/**
 * A low memory footprint [email protected] GSet} implementation,
 * which uses an array for storing the elements
 * and linked lists for collision resolution.
 *
 * No rehash will be performed.
 * Therefore, the internal array will never be resized.
 *
 * This class does not support null element.
 *
 * This class is not thread safe.
 *
 * @param <K> Key type for looking up the elements
 * @param <E> Element type, which must be
 *       (1) a subclass of K, and
 *       (2) implementing [email protected]

裏面各組件都很好理解，唯一不好理解的是Iterator,

public class SetIterator implements Iterator<E> {
    /** The starting modification for fail-fast. */
    private int iterModification = modification;
    /** The current index of the entry array. */
    private int index = -1;
    private LinkedElement cur = null;//
    private LinkedElement next = nextNonemptyEntry();//next總是指向下一個元素，在初始化時就完畢指下第一個元素。
 

                                    //在調用next()方法之後,next置為空。直到調用 ensureNext()方法。
 private boolean trackModification = true;

    /** Find the next nonempty entry starting at (index + 1). */
    private LinkedElement nextNonemptyEntry() {
      for(index++; index < entries.length && entries[index] == null; index++);
      return index < entries.length? entries[index]: null;
    }

    private void ensureNext() {
      if (trackModification && modification != iterModification) {
        throw new ConcurrentModificationException("modification=" + modification
            + " != iterModification = " + iterModification);
      }
      if (next != null) {
        return;
      }
      if (cur == null) {
        return;
      }
      next = cur.getNext();
      if (next == null) {
        next = nextNonemptyEntry();
      }
    }

    @Override
    public boolean hasNext() {
      ensureNext();
      return next != null;
    }

    @Override
    public E next() {
      ensureNext();
      if (next == null) {
        throw new IllegalStateException("There are no more elements");
      }
      cur = next;
      next = null;
      return convert(cur);
    }

    @SuppressWarnings("unchecked")
    @Override
    public void remove() {
      ensureNext();
      if (cur == null) {
        throw new IllegalStateException("There is no current element " +
            "to remove");
      }
      LightWeightGSet.this.remove((K)cur);
      iterModification++;
      cur = null;
    }

    public void setTrackModification(boolean trackModification) {
      this.trackModification = trackModification;
    }
  }</span>
 

computeCapacity()是一個工具方法，用於一定比例的內存的容器，能夠存儲多少對象。

參數，第一個是占最大內存的百分比，第二個是名稱。沒有什麽用，僅僅用作日誌輸出。

/**
   * Let t = percentage of max memory.
   * Let e = round(log_2 t).
   * Then, we choose capacity = 2^e/(size of reference),
   * unless it is outside the close interval [1, 2^30].
   */
  public static int computeCapacity(double percentage, String mapName) {
    return computeCapacity(Runtime.getRuntime().maxMemory(), percentage,
        mapName);
  }
  
  @VisibleForTesting
  static int computeCapacity(long maxMemory, double percentage,
      String mapName) {
    if (percentage > 100.0 || percentage < 0.0) {
      throw new HadoopIllegalArgumentException("Percentage " + percentage
          + " must be greater than or equal to 0 "
          + " and less than or equal to 100");
    }
    if (maxMemory < 0) {
      throw new HadoopIllegalArgumentException("Memory " + maxMemory
          + " must be greater than or equal to 0");
    }
    if (percentage == 0.0 || maxMemory == 0) {
      return 0;
    }
    //VM detection
    //See http://java.sun.com/docs/hotspot/HotSpotFAQ.html#64bit_detection
    final String vmBit = System.getProperty("sun.arch.data.model");

    //Percentage of max memory
    final double percentDivisor = 100.0/percentage;
    final double percentMemory = maxMemory/percentDivisor;
    
    //compute capacity
  /*
  具體描寫敘述例如以下：e1應該是以2為base的對數。如percentMemory為1024,結果為10。由於Math類不提供以2為base的對數，
所以採用了間接的方法，先求自然對數，再除以2的自然對數。例System.out.println(Math.log(1024)/Math.log(2));結果為10。
＋0.5是為了四舍五入。
假設占用內存為1G,則e1為30.
e2的值，假設系統為32位，則減2,由於2，e2為28,c為2的28次方，為256M個對象，每一個對象指針在32位系統中占4字節。總共1G.
假設系統為64位，e2為27，即對象個數為128M，每一個對象指針為8字節。所以共占1G.
*/ 
 final int e1 = (int)(Math.log(percentMemory)/Math.log(2.0) + 0.5);
    final int e2 = e1 - ("32".equals(vmBit)?
 2: 3);
    final int exponent = e2 < 0? 0: e2 > 30? 30: e2;
    final int c = 1 << exponent;

    LOG.info("Computing capacity for map " + mapName);
    LOG.info("VM type       = " + vmBit + "-bit");
    LOG.info(percentage + "% max memory "
        + StringUtils.TraditionalBinaryPrefix.long2String(maxMemory, "B", 1)
        + " = "
        + StringUtils.TraditionalBinaryPrefix.long2String((long) percentMemory,
            "B", 1));
    LOG.info("capacity      = 2^" + exponent + " = " + c + " entries");
    return c;
  }

hadoop 2.6.0 LightWeightGSet源碼分析