Since the Log4J API is still Enumeration-based, there are some ugly old-school loops in there. All types are from the java.util or org.apache.log4j packages.

@SuppressWarnings("unchecked")
private static Map<String,String> getLogLocations() {

  Collection<Logger> allLoggers = new ArrayList<Logger>();

  Logger rootLogger = Logger.getRootLogger();

  allLoggers.add(rootLogger);
  for (Enumeration<Logger> loggers =
          rootLogger.getLoggerRepository().getCurrentLoggers() ;
          loggers.hasMoreElements() ; ) {

    allLoggers.add(loggers.nextElement());
  }

  Set<FileAppender> fileAppenders =
        new LinkedHashSet<FileAppender>();

  for (Logger logger : allLoggers) {
    for (Enumeration<Appender> appenders =
            logger.getAllAppenders() ;
            appenders.hasMoreElements() ; ) {

      Appender appender = appenders.nextElement();
      if (appender instanceof FileAppender) {

        fileAppenders.add((FileAppender) appender);
      }
    }
  }

  Map<String, String> locations =
        new LinkedHashMap<String,String>();

  for (FileAppender appender : fileAppenders) {
    locations.put(appender.getName(), appender.getFile());
  }

  return locations;
}

The code just grabs the root logger, finds all the other loggers via its LoggerRepository, and then finds the set of all FileAppenders attached to them. It returns a Map of appender name to logfile location (both Strings), in the order they are declared in your configuration.

Not really the world’s most elegant code. Despite the clunky Log4J API, I can’t help thinking that would take no more than two or three lines in Python or Ruby.