How to make Databinding type safe and support refactoring?
Note this answer uses WinForm and was written before C# had 'NameOf()'
Thanks to Oliver for getting me started I now have a solution that both supports refactoring and is type safe. It also let me implement INotifyPropertyChanged so it copes with properties being renamed.
It’s usage looks like:
checkBoxCanEdit.Bind(c => c.Checked, person, p => p.UserCanEdit);
textBoxName.BindEnabled(person, p => p.UserCanEdit);
checkBoxEmployed.BindEnabled(person, p => p.UserCanEdit);
trackBarAge.BindEnabled(person, p => p.UserCanEdit);
textBoxName.Bind(c => c.Text, person, d => d.Name);
checkBoxEmployed.Bind(c => c.Checked, person, d => d.Employed);
trackBarAge.Bind(c => c.Value, person, d => d.Age);
labelName.BindLabelText(person, p => p.Name);
labelEmployed.BindLabelText(person, p => p.Employed);
labelAge.BindLabelText(person, p => p.Age);
The person class shows how to implemented INotifyPropertyChanged in a type safe way (or see this answer for a other rather nice way of implementing INotifyPropertyChanged, ActiveSharp - Automatic INotifyPropertyChanged also looks good ):
public class Person : INotifyPropertyChanged
{
private bool _employed;
public bool Employed
{
get { return _employed; }
set
{
_employed = value;
OnPropertyChanged(() => c.Employed);
}
}
// etc
private void OnPropertyChanged(Expression<Func<object>> property)
{
if (PropertyChanged != null)
{
PropertyChanged(this,
new PropertyChangedEventArgs(BindingHelper.Name(property)));
}
}
public event PropertyChangedEventHandler PropertyChanged;
}
The WinForms binding helper class has the meat in it that makes it all work:
namespace TypeSafeBinding
{
public static class BindingHelper
{
private static string GetMemberName(Expression expression)
{
// The nameof operator was implemented in C# 6.0 with .NET 4.6
// and VS2015 in July 2015.
// The following is still valid for C# < 6.0
switch (expression.NodeType)
{
case ExpressionType.MemberAccess:
var memberExpression = (MemberExpression) expression;
var supername = GetMemberName(memberExpression.Expression);
if (String.IsNullOrEmpty(supername)) return memberExpression.Member.Name;
return String.Concat(supername, '.', memberExpression.Member.Name);
case ExpressionType.Call:
var callExpression = (MethodCallExpression) expression;
return callExpression.Method.Name;
case ExpressionType.Convert:
var unaryExpression = (UnaryExpression) expression;
return GetMemberName(unaryExpression.Operand);
case ExpressionType.Parameter:
case ExpressionType.Constant: //Change
return String.Empty;
default:
throw new ArgumentException("The expression is not a member access or method call expression");
}
}
public static string Name<T, T2>(Expression<Func<T, T2>> expression)
{
return GetMemberName(expression.Body);
}
//NEW
public static string Name<T>(Expression<Func<T>> expression)
{
return GetMemberName(expression.Body);
}
public static void Bind<TC, TD, TP>(this TC control, Expression<Func<TC, TP>> controlProperty, TD dataSource, Expression<Func<TD, TP>> dataMember) where TC : Control
{
control.DataBindings.Add(Name(controlProperty), dataSource, Name(dataMember));
}
public static void BindLabelText<T>(this Label control, T dataObject, Expression<Func<T, object>> dataMember)
{
// as this is way one any type of property is ok
control.DataBindings.Add("Text", dataObject, Name(dataMember));
}
public static void BindEnabled<T>(this Control control, T dataObject, Expression<Func<T, bool>> dataMember)
{
control.Bind(c => c.Enabled, dataObject, dataMember);
}
}
}
This makes use of a lot of the new stuff in C# 3.5 and shows just what is possible. Now if only we had hygienic macros lisp programmer may stop calling us second class citizens)
The nameof
operator was implemented in C# 6.0 with .NET 4.6 and VS2015 in July 2015. The following is still valid for C# < 6.0
To avoid strings that contain property names, I've written a simple class using expression trees to return the name of the member:
using System;
using System.Linq.Expressions;
using System.Reflection;
public static class Member
{
private static string GetMemberName(Expression expression)
{
switch (expression.NodeType)
{
case ExpressionType.MemberAccess:
var memberExpression = (MemberExpression) expression;
var supername = GetMemberName(memberExpression.Expression);
if (String.IsNullOrEmpty(supername))
return memberExpression.Member.Name;
return String.Concat(supername, '.', memberExpression.Member.Name);
case ExpressionType.Call:
var callExpression = (MethodCallExpression) expression;
return callExpression.Method.Name;
case ExpressionType.Convert:
var unaryExpression = (UnaryExpression) expression;
return GetMemberName(unaryExpression.Operand);
case ExpressionType.Parameter:
return String.Empty;
default:
throw new ArgumentException("The expression is not a member access or method call expression");
}
}
public static string Name<T>(Expression<Func<T, object>> expression)
{
return GetMemberName(expression.Body);
}
public static string Name<T>(Expression<Action<T>> expression)
{
return GetMemberName(expression.Body);
}
}
You can use this class as follows. Even though you can use it only in code (so not in XAML), it is quite helpful (at least for me), but your code is still not typesafe. You could extend the method Name with a second type argument which defines the return value of the function, which would constrain the type of the property.
