.
var blobBytes = DownloadBytes(containerName, blobName, byteRange.BeginOffset,
byteRange.EndOffset, parallelFactor);
newStream.Seek(sig.Offset, SeekOrigin.Begin);
newStream.Write(blobBytes, 0, (int)(byteRange.EndOffset - byteRange.BeginOffset + 1));
offset += (byteRange.EndOffset - byteRange.BeginOffset + 1);
}
}
}
}
// rename .new file to original
File.Replace(localFilePath + ".new", localFilePath,null);
}
public static void WriteBinarySizedBasedSignature(SizeBasedCompleteSignature sig, Stream s)
{
var writer = new BinaryWriter(s);
int numberOfSizes = sig.Signatures.Keys.Count;
// 4 bytes. Number of key sizes.
writer.Write(numberOfSizes);
foreach (int keySize in sig.Signatures.Keys)
{
// write key size.
writer.Write(keySize);
var completeSigForKeySize = sig.Signatures[keySize];
int numberOfEntries = completeSigForKeySize.SignatureList.Length;
// number of entries for this key size.
writer.Write(numberOfEntries);
foreach (var i in completeSigForKeySize.SignatureList)
{
// 8 bytes
writer.Write(i.Offset);
// 4 bytes
writer.Write(i.Size);
// 4 bytes.
writer.Write(i.BlockNumber);
// 8 bytes.
writer.Write(i.RollingSig.Sig1);
// 8 bytes.
writer.Write(i.RollingSig.Sig2);
// should be 16 bytes.
foreach (byte b in i.MD5Signature)
{
writer.Write(b);
}
}
}
}
