Node Type Reference

Relation Nodes

Relation nodes control how child nodes are executed and how results are aggregated. Each type has both serial and parallel versions.

Serial Relation Nodes

Type	Execution	Child returns true	Child returns false	No children
AND	Sequential, short-circuits on false	All true -> true	Any false -> false (returns immediately)	none
ANY	Sequential, short-circuits on true	Any true -> true (returns immediately)	All false -> false	none
ALL	Executes all, no short-circuit	Has true and no false -> true	Any false -> false	none
NONE	Executes all	Always returns none	Always returns none	none
TRUE	Executes all	Always returns true	Always returns true	true

Return Value Notes

When all child nodes return none, AND / ANY / ALL all return none
TRUE is the only relation node that returns true even with no children
none means "does not participate in evaluation" and is not equivalent to true or false

Parallel Relation Nodes

Each serial relation node has a corresponding parallel version that submits child nodes to a thread pool or goroutine pool for concurrent execution:

Type	Concurrency Strategy
ParallelAnd	Executes all children concurrently. Returns early when any completes with false
ParallelAny	Executes all children concurrently. Returns early when any completes with true
ParallelAll	Executes all children concurrently. Waits for all to complete before returning
ParallelNone	Executes all children concurrently. Waits for all to complete, always returns none
ParallelTrue	Executes all children concurrently. Waits for all to complete, always returns true

Note

Child nodes of parallel nodes should not have data dependencies on each other. Each child node uses a shallow copy of the Roam; writing to the same key may cause race conditions.

Leaf Nodes

Leaf nodes are the terminal nodes that execute business logic.

Type	Return Value	Purpose	Typical Scenarios
Flow	true / false	Conditional checks that control flow direction	Amount validation, level checks, time filtering, permission checks
Result	true / false	Execute business operations and return results	Issue coupons, deduct inventory, send notifications, call external APIs
None	none (does not affect flow)	Auxiliary operations	Query user info, logging, data assembly, cache warming

Base Class Hierarchy

Leaf nodes provide three levels of base classes, progressively unwrapping the input parameter:

BaseLeaf (IceContext)
  +-- BaseLeafPack (IcePack)
        +-- BaseLeafRoam (IceRoam)  <-- most common

Language	Flow	Result	None
Java	`BaseLeafRoamFlow`	`BaseLeafRoamResult`	`BaseLeafRoamNone`
Go	`DoRoamFlow(ctx, roam)`	`DoRoamResult(ctx, roam)`	`DoRoamNone(ctx, roam)`
Python	`do_roam_flow(roam)`	`do_roam_result(roam)`	`do_roam_none(roam)`

Common Node Configuration

All nodes (both relation and leaf) share the following configurations:

Configuration	Description	Default
Time Window	Effective time range for the node. Nodes outside the window return none and are treated as non-existent	Unrestricted
Inverse	Inverts true to false and vice versa. none is not affected	false
Forward Node	When the forward node returns false, the main node is not executed and directly returns false. Semantically equivalent to an AND connection	None
Error Handling	Behavior when a node throws an error: SHUT_DOWN (terminate flow) or return a specified state to continue	SHUT_DOWN
Debug Flag	Whether to record this node's execution information in processInfo	false

Execution Flow

Each node's process() method executes in the following order:

Time window check -> Returns none if outside the window
Forward node execution -> Refuses execution if forward node returns false
Self logic execution -> Relation node traverses children / leaf node executes business logic
Inverse processing -> If inverse is configured, flips true/false
Error handling -> Handles exceptions according to the configured strategy