Java equivalent to PHP's preg_replace_callback

Trying to emulate PHP's callback feature seems an awful lot of work when you could just use appendReplacement() and appendTail() in a loop:

StringBuffer resultString = new StringBuffer();
Pattern regex = Pattern.compile("regex");
Matcher regexMatcher = regex.matcher(subjectString);
while (regexMatcher.find()) {
  // You can vary the replacement text for each match on-the-fly
  regexMatcher.appendReplacement(resultString, "replacement");
}
regexMatcher.appendTail(resultString);

IMPORTANT: As pointed out by Kip in the comments, this class has an infinite loop bug if the matching regex matches on the replacement string. I'll leave it as an exercise to readers to fix it, if necessary.


I don't know of anything similar that's built into Java. You could roll your own without too much difficulty, using the Matcher class:

import java.util.regex.*;

public class CallbackMatcher
{
    public static interface Callback
    {
        public String foundMatch(MatchResult matchResult);
    }

    private final Pattern pattern;

    public CallbackMatcher(String regex)
    {
        this.pattern = Pattern.compile(regex);
    }

    public String replaceMatches(String string, Callback callback)
    {
        final Matcher matcher = this.pattern.matcher(string);
        while(matcher.find())
        {
            final MatchResult matchResult = matcher.toMatchResult();
            final String replacement = callback.foundMatch(matchResult);
            string = string.substring(0, matchResult.start()) +
                     replacement + string.substring(matchResult.end());
            matcher.reset(string);
        }
    }
}

Then call:

final CallbackMatcher.Callback callback = new CallbackMatcher.Callback() {
    public String foundMatch(MatchResult matchResult)
    {
        return "<img src=\"thumbs/" + matchResults.group(1) + "\"/>";
    }
};

final CallbackMatcher callbackMatcher = new CallbackMatcher("/\[thumb(\d+)\]/");
callbackMatcher.replaceMatches(articleText, callback);

Note that you can get the entire matched string by calling matchResults.group() or matchResults.group(0), so it's not necessary to pass the callback the current string state.

EDIT: Made it look more like the exact functionality of the PHP function.

Here's the original, since the asker liked it:

public class CallbackMatcher
{
    public static interface Callback
    {
        public void foundMatch(MatchResult matchResult);
    }

    private final Pattern pattern;

    public CallbackMatcher(String regex)
    {
        this.pattern = Pattern.compile(regex);
    }

    public String findMatches(String string, Callback callback)
    {
        final Matcher matcher = this.pattern.matcher(string);
        while(matcher.find())
        {
            callback.foundMatch(matcher.toMatchResult());
        }
    }
}

For this particular use case, it might be best to simply queue each match in the callback, then afterwards run through them backwards. This will prevent having to remap indexes as the string is modified.


I wasn't quite satisfied with any of the solutions here. I wanted a stateless solution. And I didn't want to end up in an infinite loop if my replacement string happened to match the pattern. While I was at it I added support for a limit parameter and a returned count parameter. (I used an AtomicInteger to simulate passing an integer by reference.) I moved the callback parameter to the end of the parameter list, to make it easier to define an anonymous class.

Here is an example of usage:

final Map<String,String> props = new HashMap<String,String>();
props.put("MY_NAME", "Kip");
props.put("DEPT", "R&D");
props.put("BOSS", "Dave");

String subjectString = "Hi my name is ${MY_NAME} and I work in ${DEPT} for ${BOSS}";
String sRegex = "\\$\\{([A-Za-z0-9_]+)\\}";

String replacement = ReplaceCallback.replace(sRegex, subjectString, new ReplaceCallback.Callback() {
  public String matchFound(MatchResult match) {
    String group1 = match.group(1);
    if(group1 != null && props.containsKey(group1))
      return props.get(group1);
    return match.group();
  }
});

System.out.println("replacement: " + replacement);

And here is my version of ReplaceCallback class:

import java.util.concurrent.atomic.AtomicInteger;
import java.util.regex.*;

public class ReplaceCallback
{
  public static interface Callback {
    /**
     * This function is called when a match is made. The string which was matched
     * can be obtained via match.group(), and the individual groupings via
     * match.group(n).
     */
    public String matchFound(MatchResult match);
  }

  /**
   * Replaces with callback, with no limit to the number of replacements.
   * Probably what you want most of the time.
   */
  public static String replace(String pattern, String subject, Callback callback)
  {
    return replace(pattern, subject, -1, null, callback);
  }

  public static String replace(String pattern, String subject, int limit, Callback callback)
  {
    return replace(pattern, subject, limit, null, callback);
  }

  /**
   * @param regex    The regular expression pattern to search on.
   * @param subject  The string to be replaced.
   * @param limit    The maximum number of replacements to make. A negative value
   *                 indicates replace all.
   * @param count    If this is not null, it will be set to the number of
   *                 replacements made.
   * @param callback Callback function
   */
  public static String replace(String regex, String subject, int limit,
          AtomicInteger count, Callback callback)
  {
    StringBuffer sb = new StringBuffer();
    Matcher matcher = Pattern.compile(regex).matcher(subject);
    int i;
    for(i = 0; (limit < 0 || i < limit) && matcher.find(); i++)
    {
      String replacement = callback.matchFound(matcher.toMatchResult());
      replacement = Matcher.quoteReplacement(replacement); //probably what you want...
      matcher.appendReplacement(sb, replacement);
    }
    matcher.appendTail(sb);

    if(count != null)
      count.set(i);
    return sb.toString();
  }
}