Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.1.0]

Plug pipeline system — Phoenix-style middleware architecture for RPC calls
- EasyRpc.Context — context struct flowing through the pipeline (like Plug.Conn)
- EasyRpc.Plug — behaviour for custom middleware plugs
- EasyRpc.Pipeline — pipeline builder and executor with default pipeline
- EasyRpc.Plugs.NodeSelector — node selection plug
- EasyRpc.Plugs.Executor — :erpc.call plug
- EasyRpc.Plugs.Retry — retry logic plug
- EasyRpc.Plugs.Logger — structured logging plug
- EasyRpc.Plugs.ErrorHandler — result/error normalisation plug
- EasyRpc.Plugs.Telemetry — telemetry events plug (optional)
- EasyRpc.Plugs.Cache — ETS-based caching plug (example)
Custom pipeline DSL — pipeline do ... end section with plug entities
- When pipeline is defined, generated functions use Pipeline.run/4
- When no pipeline is defined, original RpcCall.execute behaviour is preserved
Added :telemetry as optional dependency
Added enable_logging option to config section for controlling detailed logging
Added nodes_provider option for dynamic node discovery via MFA
Added EasyRpc.NodeSelector.select_node_safe/2 — non-raising node selection, shared by RpcCall and the NodeSelector plug

Updated documentation to reflect Spark DSL syntax (replacing old defrpc macro references)
Improved error handling documentation with clearer examples
Per-function options passed to defrpc are now validated at compile time — unknown keys (e.g. a misspelled retry) raise EasyRpc.Error instead of being silently ignored; EasyRpc.RpcWrapper applies the same validation for consistent error messages
The default plug pipeline is pre-composed at compile time; running an RPC no longer rebuilds the closure chain per call
EasyRpc.Plugs.Telemetry gates on :telemetry availability at compile time, removing per-call rescue overhead when the optional dependency is absent

Fixed inconsistencies between README examples and actual Spark DSL syntax
Corrected documented pipeline plug option syntax — Spark rejects bare keyword options like plug MyCachePlug, ttl: 5_000; plug options must be passed as plug MyCachePlug, opts: [ttl: 5_000] (README and EasyRpc.Plugs.Cache examples updated)
Fixed failing doctest for EasyRpc.version/0 (pinned a stale version)
WrapperConfig.load_config!/2 and load_from_options!/1 now honour the :enable_logging option (previously silently dropped)
Dynamic node providers returning non-atom values now fail fast with a clear %EasyRpc.Error{} instead of reaching :erpc.call with an invalid node name
In safe mode (error_handling/retry), short-circuited results (e.g. cache hits) are wrapped consistently by ErrorHandler, so callers always see {:ok, result} | {:error, error} regardless of pipeline path
Halted contexts are no longer retried by EasyRpc.Plugs.Retry — halting means "stop the pipeline", so retrying only burned the sleep time

Migrated from macro-based DSL to Spark DSL for better extensibility
Added rpc_function entities with support for argument names (better IDE support)
Added private: true option to generate private functions (defp)
Added new_name option to create function aliases
Added comprehensive architecture documentation in guides/ARCHITECTURE.md
Added support for :infinity timeout