Re: APIs that have boolean arguments defaulting to true

On Mon, Nov 11, 2013 at 11:37 AM, Boris Zbarsky <bzbarsky@mit.edu> wrote:

> Dear all,
>
> One take-away I have from the current attempts to make undefined be
> treated as missing is that we should avoid creating APIs that have
> arguments of the form "optional boolean arg = true".
>
> The reason for that is that I think the behavior of such an argument when
> combined with the conventions for undefined is very surprising. Consider a
> function declared as:
>
>   void foo(optional boolean arg = foo);
>
> and these two patterns of calling it:
>
>   if (myValue) {
>     foo(true);
>   } else {
>     foo(false);
>   }
>
> and
>
>   foo(myValue);
>
> These two patterns have different behavior, which seems very unintuitive
> to me and to everyone else I've pointed this out to...
>
> This does mean that we can't have "foo()" with no arguments default to
> true, so we should probably design APIs such that "false" gives the sane
> default behavior if there's an optional boolean argument at all.
>

I would indeed recommend that as a best practice but not a requirement. For
exactly the reasons you state.



>
> Thoughts?
>
> -Boris
>
>


-- 
    Cheers,
    --MarkM

Received on Monday, 11 November 2013 20:13:32 UTC