Orchid.Scheduler.Context (Orchid v0.6.1)

The state machine for workflow execution.

It tracks the progress of a Recipe, including which steps are pending, running, or completed, and holds the data (params) produced so far.

Data Structure Decisions

You might notice a mix of Lists and MapSets. This is intentional to balance deterministic execution with scheduling efficiency:

:pending_steps (List) - Kept as a list to preserve the definition order from the Recipe. When multiple steps are ready simultaneously, Orchid prefers to execute them in the order they were written. This ensures predictable behavior, especially for the Serial executor.
:running_steps (MapSet) - Used for O(1) lookups. The scheduler frequently checks member?/2 inside loops; a MapSet prevents this from becoming a performance bottleneck.
:available_keys (MapSet) - Optimized for Set Algebra. Dependency resolution relies heavily on subset checks (subset?/2), which MapSets handle efficiently.
:recipe (Struct) - The Recipe being executed.
:params (Map) - A map of parameters available in the current context.
:history (List) - A list of tuples recording the execution history of steps, including their indices and output keys.
:workflow_ctx (Struct) - The Orchid.WorkflowCtx struct associated with the execution.
:assigns (Map) - A map for storing additional context-specific data.

Summary

Types

param_map()

step_index()

t()

Types

param_map()

@type param_map() :: %{optional(atom() | binary()) => Orchid.Param.t()}

step_index()

@type step_index() :: non_neg_integer()

t()

@type t() :: %Orchid.Scheduler.Context{
  assigns: %{required(any()) => any()},
  available_keys: MapSet.t(Orchid.Step.io_key()),
  history: [{Orchid.Step.t(), MapSet.t(Orchid.Step.output_keys())}],
  params: param_map(),
  pending_steps: [{Orchid.Step.t(), step_index()}],
  recipe: Orchid.Recipe.t(),
  running_steps: MapSet.t(step_index()),
  workflow_ctx: Orchid.WorkflowCtx.t()
}