Write-only variables considered harmful? Or beneficial?

re: Write-only variables considered harmful? Or beneficial?

Nick — Mon, 27 Feb 2012 22:45:39 GMT

#if DEBUG

bool b =

#endif

DoSomething(); // Apparently, this works in C# 4.

re: Write-only variables considered harmful? Or beneficial?

GregM — Sun, 06 Mar 2011 21:46:09 GMT

Adam, that depends on your environment. In my environment, for example, warnings do prevent compilation, as we use the warnings as errors option.

However, I do have the VC++ warning "variable initialized but never used" turned on for the reasons you describe.

So in my VC++ environment, this is an error

bool b = DoSomething();

but this is not

bool b;

b = DoSomething();

So VC++ does allow write-only variables, but warns about initialize-only variables.

re: Write-only variables considered harmful? Or beneficial?

Adam M. — Thu, 03 Mar 2011 21:00:39 GMT

I'm not particularly convinced. A warning doesn't prevent compilation, so if a dummy variable is added to aid debugging, it seems perfectly reasonable to issue a warning about it to remind the programmer to remove the unnecessary variable. And if they really want to leave debug code in the source forever, they can add an exclusion for the warning easily enough. I'd much rather have the compiler tell me that I forgot to do something with the variable I assigned than keep quiet, assuming that 1) I'm just adding it for debugging purposes, and 2) I intend to not remove it.

re: Write-only variables considered harmful? Or beneficial?

Eric Lippert — Tue, 24 Apr 2007 19:45:58 GMT

Since that "poor soul" is often oneself, it's not entirely altruistic!

re: Write-only variables considered harmful? Or beneficial?

John — Tue, 24 Apr 2007 19:39:41 GMT

You are getting old and wise when you start to think about the poor sole working on the code later. Thanks.

re: Write-only variables considered harmful? Or beneficial?

Alun Jones — Mon, 23 Apr 2007 23:25:59 GMT

You probably can put breakpoints on the closing brace in Visual Studio. However, I spend my day in at least 4 programming environments each with different debuggers and I won't even relate how many different C IDE's I've put up with over the years... Some habits are more portable than others.

re: Write-only variables considered harmful? Or beneficial?

Alun Jones — Mon, 23 Apr 2007 23:18:03 GMT

Never, never, never (what never? hardly ever) put functional code inside of #ifdef DEBUG sections!

It's a really bad habit to get into, and it leads to things like:

#if DEBUG

bool x = Divide(1,1);

#else

Divide(1,0);

#endif

"It always works in the debug version, but crashes every time in the release build!"

As for the previous notes about adding "return" or "continue" into a function or loop to allow a breakpoint at the end of the function or as the loop is about to loop, I seem to remember that you can actually put a breakpoint on the line with the closing brace.

re: Write-only variables considered harmful? Or beneficial?

Stefan Wenig — Mon, 23 Apr 2007 22:52:26 GMT

I know a few people who would definately answer that with: you should do test-driven development, not debugging. And I tend to agree with them more and more...

re: Write-only variables considered harmful? Or beneficial?

Chris Moorhouse — Mon, 23 Apr 2007 21:57:40 GMT

So, then, would supporting the Command - Query Separation Principle by not writing or calling functions with the TrySomething pattern be desirable, in your opinion?

If the example returned something more significant than a success flag in addition to the out parameter, would your reasoning regarding write-only variables to support debugging still stand?

re: Write-only variables considered harmful? Or beneficial?

Shog9 — Mon, 23 Apr 2007 19:47:17 GMT

> This is because there is no good way in the Visual Studio debugger to say "show me the return value of the last function call".

And of course, the VC++ debugger *did* show me just that - which made that nested function call example a very easy thing to watch. 'Till reading this post, i'd assumed that the C# debugger also supported it, and that i'd just screwed up my installation horribly in some improbable way... thanks for simultaneously breaking my heart and restoring my confidence in my own sanity.