I'm using an OnRead grid with ExpandoObjects. The error happens on the backend when trying to process the result of ToDataSourceResult(...)
When trying to access the Items property of AggregateFunctionsGroup in a multi level grouping scenario, using ExpandoObject, and the top level group has a null key, trying to access the Items property will result in a NullReferenceException. I've made this helper, but it cannot process the subgroups because of this error.
private static void FlattenGroup<T>(AggregateFunctionsGroup group, List<T> result)
{
if (group == null)
{
return;
}
if (group.HasSubgroups)
{
foreach (var sub in group.Items.OfType<AggregateFunctionsGroup>())
{
FlattenGroup(sub, result);
}
}
else
{
result.AddRange(group.Items.OfType<T>());
}
}In a scenario where I have an ExpandoObject with properties A and B, where A = 1 and B = null; grouping by A then B works. Grouping by A or B alone also works. But grouping by B then A causes the NullReferenceException when trying to access the group.Items.
Hi Tsvetomir,
Thanks for your reply. I did also find some similar posts about grouping and null key issues. However, they all seemed to be using LoadGroupsOnDemand, which we do not in our application. We do use OnRead and we cannot avoid ExpandoObjects. Additionally, the way the exception is thrown is not catchable by making a null check like you propose. The null reference is not thrown until you try to enumerate the collection. The same exception is also thrown when trying to enumerate the SubGroups property.
I don't think the REPL example is worth much, because its behaviour is different from the exact same code when I run it locally. In the REPL I can't even group by the non-null column. Running this locally produces the error, but in a slightly different way, than when the error occours on the backend. If possible, I recommend trying this locally, setting a breakpoint at the loop in FlattenGroup, then inspecting the Items and SubGroups property of the null group REPL
The problems with the workaround for us is that a null value is a valuable piece of information, because it essentially means the users have not yet decided about the value of that particular field, on that particular row. The problem with setting a placeholder value, is that we have multiple data types where anything but null will give users a wrong impression of what the data is. For instance, using a default value for a nullable int, would be 0, and would give false information to users.
We also cannot tell users that they cannot use grouping on columns where there might be null values - it's essential to the users that they can group on everything.
Here's a little snippet of how the error occours.
