Sun, 15 Sep 2024 09:47:36 +0200
improve sessionhandler trace logging
/* * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS HEADER. * * Copyright 2021 Mike Becker, Olaf Wintermann All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. */ #include "cx/hash_map.h" #include "cx/utils.h" #include <string.h> #include <assert.h> struct cx_hash_map_element_s { /** A pointer to the next element in the current bucket. */ struct cx_hash_map_element_s *next; /** The corresponding key. */ CxHashKey key; /** The value data. */ char data[]; }; static void cx_hash_map_clear(struct cx_map_s *map) { struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map; cx_for_n(i, hash_map->bucket_count) { struct cx_hash_map_element_s *elem = hash_map->buckets[i]; if (elem != NULL) { do { struct cx_hash_map_element_s *next = elem->next; // invoke the destructor cx_invoke_destructor(map, elem->data); // free the key data cxFree(map->allocator, (void *) elem->key.data); // free the node cxFree(map->allocator, elem); // proceed elem = next; } while (elem != NULL); // do not leave a dangling pointer hash_map->buckets[i] = NULL; } } map->size = 0; } static void cx_hash_map_destructor(struct cx_map_s *map) { struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map; // free the buckets cx_hash_map_clear(map); cxFree(map->allocator, hash_map->buckets); // free the map structure cxFree(map->allocator, map); } static int cx_hash_map_put( CxMap *map, CxHashKey key, void *value ) { struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map; CxAllocator const *allocator = map->allocator; unsigned hash = key.hash; if (hash == 0) { cx_hash_murmur(&key); hash = key.hash; } size_t slot = hash % hash_map->bucket_count; struct cx_hash_map_element_s *elm = hash_map->buckets[slot]; struct cx_hash_map_element_s *prev = NULL; while (elm != NULL && elm->key.hash < hash) { prev = elm; elm = elm->next; } if (elm != NULL && elm->key.hash == hash && elm->key.len == key.len && memcmp(elm->key.data, key.data, key.len) == 0) { // overwrite existing element if (map->store_pointer) { memcpy(elm->data, &value, sizeof(void *)); } else { memcpy(elm->data, value, map->item_size); } } else { // allocate new element struct cx_hash_map_element_s *e = cxMalloc( allocator, sizeof(struct cx_hash_map_element_s) + map->item_size ); if (e == NULL) { return -1; } // write the value if (map->store_pointer) { memcpy(e->data, &value, sizeof(void *)); } else { memcpy(e->data, value, map->item_size); } // copy the key void *kd = cxMalloc(allocator, key.len); if (kd == NULL) { return -1; } memcpy(kd, key.data, key.len); e->key.data = kd; e->key.len = key.len; e->key.hash = hash; // insert the element into the linked list if (prev == NULL) { hash_map->buckets[slot] = e; } else { prev->next = e; } e->next = elm; // increase the size map->size++; } return 0; } static void cx_hash_map_unlink( struct cx_hash_map_s *hash_map, size_t slot, struct cx_hash_map_element_s *prev, struct cx_hash_map_element_s *elm ) { // unlink if (prev == NULL) { hash_map->buckets[slot] = elm->next; } else { prev->next = elm->next; } // free element cxFree(hash_map->base.allocator, (void *) elm->key.data); cxFree(hash_map->base.allocator, elm); // decrease size hash_map->base.size--; } /** * Helper function to avoid code duplication. * * @param map the map * @param key the key to look up * @param remove flag indicating whether the looked up entry shall be removed * @param destroy flag indicating whether the destructor shall be invoked * @return a pointer to the value corresponding to the key or \c NULL */ static void *cx_hash_map_get_remove( CxMap *map, CxHashKey key, bool remove, bool destroy ) { struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map; unsigned hash = key.hash; if (hash == 0) { cx_hash_murmur(&key); hash = key.hash; } size_t slot = hash % hash_map->bucket_count; struct cx_hash_map_element_s *elm = hash_map->buckets[slot]; struct cx_hash_map_element_s *prev = NULL; while (elm && elm->key.hash <= hash) { if (elm->key.hash == hash && elm->key.len == key.len) { if (memcmp(elm->key.data, key.data, key.len) == 0) { void *data = NULL; if (destroy) { cx_invoke_destructor(map, elm->data); } else { if (map->store_pointer) { data = *(void **) elm->data; } else { data = elm->data; } } if (remove) { cx_hash_map_unlink(hash_map, slot, prev, elm); } return data; } } prev = elm; elm = prev->next; } return NULL; } static void *cx_hash_map_get( CxMap const *map, CxHashKey key ) { // we can safely cast, because we know the map stays untouched return cx_hash_map_get_remove((CxMap *) map, key, false, false); } static void *cx_hash_map_remove( CxMap *map, CxHashKey key, bool destroy ) { return cx_hash_map_get_remove(map, key, true, destroy); } static void *cx_hash_map_iter_current_entry(void const *it) { struct cx_iterator_s const *iter = it; // struct has to have a compatible signature return (struct cx_map_entry_s *) &(iter->kv_data); } static void *cx_hash_map_iter_current_key(void const *it) { struct cx_iterator_s const *iter = it; struct cx_hash_map_element_s *elm = iter->elem_handle; return &elm->key; } static void *cx_hash_map_iter_current_value(void const *it) { struct cx_iterator_s