universe@549: /*
universe@549:  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS HEADER.
universe@549:  *
universe@549:  * Copyright 2021 Mike Becker, Olaf Wintermann All rights reserved.
universe@549:  *
universe@549:  * Redistribution and use in source and binary forms, with or without
universe@549:  * modification, are permitted provided that the following conditions are met:
universe@549:  *
universe@549:  *   1. Redistributions of source code must retain the above copyright
universe@549:  *      notice, this list of conditions and the following disclaimer.
universe@549:  *
universe@549:  *   2. Redistributions in binary form must reproduce the above copyright
universe@549:  *      notice, this list of conditions and the following disclaimer in the
universe@549:  *      documentation and/or other materials provided with the distribution.
universe@549:  *
universe@549:  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
universe@549:  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
universe@549:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
universe@549:  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
universe@549:  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
universe@549:  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
universe@549:  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
universe@549:  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
universe@549:  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
universe@549:  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
universe@549:  * POSSIBILITY OF SUCH DAMAGE.
universe@549:  */
universe@549: 
universe@550: #include <errno.h>
universe@549: #include <string.h>
universe@549: #include "cx/hash_map.h"
universe@550: #include "cx/utils.h"
universe@549: 
universe@549: /**
universe@549:  * Computes a murmur hash-2.
universe@549:  *
universe@549:  * @param data the data to hash
universe@549:  * @param len the length of the data
universe@549:  * @return the murmur hash-2 of the data
universe@549:  */
universe@549: static unsigned cx_hash_map_murmur(
universe@549:         unsigned char const *data,
universe@549:         size_t len
universe@549: ) {
universe@549:     unsigned m = 0x5bd1e995;
universe@549:     unsigned r = 24;
universe@549:     unsigned h = 25 ^ len;
universe@549:     unsigned i = 0;
universe@549:     while (len >= 4) {
universe@549:         unsigned k = data[i + 0] & 0xFF;
universe@549:         k |= (data[i + 1] & 0xFF) << 8;
universe@549:         k |= (data[i + 2] & 0xFF) << 16;
universe@549:         k |= (data[i + 3] & 0xFF) << 24;
universe@549: 
universe@549:         k *= m;
universe@549:         k ^= k >> r;
universe@549:         k *= m;
universe@549: 
universe@549:         h *= m;
universe@549:         h ^= k;
universe@549: 
universe@549:         i += 4;
universe@549:         len -= 4;
universe@549:     }
universe@549: 
universe@549:     switch (len) {
universe@549:         case 3:
universe@549:             h ^= (data[i + 2] & 0xFF) << 16;
universe@549:                     __attribute__((__fallthrough__));
universe@549:         case 2:
universe@549:             h ^= (data[i + 1] & 0xFF) << 8;
universe@549:                     __attribute__((__fallthrough__));
universe@549:         case 1:
universe@549:             h ^= (data[i + 0] & 0xFF);
universe@549:             h *= m;
universe@549:                     __attribute__((__fallthrough__));
universe@549:         default:
universe@549:             /* do nothing */;
universe@549:     }
universe@549: 
universe@549:     h ^= h >> 13;
universe@549:     h *= m;
universe@549:     h ^= h >> 15;
universe@549: 
universe@549:     return h;
universe@549: }
universe@549: 
universe@550: static void cx_hash_map_clear(struct cx_map_s *map) {
universe@550:     struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map;
universe@550:     cx_for_n(i, hash_map->bucket_count) {
universe@550:         struct cx_hash_map_element_s *elem = hash_map->buckets[i];
universe@550:         if (elem != NULL) {
universe@550:             do {
universe@550:                 struct cx_hash_map_element_s *next = elem->next;
universe@550:                 // free the key data
