//%/////////////////////////////////////////////////////////////////////////////
//
// Copyright (c) 2000, 2001 The Open group, BMC Software, Tivoli Systems, IBM
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to 
// deal in the Software without restriction, including without limitation the 
// rights to use, copy, modify, merge, publish, distribute, sublicense, and/or 
// sell copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
// 
// THE ABOVE COPYRIGHT NOTICE AND THIS PERMISSION NOTICE SHALL BE INCLUDED IN 
// ALL COPIES OR SUBSTANTIAL PORTIONS OF THE SOFTWARE. THE SOFTWARE IS PROVIDED
// "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT
// LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR 
// PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT 
// HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 
// ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
// WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
//
//==============================================================================
//
// Author: Mike Brasher (mbrasher@bmc.com)
//
// Modified By:
//
//%/////////////////////////////////////////////////////////////////////////////

#ifndef Pegasus_HashTable_h
#define Pegasus_HashTable_h

#include <Pegasus/Common/Config.h>
#include <Pegasus/Common/String.h>

PEGASUS_NAMESPACE_BEGIN

/*  This is the default hash function object used by the HashTable template.
    Specializations are provided for common types.
*/
template<class K>
struct HashFunc
{
};

PEGASUS_TEMPLATE_SPECIALIZATION struct PEGASUS_COMMON_LINKAGE HashFunc<String>
{
    static Uint32 hash(const String& str);
};

PEGASUS_TEMPLATE_SPECIALIZATION struct HashFunc<Uint32>
{
    static Uint32 hash(Uint32 x) { return x + 13; }
};

/*  This is a function object used by the HashTable to compare keys. This is
    the default implementation. Others may be defined and passed in the 
    template argument list to perform other kinds of comparisons.
*/
template<class K>
struct EqualFunc
{
    static Boolean equal(const K& x, const K& y)
    {
	return x == y;
    }
};

/*  Representation for a bucket. The HashTable class derives from this
    bucket to append a key and value. This base class just defines
    the pointer to the next bucket in the chain.
*/
class PEGASUS_COMMON_LINKAGE _BucketBase
{
public:

    /* Default constructor. */
    _BucketBase() : next(0) { }

    /* Virtual destructor to ensure destruction of derived class 
	elements.
    */
    virtual ~_BucketBase();

    /* returns true if the key pointed to by the key argument is equal
	to the internal key of this bucket. This method must be overridden
	by the derived class.
    */
    virtual Boolean equal(const void* key) const = 0;

    /* Clone this bucket. */
    virtual _BucketBase* clone() const = 0;

    _BucketBase* next;
};

class _HashTableRep;

/* This class implements a simple hash table forward iterator. */
class PEGASUS_COMMON_LINKAGE _HashTableIteratorBase
{
public:
    
    _HashTableIteratorBase() : _first(0), _last(0), _bucket(0) { }

    operator int() const { return _bucket != 0; }

    _HashTableIteratorBase operator++(int);

    _HashTableIteratorBase& operator++();

    _HashTableIteratorBase(_BucketBase** first, _BucketBase** last);

protected:

    _BucketBase** _first;
    _BucketBase** _last;
    _BucketBase* _bucket;
    friend class _HashTableRep;
};

// ATTN: reorganization not supported yet.

/*- The _HashTableRep class is the representation class used by HashTable.

    This code is primarily an internal class used to implement the HashTable.
    But there may be occasions to use it directly.

    _HashTableRep parcels out much of the large code so that that code is not 
    instantiated by the HashTable template class many times. This scheme helps 
    reduce code bloat caused by templates. The HashTable template class below
    acts as kind of a wrapper around this class.

    _HashTableRep is implemented as an array of pointers to chains of hash
    buckets. The table initially allocates some number of chains (which can
    be controlled by the constructor) and then may increase the number of
    chains later (resulting in a reorganization of the hash table).
*/
class PEGASUS_COMMON_LINKAGE _HashTableRep
{
public:

    /*- This constructor allocates an array of pointers to chains of buckets,
	which of course are all empty at this time. The numChains argument
	If the numChains argument is less than eight, then eight chains will
	be created.
	@param numChains specifies the initial number of chains.
    */
    _HashTableRep(Uint32 numChains);

    /*- Copy constructor. */
    _HashTableRep(const _HashTableRep& x);

    /*- Destructor. */
    ~_HashTableRep();

    /*- Assignment operator. */
    _HashTableRep& operator=(const _HashTableRep& x);

    /*- Returns the size of this hash table (the number of entries). */
    Uint32 size() const { return _size; }

    /*- Clears the contents of this hash table. After this is called, the
	size() method returns zero.
    */
    void clear();

