/*
   Copyright (C) 2011  Equinor ASA, Norway.

   The file 'hash.c' is part of ERT - Ensemble based Reservoir Tool.

   ERT is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation, either version 3 of the License, or
   (at your option) any later version.

   ERT is distributed in the hope that it will be useful, but WITHOUT ANY
   WARRANTY; without even the implied warranty of MERCHANTABILITY or
   FITNESS FOR A PARTICULAR PURPOSE.

   See the GNU General Public License at <http://www.gnu.org/licenses/gpl.html>
   for more details.
*/
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <stdint.h>
#include <errno.h>

#include <ert/util/hash.hpp>
#include <ert/util/hash_sll.hpp>
#include <ert/util/hash_node.hpp>
#include <ert/util/node_data.hpp>
#include <ert/util/util.h>
#include <ert/util/stringlist.hpp>


#ifdef __cplusplus
extern "C" {
#endif

#define HASH_DEFAULT_SIZE 16
#define HASH_TYPE_ID      771065

/**
   This is **THE** hash function - which actually does the hashing.
*/

static uint32_t hash_index(const char *key, size_t len) {
  uint32_t hash = 0;
  size_t i;

  for (i=0; i < len; i++) {
    hash += key[i];
    hash += (hash << 10);
    hash ^= (hash >>  6);
  }
  hash += (hash << 3);
  hash ^= (hash >> 11);
  hash += (hash << 15);
  return hash;
}


struct hash_struct {
  UTIL_TYPE_ID_DECLARATION;
  uint32_t          size;            /* This is the size of the internal table **NOT**NOT** the number of elements in the table. */
  uint32_t          elements;        /* The number of elements in the hash table. */
  double            resize_fill;
  hash_sll_type   **table;
  hashf_type       *hashf;
};



typedef struct hash_sort_node {
  char     *key;
  int       cmp_value;
} hash_sort_type;


/*****************************************************************/
/*                    Low level access functions                 */
/*****************************************************************/


static void * __hash_get_node(const hash_type *__hash , const char *key, bool abort_on_error) {
  hash_type * hash = (hash_type *) __hash;  /* The net effect is no change - but .... ?? */
  hash_node_type * node = NULL;
  {
    const uint32_t global_index = hash->hashf(key , strlen(key));
    const uint32_t table_index  = (global_index % hash->size);

    node = hash_sll_get(hash->table[table_index] , global_index , key);
    if (node == NULL && abort_on_error)
      util_abort("%s: tried to get from key:%s which does not exist - aborting \n",__func__ , key);

  }
  return node;
}




static node_data_type * hash_get_node_data(const hash_type *hash , const char *key) {
  hash_node_type * node = (hash_node_type*)__hash_get_node(hash , key , true);
  return hash_node_get_data(node);
}




/**
   This function resizes the hash table when it has become to full.
   The table only grows - this function is called from
   __hash_insert_node().

   If you know in advance (roughly) how large the hash table will be
   it can be advantageous to call hash_resize() manually, to avoid
   repeated internal calls to hash_resize().
*/

void hash_resize(hash_type *hash, int new_size) {
  hash_sll_type ** new_table = hash_sll_alloc_table( new_size );
  hash_node_type * node;
  uint32_t i;

  for (i=0; i < hash->size; i++) {
    node = hash_sll_get_head(hash->table[i]);
    while (node != NULL) {
      uint32_t new_table_index  = hash_node_set_table_index(node , new_size);
      hash_node_type *next_node = hash_node_get_next(node);
      hash_sll_add_node(new_table[new_table_index] , node);
      node = next_node;
    }
  }

  /*
     Only freeing the table structure, *NOT* calling the node_free()
     functions, which happens when hash_sll_free() is called.
  */