universe@550:                 cxFree(map->allocator, elem->key.data);
universe@550:                 // free the node
universe@550:                 cxFree(map->allocator, elem);
universe@550:                 // proceed
universe@550:                 elem = next;
universe@550:             } while (elem != NULL);
universe@550: 
universe@550:             // do not leave a dangling pointer
universe@550:             hash_map->buckets[i] = NULL;
universe@550:         }
universe@550:     }
universe@550:     map->size = 0;
universe@550: }
universe@550: 
universe@550: static void cx_hash_map_destructor(struct cx_map_s *map) {
universe@550:     struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map;
universe@550: 
universe@550:     // free the buckets
universe@550:     cx_hash_map_clear(map);
universe@550:     cxFree(map->allocator, hash_map->buckets);
universe@550: 
universe@550:     // free the map structure
universe@550:     cxFree(map->allocator, map);
universe@550: }
universe@550: 
universe@550: static int cx_hash_map_put(
universe@550:         CxMap *map,
universe@550:         CxDataPtr key,
universe@550:         void *value
universe@550: ) {
universe@550:     struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map;
universe@550:     CxAllocator *allocator = map->allocator;
universe@550: 
universe@550:     unsigned hash = cx_hash_map_murmur(key.data, key.len);
universe@550: 
universe@550:     size_t slot = hash % hash_map->bucket_count;
universe@550:     struct cx_hash_map_element_s *elm = hash_map->buckets[slot];
universe@550:     struct cx_hash_map_element_s *prev = NULL;
universe@550: 
universe@550:     while (elm != NULL && elm->key.hash < hash) {
universe@550:         prev = elm;
universe@550:         elm = elm->next;
universe@550:     }
universe@550: 
universe@550:     if (elm == NULL || elm->key.hash != hash) {
universe@550:         struct cx_hash_map_element_s *e = cxMalloc(allocator, sizeof(struct cx_hash_map_element_s));
universe@550:         if (e == NULL) {
universe@550:             return -1;
universe@550:         }
universe@550: 
universe@550:         // write the value
universe@550:         // TODO: depending on future map features, we may want to copy here
universe@550:         e->data = value;
universe@550: 
universe@550:         // copy the key
universe@550:         void *kd = cxMalloc(allocator, key.len);
universe@550:         if (kd == NULL) {
universe@550:             return -1;
universe@550:         }
universe@550:         memcpy(kd, key.data, key.len);
universe@550:         e->key.data = kd;
universe@550:         e->key.len = key.len;
universe@550:         e->key.hash = hash;
universe@550: 
universe@550:         // insert the element into the linked list
universe@550:         if (prev == NULL) {
universe@550:             hash_map->buckets[slot] = e;
universe@550:         } else {
universe@550:             prev->next = e;
universe@550:         }
universe@550:         e->next = elm;
universe@550: 
universe@550:         // increase the size
universe@550:         map->size++;
universe@550:     } else {
universe@550:         // (elem != NULL && elem->key.hash == hash) - overwrite value of existing element
universe@550:         elm->data = value;
universe@550:     }
universe@550: 
universe@550:     return 0;
universe@550: }
universe@549: 
universe@551: static void cx_hash_map_unlink(
universe@551:         struct cx_hash_map_s *hash_map,
universe@551:         size_t slot,
universe@551:         struct cx_hash_map_element_s *prev,
universe@551:         struct cx_hash_map_element_s *elm
universe@551: ) {
universe@551:     // unlink
universe@551:     if (prev == NULL) {
universe@551:         hash_map->buckets[slot] = elm->next;
universe@551:     } else {
universe@551:         prev->next = elm->next;
universe@551:     }
universe@551:     // free element
universe@551:     cxFree(hash_map->base.allocator, elm->key.data);
universe@551:     cxFree(hash_map->base.allocator, elm);
universe@551:     // decrease size
universe@551:     hash_map->base.size--;
universe@551: }
universe@551: 
universe@549: /**
universe@550:  * Helper function to avoid code duplication.
