UPCGBlueprintElement是PCGGraph中自定义节点的基类，但官方目前还没有给出详细的文档，所以从源代码里找点答案。
UPCGBlueprintElement是PCGGraph中所有自定义节点的基类，有如下几个可供蓝图覆盖的函数：
其中 Excute、Excute with Context、NodeTitleOverride、NodeColorOverride、NodeTypeOverride是覆盖后被自动触发的；而剩余的几个“Loop Body函数”则需要在蓝图中显示调用对应的“Loop函数”才可以执行循环，比如：要调用PointLoop才可以执行Point Loop Body
这两个函数可以理解为自定义节点的“主函数”,区别就是一个有PCGContext输入另一个没有，ExecuteWithContext会调用Execute
Excute
UFUNCTION(BlueprintImplementableEvent, BlueprintCallable, Category = "PCG|Execution")
void Execute(const FPCGDataCollection& Input, FPCGDataCollection& Output);
Excute with Context
UFUNCTION(BlueprintNativeEvent, Category = "PCG|Execution")
void ExecuteWithContext(UPARAM(ref)FPCGContext& InContext, const FPCGDataCollection& Input, FPCGDataCollection& Output);
void UPCGBlueprintElement::ExecuteWithContext_Implementation(FPCGContext& InContext, const FPCGDataCollection& Input, FPCGDataCollection& Output)
{
Execute(Input, Output);
}
其实还有个C++ 函数ExecuteInternal，Excute with Context是由它调用的。
virtual bool ExecuteInternal(FPCGContext* Context) const override;
/** Finally, execute the actual blueprint */
Context->BlueprintElementInstance->ExecuteWithContext(*Context, Context->InputData, Context->OutputData);
Loop Body函数都是在Loop函数中定义的lambda表达式调用的。
Point Loop Body、Nested Loop Body、Iteration Loop Body通过FPCGAsync::AsyncPointProcessing调用；Variable Loop Body通过FPCGAsync::AsyncMultiPointProcessing调用PCGPointData类型，需要的话可以在调用Loop函数的时候传入，它们会直接传入Loop Body函数Point Loop Body需要通过PointLoop执行循环
Point Loop Body
PointLoop
FPCGAsync::AsyncPointProcessing(&InContext, InPoints.Num(), OutPoints, [this, &InContext, InData, OutData, &InPoints](int32 Index, FPCGPoint& OutPoint)
{
return PointLoopBody(InContext, InData, InPoints[Index], OutPoint, OutData->Metadata);
});
Variable Loop Body需要通过VariableLoop执行循环
VariableLoop
FPCGAsync::AsyncMultiPointProcessing(&InContext, InPoints.Num(), OutPoints, [this, &InContext, InData, OutData, &InPoints](int32 Index)
{
return VariableLoopBody(InContext, InData, InPoints[Index], OutData->Metadata);
});
}
FPCGAsync::AsyncPointProcessing和FPCGAsync::AsyncMultiPointProcessing都是异步的，它们的区别就在于处理输入:输出为1:1还是1:N的循环。Point Loop Body和VariableLoopBody,前者输入一个point返回一个point；后者输入一个point可以返回一个point的数组。比如：如果你想把一个point看作一块地基，想在它的垂直上方再生成一串点作为楼层，然后把这些点返回，那么你就需要用到VariableLoopBody。
Nested Loop Body需要通过NestedLoop执行循环
关键代码如下，但使用情景没想明白。
NestedLoop
const TArray<FPCGPoint>& InOuterPoints = InOuterData->GetPoints();
const TArray<FPCGPoint>& InInnerPoints = InInnerData->GetPoints();
TArray<FPCGPoint>& OutPoints = OutData->GetMutablePoints();
FPCGAsync::AsyncPointProcessing(&InContext, InOuterPoints.Num() * InInnerPoints.Num(), OutPoints, [this, &InContext, InOuterData, InInnerData, OutData, &InOuterPoints, &InInnerPoints](int32 Index, FPCGPoint& OutPoint)
{
return NestedLoopBody(InContext, InOuterData, InInnerData, InOuterPoints[Index / InInnerPoints.Num()], InInnerPoints[Index % InInnerPoints.Num()], OutPoint, OutData->Metadata);
});
Iteration Loop Body需要通过IterationLoop执行循环
和Point Loop Body一样，但它会提供一个NumIterations用于指定迭代次数
IterationLoop
FPCGAsync::AsyncPointProcessing(&InContext, NumIterations, OutPoints, [this, &InContext, InA, InB, OutData](int32 Index, FPCGPoint& OutPoint)
{
return IterationLoopBody(InContext, Index, InA, InB, OutPoint, OutData->Metadata);
});