C# equivalent of XAML compiled binding

[tranb3r]

Compiled bindings improve data binding performance in .NET MAUI applications.
But all code examples for Compiled bindings are written in XAML.
Is Compiled bindings a XAML feature? What is the equivalent in C#?

[brminnick]

For Compiled Bindings in C#, you need to install the CommunityToolkit.Maui.Markup NuGet Package.

Docs

https://learn.microsoft.com/dotnet/communitytoolkit/maui/markup/extensions/bindable-object-extensions#one-way-binding

Example

using static CommunityToolkit.Maui.Markup.GridRowsColumns;

class SampleContentPage : ContentPage
{
    public SampleContentPage()
    {
        Content = new Grid
        {
            RowDefinitions = Rows.Define(
                (Row.TextEntry, 36)),

            ColumnDefinitions = Columns.Define(
                (Column.Description, Star),
                (Column.Input, Stars(2))),

            Children =
            {
                new Label()
                    .Text("Code:")
                    .Row(Row.TextEntry).Column(Column.Description),

                new Entry
                {
                    Keyboard = Keyboard.Numeric,
                    BackgroundColor = Colors.AliceBlue,
                }.Row(Row.TextEntry).Column(Column.Input)
                 .FontSize(15)
                 .Placeholder("Enter number")
                 .TextColor(Colors.Black)
                 .Height(44)
                 .Margin(5, 5)
                 .Bind(Entry.TextProperty, static (ViewModel vm) vm => vm.RegistrationCode)
            }
        };
    }

    enum Row { TextEntry }
    enum Column { Description, Input }
}

[tranb3r]

Ok, this code is indeed "compiled", and not using a string path.
However, under the hood, is it 100% equivalent to the code generated by XAML compiled bindings?
How do they compare in terms of performance? I have the impression that XAML compiled bindings is a lot faster than this C# extension, but maybe I'm wrong.

[brminnick]

However, under the hood, is it 100% equivalent to the code generated by XAML compiled bindings

Yes, I wrote the code myself. Both use TypedBindings under the hood. It's all open source: https://github.com/CommunityToolkit/Maui.Markup/blob/main/src/CommunityToolkit.Maui.Markup/TypedBindingExtensions.cs

The performance is much better because it doesn't use reflection.

[tranb3r]

Actually, I don't see where it's using TypedBinding. Maybe you copied the code from dotnet maui and renamed it?
Also, are you saying it's much better than standard C# bindings (or standard XAML bindings), or is it much better than XAML compiled bindings?
I will probably need to run a benchmark to be convinced... ;)

[tranb3r]

Hi Brandon,

I've done more testing of compiled/typed bindings, both with xaml and c#.

Here are my conclusions:

• as expected, when setting a view model property, the performance of bindings is the same between xaml compiled bindings and c# typed bindings, and it's better than when not using compiled/typed bindings.
• however, the performance of view creation is a lot worse when using c# typed bindings compared to any other option (xaml without compiled bindings, xaml with compiled bindings, c# without typed bindings).

Here is the code I've used to benchmark the different cases: https://github.com/tranb3r/Issues/tree/main/BenchmarkBindings

Here are the results of my benchmarks (higher is better):

So, here are my questions:

• am I doing something wrong?
• why is the performance of c# typed bindings so bad on view creation?
• can we get the best of both worlds? (the performance of xaml compiled bindings, with the c# syntax)

[pictos]

@tranb3r can you share your benchmark? Looks like the repo you linked isn't using the BenchmarkDotnet

[tranb3r]

Hi Pedro,

I've updated my repro with BenchmarkDotNet in the Maui app.

For the CreateView test, the results are the same between StopWatch and BenchmarkDotNet.
This confirms the performance issue with C# Typed Bindings on View instantiation.

Results for CreateView (StopWatch):

Method	Mean
RunXaml	280 us
RunXamlCompiledBindings	260 us
RunCSharp	280 us
RunCSharpTypedBindings	600 us

Results for CreateView (BenchmarkDotNet):

Method	Mean	Error	StdDev
RunXaml	140.8 us	2.66 us	5.62 us
RunXamlCompiledBindings	126.3 us	2.52 us	3.62 us
RunCSharp	139.2 us	2.73 us	4.09 us
RunCSharpTypedBindings	685.5 us	13.07 us	13.42 us

For the UpdateViewModel test, the results are different between StopWatch and BenchmarkDotNet.
The StopWatch results seem coherent with the expected behavior of bindings (compiled or not, xaml or c#).
But the BenchmarkDotNet results seem wrong (no difference between the different tests).
So, I'm wondering if I'm measuring things correctly. Probably I'm not capturing the View update correctly.
What do you think? How could I improve this benchmark?

Results for UpdateViewModel (StopWatch):

Method	Mean
RunXaml	20 us
RunXamlCompiledBindings	16 us
RunCSharp	18 us
RunCSharpTypedBindings	16 us

Results for UpdateViewModel (BenchmarkDotnet):

Method	Mean	Error	StdDev
RunXaml	42.19 us	0.839 us	1.595 us
RunXamlCompiledBindings	41.86 us	0.836 us	2.002 us
RunCSharp	41.55 us	0.830 us	1.049 us
RunCSharpTypedBindings	41.93 us	0.831 us	1.943 us

[tranb3r]

@pictos
Did you have the time to look at my benchmark?
Any suggestion on how to improve it?
Thanks

[pictos]

@tranb3r not yet, I'm afraid that I'll have time to take a good look on it after .net conf

[tranb3r]

@pictos I hope my follow-up isn't too bothersome.
Your insights on this question would be highly valued whenever you have a moment.
Thank you!

[CliffAgius]

Closed as answered

[tranb3r]

@CliffAgius
Why are you closing this discussion?
I'm still needing help to figure out how to properly benchmark bindings.

[CliffAgius]

Sorry I am cleaning up the Discussions board as per the Community Stand-up discussion here: https://youtu.be/N9wMcBP4jtg?t=2889 and I see this as answered and it looked like the discussion was over. I have re-opened again.

[CliffAgius]

Coming up to a year for this discussion is it ok to close?

[bijington]

I think @brminnick did something recently to improve the performance in C#

[brminnick]

This added benchmarking, but I haven't taken the time to dig into performance improvements for CommunityToolkit.Maui.Markup yet: CommunityToolkit/Maui.Markup#287

[brminnick]

But yea, I think it’s fine to close this Discussion since the code in question is not in this repo.

Folks can contribute over on https://github.com/CommunityToolkit/Maui.Markup/.