  {
    for (i=0; i < hash->size; i++)
      free( hash->table[i] );
    free( hash->table );
  }

  hash->size     = new_size;
  hash->table    = new_table;
}


/**
   This is the low-level function for inserting a hash node.
*/

static void __hash_insert_node(hash_type *hash , hash_node_type *node) {
  uint32_t table_index = hash_node_get_table_index(node);
  {
    /*
      If a node with the same key already exists in the table
      it is removed.
    */
    hash_node_type *existing_node = (hash_node_type*)__hash_get_node(hash , hash_node_get_key(node) , false);
    if (existing_node != NULL) {
      hash_sll_del_node(hash->table[table_index] , existing_node);
      hash->elements--;
    }
  }

  hash_sll_add_node(hash->table[table_index] , node);
  hash->elements++;
  if ((1.0 * hash->elements / hash->size) > hash->resize_fill)
    hash_resize(hash , hash->size * 2);
}



/**
   This function deletes a node from the hash_table.
*/


static void hash_del__(hash_type *hash , const char *key) {
  const uint32_t global_index = hash->hashf(key , strlen(key));
  const uint32_t table_index  = (global_index % hash->size);
  hash_node_type *node        = hash_sll_get(hash->table[table_index] , global_index , key);

  if (node == NULL)
    util_abort("%s: hash does not contain key:%s - aborting \n",__func__ , key);
  else
    hash_sll_del_node(hash->table[table_index] , node);

  hash->elements--;
}



/**
   This functions takes a hash_node and finds the "next" hash node by
   traversing the internal hash structure. Should NOT be confused with
   the other functions providing iterations to user-space.
*/

static hash_node_type * hash_internal_iter_next(const hash_type *hash , const hash_node_type * node) {
  hash_node_type *next_node = hash_node_get_next(node);
  if (next_node == NULL) {
    const uint32_t table_index = hash_node_get_table_index(node);
    if (table_index < hash->size) {
      uint32_t i = table_index + 1;
      while (i < hash->size && hash_sll_empty(hash->table[i]))
        i++;

      if (i < hash->size)
        next_node = hash_sll_get_head(hash->table[i]);
    }
  }
  return next_node;
}



/**
   This is the low level function which traverses a hash table and
   allocates a char ** list of keys.

   If the hash table is empty NULL is returned.
*/

static char ** hash_alloc_keylist__(const hash_type *hash) {
  char **keylist;
  if (hash->elements > 0) {
    int i = 0;
    hash_node_type *node = NULL;
    keylist = (char**)calloc(hash->elements , sizeof *keylist);
    {
      uint32_t i = 0;
      while (i < hash->size && hash_sll_empty(hash->table[i]))
        i++;

      if (i < hash->size)
        node = hash_sll_get_head(hash->table[i]);
    }

    while (node != NULL) {
      const char *key = hash_node_get_key(node);
      keylist[i] = util_alloc_string_copy(key);
      node = hash_internal_iter_next(hash , node);
      i++;
    }
  } else
    keylist = NULL;
  return keylist;
}






/*****************************************************************/
/**
   The fundamental functions above relate the hash_node
   structure. Here comes a list of functions for inserting managed
   copies of various types.
*/

void hash_insert_string(hash_type * hash , const char * key , const char * value) {
  node_data_type * node_data = node_data_alloc_string( value );
  hash_node_type * hash_node = hash_node_alloc_new(key , node_data , hash->hashf , hash->size);
  __hash_insert_node(hash , hash_node);
}


char * hash_get_string(const hash_type * hash , const char * key) {
  node_data_type * node_data = hash_get_node_data(hash , key);
  return node_data_get_string( node_data );
}


void hash_insert_int(hash_type * hash , const char * key , int value) {
  node_data_type * node_data = node_data_alloc_int( value );
  hash_node_type * hash_node = hash_node_alloc_new(key , node_data , hash->hashf , hash->size);
  __hash_insert_node(hash , hash_node);
}


int hash_get_int(const hash_type * hash , const char * key) {
  node_data_type * node_data = hash_get_node_data(hash , key);
  return node_data_get_int( node_data );
}

/**
   Small wrapper around hash_get_int() / hash_insert_int() which
   implements a zero based counter.

      hash_inc_counter()

