Re: More about ADTs

Yes, it could be problematic if we extend the ambigious case in 9305871. Begin with:

x ? [y] - z

Ok, using the "precedence" rule, we reduce this to:

(x?[y]) - z

and go on:

x ? [y] - z : v

Oops, it was a ? : operator afterall. Too late now; have to report the ':' as a syntax error.

It'd be necessary to use a parenthesis:

x ? ([y] - z) : v

Essentially we get the rule: If the second expression in a ? : operator begins with a type cast, it has to be surrounded by parenthesis. Not that that will happen often, but it's a very ugly special case to document in the manual.

/ Martin Stjernholm, Roxen IS

Previous text:

2002-11-12 21:52: Subject: Re: More about ADTs

---

It feels like the beginning of foo?[bar]=zonk():zerblat().

/ Martin Nilsson (Fake Build Master)