Change redis-cache.adoc document title from '==' to '=' to appear as section 10.15 instead of 10.14.1,
which currently places the 'Redis Cache' ref doc under 'Redis Scripting'.
Closes#2458
Original pull request #2459
For the Jedis-based implementations of bzpopmin and bzpopmax, a NPE is
thrown if the provided timeout elapses and the server responds with a
null array.
Closes#2324
We now use correctly null to indicate absent keys when using reactive MGET. Previously, we used an empty byte buffer that could be incorrectly translated to an empty string when using the string codec. An empty byte buffer can also be returned if the value length is zero leading to a state that doesn't allow distinguishing between absence and empty value.
Closes#2402
We now unsubscribe from Redis only once when closing a Subscription object to avoid duplicate traffic and unexpected Redis responses.
Jedis does not have a mechanism to consume the additional unsubscribe response which leaves protocol frames on the InputStream leading to a corrupt state.
Closes#2355
This commit makes sure to avoid typing of primitive wrapper types wrapped inside what is considered a jackson ReferenceType (such as Optional, AtomicReference,...).
Original Pull Request: #2364
We now ensure proper exception handling in connection close methods to avoid resource leaks. Also, exceptions during connection close are no longer thrown to ensure proper resource cleanup behavior and API design.
See #2356
We now ensure proper exception handling in close methods to avoid resource leaks. Also, exceptions during connection close are no longer thrown to ensure proper resource cleanup behavior and API design.
Closes#2356
We now support configuration of a bean reference to RedisMessageListenerContainer that should be used with `RedisKeyValueAdapter` for easier configuration of the listener container.
Closes#1827
Jedis Cluster sort now considers if the destination key is sharing the same slot as the source key to use same-slot sorting.
Additionally, sort results that do not map to the same slot replace the destination key with a list instead of checking the key type and appending results.
Closes#2341
We now consider the actual value type when determining whether a value should be considered an entity. Previously, we relied on PersistentProperty type information only which made it impossible to use generic-typed (or Object) properties for entities.
Closes#2349
LettuceConnection.switchToPubSub now correctly releases its underlying connection when switching to Pub/Sub. Also, we improved safeguards to avoid using closed connections.
Closes#2331
We now close the cluster command executor before cleaning up the connection pools so that we first release all held connections before pruning the connection pools.
Previously, the pools were pruned first leading to an attempt to return the connection held by the cluster command executor causing a PoolException.
Closes#2330
We now try to eagerly apply a custom converter when reading Redis properties in the MappingRedisConverter before considering type hints from the hash.
We require type hints to properly restore the target type for a hash entry as Redis data is all byte arrays, so a simple String needs a type hint before we can load it back into an e.g. Object-typed property. Subtypes of properties (such as ZoneRegion for ZoneId) are sometimes not associated with a converter as only the parent type (ZoneId) is associated with a converter. Trying to convert the value into the subtype directly through our ConversionService would fail in that case.
Closes#2307