   Will increment the integer value stored in the node_data instance,
   and return the updated value. If the key is not present in the hash
   it will be inserted as an integer with value 0, and 0 will be
   returned.
*/

int hash_inc_counter(hash_type * hash , const char * counter_key) {
  if (hash_has_key( hash , counter_key)) {
    node_data_type * node_data = hash_get_node_data(hash , counter_key);
    return node_data_fetch_and_inc_int( node_data );
  } else {
    hash_insert_int(hash , counter_key , 0);
    return 0;
  }
}

/**
   Will return 0 if the key is not in the hash.
*/

int hash_get_counter(const hash_type * hash , const char * key) {
  if (hash_has_key( hash , key ))
    return hash_get_int( hash , key );
  else
    return 0;
}


void hash_insert_double(hash_type * hash , const char * key , double value) {
  node_data_type * node_data = node_data_alloc_double( value );
  hash_node_type * hash_node = hash_node_alloc_new(key , node_data , hash->hashf , hash->size);
  __hash_insert_node(hash , hash_node);
}

double hash_get_double(const hash_type * hash , const char * key) {
  node_data_type * node_data = hash_get_node_data(hash , key);
  return node_data_get_double( node_data );
}


/*****************************************************************/

void hash_del(hash_type *hash , const char *key) {
  hash_del__(hash , key);
}

/**
   This function will delete the key if it exists in the hash, but it
   will NOT fail if the key is not already in the hash table.
*/

void hash_safe_del(hash_type * hash , const char * key) {
  if (__hash_get_node(hash , key , false))
    hash_del__(hash , key);
}


void hash_clear(hash_type *hash) {
  int old_size = hash_get_size(hash);
  if (old_size > 0) {
    char **keyList = hash_alloc_keylist__( hash );
    for (int i=0; i < old_size; i++) {
      hash_del__(hash , keyList[i]);
      free(keyList[i]);
    }
    free(keyList);
  }
}


void * hash_get(const hash_type *hash , const char *key) {
  hash_node_type * hash_node = (hash_node_type*)__hash_get_node(hash , key , true);
  node_data_type * data_node = (node_data_type*)hash_node_get_data( hash_node );
  return node_data_get_ptr( data_node );
}


/**
   This function will return NULL if the hash does not
   contain 'key'.
*/
void * hash_safe_get( const hash_type * hash , const char * key ) {
  hash_node_type * hash_node = (hash_node_type*)__hash_get_node(hash , key , false);
  if (hash_node != NULL) {
    node_data_type * data_node = hash_node_get_data( hash_node );
    return node_data_get_ptr( data_node );
  } else
    return NULL;
}


/**
   This function will:

    1. Return an object from the hash table.
    2. Remove it from the table.

   Observe that if the object has been installed with a destructor,
   the object will be destroyed by the hash_del() operation, and the
   return value is complete gibberish - i.e. this function can NOT be
   used on hash-owned references.
*/

void * hash_pop( hash_type * hash , const char * key) {
  void * value = hash_get( hash , key);
  hash_del( hash , key );
  return value;
}




/******************************************************************/
/*                     Allocators                                 */
/******************************************************************/


static hash_type * __hash_alloc(int size, double resize_fill , hashf_type *hashf) {
  hash_type* hash;
  hash = (hash_type*)util_malloc(sizeof *hash );
  UTIL_TYPE_ID_INIT(hash , HASH_TYPE_ID);
  hash->size      = size;
  hash->hashf     = hashf;
  hash->table     = hash_sll_alloc_table(hash->size);
  hash->elements  = 0;
  hash->resize_fill  = resize_fill;
  return hash;
}


hash_type * hash_alloc() {
  return __hash_alloc(HASH_DEFAULT_SIZE , 0.50 , hash_index);
}



