path: root/src/library/scala/collection/mutable/HashTable.scala

/*                     __                                               *\
**     ________ ___   / /  ___     Scala API                            **
**    / __/ __// _ | / /  / _ |    (c) 2003-2010, LAMP/EPFL             **
**  __\ \/ /__/ __ |/ /__/ __ |    http://www.scala-lang.org/           **
** /____/\___/_/ |_/____/_/ | |                                         **
**                          |/                                          **
\*                                                                      */



package scala.collection
package mutable

/** This class can be used to construct data structures that are based
 *  on hashtables. Class `HashTable[A]` implements a hashtable
 *  that maps keys of type `A` to values of the fully abstract
 *  member type `Entry`. Classes that make use of `HashTable`
 *  have to provide an implementation for `Entry`.
 *
 *  There are mainly two parameters that affect the performance of a hashtable:
 *  the <i>initial size</i> and the <i>load factor</i>. The <i>size</i>
 *  refers to the number of <i>buckets</i> in the hashtable, and the <i>load
 *  factor</i> is a measure of how full the hashtable is allowed to get before
 *  its size is automatically doubled. Both parameters may be changed by
 *  overriding the corresponding values in class `HashTable`.
 *
 *  @author  Matthias Zenger
 *  @author  Martin Odersky
 *  @version 2.0, 31/12/2006
 *  @since   1
 *
 *  @tparam A     type of the elements contained in this hash table.
 */
trait HashTable[A, Entry >: Null <: HashEntry[A, Entry]] {
  import HashTable._

  @transient protected var _loadFactor = defaultLoadFactor

  /** The actual hash table.
   */
  @transient protected var table: Array[HashEntry[A, Entry]] = new Array(initialCapacity)

  /** The number of mappings contained in this hash table.
   */
  @transient protected var tableSize: Int = 0

  /** The next size value at which to resize (capacity * load factor).
   */
  @transient protected var threshold: Int = initialThreshold(_loadFactor)

  protected def initialSize: Int = HashTable.initialSize

  /**
   * Initializes the collection from the input stream. `f` will be called for each key/value pair
   * read from the input stream in the order determined by the stream. This is useful for
   * structures where iteration order is important (e.g. LinkedHashMap).
   */
  private[collection] def init[B](in: java.io.ObjectInputStream, f: (A, B) => Entry) {
    in.defaultReadObject

    _loadFactor = in.readInt
    assert(_loadFactor > 0)

    val size = in.readInt
    assert(size >= 0)

    val smDefined = in.readBoolean

    table = new Array(capacity(sizeForThreshold(_loadFactor, size)))
    threshold = newThreshold(_loadFactor, table.size)

    if (smDefined) sizeMapInit(table.size)

    var index = 0
    while (index < size) {
      addEntry(f(in.readObject.asInstanceOf[A], in.readObject.asInstanceOf[B]))
      index += 1
    }
  }

  /**
   * Serializes the collection to the output stream by saving the load factor, collection
   * size, collection keys and collection values. `value` is responsible for providing a value
   * from an entry.
   *
   * `foreach` determines the order in which the key/value pairs are saved to the stream. To
   * deserialize, `init` should be used.
   */
  private[collection] def serializeTo[B](out: java.io.ObjectOutputStream, value: Entry => B) {
    out.defaultWriteObject
    out.writeInt(_loadFactor)
    out.writeInt(tableSize)
    out.writeBoolean(isSizeMapDefined)
    foreachEntry { entry =>
      out.writeObject(entry.key)
      out.writeObject(value(entry))
    }
  }

  /** Find entry with given key in table, null if not found.
   */
  protected def findEntry(key: A): Entry = {
    val h = index(elemHashCode(key))
    var e = table(h).asInstanceOf[Entry]
    while (e != null && !elemEquals(e.key, key)) e = e.next
    e
  }

  /** Add entry to table
   *  pre: no entry with same key exists
   */
  protected def addEntry(e: Entry) {
    val h = index(elemHashCode(e.key))
    e.next = table(h).asInstanceOf[Entry]
    table(h) = e
    tableSize = tableSize + 1
    nnSizeMapAdd(h)
    if (tableSize > threshold)
      resize(2 * table.length)
  }

