在

In How do I change color of a particular word document using apache poi? I have shown an algorithm to split XWPFRuns for formatting reasons. This is only for formatting one character and it does not clone the run properties. But the basic is shown. We have to look at the entire paragraph since only there are methods for inserting runs. And we need looping over the run texts character wise since all methods for split into words will lead to problems with punctuation marks while reassembling the words to a paragraph then.

缺少一种将运行属性从原始运行克隆到新添加的运行属性的方法.这可以通过克隆基础 w:rPr 元素来完成.

What lacks is a method for cloning the run properties from original run to the new added ones. This could be done by cloning the underlying w:rPr element.

然后整个方法是遍历所有段落.如果其中包含带有关键字的运行，则将运行文本拆分为字符.然后遍历该运行中的所有字符并缓冲它们.如果缓冲的字符流以关键字结尾，则将除关键字之外的所有当前缓冲字符设置为实际运行的文本.然后为formatted关键字插入新的运行，并从原始运行中克隆运行属性.将关键字设置为运行并进行其他格式化.然后为下一个字符插入新的运行，并从原始运行中克隆运行属性.在该段中的每次运行中都如此.

Then the whole approach is to go through all runs in paragraph. If we have a run with keyword in it, then split run text into characters. Then go through all characters in that run and buffer them. If the buffered character stream ends with the keyword, then set all chars, which are current buffered, except the keyword, as the text of the actual run. Then insert new run for the formatted keyword and clone the run properties from original run. Set the keyword into the run and do the additional formatting. Then insert a new run for the next characters and also clone the run properties from original run. So on for each run in the paragraph.

完整示例:

import java.io.*;
import org.apache.poi.xwpf.usermodel.*;
import org.openxmlformats.schemas.wordprocessingml.x2006.main.*;

import org.apache.xmlbeans.XmlObject;
import org.apache.xmlbeans.XmlCursor;

import java.util.*;
import java.awt.Desktop;

public class WordFormatWords {

 static void cloneRunProperties(XWPFRun source, XWPFRun dest) { // clones the underlying w:rPr element
  CTR tRSource = source.getCTR();
  CTRPr rPrSource = tRSource.getRPr();
  if (rPrSource != null) {
   CTRPr rPrDest = (CTRPr)rPrSource.copy();
   CTR tRDest = dest.getCTR();
   tRDest.setRPr(rPrDest);
  }
 }

 static void formatWord(XWPFParagraph paragraph, String keyword, Map<String, String> formats) {
  int runNumber = 0;
  while (runNumber < paragraph.getRuns().size()) { //go through all runs, we cannot use for each since we will possibly insert new runs
   XWPFRun run = paragraph.getRuns().get(runNumber);
   XWPFRun run2 = run;
   String runText = run.getText(0);
   if (runText != null && runText.contains(keyword)) { //if we have a run with keyword in it, then

    // This code part is to manage comment ranges.
    // Do we have commentRangeEnd immediately after the run?
    // If so then remember that in a cursor.
    XmlCursor commentRangeEndCursor = null; 
    XmlCursor cursor = run.getCTR().newCursor();
    cursor.toEndToken();
    if (cursor.hasNextToken()) {
     cursor.toNextToken();
     XmlObject commentRangeEnd = cursor.getObject();
     if (commentRangeEnd != null && commentRangeEnd instanceof CTMarkupRange) {
      commentRangeEndCursor = cursor;
     }
    }

    char[] runChars = runText.toCharArray(); //split run text into characters
    StringBuffer sb = new StringBuffer();
    for (int charNumber = 0; charNumber < runChars.length; charNumber++) { //go through all characters in that run
     sb.append(runChars[charNumber]); //buffer all characters
     runText = sb.toString();
     if (runText.endsWith(keyword)) { //if the bufferend character stream ends with the keyword  
      //set all chars, which are current buffered, except the keyword, as the text of the actual run
      run.setText(runText.substring(0, runText.length() - keyword.length()), 0); 
      run2 = paragraph.insertNewRun(++runNumber); //insert new run for the formatted keyword
      cloneRunProperties(run, run2); // clone the run properties from original run
      run2.setText(keyword, 0); // set the keyword in run
      for (String toSet : formats.keySet()) { // do the additional formatting
       if ("color".equals(toSet)) {
        run2.setColor(formats.get(toSet));
       } else if ("bold".equals(toSet)) {
        run2.setBold(Boolean.valueOf(formats.get(toSet)));
       }
      }
      run2 = paragraph.insertNewRun(++runNumber); //insert a new run for the next characters
      cloneRunProperties(run, run2); // clone the run properties from original run
      run = run2;
      sb = new StringBuffer(); //empty the buffer
     } 
    }
    run.setText(sb.toString(), 0); //set all characters, which are currently buffered, as the text of the actual run

    // This code part is to manage comment ranges.
    // If we had remembered commentRangeEnd, then move this to here now.
    if(commentRangeEndCursor != null) {
     cursor = run.getCTR().newCursor();
     cursor.toEndToken();
     if (cursor.hasNextToken()) {
      cursor.toNextToken();
      commentRangeEndCursor.moveXml(cursor);
     }
     cursor.dispose();
     commentRangeEndCursor.dispose();
    }

   }
   runNumber++;
  }
 }


 public static void main(String[] args) throws Exception {

  XWPFDocument doc = new XWPFDocument(new FileInputStream("source.docx"));

  String[] keywords = new String[]{"fox", "dog"};
  Map<String, String> formats = new HashMap<String, String>();
  formats.put("bold", "true");
  formats.put("color", "DC143C");

  for (XWPFParagraph paragraph : doc.getParagraphs()) { //go through all paragraphs
   for (String keyword : keywords) {
    formatWord(paragraph, keyword, formats);
   }
  }

  FileOutputStream out = new FileOutputStream("result.docx");
  doc.write(out);
  out.close();
  doc.close();

  System.out.println("Done");
  Desktop.getDesktop().open(new File("result.docx"));

 }
}

此代码还考虑了 XML 标记范围元素，例如 commentRangeEnd ，它们位于运行的 r 元素之后.这样的标记范围元素用于标记其他元素组的开始和结束.例如，在 commentRangeStart 和 commentRangeEnd 之间具有相同的 id 的一组文本运行元素.

This code also takes care about XML markup range elements such as commentRangeEnd which are immediately after the run's r element. Such markup range elements are used to mark start and end of groups of other elements. For example a group of text run elements to those a comment is applied is between commentRangeStart and commentRangeEnd having same id.

如果需要分割的运行紧跟在 commentRangeEnd 之后，那么我们会在游标中记住这一点.然后，在分割运行之后，将此 commentRangeEnd 移至最后一个新插入的运行之后.因此，评论应保持正确.

If immediately after the run which needs to be split follows a commentRangeEnd, then we remember that in a cursor. Then after splitting the run we move this commentRangeEnd immediately behind the last new inserted run. So comments should stay correct.

当然，即使这样也会有一些缺点，因为有时 Microsoft Word 如何在文本运行中存储文本很笨拙.当以 Microsoft Word 为源时，没有唯一且通用的解决方案.

Of course even this will have some disadvantages because of the clumsy kind on how Microsoft Word stores text in text runs sometimes. There is not the one and only general solution for this when Microsoft Word is the source.