Core Java

Generic Text Comparison Tool with LCS Approach

Detecting and showing differences of two texts (especially having hundreds or thousands of lines) is a common problem. Using pure java.lang.String class methods may be a solution, but the most important issue for that kind of operations, “performance” will not be satisfactory. We want an efficient solution which may have a view as below:

Text Difference Tool Example

The problem contains two parts:

  • Detecting differences of two texts: For detecting differences, an efficient dynamic algorithm of LCS (Longest Common Subsequence) used in this solution. This solution has O(text1WordCount * text2WordCount) complexity and coded as “longestCommonSubsequence” method below.
  • Visualizing the differences: For visualizing, an HTML tag based approach is used, which marks new words of text2 with green color and old words of text1 with red color. This solution has O(changedWordsCount * (text1WordCount+text2WordCount)) complexity and coded as “markTextDifferences” method below.

Note1: For simplicity, “normalizeText” method is used for removing \n, \t and multiple space characters. Note2: This class was created as a Vaadin component. But “longestCommonSubsequence” is pure generic and “markTextDifferences” method is generic on HTML based visual components, so they can also be used with different frameworks.

import java.util.ArrayList;
import com.vaadin.ui.CustomComponent;
import com.vaadin.ui.Label;
import com.vaadin.ui.Layout;
import com.vaadin.ui.VerticalLayout;

/**
* Text comparison component which marks differences of two texts with colors.
* 
* @author cb
*/
public class TextCompareComponent extends CustomComponent {

    private Layout mainLayout = new VerticalLayout();
    private ArrayList<String> longestCommonSubsequenceList;
    private static final String INSERT_COLOR = "#99FFCC";
    private static final String DELETE_COLOR = "#CB6D6D";

    public TextCompareComponent(String text1, String text2) {

        text1 = normalizeText(text1);
        text2 = normalizeText(text2);

        this.longestCommonSubsequenceList = longestCommonSubsequence(text1, text2);
        String result = markTextDifferences(text1, text2, 
            longestCommonSubsequenceList, INSERT_COLOR, DELETE_COLOR);

        Label label = new Label(result, Label.CONTENT_XHTML);
        mainLayout.addComponent(label);
        setCompositionRoot(mainLayout);
    }

     /**
      * Finds a list of longest common subsequences (lcs) of given two texts.
      * 
      * @param text1
      * @param text2
      * @return - longest common subsequence list
      */
     private ArrayList<String> longestCommonSubsequence(String text1,String text2){

         String[] text1Words = text1.split(" ");
         String[] text2Words = text2.split(" ");

         int text1WordCount = text1Words.length;
         int text2WordCount = text2Words.length;

         int[][] solutionMatrix = new int[text1WordCount + 1][text2WordCount + 1];

         for (int i = text1WordCount - 1; i >=0; i--) {
             for (int j = text2WordCount - 1; j >= 0; j--) {
                 if (text1Words[i].equals(text2Words[j])){
                     solutionMatrix[i][j] = solutionMatrix[i + 1][j + 1] + 1;
                 }
                 else {
                     solutionMatrix[i][j] = Math.max(solutionMatrix[i + 1][j],
                         solutionMatrix[i][j + 1]);
                 }
              }
          }

          int i = 0, j = 0;
          ArrayList<String> lcsResultList =new ArrayList<String>();

          while (i < text1WordCount && j < text2WordCount) {
              if (text1Words[i].equals(text2Words[j])) {
                  lcsResultList.add(text2Words[j]);
                  i++;
                  j++;
              } 
              else if (solutionMatrix[i + 1][j] >= solutionMatrix[i][j + 1]) {
                  i++;
              }
              else {
                  j++;
              }
          }
          return lcsResultList;
      }

      /**
       * Normalizes given string by deleting \n, \t and extra spaces.
       * 
       * @param text - initial string
       * @return - normalized string
       */
       private String normalizeText(String text) {

           text = text.trim();
           text = text.replace("\n", " ");
           text = text.replace("\t", " ");

           while (text.contains("  ")) {
               text = text.replace("  ", " ");
           }
           return text;
       }

      /**
       * Returns colored inserted/deleted text representation of given two texts.
       * Uses longestCommonSubsequenceList to determine colored sections.
       *
       * @param text1
       * @param text2
       * @param lcsList
       * @param insertColor
       * @param deleteColor
       * @return - colored result text
       */
      private String markTextDifferences(String text1, String text2,
        ArrayList<String> lcsList, String insertColor, String deleteColor) {

        StringBuffer stringBuffer = new StringBuffer();

        if (text1 != null && lcsList != null) {

            String[] text1Words = text1.split(" ");
            String[] text2Words = text2.split(" ");
            int i = 0, j = 0, word1LastIndex = 0, word2LastIndex = 0;

            for (int k = 0; k < lcsList.size(); k++) {
                for (i = word1LastIndex, j = word2LastIndex;
                    i < text1Words.length && j < text2Words.length;) {
                    if (text1Words[i].equals(lcsList.get(k)) &&
                        text2Words[j].equals(lcsList.get(k))) {
                        stringBuffer.append("<SPAN>" + lcsList.get(k) + " </SPAN>");
                        word1LastIndex = i + 1;
                        word2LastIndex = j + 1;
                        i = text1Words.length;
                        j = text2Words.length;
                    }
                    else if (!text1Words[i].equals(lcsList.get(k))) {
                        for (; i < text1Words.length &&
                            !text1Words[i].equals(lcsList.get(k)); i++) {
                            stringBuffer.append("<SPAN style='BACKGROUND-COLOR:" +
                                deleteColor + "'>" + text1Words[i] + " </SPAN>");
                        }
                    } else if (!text2Words[j].equals(lcsList.get(k))) {
                        for (; j < text2Words.length &&
                            !text2Words[j].equals(lcsList.get(k)), j++) {
                            stringBuffer.append("<SPAN style='BACKGROUND-COLOR:" +
                                insertColor + "'>" + text2Words[j] + " </SPAN>");
                        }
                    }
                }
            }
            for (; word1LastIndex < text1Words.length; word1LastIndex++) {
                stringBuffer.append("<SPAN style='BACKGROUND-COLOR:" +
                    deleteColor + "'>" + text1Words[word1LastIndex] + " </SPAN>");
            }
            for (; word2LastIndex < text2Words.length; word2LastIndex++) {
                stringBuffer.append("<SPAN style='BACKGROUND-COLOR:" +
                    insertColor + "'>" + text2Words[word2LastIndex] + " </SPAN>");
            }
        }
        return stringBuffer.toString();
    }
}

Reference: A Generic Text Comparison Tool Implementation with LCS Approach from our JCG partner Cagdas Basaraner at the CodeBuild blog.

Cagdas Basaraner

Cagdas Basaraner is a software engineer graduated from Hacettepe University Computer Engineering department (Turkey), having 5 years professional experience. He is working on JEE web technologies, and also a former developer of information systems using Microsoft technologies and Command & Control (C4I) systems with Java technologies.
Subscribe
Notify of
guest

This site uses Akismet to reduce spam. Learn how your comment data is processed.

0 Comments
Inline Feedbacks
View all comments
Back to top button