  /** Remove entry from table if present.
   */
  protected def removeEntry(key: A) : Entry = {
    val h = index(elemHashCode(key))
    var e = table(h).asInstanceOf[Entry]
    if (e != null) {
      if (elemEquals(e.key, key)) {
        table(h) = e.next
        tableSize = tableSize - 1
        nnSizeMapRemove(h)
        return e
      } else {
        var e1 = e.next
        while (e1 != null && !elemEquals(e1.key, key)) {
          e = e1
          e1 = e1.next
        }
        if (e1 != null) {
          e.next = e1.next
          tableSize = tableSize - 1
          nnSizeMapRemove(h)
          return e1
        }
      }
    }
    null
  }

  /** An iterator returning all entries.
   */
  protected def entriesIterator: Iterator[Entry] = new Iterator[Entry] {
    val iterTable = table
    var idx = table.length - 1
    var es = iterTable(idx).asInstanceOf[Entry]
    scan()
    def hasNext = es != null
    def next = {
      val res = es
      es = es.next
      scan()
      res
    }
    def scan() {
      while (es == null && idx > 0) {
        idx = idx - 1
        es = iterTable(idx).asInstanceOf[Entry]
      }
    }
  }

  /*
   * We should implement this as a primitive operation over the underlying array, but it can
   * cause a behaviour change in edge cases where:
   * - Someone modifies a map during iteration
   * - The insertion point is close to the iteration point.
   *
   * The reason this happens is that the iterator prefetches the following element before
   * returning from next (to simplify the implementation of hasNext) while the natural
   * implementation of foreach does not.
   *
   * It should be mentioned that modifying a map during iteration leads to unpredictable
   * results with either implementation.
   */
  protected final def foreachEntry[C](f: Entry => C) { entriesIterator.foreach(f) }

  /** An iterator returning all entries */
  @deprecated("use entriesIterator instead")
  protected def entries: Iterator[Entry] = entriesIterator

  /** Remove all entries from table
   */
  protected def clearTable() {
    var i = table.length - 1
    while (i >= 0) { table(i) = null; i = i - 1 }
    tableSize = 0
    nnSizeMapReset(0)
  }

  private def resize(newSize: Int) {
    val oldTable = table
    table = new Array(newSize)
    nnSizeMapReset(table.length)
    var i = oldTable.length - 1
    while (i >= 0) {
      var e = oldTable(i)
      while (e != null) {
        val h = index(elemHashCode(e.key))
        val e1 = e.next
        e.next = table(h).asInstanceOf[Entry]
        table(h) = e
        e = e1
        nnSizeMapAdd(h)
      }
      i = i - 1
    }
    threshold = newThreshold(_loadFactor, newSize)
  }

  @transient protected var sizemap: Array[Int] = null
  protected final def sizeMapBucketBitSize = 5
  // so that:
  protected final def sizeMapBucketSize = 1 << sizeMapBucketBitSize
  protected final def totalSizeMapBuckets = if (sizeMapBucketSize < table.length) 1 else table.length / sizeMapBucketSize

  /*
   * The following three sizeMap* functions (Add, Remove, Reset)
   * are used to update the size map of the hash table.
   *
   * The size map logically divides the hash table into `sizeMapBucketSize` element buckets
   * by keeping an integer entry for each such bucket. Each integer entry simply denotes
   * the number of elements in the corresponding bucket.
   * Best understood through an example, see:
   * table   = [/, 1, /, 6, 90, /, -3, 5]    (8 entries)
   * sizemap = [     2     |     3      ]    (2 entries)
   * where sizeMapBucketSize == 4.
   *
   * By default the size map is not initialized, so these methods don't do anything, thus,
   * their impact on hash table performance is negligible. However, if the hash table
   * is converted into a parallel hash table, the size map is initialized, as it will be needed
   * there.
   */
  protected def nnSizeMapAdd(h: Int) = if (sizemap ne null) {
    sizemap(h >> sizeMapBucketBitSize) += 1
  }

  protected def nnSizeMapRemove(h: Int) = if (sizemap ne null) {
    sizemap(h >> sizeMapBucketBitSize) -= 1
  }

  protected def nnSizeMapReset(tableLength: Int) = if (sizemap ne null) {
    val nsize = calcSizeMapSize(tableLength)
    if (sizemap.length != nsize) sizemap = new Array[Int](nsize)
    else java.util.Arrays.fill(sizemap, 0)
  }

  protected def calcSizeMapSize(tableLength: Int) = (tableLength >> sizeMapBucketBitSize) + 1

