I have solved the problem of extracting the value from the BindingExpression
, with a minor limitation.
First, some more complete XAML:
<Window x:Class="ValidationRuleTest.MainWindow"
xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
xmlns:local="clr-namespace:ValidationRuleTest"
Title="MainWindow" Height="100" Width="525">
<Window.DataContext>
<local:MainWindowViewModel/>
</Window.DataContext>
<Grid>
<Grid.RowDefinitions>
<RowDefinition Height="Auto"/>
<RowDefinition Height="Auto"/>
</Grid.RowDefinitions>
<Grid.ColumnDefinitions>
<ColumnDefinition Width="50"/>
<ColumnDefinition Width="*"/>
</Grid.ColumnDefinitions>
<TextBlock Text="String 1"/>
<TextBox Grid.Column="1">
<TextBox.Text>
<Binding Path="String1" UpdateSourceTrigger="PropertyChanged">
<Binding.ValidationRules>
<local:RequiredRule ValidationStep="RawProposedValue"/>
</Binding.ValidationRules>
</Binding>
</TextBox.Text>
</TextBox>
<TextBlock Text="String 2" Grid.Row="1"/>
<TextBox Grid.Column="1" Grid.Row="1">
<TextBox.Text>
<Binding Path="String2" UpdateSourceTrigger="PropertyChanged">
<Binding.ValidationRules>
<local:RequiredRule ValidationStep="UpdatedValue"/>
</Binding.ValidationRules>
</Binding>
</TextBox.Text>
</TextBox>
</Grid>
</Window>
Note that the first TextBox uses ValidationStep="RawProposedValue"
(the default), while the second one uses ValidationStep="UpdatedValue"
, but both use the same validation rule.
A simple ViewModel (neglecting INPC and other useful stuff):
class MainWindowViewModel
{
public string String1
{ get; set; }
public string String2
{ get; set; }
}
And finally, the new RequiredRule:
class RequiredRule : ValidationRule
{
public override ValidationResult Validate(object value,
System.Globalization.CultureInfo cultureInfo)
{
// Get and convert the value
string stringValue = GetBoundValue(value) as string;
// Specific ValidationRule implementation...
if (String.IsNullOrWhiteSpace(stringValue))
{
return new ValidationResult(false, "Must not be empty");
}
else
{
return new ValidationResult(true, null);
}
}
private object GetBoundValue(object value)
{
if (value is BindingExpression)
{
// ValidationStep was UpdatedValue or CommittedValue (Validate after setting)
// Need to pull the value out of the BindingExpression.
BindingExpression binding = (BindingExpression)value;
// Get the bound object and name of the property
object dataItem = binding.DataItem;
string propertyName = binding.ParentBinding.Path.Path;
// Extract the value of the property.
object propertyValue = dataItem.GetType().GetProperty(propertyName).GetValue(dataItem, null);
// This is what we want.
return propertyValue;
}
else
{
// ValidationStep was RawProposedValue or ConvertedProposedValue
// The argument is already what we want!
return value;
}
}
}
The GetBoundValue()
method will dig out the value I care about if it gets a BindingExpression, or simply kick back the argument if it's not. The real key was finding the "Path", and then using that to get the property and its value.
The limitation: In my original question, my binding had Path="Identity.Name"
, as I was digging into sub-objects of my ViewModel. This will not work, as the code above expects the path to be directly to a property on the bound object. Fortunately, I have already flattened my ViewModel so this is no longer the case, but a workaround could be to set the control's datacontext to be the sub-object, first.
I'd like to give some credit to Eduardo Brites, as his answer and discussion got me back to digging on this, and did provide a piece to his puzzle. Also, while I was about to ditch the ValidationRules entirely and use IDataErrorInfo instead, I like his suggestion on using them together for different types and complexities of validation.