Is it OK not to handle returned value of a C# method? What is good practice in this example?

Out of curiosity...what happens when we call a method that returns some value but we don't handle/use it? And we also expect that sometimes this returned value could be really big. Where that value goes? Is it even created? If it is, are there any performance issues or other problems that can occur? (what is the best practice in this kind of situation?)

Let's say we have method that does some database operations (insert, update) and returns some data in DataTable object. And I also know that this DataTable object could be really big sometimes:

public static Datatable InsertIntoDB(...) 
{
      // executing db command, getting values, creating & returning Datatable object...
      ...
      return myDataTable;
}

And then when this method is used it is called like these:

DataTable myDataTable = InsertIntoDB(...);
// this Datatable object is handled in some way

But sometimes simply like this:

InsertIntoDB(...);
// returned value not handled; Problem???

On my first thought it think the system is smart enough to see the returned value is ignored and does not cause any problems (it is simply released) but I want to be sure and hear more detailed explanation of it from someone who is more experien开发者_如何学编程ced in this area than me.

The returned value (or reference, if it's a reference type) is pushed onto the stack and then popped off again.

No biggy.

If the return value isn't relevant, you can safely do this.

But be sure that it isn't relevant, just in case.

Here's some code:

    static string GetSomething()
    {
        return "Hello";
    }

    static void Method1()
    {
        string result = GetSomething();
    }

    static void Method2()
    {
        GetSomething();
    }

If we look at the IL:

Method1:

.locals init ([0] string result)
IL_0000:  nop
IL_0001:  call       string ConsoleApplication3.Program::GetSomething()
IL_0006:  stloc.0
IL_0007:  ret

Method2:

IL_0000:  nop
IL_0001:  call       string ConsoleApplication3.Program::GetSomething()
IL_0006:  pop
IL_0007:  ret

Exactly the same number of instructions. In Method1, the value is stored in the local string result (stloc.0), which is deleted when it goes out of scope. In Method2, the pop operation simply removes it from the stack.

In your case of returning something 'really big', that data has already been created and the method returns a reference to it; not the data itself. In Method1(), the reference is assigned to the local variable and the garbage collector will tidy it up after the variable has gone out of scope (the end of the method in this case). In Method2(), the garbage collector can get to work, any time after the reference has been popped from the stack.

By ignoring the return value, if it really isn't needed, the garbage collector can potentially get to work sooner and release any memory that's been assigned. But there's very little in it (certainly in this case), but with a long running method, hanging onto that data could be an issue.

But far-and-away the most important thing is to be sure that the return value that you're ignoring isn't something that you should be acting on.

EDIT: Softened the language very slightly, and clarified.

It's rarely a good idea to ignore the return value, in my experience - at least in cases where the return values are there to convey new information instead of simply being for convenience.

One example where I've seen it be okay:

int foo;
int.TryParse(someText, out foo);

// Keep going

Here foo will be 0 if either someText contained "0", or it couldn't be parsed. We may not care which was the case in which case the return value of the method is irrelevant to us.

Another example is in a dictionary - suppose you're trying to count the number of occurrences of each string. You can use:

int count;
dictionary.TryGetValue(word, out count);
dictionary[word] = count + 1;

If the word wasn't in the dictionary to start with, that's equivalent to there being a count of 0 - which is what will already happen as a result of calling TryGetValue.

As a counter-example, ignoring the value returned by Stream.Read (and assuming that it's managed to read all the data you asked for) is a common mistake.

If you don't need the return value and it will have taken a lot of effort to compute, it may be worth looking for something which will achieve the same desired side-effects without the extra computation - but there's no extra performance implication. I'd be more worried about the correctness of ignoring a return value than the performance.

EDIT: Other examples where it's okay to ignore the return value:

• Some fluent interfaces, including StringBuilder; while StringBuilder.Append(x).Append(y); uses the first return value for the second call, very often the return value of a call will be ignored, e.g. when appending in a loop
• Some collection calls can give return values which are sometimes ignored - e.g. HashSet<T>.Add which indicates whether the value was actually added, or was already present. Sometimes you just don't care.

But for the vast majority of the time, ignoring the return value of a method indicates that it's doing more than you need it to.

From a memory management point of view thats fine - if the calling function doesn't use it, it goes out of scope and gets garbage collected.

In this particular case DataTable does implement IDisposable so its not all 100% fine:

If the returned object implements IDisposable then its a good idea to dispose it, for example:

using (var retVal = InsertIntoDB(...))
{
    // Could leave this empty if you wanted
}

It depends on the returned value it self.

The compiler will generate that value in the caller method, So if the value is IDispolable or expose Close method or if it have resources that should be released, then you should not ignore it and dispose it properly, or else you may suffering from problems and memory leaks..

For instance, if the returned value is a FileStream and you did not close the stream, the file may not closed until your application is terminated, more over if your application try to open the file again, it may throw exception that indicates "The file is in used by another process". So you should be careful on that kind of returned object and never ignore it!

It's totally fine to ignore the return value.

However. The architectural design is, IMHO, not good. An insert method should not return anything at all (other than MAYBE true or false on success or failure). If one would need to get a new, updated, dataset then one should ask for it, i.e call some other method to do so.

The returned value is thrown away if not used, but it is created. It is perfectly reasonable not to use it ( although you should be ceratin that this is the right thing to be doing ), but if it takes a lot of resource to create, then this is wasted.

You may want to consider whether another method would be a better options, that doesn't create the return object at all.

To give a different perspective on things, I think that method should be redesigned. Take a look at the Command-Query separation.

Also, it's rarely a good idea to silently ignore a return value. Readers of the code might not have the original context of the author. They might think he just forgot to use it. If the return value is not important, better be explicit about this decision:

var ignoredReturnValue = InsertIntoDB(...);

Interestingly, Nemerle actually gives you a warning if you ignore a return value. To not get the warning, you have to be explicit about your decision and write:

_ = InsertIntoDB(...);

I'm certain that this doesn't cause any problems, otherwise C# wouldn't be a very reliable language.

I'm guessing the compiler isn't smart enough to optimize this. What most likely happens is the ordinary logic inside the function call is executed, e.g. creating the object and allocating memory for it. If a reference type is returned but not captured, garbage collection will free up the memory again.

As others have stated, from a design view ignoring the return value does indicate a problem and most likely you should be looking at the return value.

If your function do some changes to other objects (for exemple a DB), I think it's okay to not handle the returned object if you don't need it.

All this talk about whether it is okay to ignore returned types is not needed, we do it all the time anyway in C#. Lots of functions you use as if they are returning void are not returning void. Think about a common function like Button1.Focus()

Did you know that the .Focus() function returns a bool value? It returns true if it succeeded in focusing on the control. So you could test it as a bool by saying:

if (Button1.Focus == true) MessageBox.Show("Button Focused successfully."); else MessageBox.Show("Could not focus on the button, sorry.");

But normally, you don't do this. You just say: Button1.Focus();

and you're done. I could give a hundred other examples where we ignore return values, like when a function runs but also returns a reference to something it created, but you don't care about the reference, you just wanted it to do the action (or you just want to simply check whether there is a reference or if it is null)

The point is, we ignore return values all the time, even if you don't know it.