StatelessWidget vs a function returning Widgets in terms of performance

问题

Is there any difference, performance wise, in using a StatelessWidget vs a function returning a Widget?

I'm well aware of at least the differences pointed out in this flutter's repo issue which don't have a relationship with performance.

The fact is that I have some colleagues claiming that functional widgets are worst in terms of performance but after reading a little bit about the subject I cannot find any conclusive piece of documentation that can give credit to that assertion so any kind of clarification regarding this matter will be very welcome!

As far as I can see the only difference between them would be in the case of using a const Widget, which seems that would avoid the rebuilding phase.

回答1:

First of all, I'd like to note that a package is available to make a StatelessWidget from a function: functional_widget

The gain is performance is not necessarily true. It depends on how you use your widgets, mostly how you use them to manage your state.

By default, classes may degrade performances when opposed to functions in an application that doesn't utilize their power.

The real question is: What is their power?

Simple: Classes can update independently from each other. Functions cannot

It is possible for classes to partially update the widget tree.

Consider a widget that rebuilds every frame and returns its child:

class InfiniteLoop extends StatefulWidget {
  const InfiniteLoop({Key key, this.child}) : super(key: key);
  final Widget child;
  @override
  _InfiniteLoopState createState() => _InfiniteLoopState();
}

class _InfiniteLoopState extends State<InfiniteLoop> {
  @override
  Widget build(BuildContext context) {
    WidgetsBinding.instance.addPostFrameCallback((_) => setState(() {}));

    return widget.child;
  }
}

Now if we wrap our whole application in that widget, what happens?

void main() => runApp(InfiniteLoop(child: MyApp()));

Nothing

Sure, you'll have one widget that rebuilds often in your tree. But in reality, the build method of MyApp will be called only once.

That's because Flutter is able to abort the tree rebuild when the instance of a widget doesn't change.

Classes can abuse of this optimization.

Using classes it is possible to cleverly split the rebuilding of your widget tree into independent parts.

It's not reasonable to list all the potential optimization factors that a class allow, as there are too many.

The following example is a widget that takes an int and formats it into a Text. The catch is, this widget will rebuild only if the int passed change:

class Counter extends StatelessWidget {
  const Counter({Key key, this.value}) : super(key: key);

  final int value;

  @override
  Widget build(BuildContext context) {
    return Text(value.toString());
  }

  @override
  bool operator ==(Object other) =>
      identical(this, other) || (other is Counter && other.value == value);

  @override
  int get hashCode => value.hashCode;
}

This works because Flutter uses the == operator to know if a widget should update or not (hence why const constructor is a good optimization factor).

This is not the only solution, but it's a good example of something that functions can't do.

来源：https://stackoverflow.com/questions/54824933/statelesswidget-vs-a-function-returning-widgets-in-terms-of-performance

标签

performance

layout

Flutter

statelesswidget