    /*- Inserts new key-value pair into hash table. Deletes the bucket on
	failure so caller need not.
	@param hashCode hash code generated by caller's hash function.
	@param bucket bucket to be inserted.
	@param key pointer to key.
	@return true if insertion successful; false if duplicate key.
    */
    Boolean insert(Uint32 hashCode, _BucketBase* bucket, const void* key);

    /*- Finds the bucket with the given key. This method uses the
	_BucketBase::equal() method to compare keys.
	@param hashCode hash code generated by caller's hash function.
	@param key void pointer to key.
	@return pointer to bucket with that key or zero otherwise.
    */
    const _BucketBase* lookup(Uint32 hashCode, const void* key);

    /*- Removes the bucket with the given key. This method uses the
	_BucketBase::equal() method to compare keys.
	@param hashCode hash code generated by caller's hash function.
	@param key void pointer to key.
	@return true if entry found and removed and false otherwise.
    */
    Boolean remove(Uint32 hashCode, const void* key);

    _BucketBase** getChains() const { return _chains; }

    Uint32 getNumChains() const { return _numChains; }

protected:

    Uint32 _size;
    Uint32 _numChains;
    _BucketBase** _chains;
};

/* The _Bucket class is used to implement the HashTable class.
*/
template<class K, class V, class E>
class _Bucket : public _BucketBase
{
public:

    _Bucket(const K& key, const V& value) : _key(key), _value(value) { }

    virtual ~_Bucket();

    virtual Boolean equal(const void* key) const;

    virtual _BucketBase* clone() const;

    K& getKey() { return _key; }

    V& getValue() { return _value; }

private:

    K _key;
    V _value;
};

template<class K, class V, class E>
Boolean _Bucket<K, V, E>::equal(const void* key) const
{
    return E::equal(*((K*)key), _key);
}

template<class K, class V, class E>
_Bucket<K, V, E>::~_Bucket()
{

}

template<class K, class V, class E>
_BucketBase* _Bucket<K, V, E>::clone() const
{
    return new _Bucket<K, V, E>(_key, _value);
}

/* Iterator for HashTable class. */
template<class K, class V, class E>
class _HashTableIterator : public _HashTableIteratorBase
{
public:

    _HashTableIterator() 
	: _HashTableIteratorBase() { }

    _HashTableIterator(_BucketBase** first, _BucketBase** last)
	: _HashTableIteratorBase(first, last) { }

    const K& key() const { return ((_Bucket<K, V, E>*)_bucket)->getKey(); }

    const V& value() const { return ((_Bucket<K, V, E>*)_bucket)->getValue(); }
};

/** The HashTable class provides a simple hash table implementation which 
    associates key-value pairs.

    This implementation minimizes template bloat considerably by factoring out
    most of the code into a common non-template class (see _HashTableRep).
    The HashTable class is mostly a wrapper to add proper type semantics to the
    use of its representation class.

    Hashing as always is O(1).

    HashTable uses the most popular hash table implementation which utilizes 
    an array of pointers to bucket chains. This is organized as follows:

        <pre>
           +---+ 
           |   |   +-----+-------+
         0 | ----->| key | value |
           |   |   +-----+-------+
           +---+
           |   |   +-----+-------+   +-----+-------+   +-----+-------+
         1 | ----->| key | value |-->| key | value |-->| key | value |
           |   |   +-----+-------+   +-----+-------+   +-----+-------+
           +---+
             .
             .
             .
           +---+
           |   |   +-----+-------+   +-----+-------+
        N-1| ----->| key | value |-->| key | value |
           |   |   +-----+-------+   +-----+-------+
           +---+
        </pre>

    To locate an item a hash function is applied to the key to produce an 
    integer value. Then the modulo of that integer is taken with N to select 
    a chain (as shown above). Then the chain is searched for a bucket whose
    key value is the same as the target key.

    The number of chains default to DEFAULT_NUM_CHAINS but should be about
    one-third the number of expected entries (so that the average chain
    will be three long). Making the number of chains too large will waste
    space causing the hash table to be very sparse. But for optimal efficiency,
    one might set the number of chains to be the same as the expected number
    of entries.

    This implementation does have NOT an adaptive growth algorithm yet which 
    would allow it to increase the number of chains periodically based on some
    statistic (e.g., when the number of entries is more than three times the
    number of chains; this would keep the average chain length below three).

    The following example shows how to instantiate a HashTable which associates
    String keys with Uint32 values.

	<pre>
	typedef HashTable&lt;String, Uint32&gt; HT;
	HT ht;
	</pre>

    Some of the template arguments are defaulted in the above example (the 
    third and forth). The instantiation is explicitly qualified like this
    (which by the way has exactly the same effect).