var name = Member.Name<MyClass>(x => x.MyProperty); // name == "MyProperty"
Until now I haven't found anything which solves the databinding typesafety issue.
The Framework 4.5 provides us with the CallerMemberNameAttribute
, which makes passing the property name as a string unnecessary:
private string m_myProperty;
public string MyProperty
{
get { return m_myProperty; }
set
{
m_myProperty = value;
OnPropertyChanged();
}
}
private void OnPropertyChanged([CallerMemberName] string propertyName = "none passed")
{
// ... do stuff here ...
}
If you're working on Framework 4.0 with KB2468871 installed, you can install the Microsoft BCL Compatibility Pack via nuget, which also provides this attribute.
This blog article raises some good questions about the performance of this approach. You could improve upon those shortcomings by converting the expression to a string as part of some kind of static initialization.
The actual mechanics might be a little unsightly, but it would still be type-safe, and approximately equal performance to the raw INotifyPropertyChanged.
Something kind of like this:
public class DummyViewModel : ViewModelBase
{
private class DummyViewModelPropertyInfo
{
internal readonly string Dummy;
internal DummyViewModelPropertyInfo(DummyViewModel model)
{
Dummy = BindingHelper.Name(() => model.Dummy);
}
}
private static DummyViewModelPropertyInfo _propertyInfo;
private DummyViewModelPropertyInfo PropertyInfo
{
get { return _propertyInfo ?? (_propertyInfo = new DummyViewModelPropertyInfo(this)); }
}
private string _dummyProperty;
public string Dummy
{
get
{
return this._dummyProperty;
}
set
{
this._dummyProperty = value;
OnPropertyChanged(PropertyInfo.Dummy);
}
}
}
One way to get feedback if your bindings are broken, is to create a DataTemplate and declare its DataType to be the type of the ViewModel that it binds to e.g. if you have a PersonView and a PersonViewModel you would do the following:
Declare a DataTemplate with DataType = PersonViewModel and a key (e.g. PersonTemplate)
Cut all PersonView xaml and paste it into the data template (which ideally can just be at the top of the PersonView.
3a. Create a ContentControl and set the ContentTemplate = PersonTemplate and bind its Content to the PersonViewModel.
3b. Another option is to not give a key to the DataTemplate and not set the ContentTemplate of the ContentControl. In this case WPF will figure out what DataTemplate to use, since it knows what type of object you are binding to. It will search up the tree and find your DataTemplate and since it matches the type of the binding, it will automatically apply it as the ContentTemplate.
You end up with essentially the same view as before, but since you mapped the DataTemplate to an underlying DataType, tools like Resharper can give you feedback (via Color identifiers - Resharper-Options-Settings-Color Identifiers) as to wether your bindings are broken or not.
You still won't get compiler warnings, but can visually check for broken bindings, which is better than having to check back and forth between your view and viewmodel.
Another advantage of this additional information you give, is, that it can also be used in renaming refactorings. As far as I remember Resharper is able to automatically rename bindings on typed DataTemplates when the underlying ViewModel's property name is changed and vice versa.