UTIL_SAFE_CAST_FUNCTION( hash , HASH_TYPE_ID)
UTIL_SAFE_CAST_FUNCTION_CONST( hash , HASH_TYPE_ID)
UTIL_IS_INSTANCE_FUNCTION(hash , HASH_TYPE_ID)

void hash_free(hash_type *hash) {
  uint32_t i;
  for (i=0; i < hash->size; i++)
    hash_sll_free(hash->table[i]);
  free(hash->table);
  free(hash);
}


void hash_free__(void * void_hash) {
  hash_free(hash_safe_cast( void_hash ));
}


char ** hash_alloc_keylist(const hash_type *hash) {
  return hash_alloc_keylist__(hash);
}

stringlist_type * hash_alloc_stringlist(const hash_type * hash) {
  stringlist_type * stringlist = stringlist_alloc_new();
  char ** keylist = hash_alloc_keylist__(hash);
  int i;
  for (i = 0; i < hash_get_size( hash ); i++) {
    stringlist_append_copy( stringlist , keylist[i] );
    free(keylist[i]);
  }

  free( keylist );
  return stringlist;
}



/*****************************************************************/
/**
    The standard functions for inserting an entry in the hash table:

    hash_insert_copy(): The hash table uses copyc() to make a copy of
         value, which is inserted in the table. This means that the
         calling scope is free to do whatever it wants with value; and
         is also responsible for freeing it. The hash_table takes
         responsibility for freeing it's own copy.

    hash_insert_hash_owned_ref(): The hash table takes ownership of
         'value', in the sense that the hash table will call the
         'destructor' del() on value when the node is deleted.

    hash_insert_ref(): The hash table ONLY contains a pointer to
         value; the calling scope retains full ownership to
         value. When the hash node is deleted, the hash implementation
         will just drop the reference on the floor.
*/


void hash_insert_copy(hash_type *hash , const char *key , const void *value , copyc_ftype *copyc , free_ftype *del) {
  hash_node_type *hash_node;
  if (copyc == NULL || del == NULL)
    util_abort("%s: must provide copy constructer and delete operator for insert copy - aborting \n",__func__);
  {
    node_data_type * data_node = node_data_alloc_ptr( value , copyc , del );
    hash_node                  = hash_node_alloc_new(key , data_node , hash->hashf , hash->size);
    __hash_insert_node(hash , hash_node);
  }
}


/**
  This function will insert a reference "value" with key "key"; when
  the key is deleted - either by an explicit call to hash_del(), or
  when the complete hash table is free'd with hash_free(), the
  destructur 'del' is called with 'value' as argument.

  It is importand to realize that when elements are inserted into a
  hash table with this function the calling scope gives up
  responsibility of freeing the memory pointed to by value.
*/

void hash_insert_hash_owned_ref(hash_type *hash , const char *key , const void *value , free_ftype *del) {
  hash_node_type *hash_node;
  if (del == NULL)
    util_abort("%s: must provide delete operator for insert hash_owned_ref - aborting \n",__func__);
  {
    node_data_type * data_node = node_data_alloc_ptr( value , NULL , del );
    hash_node                  = hash_node_alloc_new(key , data_node , hash->hashf , hash->size);
    __hash_insert_node(hash , hash_node);
  }
}


void hash_insert_ref(hash_type *hash , const char *key , const void *value) {
  hash_node_type *hash_node;
  {
    node_data_type * data_node = node_data_alloc_ptr( value , NULL , NULL);
    hash_node                  = hash_node_alloc_new(key , data_node , hash->hashf , hash->size);
    __hash_insert_node(hash , hash_node);
  }
}



bool hash_has_key(const hash_type *hash , const char *key) {
  if (__hash_get_node(hash , key , false) == NULL)
    return false;
  else
    return true;
}


int hash_get_size(const hash_type *hash) {
  return hash->elements;
}



/******************************************************************/
/**
   Here comes a list of functions for allocating keylists which have
   been sorted in various ways.
*/


static hash_sort_type * hash_alloc_sort_list(const hash_type *hash ,
                                             const char **keylist) {

  int i; hash_sort_type * sort_list = (hash_sort_type*)calloc(hash_get_size(hash) , sizeof * sort_list);
  for (i=0; i < hash_get_size(hash); i++)
    sort_list[i].key = util_alloc_string_copy(keylist[i]);