universe@549:  *
universe@550:  * @param map the map
universe@550:  * @param key the key to look up
universe@550:  * @param remove flag indicating whether the looked up entry shall be removed
universe@550:  * @return the value corresponding to the key or \c NULL
universe@549:  */
universe@550: static void *cx_hash_map_get_remove(
universe@550:         CxMap *map,
universe@550:         CxDataPtr key,
universe@550:         bool remove
universe@550: ) {
universe@550:     struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map;
universe@550: 
universe@550:     unsigned hash = cx_hash_map_murmur(key.data, key.len);
universe@550: 
universe@550:     size_t slot = hash % hash_map->bucket_count;
universe@550:     struct cx_hash_map_element_s *elm = hash_map->buckets[slot];
universe@550:     struct cx_hash_map_element_s *prev = NULL;
universe@550:     while (elm && elm->key.hash <= hash) {
universe@550:         if (elm->key.hash == hash && elm->key.len == key.len) {
universe@550:             if (memcmp(elm->key.data, key.data, key.len) == 0) {
universe@550:                 void *data = elm->data;
universe@550:                 if (remove) {
universe@551:                     cx_hash_map_unlink(hash_map, slot, prev, elm);
universe@550:                 }
universe@550:                 return data;
universe@550:             }
universe@550:         }
universe@550:         prev = elm;
universe@550:         elm = prev->next;
universe@550:     }
universe@550: 
universe@550:     return NULL;
universe@549: }
universe@550: 
universe@550: static void *cx_hash_map_get(
universe@550:         CxMap const *map,
universe@550:         CxDataPtr key
universe@550: ) {
universe@550:     // we can safely cast, because we know when remove=false, the map stays untouched
universe@550:     return cx_hash_map_get_remove((CxMap *) map, key, false);
universe@550: }
universe@550: 
universe@550: static void *cx_hash_map_remove(
universe@550:         CxMap *map,
universe@550:         CxDataPtr key
universe@550: ) {
universe@550:     return cx_hash_map_get_remove(map, key, true);
universe@550: }
universe@550: 
universe@551: static void *cx_hash_map_iter_current_entry(CxIterator const *iter) {
universe@551:     // struct has to have a compatible signature
universe@551:     struct cx_map_entry_s *entry = (struct cx_map_entry_s *) &(iter->kv_data);
universe@551:     return entry;
universe@551: }
universe@551: 
universe@551: static void *cx_hash_map_iter_current_key(CxIterator const *iter) {
universe@551:     struct cx_hash_map_element_s *elm = iter->elem_handle;
universe@551:     return &elm->key;
universe@551: }
universe@551: 
universe@551: static void *cx_hash_map_iter_current_value(CxIterator const *iter) {
universe@551:     struct cx_hash_map_element_s *elm = iter->elem_handle;
universe@551:     // TODO: return a pointer to data if this map is storing copies
universe@551:     return elm->data;
universe@551: }
universe@551: 
universe@551: static bool cx_hash_map_iter_valid(CxIterator const *iter) {
universe@551:     return iter->elem_handle != NULL;
universe@551: }
universe@551: 
universe@551: static void cx_hash_map_iter_next(CxIterator *iter) {
universe@551:     struct cx_hash_map_s *map = iter->src_handle;
universe@551:     struct cx_hash_map_element_s *elm = iter->elem_handle;
universe@551: 
universe@551:     // remove current element, if asked
universe@551:     if (iter->remove) {
universe@551:         // clear the flag
universe@551:         iter->remove = false;
universe@551: 
universe@551:         // determine the next element
universe@551:         struct cx_hash_map_element_s *next = elm->next;
universe@551: 
universe@551:         // search the previous element
universe@551:         struct cx_hash_map_element_s *prev = NULL;
universe@551:         if (map->buckets[iter->slot] != elm) {
universe@551:             prev = map->buckets[iter->slot];
universe@551:             while (prev->next != elm) {
universe@551:                 prev = prev->next;
universe@551:             }
universe@551:         }
universe@551: 
universe@551:         // unlink
universe@551:         cx_hash_map_unlink(map, iter->slot, prev, elm);
universe@551: 
universe@551:         // advance
universe@551:         elm = next;
universe@551:     } else {
universe@551:         // just advance
universe@551:         elm = elm->next;
universe@551:     }
universe@551: 
universe@551:     // do we leave the bucket?
