// very manual writer... but want to see how small I can get the data.
public static CompleteSignature ReadBinaryCompleteSignature(Stream s)
{
var sig = new CompleteSignature();
var l = new List<BlockSignature>();
var reader = new BinaryReader(s);
int numberOfEntries = reader.ReadInt32();
for (var i = 0; i < numberOfEntries; i++)
{
var entry = new BlockSignature();
// 8 bytes. offset
long offset = reader.ReadInt64();
// 4 bytes. size
int size = reader.ReadInt32();
// 4 bytes. Block Number;
int blockNumber = reader.ReadInt32();
// 4 bytes. Rolling Signature.
decimal sig1 = reader.ReadDecimal();
decimal sig2 = reader.ReadDecimal();
RollingSignature rollingSig = new RollingSignature() { Sig1 = sig1, Sig2 = sig2 };
// should be 16 bytes.
byte[] md5 = reader.ReadBytes(16);
entry.BlockNumber = (UInt32)blockNumber;
entry.RollingSig = (RollingSignature)rollingSig;
entry.MD5Signature = md5;
entry.Offset = offset;
entry.Size = (uint)size;
l.Add(entry);
}
sig.SignatureList = l.ToArray<BlockSignature>();
return sig;
}
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;
}
private static List<RemainingBytes> SearchLocalFileForSignaturesBasedOnSize(CompleteSignature sig, MemoryMappedViewAccessor accessor, List<RemainingBytes> remainingByteList, int sigSize, long fileSize, List<BlockSignature> signaturesToReuse)
{
var windowSize = sigSize;
var newRemainingBytes = new List<RemainingBytes>();
var sigDict = GenerateBlockDict(sig);
var buffer = new byte[sigSize];
var offset = 0L;
foreach (var byteRange in remainingByteList)
{
var byteRangeSize = byteRange.EndOffset - byteRange.BeginOffset + 1;
// if byte range is large... and signature size is small (what values???) then dont check.
// We could end up with LOADS of tiny sig matching where ideally we'd use a larger new sig block.
// The exception is when the sig size exactly matches the byterange size... then we allow it to check if the sig will match
// in practice this allows small (1-2 byte sigs) to match the byte ranges.
if (byteRangeSize > 1000 && sigSize > 100 || byteRangeSize == sigSize)
{
// if byteRange is smaller than the key we're using, then there cannot be a match so add
// it to the newRemainingBytes list
if (byteRange.EndOffset - byteRange.BeginOffset + 1 >= windowSize)
{
// search this byterange for all possible keys.
offset = byteRange.BeginOffset;
var generateFreshSig = true;
var bytesRead = 0L;
RollingSignature? currentSig = null;
long oldEndOffset = byteRange.BeginOffset;
do
{
if (generateFreshSig)
{
bytesRead = accessor.ReadArray(offset, buffer, 0, windowSize);
currentSig = CreateRollingSignature(buffer, (int)bytesRead);
}
else
{
// roll existing sig.
var previousByte = accessor.ReadByte(offset - 1);
var nextByte = accessor.ReadByte(offset + windowSize - 1); // Need bounds checking?
currentSig = RollSignature(windowSize, previousByte, nextByte, currentSig.Value);
}
if (sigDict.ContainsKey(currentSig.Value))
{
// populate buffer. Potential waste of IO here.
bytesRead = accessor.ReadArray(offset, buffer, 0, windowSize);
// check md5 sig.
var md5Sig = CreateMD5Signature(buffer, (int)bytesRead);
var sigsForCurrentRollingSig = sigDict[currentSig.Value];
// have a matching md5? If so, we have a match.
var matchingSigs =
sigsForCurrentRollingSig.Where(s => s.MD5Signature.SequenceEqual(md5Sig))
.Select(n => n)
.ToList();
if (matchingSigs.Any())
{
// need to add any byte ranges between oldEndOffset and offset as bytes remaining (ie not part of any sig).
// This is for catching any bytes BEFORE the sig match we've just found.
if (oldEndOffset != offset)
{
newRemainingBytes.Add(new RemainingBytes()
{
BeginOffset = oldEndOffset,
EndOffset = offset - 1
});
}
var matchingSig = matchingSigs[0];
// when storing which existing sig to use, make sure we know the offset in the NEW file it should appear.
matchingSig.Offset = offset;
signaturesToReuse.Add(matchingSig);
offset += windowSize;
generateFreshSig = true;
oldEndOffset = offset;
}
else
{
offset++;
generateFreshSig = false;
}
}
else
{
// no match. Just increment offset and generate rolling sig.
offset++;
generateFreshSig = false;
}
} while (offset + windowSize <= byteRange.EndOffset + 1);
// add remaining bytes to newRemainingBytes list
// Possible to have single byte at end with offset at very last byte.
if (offset <= byteRange.EndOffset)
{
newRemainingBytes.Add(new RemainingBytes()
{
BeginOffset = oldEndOffset,
EndOffset = byteRange.EndOffset
});
}
// if last sig
}
else
{
newRemainingBytes.Add(byteRange);
}
}
else
{
newRemainingBytes.Add(byteRange);
}
}
return newRemainingBytes;
}
internal static Dictionary<RollingSignature, List<BlockSignature>> GenerateBlockDict(CompleteSignature sig)
{
return GenerateBlockDict(sig.SignatureList);
}
/// <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;
}