  return sort_list;
}

static void hash_free_sort_list(const hash_type *hash , hash_sort_type *sort_list) {
  int i;
  for (i=0; i < hash_get_size(hash); i++)
    free(sort_list[i].key);
  free(sort_list);
}


static int hash_sortlist_cmp(const void *_p1 , const void  *_p2) {
  const hash_sort_type *p1 = (const hash_sort_type *) _p1;
  const hash_sort_type *p2 = (const hash_sort_type *) _p2;

  if (p1->cmp_value == p2->cmp_value)
    return 0;
  else if (p1->cmp_value < p2->cmp_value)
    return -1;
  else
    return 1;
}


static char ** __hash_alloc_ordered_keylist(const hash_type *hash , int ( hash_get_cmp_value) (const void * )) {
  int i;
  char **sorted_keylist;
  char **tmp_keylist   = hash_alloc_keylist(hash);
  hash_sort_type * sort_list = hash_alloc_sort_list(hash , (const char **) tmp_keylist);

  for (i = 0; i < hash_get_size(hash); i++)
    sort_list[i].cmp_value = hash_get_cmp_value( hash_get(hash , sort_list[i].key) );

  qsort(sort_list , hash_get_size(hash) , sizeof *sort_list , &hash_sortlist_cmp);
  sorted_keylist = (char**)calloc(hash_get_size(hash) , sizeof *sorted_keylist);
  for (i = 0; i < hash_get_size(hash); i++) {
    sorted_keylist[i] = util_alloc_string_copy(sort_list[i].key);
    free(tmp_keylist[i]);
  }
  free(tmp_keylist);
  hash_free_sort_list(hash , sort_list);
  return sorted_keylist;
}


char ** hash_alloc_sorted_keylist(const hash_type *hash , int ( hash_get_cmp_value) (const void *)) {
  char ** key_list;

  key_list = __hash_alloc_ordered_keylist(hash , hash_get_cmp_value);

  return key_list;
}


static int key_cmp(const void *_s1 , const void *_s2) {
  const char * s1 = *((const char **) _s1);
  const char * s2 = *((const char **) _s2);

  return strcmp(s1 , s2);
}


static char ** __hash_alloc_key_sorted_list(const hash_type *hash, int (*cmp) (const void * , const void *)) {
  char **keylist = hash_alloc_keylist(hash);

  qsort(keylist , hash_get_size(hash) , sizeof *keylist , cmp);
  return keylist;
}



char ** hash_alloc_key_sorted_list(const hash_type * hash, int (*cmp) (const void *, const void *))
{
  char ** key_list;

  key_list =__hash_alloc_key_sorted_list(hash , cmp);

  return key_list;
}



bool hash_key_list_compare(const hash_type * hash1, const hash_type * hash2)
{
  bool has_equal_keylist;
  int i,size1, size2;
  char **keylist1, **keylist2;

  size1 = hash_get_size(hash1);
  size2 = hash_get_size(hash2);

  if(size1 != size2) return false;

  keylist1 = hash_alloc_key_sorted_list(hash1, &key_cmp);
  keylist2 = hash_alloc_key_sorted_list(hash2, &key_cmp);

  has_equal_keylist = true;
  for(i=0; i<size1; i++)
  {
    if(strcmp(keylist1[i],keylist2[i]) != 0)
    {
      has_equal_keylist = false;
      break;
    }
  }

  for(i=0; i<size1; i++) {
    free( keylist1[i] );
    free( keylist2[i] );
  }
  free( keylist1 );
  free( keylist2 );
  return has_equal_keylist;

}




/*****************************************************************/
/**
   This function will take a list of strings of type:

     ["OPT1:Value1" , "MIN:0.0001" , "MAX:1.00" , "FILE:XX"]

   and build a hash table where the element in front of ':' is used as
   key, and the element behind the ':' is used as value. The value is
   internalized as a (char *) pointer with no type conversion.  In the
   calling scope the values should be extracted with hash_get().

