Handle-with-cache.c 【2025】

pthread_mutex_lock(&cache_lock); // Double-check: another thread might have inserted it while we were loading entry = g_hash_table_lookup(handle_cache, &user_id); if (entry) { // Discard our loaded profile and use the cached one free_user_profile(profile); entry->ref_count++; pthread_mutex_unlock(&cache_lock); return entry->profile; }

// Store in cache (use user_id as key) int *key = malloc(sizeof(int)); *key = user_id; g_hash_table_insert(handle_cache, key, new_entry);

The module handle-with-cache.c exemplifies a classic design pattern: the . A "handle" is an opaque pointer or identifier to a resource, and the cache stores recently accessed handles to avoid redundant initialization or I/O operations. handle-with-cache.c

A common optimization is or using a per-key mutex:

// Background thread or called periodically void evict_stale_handles(int max_age_seconds, int max_size) { pthread_mutex_lock(&cache_lock); time_t now = time(NULL); GList *to_remove = NULL; The cache is transparent to the caller

static UserProfile* load_user_profile_from_disk(int user_id) { // Simulate expensive I/O printf("Loading user %d from disk...\n", user_id); sleep(1); // Pretend this is slow UserProfile *profile = malloc(sizeof(UserProfile)); profile->user_id = user_id; profile->name = malloc(32); profile->email = malloc(64); sprintf(profile->name, "User_%d", user_id); sprintf(profile->email, "user%d@example.com", user_id); return profile; } This is the heart of the module. The cache is transparent to the caller.

A handle cache solves this by storing active handles in a key-value store after the first access. Subsequent requests bypass the expensive operation and return the cached handle directly. A well-written handle-with-cache.c typically contains four main sections: 1. The Handle and Cache Structures First, we define our handle type (opaque to the user) and the cache entry. A well-written handle-with-cache

// handle-with-cache.c #include <stdio.h> #include <stdlib.h> #include <string.h> #include <glib.h> // Using GLib's hash table for simplicity typedef struct { int user_id; char *name; char *email; // ... other data } UserProfile;

// Improved get_handle() with double-check UserProfile* get_user_profile_handle_safe(int user_id) { pthread_mutex_lock(&cache_lock); CacheEntry *entry = g_hash_table_lookup(handle_cache, &user_id); if (entry) { entry->ref_count++; pthread_mutex_unlock(&cache_lock); return entry->profile; } pthread_mutex_unlock(&cache_lock); // Load outside lock UserProfile *profile = load_user_profile_from_disk(user_id);

// Remove stale entries for (GList *l = to_remove; l; l = l->next) { int *key = l->data; CacheEntry *entry = g_hash_table_lookup(handle_cache, key); free(entry->profile->name); free(entry->profile->email); free(entry->profile); free(entry); g_hash_table_remove(handle_cache, key); free(key); } g_list_free(to_remove);

pthread_mutex_unlock(&cache_lock); } The cache_lock mutex protects the hash table, but note that get_handle() releases the lock during the actual load_user_profile_from_disk() call. This is crucial to avoid blocking all threads during I/O. However, it introduces a race condition where two threads might simultaneously miss the cache and both load the same resource.