// SPDX-License-Identifier: BSD-2-Clause
// Copyright (C) 2010 Andrea Mazzoleni

#include "tommyhashdyn.h"
#include "tommylist.h"

/******************************************************************************/
/* hashdyn */

TOMMY_API void tommy_hashdyn_init(tommy_hashdyn* hashdyn)
{
	/* fixed initial size */
	hashdyn->bucket_bit = TOMMY_HASHDYN_BIT;
	hashdyn->bucket_max = (tommy_size_t)1 << hashdyn->bucket_bit;
	hashdyn->bucket_mask = hashdyn->bucket_max - 1;
	hashdyn->bucket = tommy_cast(tommy_hashdyn_node**, tommy_calloc(hashdyn->bucket_max, sizeof(tommy_hashdyn_node*)));

	hashdyn->count = 0;
}

TOMMY_API void tommy_hashdyn_done(tommy_hashdyn* hashdyn)
{
	tommy_free(hashdyn->bucket);
}

/**
 * Resize the bucket vector.
 */
static void tommy_hashdyn_resize(tommy_hashdyn* hashdyn, tommy_uint_t new_bucket_bit)
{
	tommy_size_t bucket_bit;
	tommy_size_t bucket_max;
	tommy_size_t new_bucket_max;
	tommy_size_t new_bucket_mask;
	tommy_hashdyn_node** new_bucket;

	bucket_bit = hashdyn->bucket_bit;
	bucket_max = hashdyn->bucket_max;

	new_bucket_max = (tommy_size_t)1 << new_bucket_bit;
	new_bucket_mask = new_bucket_max - 1;

	/* allocate the new vector using malloc() and not calloc() */
	/* because data is fully initialized in the update process */
	new_bucket = tommy_cast(tommy_hashdyn_node**, tommy_malloc(new_bucket_max * sizeof(tommy_hashdyn_node*)));

	/* reinsert all the elements */
	if (new_bucket_bit > bucket_bit) {
		tommy_size_t i;

		/* grow */
		for (i = 0; i < bucket_max; ++i) {
			tommy_hashdyn_node* j;

			/* setup the new two buckets */
			new_bucket[i] = 0;
			new_bucket[i + bucket_max] = 0;

			/* reinsert the bucket */
			j = hashdyn->bucket[i];
			while (j) {
				tommy_hashdyn_node* j_next = j->next;
				tommy_size_t pos = j->index & new_bucket_mask;
				if (new_bucket[pos])
					tommy_list_insert_tail_not_empty(new_bucket[pos], j);
				else
					tommy_list_insert_first(&new_bucket[pos], j);
				j = j_next;
			}
		}
	} else {
		tommy_size_t i;

		/* shrink */
		for (i = 0; i < new_bucket_max; ++i) {
			/* setup the new bucket with the lower bucket*/
			new_bucket[i] = hashdyn->bucket[i];

			/* concat the upper bucket */
			tommy_list_concat(&new_bucket[i], &hashdyn->bucket[i + new_bucket_max]);
		}
	}

	tommy_free(hashdyn->bucket);

	/* setup */
	hashdyn->bucket_bit = new_bucket_bit;
	hashdyn->bucket_max = new_bucket_max;
	hashdyn->bucket_mask = new_bucket_mask;
	hashdyn->bucket = new_bucket;
}

/**
 * Grow.
 */
tommy_inline void hashdyn_grow_step(tommy_hashdyn* hashdyn)
{
	/* grow if more than 50% full */
	if (hashdyn->count >= hashdyn->bucket_max / 2)
		tommy_hashdyn_resize(hashdyn, hashdyn->bucket_bit + 1);
}

/**
 * Shrink.
 */
tommy_inline void hashdyn_shrink_step(tommy_hashdyn* hashdyn)
{
	/* shrink if less than 12.5% full */
	if (hashdyn->count <= hashdyn->bucket_max / 8 && hashdyn->bucket_bit > TOMMY_HASHDYN_BIT)
		tommy_hashdyn_resize(hashdyn, hashdyn->bucket_bit - 1);
}

TOMMY_API void tommy_hashdyn_insert(tommy_hashdyn* hashdyn, tommy_hashdyn_node* node, void* data, tommy_hash_t hash)
{
	tommy_size_t pos = hash & hashdyn->bucket_mask;

	tommy_list_insert_tail(&hashdyn->bucket[pos], node, data);

	node->index = hash;

	++hashdyn->count;

	hashdyn_grow_step(hashdyn);
}

TOMMY_API void* tommy_hashdyn_remove_existing(tommy_hashdyn* hashdyn, tommy_hashdyn_node* node)
{
	tommy_size_t pos = node->index & hashdyn->bucket_mask;

	tommy_list_remove_existing(&hashdyn->bucket[pos], node);

