RadGrid row still advances after error set on DataView row during row-level validation

[Scott]

I am trying to implement row-level validation when using the RadGridView bound to a DataView. The xaml looks like this:

<telerik:RadGridView Grid.Column="1" Grid.Row="1" Name="_gridTableGrid" ItemsSource="{Binding TableDataView, Mode=TwoWay}" />

The TableDataView property is a DataView from a DataSet into which I have read a series of records, e.g.,

TableDataView = myDataSet.Tables[0].DefaultView;

I subscribe to the following events:

myDataSet.Tables[0].RowChanging += new DataRowChangeEventHandler(DataTable_RowChanging);
myDataSet.Tables[0].RowChanged += new DataRowChangeEventHandler(DataTable_RowChanged);

I am using row-level validation because the row change may be rejected by the DBMS when I go to update the backing store. With other grids, the validation can be done before the focus leaves the current row and signaling a validation failure prevents the user from leaving that row without either correcting the error or rolling the change back. I am trying to use the e.Row.RowError property that I recieve in the DataRowChangeEventArgs e argument passed to my event handlers. That does not seem to have the effect that I am looking for: specifically, the row header should show an error and the user is restricted from leaving the record until the issue is addressed.

Any suggestions on how this would be handled from my view-model/model? Maybe the change in error status needs to be propagated back to the grid in some fashion other than this because this has no effect. If I work through cell-level validation, I get results that one would expect. But, cell-level validation is not useful in this condition.

Edit: By the way, if you try to use e.Row.RejectChanges() in the DataTable_RowChanging event handler, you'll get an error that says that you should throw an exception to cancel the row operation. If I throw an InvalidOperationException, for example, it is caught by my unhandled exception handler. I think that is the Microsoft method - the exception will be caught by the grid control and the message displayed. So, no joy in that direction, either.

Edit 2: Ok, so we don't handle it from our (view-)model, we go out to the code behind. Change the xaml to include RowValidating="OnRowValidating" and  RowValidated="OnRowValidated" and add those handlers. In a stub OnRowValidating
handler, write:

private void OnRowValidating(object sender, Telerik.Windows.Controls.GridViewRowValidatingEventArgs e)
{
    e.IsValid = false;
    e.ValidationResults.Add(new Telerik.Windows.Controls.GridViewCellValidationResult() {ErrorMessage = "byte me!"} );
}

Now we get the expected behavior: it will flag the record in the header and the tool tip 'byte me!' is shown on mouse over. This brings up another problem - the cancel behavior. As expected, the first escape cancels the edit on the current cell (really the last cell with the focus before the validation event). The second escape reverts the record. Almost. Here's what happens: if the first cell contained the value '5' and it was changed to '55' before the commit, the '55' is left in that field after the cancel. So the user now sees his incorrect value(s) in the grid after he performed the escape sequence.

The frustration with this behavior is quite high. We're so close in getting the correct behavior from the view-model. Why didn't Telerik just allow the exception ala MS to signal the validation error? And, when I use what appears to be the Telerik method, it fails in the rollback. I am using the last Q2 release because my attempt to move to the Q3 found a lack of fixes for what I currently have in the Q2 code and new errors introduced in the Q3 release.

[Scott]

I've packaged up the description above and am sending it in as a bug report.

Another issue that I am dealing with is drag and drop. In Q2, initiating a DnD op results in an exception. In Q3, the exception is gone and the DnD operation never initiates. And, we have this issue. My question is, what are Telerik's priorities when they create this WPF code? Does Telerik expect that we replace all MS components with theirs in order to get the required functionality, or does Telerik endeavor to make their code compatible with the other WPF framework components *and* the modern practice of mvvm programming? I truly hope it's the latter. Also, in my view, Telerik's quarterly release schedule attempts to push too much content at the expense of code quality. Each quarter sees something break and a lack of fixes for some trivial issues. That observation lead me to ask the first question because I may be approaching the use of the Telerik controls incorrectly. Perhaps I am expected to replace all the stock MS functionality with Telerik's, but I am attempting to use Telerik's controls to extend functionality in only those areas that I need to?

[Tsvyatko]

Hi Scott,

 In Q3 we have made some improvements in the RadDragAndDropManager, as well as creating DragDropManager with more flexible API which logic is now used internally from the RadDragAndDropManager. We have provided option to which the new and the old execution mode as described here - http://www.telerik.com/help/wpf/raddragdropmanager-howto-change-drag-execution-mode.html.

So, as a first step I can suggest the try switch to legacy mode and check whether the situation improve. If, not would it be possible to share some additional information regarding the scenario the DnD fails.

All the best,
Tsvyatko
the Telerik team