universe@551:     if (elm == NULL) {
universe@551:         // search the next bucket
universe@551:         for (; elm == NULL && iter->slot < map->bucket_count; iter->slot++) {
universe@551:             elm = map->buckets[iter->slot];
universe@551:         }
universe@551:     }
universe@551: 
universe@551:     // advance the index in any case
universe@551:     iter->index++;
universe@551: 
universe@551:     // fill the struct with the current element
universe@551:     iter->elem_handle = elm;
universe@551:     if (elm == NULL) {
universe@551:         iter->kv_data.key = NULL;
universe@551:         iter->kv_data.value = NULL;
universe@551:     } else {
universe@551:         iter->kv_data.key = &elm->key;
universe@551:         // TODO: pointer to data if this map is storing copies
universe@551:         iter->kv_data.value = elm->data;
universe@551:     }
universe@551: }
universe@551: 
universe@551: static CxIterator cx_hash_map_iterator(CxMap *map) {
universe@550:     CxIterator iter;
universe@551: 
universe@551:     iter.src_handle = map;
universe@551:     iter.valid = cx_hash_map_iter_valid;
universe@551:     iter.next = cx_hash_map_iter_next;
universe@551:     iter.current = cx_hash_map_iter_current_entry;
universe@551: 
universe@551:     iter.slot = 0;
universe@551:     iter.index = 0;
universe@551:     iter.remove = false;
universe@551: 
universe@551:     if (map->size > 0) {
universe@551:         struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map;
universe@551:         struct cx_hash_map_element_s *elem = NULL;
universe@551:         for (; elem == NULL; iter.slot++) {
universe@551:             elem = hash_map->buckets[iter.slot];
universe@551:         }
universe@551:         iter.elem_handle = elem;
universe@551:         iter.kv_data.key = NULL;
universe@551:         iter.kv_data.value = NULL;
universe@551:     } else {
universe@551:         iter.elem_handle = NULL;
universe@551:         iter.kv_data.key = NULL;
universe@551:         iter.kv_data.value = NULL;
universe@551:     }
universe@551: 
universe@550:     return iter;
universe@550: }
universe@550: 
universe@551: static CxIterator cx_hash_map_iterator_keys(CxMap *map) {
universe@551:     CxIterator iter = cx_hash_map_iterator(map);
universe@551:     iter.current = cx_hash_map_iter_current_key;
universe@550:     return iter;
universe@550: }
universe@550: 
universe@551: static CxIterator cx_hash_map_iterator_values(CxMap *map) {
universe@551:     CxIterator iter = cx_hash_map_iterator(map);
universe@551:     iter.current = cx_hash_map_iter_current_value;
universe@550:     return iter;
universe@550: }
universe@550: 
universe@550: static cx_map_class cx_hash_map_class = {
universe@550:         cx_hash_map_destructor,
universe@550:         cx_hash_map_clear,
universe@550:         cx_hash_map_put,
universe@550:         cx_hash_map_get,
universe@550:         cx_hash_map_remove,
universe@550:         cx_hash_map_iterator,
universe@550:         cx_hash_map_iterator_keys,
universe@550:         cx_hash_map_iterator_values,
universe@550: };
universe@550: 
universe@550: CxMap *cxHashMapCreate(
universe@550:         CxAllocator *allocator,
universe@550:         size_t buckets
universe@550: ) {
universe@550:     if (buckets == 0) {
universe@550:         // implementation defined default
universe@550:         buckets = 16;
universe@550:     }
universe@550: 
universe@550:     struct cx_hash_map_s *map = cxMalloc(allocator, sizeof(struct cx_hash_map_s));
universe@550:     if (map == NULL) return NULL;
universe@550: 
universe@550:     // initialize hash map members
universe@550:     map->bucket_count = buckets;
universe@550:     map->buckets = cxCalloc(allocator, buckets, sizeof(struct cx_hash_map_element_s *));
universe@550:     if (map->buckets == NULL) {
universe@550:         cxFree(allocator, map);
universe@550:         return NULL;
universe@550:     }
universe@550: 
universe@550:     // initialize base members
universe@550:     map->base.cl = &cx_hash_map_class;
universe@550:     map->base.allocator = allocator;
universe@550:     map->base.size = 0;
universe@550: 
universe@550:     return (CxMap *) map;
universe@550: }