[5/5] Add Async MultiplexedClusterConnection #640
Commits on Aug 23, 2022
-
cluster_client: implement proper builder pattern
- implement proper builder pattern by adding `new` & `builder` methods to ClusterClient and `new` & `build` methods to ClusterClientBuilder - deprecate & redirect `open` methods for ClusterClient & ClusterClientBuilder - move build method to ClusterClientBuilder
-
cluster_client: use ClusterParams struct to pass params
- this is to simplify passing multiple params to & inside ClusterConnection impl
-
-
cluster: simplify TlsMode logic and add tls_insecure param
- use plain host:port node naming, even for TLS as the tls mode/insecure flag will always be the same for all connections
-
cluster: refactor connect & create_initial_connections
- move the connect method inside impl - create new Connection directly instead of creating a Client first - convert to methods & use required params from self instead of passing - set connections within create_initial_connections & call refresh_slots
-
cluster: keep all params in ClusterParams struct & use it to pass all…
… cluster related params - remove unnecessary RefCell from params - this is to simplify passing multiple params to & inside the impl
-
-
-
-
-
cluster: move slots response parsing logic to get_slots function
- rename get_slots to parse_slots_response - move coverage checks & Slot creation logic into this fn - move CLUSTER SLOTS query logic out of this fn - this to easily reuse this function in the Async impl
-
cluster: remove random connection logic
- since we only operate in 100% slot coverage scenarios, there is no point in maintaining random connection logic, which is supposed to help us find missing nodes in the cluster topology - this also allows us to remove RefCell around connections & slots
-
cluster: remove RefCell for connections/slot
- there is no point in using RefCell anymore
-
cluster: consolidate routing logic into Routable trait
- this is to reuse the routing logic within ClusterConnection and also reuse it in async impl
-
-
cluster: consolidate logic for create_initial_connections & refresh_s…
…lots - set write/read timeouts just after creating connection - consolidate logic for create_initial_connections & refresh_slots connection creating/updating part into refresh_connections
-
cluster: directly expose Slot fields instead of bare methods
- there is no point in having unnecessary indirection
-
cluster: merge send_all_commands & recv_all_commands
- there is no point in having 2 different functions
-
-
-
aio: fix ConnectionLike trait documentation & add supports_pipelining…
…/supports_transactions
-
sio: improve ConnectionLike trait documentation & add req_pipeline/su…
…pports_transactions
-
cluster: use common pipeline instead of separate ClusterPipeline
- remove ClusterPipeline & add re-exports for Pipeline/pipe - use `req_pipeline` instead of the packed version to avoid unpacking again for cluster - move cluster-specific pipeline checks into the new req_pipeline trait function
-