   Strings which can not be interpreted as KEY:VALUE are simply
   ignored.
*/


hash_type * hash_alloc_from_options(const stringlist_type * options) {
  int num_options = stringlist_get_size( options );
  hash_type * opt_hash = hash_alloc();
  int iopt;

  for (iopt = 0; iopt < num_options; iopt++) {
    char * option;
    char * value;

    util_binary_split_string( stringlist_iget(options , iopt) , ":" , true , &option , &value);
    if ((option != NULL) && (value != NULL))
      hash_insert_hash_owned_ref( opt_hash , option , util_alloc_string_copy(value) , free);
    // Warning: could not interpret string as KEY:VALUE - ignored


    free(option);
    free(value);
  }

  return opt_hash;
}


 bool hash_add_option( hash_type * hash, const char * key_value) {
   bool addOK = false;
   {
     char * value;
     char * key;

     util_binary_split_string( key_value , ":" , true , &key , &value);
     if (value != NULL) {
       hash_insert_hash_owned_ref( hash , key , value , free );
       addOK = true;
     }

     free( key );
   }
   return addOK;
}



/*****************************************************************/


/**
  This is a **VERY** simple iteration object.

  Do **NOT** use with multi-threading.
*/
struct hash_iter_struct {
  const hash_type  * hash;            /* The hash we are iterating over. */
  char            ** keylist;         /* The keys in the hash table - at the moment of hash_iter_alloc(). */
  int                num_keys;
  int                current_key_num; /* This integer retains the state. */
};


void hash_iter_restart( hash_iter_type * iter ) {
  iter->current_key_num = 0;
}




hash_iter_type * hash_iter_alloc(const hash_type * hash) {
  hash_iter_type * iter = (hash_iter_type*)util_malloc(sizeof * iter );

  iter->hash            = hash;
  iter->num_keys        = hash_get_size(hash);
  iter->keylist         = hash_alloc_keylist( (hash_type *) hash);
  hash_iter_restart( iter );
  return iter;
}



void hash_iter_free(hash_iter_type * iter) {
  util_free_stringlist(iter->keylist, iter->num_keys);
  free(iter);
}



bool hash_iter_is_complete(const hash_iter_type * iter) {
  if(iter->current_key_num == iter->num_keys)
    return true;
  else
    return false;
}



/**
  Get the next key.

  Returns NULL if the iteration has ended.
*/
const char * hash_iter_get_next_key(hash_iter_type * iter) {
  const char * key;

  if(iter->current_key_num == iter->num_keys)
    return NULL;

  key = iter->keylist[iter->current_key_num];
  iter->current_key_num++;

  if(!hash_has_key(iter->hash, key))
    util_abort("%s: Programming error. Using hash_iter with multi-threading??\n", __func__);

  return key;
}



void * hash_iter_get_next_value(hash_iter_type * iter) {
  const char * key = hash_iter_get_next_key(iter);

  if(key != NULL)
    return hash_get(iter->hash, key);
  else
    return NULL;
}


/*****************************************************************/

/**
   This function will iterate through the hash table, and call the
   function func on each of the elements in the table inplace:

       ....
       value = hash_get( hash , key );
       func( value );                      <-- The call is inplace - with no arguments,
       ....                                    and no return value. The content of 'value'
                                               can obviously change, but 'value' itself
                                               must still be a valid reference!
*/


void hash_apply( hash_type * hash , hash_apply_ftype * func) {
  hash_iter_type * iter = hash_iter_alloc( hash );
  while (!hash_iter_is_complete( iter )) {
    const char * key = hash_iter_get_next_key( iter );
    void * value     = hash_get( hash , key );

    func( value );
  }
  hash_iter_free( iter );
}



#undef HASH_GET_SCALAR
#undef HASH_INSERT_SCALAR
#undef HASH_INSERT_ARRAY
#undef HASH_GET_ARRAY_PTR
#undef HASH_NODE_AS
#undef HASH_DEFAULT_SIZE

#ifdef __cplusplus
}
#endif