public static byte[] Create(byte[] origin, byte[] target) {
var zDelta = new Writer();
int lenOut = target.Length;
int lenSrc = origin.Length;
int i, lastRead = -1;
zDelta.PutInt((uint) lenOut);
zDelta.PutChar('\n');
// If the source is very small, it means that we have no
// chance of ever doing a copy command. Just output a single
// literal segment for the entire target and exit.
if (lenSrc <= NHASH) {
zDelta.PutInt((uint) lenOut);
zDelta.PutChar(':');
zDelta.PutArray(target, 0, lenOut);
zDelta.PutInt(Checksum(target));
zDelta.PutChar(';');
return zDelta.ToArray();
}
// Compute the hash table used to locate matching sections in the source.
int nHash = (int) lenSrc / NHASH;
int[] collide = new int[nHash];
int[] landmark = new int[nHash];
for (i = 0; i < collide.Length; i++) collide[i] = -1;
for (i = 0; i < landmark.Length; i++) landmark[i] = -1;
int hv;
RollingHash h = new RollingHash();
for (i = lenSrc - lenSrc % NHASH - NHASH; i >= 0; i -= NHASH)
h.Init(origin, i);
hv = (int) (h.Value() % nHash);
collide[i/NHASH] = landmark[hv];
landmark[hv] = i/NHASH;
}
public static byte[] Create(byte[] origin, byte[] target)
{
var zDelta = new Writer();
int lenOut = target.Length;
int lenSrc = origin.Length;
int i, lastRead = -1;
zDelta.PutInt((uint)lenOut);
zDelta.PutChar('\n');
// If the source is very small, it means that we have no
// chance of ever doing a copy command. Just output a single
// literal segment for the entire target and exit.
if (lenSrc <= NHASH)
{
zDelta.PutInt((uint)lenOut);
zDelta.PutChar(':');
zDelta.PutArray(target, 0, lenOut);
zDelta.PutInt(Checksum(target));
zDelta.PutChar(';');
return(zDelta.ToArray());
}
// Compute the hash table used to locate matching sections in the source.
int nHash = (int)lenSrc / NHASH;
int[] collide = new int[nHash];
int[] landmark = new int[nHash];
for (i = 0; i < collide.Length; i++)
{
collide[i] = -1;
}
for (i = 0; i < landmark.Length; i++)
{
landmark[i] = -1;
}
int hv;
RollingHash h = new RollingHash();
for (i = 0; i < lenSrc - NHASH; i += NHASH)
{
h.Init(origin, i);
hv = (int)(h.Value() % nHash);
collide[i / NHASH] = landmark[hv];
landmark[hv] = i / NHASH;
}
int _base = 0;
int iSrc, iBlock;
int bestCnt, bestOfst = 0, bestLitsz = 0;
while (_base + NHASH < lenOut)
{
bestOfst = 0;
bestLitsz = 0;
h.Init(target, _base);
i = 0; // Trying to match a landmark against zOut[_base+i]
bestCnt = 0;
while (true)
{
int limit = 250;
hv = (int)(h.Value() % nHash);
iBlock = landmark[hv];
while (iBlock >= 0 && (limit--) > 0)
{
//
// The hash window has identified a potential match against
// landmark block iBlock. But we need to investigate further.
//
// Look for a region in zOut that matches zSrc. Anchor the search
// at zSrc[iSrc] and zOut[_base+i]. Do not include anything prior to
// zOut[_base] or after zOut[outLen] nor anything after zSrc[srcLen].
//
// Set cnt equal to the length of the match and set ofst so that
// zSrc[ofst] is the first element of the match. litsz is the number
// of characters between zOut[_base] and the beginning of the match.
// sz will be the overhead (in bytes) needed to encode the copy
// command. Only generate copy command if the overhead of the
// copy command is less than the amount of literal text to be copied.
//
int cnt, ofst, litsz;
int j, k, x, y;
int sz;
// Beginning at iSrc, match forwards as far as we can.
// j counts the number of characters that match.
iSrc = iBlock * NHASH;
for (j = 0, x = iSrc, y = _base + i; x < lenSrc && y < lenOut; j++, x++, y++)
{
if (origin[x] != target[y])
{
break;
}
}
j--;
// Beginning at iSrc-1, match backwards as far as we can.
// k counts the number of characters that match.
for (k = 1; k < iSrc && k <= i; k++)
{
if (origin[iSrc - k] != target[_base + i - k])
{
break;
}
}
k--;
// Compute the offset and size of the matching region.
ofst = iSrc - k;
cnt = j + k + 1;
litsz = i - k; // Number of bytes of literal text before the copy
// sz will hold the number of bytes needed to encode the "insert"
// command and the copy command, not counting the "insert" text.
sz = DigitCount(i - k) + DigitCount(cnt) + DigitCount(ofst) + 3;
if (cnt >= sz && cnt > bestCnt)
{
// Remember this match only if it is the best so far and it
// does not increase the file size.
bestCnt = cnt;
bestOfst = iSrc - k;
bestLitsz = litsz;
}
// Check the next matching block
iBlock = collide[iBlock];
}
// We have a copy command that does not cause the delta to be larger
// than a literal insert. So add the copy command to the delta.
if (bestCnt > 0)
{
if (bestLitsz > 0)
{
// Add an insert command before the copy.
zDelta.PutInt((uint)bestLitsz);
zDelta.PutChar(':');
zDelta.PutArray(target, _base, _base + bestLitsz);
_base += bestLitsz;
}
_base += bestCnt;
zDelta.PutInt((uint)bestCnt);
zDelta.PutChar('@');
zDelta.PutInt((uint)bestOfst);
zDelta.PutChar(',');
if (bestOfst + bestCnt - 1 > lastRead)
{
lastRead = bestOfst + bestCnt - 1;
}
bestCnt = 0;
break;
}
// If we reach this point, it means no match is found so far
if (_base + i + NHASH >= lenOut)
{
// We have reached the end and have not found any
// matches. Do an "insert" for everything that does not match
zDelta.PutInt((uint)(lenOut - _base));
zDelta.PutChar(':');
zDelta.PutArray(target, _base, _base + lenOut - _base);
_base = lenOut;
break;
}
// Advance the hash by one character. Keep looking for a match.
h.Next(target[_base + i + NHASH]);
i++;
}
}
// Output a final "insert" record to get all the text at the end of
// the file that does not match anything in the source.
if (_base < lenOut)
{
zDelta.PutInt((uint)(lenOut - _base));
zDelta.PutChar(':');
zDelta.PutArray(target, _base, _base + lenOut - _base);
}
// Output the final checksum record.
zDelta.PutInt(Checksum(target));
zDelta.PutChar(';');
return(zDelta.ToArray());
}