	<pre>
	typedef HashTable&lt;String, Uint32, EqualFunc&lt;String&gt;, HashFunc&lt;String&gt;&gt; HT;
	</pre>

    The third and forth arguments are described more in detail later.

    Then, entries may be inserted like this:

	<pre>
	ht.insert("Red", 111);
	ht.insert("Green", 222);
	ht.insert("Blue", 222);
	</pre>

    And entries may be looked up as follows:

	<pre>
	Uint32 value;
	ht.lookup("Red", value);
	</pre>

    And entries may be removed like this:

	<pre>
	h.remove("Red");
	</pre>

    Iteration is done like this:

	<pre>
	for (HT::Iterator i = ht.start(); i; i++)
	{
	    // To access the key call i.key()!
	    // To access the value call i.value()!
	}
	</pre>

    Note that only forward iteration is supported (no backwards iteration).

    Equality of keys is determined using the EqualFunc class which is
    the default third argument of the template argument list. A new function 
    object may be defined and passed to modify the behavior (for example, one 
    might define equality of strings to ignore whitespace). Here is how to 
    define and use a new equality function object:

	<pre>
	struct MyEqualFunc
	{
	    static Boolean equal(const String& x, const String& y)
	    {
		// do something here to test for equality!
	    }
	};

	...

	EqualFunc&lt;String, Uint32, MyEqualFunc&gt; ht;
	</pre>

    When the lookup(), insert(), and remove() methods are called, the 
    MyEqualFunc::equal() method will be used to determine equality.

    Hash functions are provided for common types (as part of the default
    HashFunc class). For other types it is possible to define a custom function
    object as follows:

	<pre>
	struct MyHashFunc
	{
	    static Uint32 hash(const String& x)
	    {
		// Do some hashing here!
	    }
	};

	...

	EqualFunc&lt;String, Uint32, MyEqualFunc, MyHashFunc&gt; ht;
	</pre>

    As always, the hash function should provide a reasonably uniform 
    distrubtion so that all of the entries don't get crowded into a few
    chains. Note that a hash function which returns zero, would force
    the pathalogical case in which all entries are placed in the first
    chain.
*/
template<class K, class V, class E , class H >
class HashTable
{
public:

    typedef _HashTableIterator<K, V, E> Iterator;

    /* By default, we create this many chains initially */
    enum { DEFAULT_NUM_CHAINS = 32 };

    /** Constructor.
	@param numChains number of chains to create.
    */
    HashTable(Uint32 numChains = DEFAULT_NUM_CHAINS) : _rep(numChains)
    {

    }

    /** Copy constructor. */
    HashTable(const HashTable<K,V,E,H>& x) : _rep(x._rep)
    {

    }

    /** Assignment operator. */
    HashTable<K,V,E,H>& operator=(const HashTable<K,V,E,H>& x)
    {
	if (this != &x)
	    _rep = x._rep;
	return *this;
    }

    /** Returns the size of this hash table (the number of entries). */
    Uint32 size() const { return _rep.size(); }

    /** Clears the contents of this hash table. After this is called, the
	size() method returns zero.
    */
    void clear() { _rep.clear(); }

    /** Inserts new key-value pair into hash table.
	@param key key component.
	@param value value component.
	@return true on success; false if duplicate key.
    */
    Boolean insert(const K& key, const V& value)
    {
	return _rep.insert(
	    H::hash(key), new _Bucket<K, V, E>(key, value), &key);
    }

    /** Checks to see if hash table contains an entry with the given key.
	@param key key to be searched for
	@return true if hash table contains an entry with the given key.
    */
    Boolean contains(const K& key)
    {
	V value;
	return lookup(key, value);
    }

    /** Looks up the entry with the given key.
	@param key key of entry to be located.
	@param value output value.
	@return true if found; false otherwise.
    */
    Boolean lookup(const K& key, V& value);

    /** Removes the entry with the given key.
	@param key key of entry to be removed.
	@return true on success; false otherwise.
    */
    Boolean remove(const K& key)
    {
	return _rep.remove(H::hash(key), &key);
    }

    /** Obtains an iterator for this object. */
    Iterator start() const 
    {
	return Iterator(
	    _rep.getChains(), _rep.getChains() + _rep.getNumChains());
    }

private:

    _HashTableRep _rep;
};

template<class K, class V, class E, class H>
inline Boolean HashTable<K, V, E, H>::lookup(const K& key, V& value)
{
    _Bucket<K, V, E>* bucket 
	= (_Bucket<K, V, E>*)_rep.lookup(H::hash(key), &key);

    if (bucket)
    {
	value = bucket->getValue();
	return true;
    }

    return false;
}

PEGASUS_NAMESPACE_END

#endif /* Pegasus_HashTable_h */