const *iter = it; struct cx_hash_map_s const *map = iter->src_handle; struct cx_hash_map_element_s *elm = iter->elem_handle; if (map->base.store_pointer) { return *(void **) elm->data; } else { return elm->data; } } static bool cx_hash_map_iter_valid(void const *it) { struct cx_iterator_s const *iter = it; return iter->elem_handle != NULL; } static void cx_hash_map_iter_next(void *it) { struct cx_iterator_s *iter = it; struct cx_hash_map_element_s *elm = iter->elem_handle; // remove current element, if asked if (iter->base.remove) { // obtain mutable pointer to the map struct cx_mut_iterator_s *miter = it; struct cx_hash_map_s *map = miter->src_handle; // clear the flag iter->base.remove = false; // determine the next element struct cx_hash_map_element_s *next = elm->next; // search the previous element struct cx_hash_map_element_s *prev = NULL; if (map->buckets[iter->slot] != elm) { prev = map->buckets[iter->slot]; while (prev->next != elm) { prev = prev->next; } } // destroy cx_invoke_destructor((struct cx_map_s *) map, elm->data); // unlink cx_hash_map_unlink(map, iter->slot, prev, elm); // advance elm = next; } else { // just advance elm = elm->next; iter->index++; } // search the next bucket, if required struct cx_hash_map_s const *map = iter->src_handle; while (elm == NULL && ++iter->slot < map->bucket_count) { elm = map->buckets[iter->slot]; } // fill the struct with the next element iter->elem_handle = elm; if (elm == NULL) { iter->kv_data.key = NULL; iter->kv_data.value = NULL; } else { iter->kv_data.key = &elm->key; if (map->base.store_pointer) { iter->kv_data.value = *(void **) elm->data; } else { iter->kv_data.value = elm->data; } } } static bool cx_hash_map_iter_flag_rm(void *it) { struct cx_iterator_base_s *iter = it; if (iter->mutating) { iter->remove = true; return true; } else { return false; } } static CxIterator cx_hash_map_iterator( CxMap const *map, enum cx_map_iterator_type type ) { CxIterator iter; iter.src_handle = map; iter.base.valid = cx_hash_map_iter_valid; iter.base.next = cx_hash_map_iter_next; switch (type) { case CX_MAP_ITERATOR_PAIRS: iter.base.current = cx_hash_map_iter_current_entry; break; case CX_MAP_ITERATOR_KEYS: iter.base.current = cx_hash_map_iter_current_key; break; case CX_MAP_ITERATOR_VALUES: iter.base.current = cx_hash_map_iter_current_value; break; default: assert(false); } iter.base.flag_removal = cx_hash_map_iter_flag_rm; iter.base.remove = false; iter.base.mutating = false; iter.slot = 0; iter.index = 0; if (map->size > 0) { struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map; struct cx_hash_map_element_s *elm = hash_map->buckets[0]; while (elm == NULL) { elm = hash_map->buckets[++iter.slot]; } iter.elem_handle = elm; iter.kv_data.key = &elm->key; if (map->store_pointer) { iter.kv_data.value = *(void **) elm->data; } else { iter.kv_data.value = elm->data; } } else { iter.elem_handle = NULL; iter.kv_data.key = NULL; iter.kv_data.value = NULL; } return iter; } static cx_map_class cx_hash_map_class = { cx_hash_map_destructor, cx_hash_map_clear, cx_hash_map_put, cx_hash_map_get, cx_hash_map_remove, cx_hash_map_iterator, }; CxMap *cxHashMapCreate( CxAllocator const *allocator, size_t itemsize, size_t buckets ) { if (buckets == 0) { // implementation defined default buckets = 16; } struct cx_hash_map_s *map = cxCalloc(allocator, 1, sizeof(struct cx_hash_map_s)); if (map == NULL) return NULL; // initialize hash map members map->bucket_count = buckets; map->buckets = cxCalloc(allocator, buckets, sizeof(struct cx_hash_map_element_s *)); if (map->buckets == NULL) { cxFree(allocator, map); return NULL; } // initialize base members map->base.cl = &cx_hash_map_class; map->base.allocator = allocator; if (itemsize > 0) { map->base.store_pointer = false; map->base.item_size = itemsize; } else { map->base.store_pointer = true; map->base.item_size = sizeof(void *); } return (CxMap *) map; } int cxMapRehash(CxMap *map) { struct cx_hash_map_s *hash_map = (struct cx_hash_map_s *) map; if (map->size > ((hash_map->bucket_count * 3) >> 2)) { size_t new_bucket_count = (map->size * 5) >> 1; struct cx_hash_map_element_s **new_buckets = cxCalloc( map->allocator, new_bucket_count, sizeof(struct cx_hash_map_element_s *) ); if (new_buckets == NULL) { return 1; } // iterate through the elements and assign them to their new slots cx_for_n(slot, hash_map->bucket_count) { struct cx_hash_map_element_s *elm = hash_map->buckets[slot]; while (elm != NULL) { struct cx_hash_map_element_s *next = elm->next; size_t new_slot = elm->key.hash % new_bucket_count; // find position where to insert struct cx_hash_map_element_s *bucket_next = new_buckets[new_slot]; struct cx_hash_map_element_s *bucket_prev = NULL; while (bucket_next != NULL && bucket_next->key.hash < elm->key.hash) { bucket_prev = bucket_next; bucket_next = bucket_next->next; } // insert if (bucket_prev == NULL) { elm->next = new_buckets[new_slot]; new_buckets[new_slot] = elm; } else { bucket_prev->next = elm; elm->next = bucket_next; } // advance elm = next; } } // assign result to the map hash_map->bucket_count = new_bucket_count; cxFree(map->allocator, hash_map->buckets); hash_map->buckets = new_buckets; } return 0; }