} // 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(DiffData dataA, int lowerA, int upperA, DiffData dataB, int lowerB, int upperB, int[] downVector, int[] upVector) { // 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] == dataB.Data[lowerB]) { lowerA++; lowerB++; } // Fast walkthrough equal lines at the end while (lowerA < upperA && lowerB < upperB && dataA.Data[upperA - 1] == dataB.Data[upperB - 1]) { --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 var 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()
} // DiffText /// <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(DiffData Data) { int StartPos, EndPos; StartPos = 0; while (StartPos < Data.Length) { while ((StartPos < Data.Length) && (Data.modified[StartPos] == false)) StartPos++; EndPos = StartPos; while ((EndPos < Data.Length) && (Data.modified[EndPos] == true)) EndPos++; if ((EndPos < Data.Length) && (Data.data[StartPos] == Data.data[EndPos])) { Data.modified[StartPos] = false; Data.modified[EndPos] = true; } else { StartPos = EndPos; } // if } // while } // Optimize
} // 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(DiffData DataA, int LowerA, int UpperA, DiffData DataB, int LowerB, int UpperB, int[] DownVector, int[] UpVector) { // 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] == DataB.data[LowerB]) { LowerA++; LowerB++; } // Fast walkthrough equal lines at the end while (LowerA < UpperA && LowerB < UpperB && DataA.data[UpperA - 1] == DataB.data[UpperB - 1]) { --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()
} // DiffText /// <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(DiffData data) { var startPos = 0; while (startPos < data.Length) { while (startPos < data.Length && !data.Modified[startPos]) startPos++; var endPos = startPos; while (endPos < data.Length && data.Modified[endPos]) endPos++; if (endPos < data.Length && data.Data[startPos] == data.Data[endPos]) { data.Modified[startPos] = false; data.Modified[endPos] = true; } else { startPos = endPos; } // if } // while } // Optimize
/// <summary> /// Find the difference in 2 text documents, comparing by textlines. /// The algorithm itself is comparing 2 arrays of numbers so when comparing 2 text documents /// each line is converted into a (hash) number. This hash-value is computed by storing all /// textlines into a common hashtable so i can find dublicates in there, and generating a /// new number each time a new textline is inserted. /// </summary> /// <param name="textA">A-version of the text (usualy the old one)</param> /// <param name="textB">B-version of the text (usualy the new one)</param> /// <param name="trimSpace">When set to true, all leading and trailing whitespace characters are stripped out before the comparation is done.</param> /// <param name="ignoreSpace">When set to true, all whitespace characters are converted to a single space character before the comparation is done.</param> /// <param name="ignoreCase">When set to true, all characters are converted to their lowercase equivivalence before the comparation is done.</param> /// <returns>Returns a array of Items that describe the differences.</returns> public static Item[] DiffText(string[] textA, string[] textB, bool trimSpace, bool ignoreSpace, bool ignoreCase) { // prepare the input-text and convert to comparable numbers. var h = new Hashtable(textA.Length + textB.Length); // The A-Version of the data (original data) to be compared. var dataA = new DiffData(DiffCodes(textA, h, trimSpace, ignoreSpace, ignoreCase)); // The B-Version of the data (modified data) to be compared. var dataB = new DiffData(DiffCodes(textB, h, trimSpace, ignoreSpace, ignoreCase)); var max = dataA.Length + dataB.Length + 1; // vector for the (0,0) to (x,y) search var downVector = new int[2 * max + 2]; // vector for the (u,v) to (N,M) search var upVector = new int[2 * max + 2]; Lcs(dataA, 0, dataA.Length, dataB, 0, dataB.Length, downVector, upVector); Optimize(dataA); Optimize(dataB); return CreateDiffs(dataA, dataB); } // DiffText
/// <summary> /// Find the difference in 2 text documents, comparing by textlines. /// The algorithm itself is comparing 2 arrays of numbers so when comparing 2 text documents /// each line is converted into a (hash) number. This hash-value is computed by storing all /// textlines into a common hashtable so i can find dublicates in there, and generating a /// new number each time a new textline is inserted. /// </summary> /// <param name="TextA">A-version of the text (usualy the old one)</param> /// <param name="TextB">B-version of the text (usualy the new one)</param> /// <param name="trimSpace">When set to true, all leading and trailing whitespace characters are stripped out before the comparation is done.</param> /// <param name="ignoreSpace">When set to true, all whitespace characters are converted to a single space character before the comparation is done.</param> /// <param name="ignoreCase">When set to true, all characters are converted to their lowercase equivivalence before the comparation is done.