4.5.4

Bug Fix

• Fix leftover default connection pool construction in
  blocking cluster connection pool

Chores

• Reduce excessive matrix in default CI
• Add scheduled compatibility CI run
• Cleanup unnecessary asyncio markers in tests
• Refactor readonly command detection to use
  command flags from redis documentation
• Issue warning if Redis.select is
  called directly

4.5.3

Bug Fix

• Ensure default cluster connection pools are not recreated upon access. #92

4.5.2

Bug Fix

• Implicitly initialize cluster connection pool when pubsub subscribe is called (#88)
• Fix handling of sharded pubsub unsubscribe message
• Fix unsubscribe all for sharded pubsub

Compatibility

• Improve surfacing underlying errors when initializing cluster (#89)

4.5.1

Bug Fix

• Fix context leak when commands issued within ensure_replication and ignore_replies context managers fail

Recipes

• Fix LUA lock recipe to work with clusters with no replicas.
• Ensure LUA lock recipe waits on replication of lock to n/2 replicas if replicas exist in the cluster

4.5.0

Compatibility

• Bring back python 3.7 support

4.4.0

Breaking changes

• Default nodemanager_follow_cluster to True

Deprecations

• Deprecate readonly constructor argument in
  cluster client in favor of read_from_replicas

• Remove invalid property setter for noreply mode

Bug Fixes

• Fix incorrect behavior of ignore_replies context manager
  as it was not actually setting CLIENT REPLY and simply
  discarding connections
• Ensure fetching a random connection doesn't deplete the
  node list in the connection pool
• Ensure connection pools are disconnected on finalization
  to avoid leaking connections

4.3.1

Bug Fix

• Fix incorrect calculation of per node connection pool size in BlockingClusterConnectionPool when readonly=False
• Ensure max_connection is at least equal to the number of nodes in the cluster and raise a warning when it is not

4.3.0

Features

• Allow passing connection_pool_cls to Redis and RedisCluster constructors to pick
  the connection pool implementation during client construction

Breaking Changes

• Raise coredis.exceptions.ConnectionError instead of coredis.exceptions.RedisClusterError
  when a connection can't be acquired due to max_connections being hit.

4.2.1

Compatibility

• Add support and test coverage for PyPy 3.8.

Bug Fix

• Ensure coredis.RedisCluster.ensure_replication can be used
  with ~coredis.RedisCluster.ensure_replication.replicas <
  total number of replicas

4.2.0

Bug Fix

• Fix routing of coredis.Redis.script_kill and coredis.Redis.function_kill to only route to primaries
• Ensure all arguments expecting collections consistently use coredis.typing.Parameters

Chores

• Fix ordering of keyword arguments of coredis.Redis.set to be consistent with command documentation
• Improve documentation regarding request routing and repsonse merging for cluster multi node and multi shard commands
• Sort all literal annotations