} // SMS
/// <summary>
/// This is the divide-and-conquer implementation of the longes common-subsequence (LCS)
/// algorithm.
/// The published algorithm passes recursively parts of the A and B sequences.
/// To avoid copying these arrays the lower and upper bounds are passed while the sequences stay constant.
/// </summary>
/// <param name="dataA">sequence A</param>
/// <param name="lowerA">lower bound of the actual range in DataA</param>
/// <param name="upperA">upper bound of the actual range in DataA (exclusive)</param>
/// <param name="dataB">sequence B</param>
/// <param name="lowerB">lower bound of the actual range in DataB</param>
/// <param name="upperB">upper bound of the actual range in DataB (exclusive)</param>
/// <param name="downVector">a vector for the (0,0) to (x,y) search. Passed as a parameter for speed reasons.</param>
/// <param name="upVector">a vector for the (u,v) to (N,M) search. Passed as a parameter for speed reasons.</param>
private static void Lcs<T>(DiffData<T> dataA, int lowerA, int upperA, DiffData<T> dataB, int lowerB, int upperB, int[] downVector, int[] upVector) where T : IComparable
{
// Debug.Write(2, "LCS", String.Format("Analyse the box: A[{0}-{1}] to B[{2}-{3}]", LowerA, UpperA, LowerB, UpperB));
// Fast walkthrough equal lines at the start
while (lowerA < upperA && lowerB < upperB && dataA.Data[lowerA].CompareTo(dataB.Data[lowerB]) == 0)
{
lowerA++; lowerB++;
}
// Fast walkthrough equal lines at the end
while (lowerA < upperA && lowerB < upperB && dataA.Data[upperA - 1].CompareTo(dataB.Data[upperB - 1]) == 0)
{
--upperA; --upperB;
}
if (lowerA == upperA)
{
// mark as inserted lines.
while (lowerB < upperB)
dataB.Modified[lowerB++] = true;
}
else if (lowerB == upperB)
{
// mark as deleted lines.
while (lowerA < upperA)
dataA.Modified[lowerA++] = true;
}
else
{
// Find the middle snakea and length of an optimal path for A and B
Smsrd smsrd = Sms(dataA, lowerA, upperA, dataB, lowerB, upperB, downVector, upVector);
// Debug.Write(2, "MiddleSnakeData", String.Format("{0},{1}", smsrd.x, smsrd.y));
// The path is from LowerX to (x,y) and (x,y) to UpperX
Lcs(dataA, lowerA, smsrd.X, dataB, lowerB, smsrd.Y, downVector, upVector);
Lcs(dataA, smsrd.X, upperA, dataB, smsrd.Y, upperB, downVector, upVector); // 2002.09.20: no need for 2 points
}
} // LCS()
/// <summary>
/// If a sequence of modified lines starts with a line that contains the same content
/// as the line that appends the changes, the difference sequence is modified so that the
/// appended line and not the starting line is marked as modified.
/// This leads to more readable diff sequences when comparing text files.
/// </summary>
/// <param name="data">A Diff data buffer containing the identified changes.</param>
private static void Optimize<T>(DiffData<T> data) where T : IComparable
{
int startPos, endPos;
startPos = 0;
while (startPos < data.Length)
{
while ((startPos < data.Length) && (!data.Modified[startPos]))
startPos++;
endPos = startPos;
while ((endPos < data.Length) && (data.Modified[endPos]))
endPos++;
if ((endPos < data.Length) && (data.Data[startPos].CompareTo(data.Data[endPos]) == 0))
{
data.Modified[startPos] = false;
data.Modified[endPos] = true;
}
else
{
startPos = endPos;
} // if
} // while
} // Optimize
} // Diff
/// <summary>
/// This is the algorithm to find the Shortest Middle Snake (SMS).
