1.16.0

Added:
- HostInfo.ScyllaNrShards function

1.15.0

Added:
- ScyllaToken and ScyllaShardID to Query/BatchObserver

1.14.0

Added:
- ClusterConfig.MaxRequestsPerConn to control maximum streams (concurrency) used per connection.
- Support for rate limit error CQL extension.
- ClusterConfig.HeavyLoadedConnectionThreshold and ClusterConfig.HeavyLoadedSwitchConnectionPercentage to
  control heavy loaded optimization.

Fixed:
- Heavy loaded optimization was never triggering.

Release 1.13.0

Added:
- rack penalty to TokenAwareLatencyHostPolicy

Release 1.12.0

Added:

- Added a RackAwareRoundRobinPolicy that attempts to keep client->server traffic in the same rack when possible.

Changed:

- GetCustomPayload now returns nil instead of panicking in case of query error.
- Merge shard awareness with original behaviour.

Fixed:

- Nil pointer dereference in events.go when handling node removal.
- Reading peers from DataStax Enterprise clusters.
- Unmarshaling maps did not pre-allocate the map.
- Use ScyllaShardAwareDialer as a default dialer.
- Use ConnectionConfig values when establishing session to initial contact point.
- Run go vet and tests in all subpackages in CI.
- When host information was missing, driver used resolved IP address as
  TLS.ServerName. Instead it should connect to Server specified in
  ConnectionConfig and use NodeDomain as SNI.

Release 1.11.0

Added:

- The driver now supports connecting to nodes through proxy.

Changed:

- The driver now uses host_id to identify nodes instead of IP address.
- Pick least busy connection instead of the target shard if the cluster is under heavy load.
- Return an error if policies are reused between sessions.

Fixed:

- DESC constant changed to ColumnOrder instead of bool.
- Events are now ignored if they are disabled.
- HostFilter is now applied for control connection too.
- Race condition in HostInfo.HostnameAndPort.
- Panic in unmarshalUDT if not passed a pointer.
- Data corruption bug when marshaling UDTs with missing fields.
- Marshaling of maps with missing fields failed with error instead of marshaling nulls.

Release 1.10.0

Added:

- Expose predefined raw protocol error codes
- Add SingleHostQueryExecutor
- Document that NumConns has no effect when working with Scylla
- Add AllowedAuthenticators to PasswordAuthenticator

Release 1.9.0

Added:

- Documentation about host policies
- lz4 compressor module
- Method to allow inspecting host state of TokenAwareLatencyHostPolicy
- Tracer support for batches

Fixed:

- Deadlock during Session.Close

Release 1.8.1

Fixed:

- goroutine could stuck forever in closeWithError
- connection mutex could be held relatively long
  in closeWithError

Release 1.8.0

Added:

- ObservedQuery/ObservedBatch now contain bound query values
- com.scylladb.auth.SaslauthdAuthenticator and
  com.scylladb.auth.TransitionalAuthenticator are now approved
  authenticators

Fixed:

- Reconnecting to cluster after all connections are severed