README.md

Bin-Stash

Library for caching data:

• In memory (using Caffeine) or with a centralized cache (Redis using Lettuce Reactive) making a single tier cache.
• Or using both as a two tier stage cache.

Getting Started

1. Installation process

For local cache only

dependencies {
    implementation 'com.github.bancolombia:bin-stash-local:<version>'
}

For a centralized (redis) cache only

dependencies {
    implementation 'com.github.bancolombia:bin-stash-centralized:<version>'
}

For an hybrid (local and centralized) cache

dependencies {
    implementation 'com.github.bancolombia:bin-stash-hybrid:<version>'
}

2. Supported Configuration

stash:
  memory:
    expireTime: 60 # 1 minute
    maxSize: 10_000
  redis:
    host: myredis.host
    # Only when connecting to a master/replica
    #replicas: replica.myredis.host, replica2.myredis.host
    port: 6379
    database: 0
    # for rbac access
    username: myuser
    password: mypwd
    useSsl: true
    expireTime: 3600 # 1 hour

Configuration	Description
stash.memory.maxSize	maximum allowed bytes to store in memory cache
stash.memory.expireTime	set maximum time to hold keys in cache (in seconds). If not defined, a value of 300 seconds is used as default. Note that save() methods that receive a TTL argument, will ignore such value if its greater than expireTime.
stash.redis.host	host to connect to (when connecting to a master-replica cluster this is the master host)
stash.redis.replicas	host names of replicas, comma separated. (when connecting to a master-replica cluster)
stash.redis.port	redis port (when connecting to master-replicas will use same port for all hosts)
stash.redis.database	database number to use on single node (0 default)
stash.redis.username	username (when using RBAC)
stash.redis.password	password when using AUTH or RBAC
stash.redis.useSsl	true or false. Indicates the client to connect to redis via secure connection
stash.redis.expireTime	default TTL time (in seconds) to hold every key stored in redis. If this parameter is not defined a default value of 300 seconds is used. This value can be overriden for an specific key, with the TTL argument in the save() methods.

3. Usage

   3.1. For Local Cache

   @Configuration
   public class LocalExampleConfiguration {
   
       @Bean
       public ObjectMapper objectMapper() {
           return new ObjectMapper();
       }
   
       @Bean
       public ObjectCache<Person> objectCache(LocalCacheFactory<Person> localCacheFactory) {
           return localCacheFactory.newObjectCache();
       }
   }

   3.2. For Centralized Cache

   @Configuration
   public class CentralizedExampleConfiguration {
   
       [...]
   
       @Bean
       public ObjectCache<Person> objectCache(CentralizedCacheFactory<> centralizedCacheFactory) {
           return centralizedCacheFactory.newObjectCache();
       }
   }

   3.3. For Hybrid Cache

   @Configuration
   public class HybridExampleConfiguration {
   
       [...]
   
       @Bean
       public List<SyncRule> cacheSyncRules() {
           // Just one rule in this demo. Push local cache key-values to centralized cache, disregarding syncType 
           // (UPSTREAM or DOWNSTREAM) and pull centralized cache key-values to local cache when affected 
           // key (keyArg) is ANY string.
           SyncRule simpleSyncRule = (keyArg, syncType) -> true;
           return Collections.singletonList(simpleSyncRule);
       }  
   
       @Bean
       public ObjectCache<Person> objectCache(HybridCacheFactory<Person> cacheFactory,
                                             List<SyncRule> cacheSyncRules) {
          return cacheFactory.newObjectCache(cacheSyncRules);
       }
   }

You can now use ObjectCache<> in your app:

@Component
public class PersonCachedHandler {

    public final DummyRepo dummyRepo;
    public final ObjectCache<Person> cache;

    @Autowired
    public PersonCachedHandler(DummyRepo dummyRepo,
                               ObjectCache<Person> cache) {
        this.dummyRepo = dummyRepo;
        this.cache = cache;
    }

    public Mono<ServerResponse> fetchWithCache(ServerRequest request) {

        String key = Optional.of(request.pathVariable("id"))
                .orElse("JhonSmith1");

        // Usage with reactor-extra CacheMono
        Mono<Person> cached = CacheMono
            .lookup(k -> cache.get(k, Person.class) // Get from cache
                     .map(Signal::next), key)
                .onCacheMissResume(dummyRepo.findByName(key)) // Get from some db repo
                .andWriteWith((k, sig) ->
                        cache.save(k, sig.get()).then() // Save to cache
                );

        return cached
                .flatMap(person -> ServerResponse.ok().contentType(MediaType.APPLICATION_JSON)
                    .body(BodyInserters.fromValue(person))
        );
    }
}

3. Latest releases
4. API references

How two tier cache works

When using the cache in hybrid mode (two tier cache), requests works as described:

GET Operation

1. When a lookup misses local cache, and upstream sync is allowed, bin-stash tries to lookup in the centralized cache.
2. If the lookup operation in the centralized cache hits a key, then return the value. Also if downtream sync is allowed, the key-value is replicated in local cache.

SAVE operation

1. Writes are performed in local cache, and if upstream sync is allowed, bin-stash tries to write key-value in the centralized cache, given this key doesn't previously exists.

EVICT operation

1. Evict(key) is performed in local cache, and if upstream sync is allowed, bin-stash tries to evict key-value in the centralized cache.
2. EvictAll operation it's performed on local and is not syncronized upstream.

Sync Rules

DoubleTierCache uses a collection of SyncRule to determine if upstream/downstream sync takes place. This functional interface takes keyname and SyncType (UPSTREAM or DOWNSTREAM) as arguments, and should return a boolean.

Build and Test

TODO: Describe and show how to build your code and run the tests.

Contribute

TODO: Explain how other users and developers can contribute to make your code better.