Efficiency of (multiple) MultiDataTrigger vs. Conv

问题:

I'm currently analyzing some XAML that uses style that make extensive use of MultiDataTriggers (8-10 multi data triggers per style, with 4-6 conditions per trigger). When looking at this I'm considering whether it would be more efficient to use a converter (or multi value converter), especially as MultiDataTriggers cannot be debugged.

Can anyone authoritatively state how MultiDataTriggers are compiled? I understand that the Conditions are ANDed together, is this compiled in such a way that short cutting is featured?

What about multiple MultiDataTriggers? Are they short circuited so that the first one fully satisfied causes evaluation to stop? Or are they all evaluated with the last one winning if several are satisfied?

回答1:

Triggers are evaluated from top to bottom. It holds true for all sorts of triggers (Trigger, DataTrigger, MultiTrigger and MutliDataTrigger).

What about multiple MultiDataTriggers? Are they short circuited so that the first one fully satisfied causes evaluation to stop? Or are they all evaluated with the last one winning if several are satisfied?

As stated triggers are evaluated from top to bottom. So, in case first one satisfy all conditions doesn't mean further triggers won't be evaluated. All triggers applied on changed property are evaluated and in case any two of them are setting same property inside a trigger then last trigger always won and overrides the property set by first trigger.

<TextBlock>
   <TextBlock.Style>
      <Style TargetType="TextBlock">
          <Style.Triggers>
             <DataTrigger Binding="{Binding IsEnable}" Value="True">
                <Setter Property="Text" Value="Test1"/>
             </DataTrigger>
             <DataTrigger Binding="{Binding IsEnable}" Value="True">
                <Setter Property="Text" Value="Test2"/>
             </DataTrigger>
           </Style.Triggers>
      </Style>
  </TextBlock.Style>
</TextBlock>

Text will always be Test2 when IsEnable evaluates out to be true.

---

Can anyone authoritatively state how MultiDataTriggers are compiled? I understand that the Conditions are ANDed together, is this compiled in such a way that short cutting is featured?

Yeah, short cutting is featured in MultiDataTrigger i.e. if first condition evaluate to be false, second condition won't be checked. This sample validates this -

<TextBlock>
   <TextBlock.Style>
       <Style TargetType="TextBlock">
           <Style.Triggers>
               <MultiDataTrigger>
                   <MultiDataTrigger.Conditions>
                       <Condition Binding="{Binding IsEnable,
                             Converter={StaticResource SingleValueConverter}}" 
                                  Value="True"/>
                       <Condition Binding="{Binding IsEnable,
                             Converter={StaticResource SingleValueConverter}}"
                                  Value="True"/>
                    </MultiDataTrigger.Conditions>
                    <Setter Property="Text" Value="Test"/>
                </MultiDataTrigger>
          </Style.Triggers>
       </Style>
   </TextBlock.Style>
</TextBlock>

On both conditions a converter is applied but in case IsEnabled is false, converter gets hit only once because first condition evaluates out to be false. But in case IsEnabled is true, converter gets hit twice since first condition is meet successfully.

回答2:

Converters allow debugging and more complex logics but they also have to check all their Bindings and call an external function each time one of their Bindings change. So they are slower than Triggers in almost all cases. Triggers stop at the first not-met-condition.

So my answer is to go with MutiDataTrigger as much as you can, when you needed more logic, depending on whether it's possible that some conditions are repeated elsewhere or not, you can choose to implement an extra DependencyProperty (which changes when some few other properties change) or to use a converter.

For example I have 5 properties which I want to bind to:

IsChecked = A && B && (C || D || !E)
IsReadonly = !A && !B && (C || D || !E)

so I would create a new property F equal to C || D || !E and when one of these three changes, update F. now I can use F as the third trigger binding path.