	--hashdyn->count;

	hashdyn_shrink_step(hashdyn);

	return node->data;
}

TOMMY_API void* tommy_hashdyn_remove(tommy_hashdyn* hashdyn, tommy_search_func* cmp, const void* cmp_arg, tommy_hash_t hash)
{
	tommy_size_t pos = hash & hashdyn->bucket_mask;
	tommy_hashdyn_node* node = hashdyn->bucket[pos];

	while (node) {
		/* we first check if the hash matches, as in the same bucket we may have multiples hash values */
		if (node->index == hash && cmp(cmp_arg, node->data) == 0) {
			tommy_list_remove_existing(&hashdyn->bucket[pos], node);

			--hashdyn->count;

			hashdyn_shrink_step(hashdyn);

			return node->data;
		}
		node = node->next;
	}

	return 0;
}

TOMMY_API void tommy_hashdyn_foreach(tommy_hashdyn* hashdyn, tommy_foreach_func* func)
{
	tommy_size_t bucket_max = hashdyn->bucket_max;
	tommy_hashdyn_node** bucket = hashdyn->bucket;
	tommy_size_t pos;

	for (pos = 0; pos < bucket_max; ++pos) {
		tommy_hashdyn_node* node = bucket[pos];

		while (node) {
			void* data = node->data;
			node = node->next;
			func(data);
		}
	}
}

TOMMY_API void tommy_hashdyn_foreach_arg(tommy_hashdyn* hashdyn, tommy_foreach_arg_func* func, void* arg)
{
	tommy_size_t bucket_max = hashdyn->bucket_max;
	tommy_hashdyn_node** bucket = hashdyn->bucket;
	tommy_size_t pos;

	for (pos = 0; pos < bucket_max; ++pos) {
		tommy_hashdyn_node* node = bucket[pos];

		while (node) {
			void* data = node->data;
			node = node->next;
			func(arg, data);
		}
	}
}

TOMMY_API tommy_size_t tommy_hashdyn_memory_usage(tommy_hashdyn* hashdyn)
{
	return hashdyn->bucket_max * (tommy_size_t)sizeof(hashdyn->bucket[0])
	       + tommy_hashdyn_count(hashdyn) * (tommy_size_t)sizeof(tommy_hashdyn_node);
}

/**
 * \brief Transfers all elements from the dynamic hashtable into a tommy_list.
 *
 * Removes every element from the \p hashdyn hashtable and inserts them
 * into the provided \p list (at the tail), preserving the per-bucket order
 * but not guaranteeing any particular global order.
 *
 * After the call:
 * - the hashtable is left completely empty (\c count = 0, all buckets = NULL)
 * - the target list contains all the elements that were previously in the hashtable
 *
 * This function is useful when you need to:
 * - extract all elements to process/sort them outside the hash table
 * - convert the hashtable into a list for sequential iteration
 * - prepare for a full clear + re-insertion with different hash/ordering
 * - move ownership of the nodes to a list-based container
 *
 * \note The operation is O(n) where n is the number of elements.
 * \note No memory allocation is performed.
 * \note The relative order of elements that were in the same bucket is preserved,
 *       but the order among different buckets is bucket-order dependent.
 *
 * Typical usage pattern:
 * \code
 * tommy_list all_elements;
 * tommy_list_init(&all_elements);
 *
 * // move everything out of the hashtable into the list
 * tommy_hashdyn_to_list(&hashdyn, &all_elements);
 *
 * // now you can sort, filter, process sequentially, etc.
 * tommy_list_sort(&all_elements, compare_by_value);
 * \endcode
 *
 * \param hashdyn The dynamic hashtable to drain
 * \param list    The list that will receive all the elements
 */
TOMMY_API void tommy_hashdyn_to_list(tommy_hashdyn* hashdyn, tommy_list* list)
{
	tommy_size_t bucket_max = hashdyn->bucket_max;
	tommy_hashdyn_node** bucket = hashdyn->bucket;
	tommy_size_t pos;

	/* move everything to the list */
	for (pos = 0; pos < bucket_max; ++pos) {
		tommy_hashdyn_node* node = bucket[pos];

		while (node) {
			tommy_node* node_next = node->next;
			tommy_list_insert_tail(list, node, node->data);
			node = node_next;
		}
	}

	/* clear all */
	memset(hashdyn->bucket, 0, hashdyn->bucket_max * sizeof(tommy_hashdyn_node*));
	hashdyn->count = 0;
}