[If someone has counterexamples to this -- cases where a "?" postfix operator that _isn't_ immediately followed by "." or "[" is a significant convenience -- please post them!]
The version of the hack I'm using myself is extended to allow '?(' -- which works exactly like the function call version of CoffeeScript's existential operator. It is occasionally useful -- allowing things like
drawfun[type]?(record)
to evaluate as
(drawfun[type] or _nullop)(record)
Like the ?. and ?[ patches, it can be implemented cheaply, as a small patch to primaryexp(). But my current implementation is a fairly dirty kludge. In the case that drawfun[type] is missing, I'll actually evaluate the symbol '_nullop'. This works great as long as we can trust _nullop to be defined, but has the potential to cause all sorts of trouble otherwise. I'm sure there are cleaner ways of writing it up -- but all the options I can think of are either less efficient, or involve much larger diffs. And even with a clean implementation, I'm not entirely convinced ?( is a good feature to add to a language. The obscurity/usefulness tradeoff doesn't seem nearly as favorable as the safe navigation operators'.
-Sven