</param> /// <returns>Returns a array of Items that describe the differences.</returns> public static Item[] DiffText(string[] TextA, string[] TextB, bool trimSpace, bool ignoreSpace, bool ignoreCase) { // prepare the input-text and convert to comparable numbers. Hashtable h = new Hashtable(TextA.Length + TextB.Length); // The A-Version of the data (original data) to be compared. DiffData DataA = new DiffData(DiffCodes(TextA, h, trimSpace, ignoreSpace, ignoreCase)); // The B-Version of the data (modified data) to be compared. DiffData DataB = new DiffData(DiffCodes(TextB, h, trimSpace, ignoreSpace, ignoreCase)); h = null; // free up hashtable memory (maybe) int MAX = DataA.Length + DataB.Length + 1; /// vector for the (0,0) to (x,y) search int[] DownVector = new int[2 * MAX + 2]; /// vector for the (u,v) to (N,M) search int[] UpVector = new int[2 * MAX + 2]; LCS(DataA, 0, DataA.Length, DataB, 0, DataB.Length, DownVector, UpVector); Optimize(DataA); Optimize(DataB); return CreateDiffs(DataA, DataB); } // DiffText
} // DiffCodes /// <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(DiffData dataA, int lowerA, int upperA, DiffData dataB, int lowerB, int upperB, int[] downVector, int[] upVector) { if (downVector == null) throw new ArgumentNullException(nameof(downVector)); if (upVector == null) throw new ArgumentNullException(nameof(upVector)); var max = dataA.Length + dataB.Length + 1; var downK = lowerA - lowerB; // the k-line to start the forward search var upK = upperA - upperB; // the k-line to start the reverse search var delta = upperA - lowerA - (upperB - lowerB); var 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 var downOffset = max - downK; var upOffset = max - upK; var 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 (var d = 0; d <= maxD; d++) { // Extend the forward path. Smsrd ret; for (var 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; 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 } var y = x - k; // find the end of the furthest reaching forward D-path in diagonal k. while (x < upperA && y < upperB && dataA.Data[x] == dataB.Data[y]) { 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 (var 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; 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 var y = x - k; while (x > lowerA && y > lowerB && dataA.Data[x - 1] == dataB.Data[y - 1]) { 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
} // LCS() /// <summary> /// Scan the tables of which lines are inserted and deleted, /// producing an edit script in forward order. /// </summary> private static Item[] CreateDiffs(DiffData DataA, DiffData DataB) { ArrayList a = new ArrayList(); Item aItem; Item[] result; int StartA, StartB; int LineA, LineB; LineA = 0; LineB = 0; while (LineA < DataA.Length || LineB < DataB.Length) { if ((LineA < DataA.Length) && (!DataA.modified[LineA]) && (LineB < DataB.Length) && (!DataB.modified[LineB])) { // equal lines LineA++; LineB++; } else { // maybe deleted and/or inserted lines StartA = LineA; StartB = LineB; while (LineA < DataA.Length && (LineB >= DataB.Length || DataA.modified[LineA])) // while (LineA < DataA.Length && DataA.modified[LineA]) LineA++; while (LineB < DataB.Length && (LineA >= DataA.Length || DataB.modified[LineB])) // while (LineB < DataB.Length && DataB.modified[LineB]) LineB++; if ((StartA < LineA) || (StartB < LineB)) { // store a new difference-item aItem = new Item(); aItem.StartA = StartA; aItem.StartB = StartB; aItem.deletedA = LineA - StartA; aItem.insertedB = LineB - StartB; a.Add(aItem); } // if } // if } // while result = new Item[a.Count]; a.CopyTo(result); return (result); }
} // DiffCodes /// <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(DiffData DataA, int LowerA, int UpperA, DiffData DataB, int LowerB, int UpperB, int[] DownVector, int[] UpVector) { 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] == DataB.data[y])) { 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] == DataB.data[y - 1])) { 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
} // Optimize /// <summary> /// Find the difference in 2 arrays of integers. /// </summary> /// <param name="ArrayA">A-version of the numbers (usualy the old one)</param> /// <param name="ArrayB">B-version of the numbers (usualy the new one)</param> /// <returns>Returns a array of Items that describe the differences.</returns> public static Item[] DiffInt(int[] ArrayA, int[] ArrayB) { // The A-Version of the data (original data) to be compared. DiffData DataA = new DiffData(ArrayA); // The B-Version of the data (modified data) to be compared. DiffData DataB = new DiffData(ArrayB); int MAX = DataA.Length + DataB.Length + 1; /// vector for the (0,0) to (x,y) search int[] DownVector = new int[2 * MAX + 2]; /// vector for the (u,v) to (N,M) search int[] UpVector = new int[2 * MAX + 2]; LCS(DataA, 0, DataA.Length, DataB, 0, DataB.Length, DownVector, UpVector); return CreateDiffs(DataA, DataB); } // Diff