  // discards the previous sizemap and only allocates a new one
  protected def sizeMapInit(tableLength: Int) {
    sizemap = new Array[Int](calcSizeMapSize(tableLength))
  }

  // discards the previous sizemap and populates the new one
  protected def sizeMapInitAndRebuild {
    sizeMapInit(table.length)

    // go through the buckets, count elements
    var tableidx = 0
    var bucketidx = 0
    val tbl = table
    var tableuntil = 0
    if (tbl.length < sizeMapBucketSize) tableuntil = tbl.length else tableuntil = sizeMapBucketSize
    val totalbuckets = totalSizeMapBuckets
    while (bucketidx < totalbuckets) {
      var currbucketsize = 0
      while (tableidx < tableuntil) {
        var e = tbl(tableidx)
        while (e ne null) {
          currbucketsize += 1
          e = e.next
        }
        tableidx += 1
      }
      sizemap(bucketidx) = currbucketsize
      tableuntil += sizeMapBucketSize
      bucketidx += 1
    }
  }

  def printSizeMap {
    println(sizemap.toList)
  }

  protected def sizeMapDisable = sizemap = null

  protected def isSizeMapDefined = sizemap ne null

  protected def elemEquals(key1: A, key2: A): Boolean = (key1 == key2)

  protected def elemHashCode(key: A) = if (key == null) 0 else key.##

  protected final def improve(hcode: Int) = {
    var h: Int = hcode + ~(hcode << 9)
    h = h ^ (h >>> 14)
    h = h + (h << 4)
    h ^ (h >>> 10)
  }

  // Note:
  // we take the most significant bits of the hashcode, not the lower ones
  // this is of crucial importance when populating the table in parallel
  protected final def index(hcode: Int) = {
    val ones = table.length - 1
    (improve(hcode) >> (32 - java.lang.Integer.bitCount(ones))) & ones
  }

  protected def initWithContents(c: HashTable.Contents[A, Entry]) = if (c != null) {
    _loadFactor = c.loadFactor
    table = c.table
    tableSize = c.tableSize
    threshold = c.threshold
    sizemap = c.sizemap
  }

  private[collection] def hashTableContents = new HashTable.Contents(
    _loadFactor,
    table,
    tableSize,
    threshold,
    sizemap
  )
}

private[collection] object HashTable {
  /** The load factor for the hash table (in 0.001 step).
   */
  private[collection] final def defaultLoadFactor: Int = 750 // corresponds to 75%
  private[collection] final def loadFactorDenum = 1000;

  /** The initial size of the hash table.
   */
  private[collection] final def initialSize: Int = 16

  /** The initial threshold.
   */
  private[collection] final def initialThreshold(_loadFactor: Int): Int = newThreshold(_loadFactor, initialCapacity)

  private[collection] final def initialCapacity = capacity(initialSize)

  private[collection] final def newThreshold(_loadFactor: Int, size: Int) = ((size.toLong * _loadFactor) / loadFactorDenum).toInt

  private[collection] final def sizeForThreshold(_loadFactor: Int, thr: Int) = thr * loadFactorDenum / _loadFactor

  private[collection] final def capacity(expectedSize: Int) = if (expectedSize == 0) 1 else powerOfTwo(expectedSize)

  /**
   * Returns a power of two >= `target`.
   */
  private[collection] def powerOfTwo(target: Int): Int = {
    /* See http://bits.stephan-brumme.com/roundUpToNextPowerOfTwo.html */
    var c = target - 1;
    c |= c >>>  1;
    c |= c >>>  2;
    c |= c >>>  4;
    c |= c >>>  8;
    c |= c >>> 16;
    c + 1;
  }

  class Contents[A, Entry >: Null <: HashEntry[A, Entry]](
    val loadFactor: Int,
    val table: Array[HashEntry[A, Entry]],
    val tableSize: Int,
    val threshold: Int,
    val sizemap: Array[Int]
  )

}