Question

I am developing a library that initializes plain-old data objects from data retrieved from SQL Server stored procedures. One of its methods is the following one:

public static T RetrieveCompound<T, U>(string cmdText, params Parameter[] parameters)
    where T : Header<U>, new()
    where U : class, new() {

    Box<T> headerBox = new Box<T>();
    List<U> details;
    Execute(cmdText, parameters, new Action<Row>[] {
        SetReferenceAction<T>(headerBox) +
        delegate(Row row) { details = headerBox.Value.Details; },
        delegate(Row row) { details.Add(row.ToObject<U>()); }
    });
    return headerBox.Value;
}

The third parameter of Execute is an array of Action<Row>s. Although no static analyzer can programmatically prove it, because of the way the Execute method was programmed, no delegate can be run before the ones preceding it in the array. That means that the delegate

delegate(Row row) { details.Add(row.ToObject<U>()); } // use

must necessarily run after the delegate

delegate(Row row) { details = headerBox.Value.Details; } // initialize

And, thus, the variable details will necessarily be initialized before it is used.

But C# still complains: “Use of unassigned variable ‘details’.”

Is there any way to make C# not complain about uninitialized variables that are actually not?

---

I am beginning to think C# is not for me.

Answer 1

This is the usual workaround:

List<U> details = null;