/// </summary>
/// <param name="dataA">sequence A</param>
/// <param name="lowerA">lower bound of the actual range in DataA</param>
/// <param name="upperA">upper bound of the actual range in DataA (exclusive)</param>
/// <param name="dataB">sequence B</param>
/// <param name="lowerB">lower bound of the actual range in DataB</param>
/// <param name="upperB">upper bound of the actual range in DataB (exclusive)</param>
/// <param name="downVector">a vector for the (0,0) to (x,y) search. Passed as a parameter for speed reasons.</param>
/// <param name="upVector">a vector for the (u,v) to (N,M) search. Passed as a parameter for speed reasons.</param>
/// <returns>a MiddleSnakeData record containing x,y and u,v</returns>
private static Smsrd Sms<T>(DiffData<T> dataA, int lowerA, int upperA, DiffData<T> dataB, int lowerB, int upperB, int[] downVector, int[] upVector) where T : IComparable
{
Smsrd ret;
int max = dataA.Length + dataB.Length + 1;
int downK = lowerA - lowerB; // the k-line to start the forward search
int upK = upperA - upperB; // the k-line to start the reverse search
int delta = (upperA - lowerA) - (upperB - lowerB);
bool oddDelta = (delta & 1) != 0;
// The vectors in the publication accepts negative indexes. the vectors implemented here are 0-based
// and are access using a specific offset: UpOffset UpVector and DownOffset for DownVektor
int downOffset = max - downK;
int upOffset = max - upK;
int maxD = ((upperA - lowerA + upperB - lowerB) / 2) + 1;
// Debug.Write(2, "SMS", String.Format("Search the box: A[{0}-{1}] to B[{2}-{3}]", LowerA, UpperA, LowerB, UpperB));
// init vectors
downVector[downOffset + downK + 1] = lowerA;
upVector[upOffset + upK - 1] = upperA;
for (int d = 0; d <= maxD; d++)
{
// Extend the forward path.
for (int k = downK - d; k <= downK + d; k += 2)
{
// Debug.Write(0, "SMS", "extend forward path " + k.ToString());
// find the only or better starting point
int x, y;
if (k == downK - d)
{
x = downVector[downOffset + k + 1]; // down
}
else
{
x = downVector[downOffset + k - 1] + 1; // a step to the right
if ((k < downK + d) && (downVector[downOffset + k + 1] >= x))
x = downVector[downOffset + k + 1]; // down
}
y = x - k;
// find the end of the furthest reaching forward D-path in diagonal k.
while ((x < upperA) && (y < upperB) && (dataA.Data[x].CompareTo(dataB.Data[y]) == 0))
{
x++; y++;
}
downVector[downOffset + k] = x;
// overlap ?
if (oddDelta && (upK - d < k) && (k < upK + d))
{
if (upVector[upOffset + k] <= downVector[downOffset + k])
{
ret.X = downVector[downOffset + k];
ret.Y = downVector[downOffset + k] - k;
// ret.u = UpVector[UpOffset + k]; // 2002.09.20: no need for 2 points
// ret.v = UpVector[UpOffset + k] - k;
return (ret);
} // if
} // if
} // for k
// Extend the reverse path.
for (int k = upK - d; k <= upK + d; k += 2)
{
// Debug.Write(0, "SMS", "extend reverse path " + k.ToString());
// find the only or better starting point
int x, y;
if (k == upK + d)
{
x = upVector[upOffset + k - 1]; // up
}
else
{
x = upVector[upOffset + k + 1] - 1; // left
if ((k > upK - d) && (upVector[upOffset + k - 1] < x))
x = upVector[upOffset + k - 1]; // up
} // if
y = x - k;
while ((x > lowerA) && (y > lowerB) && (dataA.Data[x - 1].CompareTo(dataB.Data[y - 1]) == 0))
{
x--; y--; // diagonal
}
upVector[upOffset + k] = x;
// overlap ?
if (!oddDelta && (downK - d <= k) && (k <= downK + d))
{
if (upVector[upOffset + k] <= downVector[downOffset + k])
{
ret.X = downVector[downOffset + k];
ret.Y = downVector[downOffset + k] - k;
// ret.u = UpVector[UpOffset + k]; // 2002.09.20: no need for 2 points
// ret.v = UpVector[UpOffset + k] - k;
return (ret);
} // if
} // if
} // for k
} // for D
throw new ApplicationException("the algorithm should never come here.");
} // SMS