On Tue, 2009-12-01 at 14:22 +0000, Ian Lynagh wrote:
> On Tue, Dec 01, 2009 at 12:40:33PM +0000, Duncan Coutts wrote:
> > On Mon, 2009-11-30 at 23:30 +0000, Ian Lynagh wrote:
> > > On Mon, Nov 30, 2009 at 11:05:21PM +0000, Duncan Coutts wrote:
> > > > On Mon, 2009-11-30 at 17:26 +0000, Ian Lynagh wrote:
> > > > 
> > > > > f2 x = x + case () of
> > > > >            () -> x
> > > > >            where x = 5
> > > > 
> > > > 10, same reason.
> > > 
> > > So if you look at f2 from the perspective of someone new to the
> > > language, given they understand the scoping of f1 and f3, would they
> > > expect that where binding to scope over the entire RHS rather than just
> > > the case expression? Maybe it's just me, but I don't think that would
> > > match my intuition.
> > 
> > They would know that it has to be one or the other. The new person would
> > reason that the language designers either picked > or >= for column
> > offset. Which one they picked is something a new person would have to
> > look up, or just try, or avoid.
> > 
> > >From a casual perspective there's nothing subtle about this case, it
> > does just look like a question of an arbitrary choice between >= vs >. 
> > Programmers don't (and do not have to) realise that in reality parsing
> > this relies on the error rule.
> 
> Let me try asking it another way: If in H98 the above was an error, and
> we were discussing what to change it to, how would you rank the options?
> 
> My answer would be:
> 
> 1: Keep it as an error; no real need/benefit to make it a special case
> 2: The bindings should scope over the case expression only, as the where
>    clause looks like it "belongs" with the case expression
> 3: Scope over the whole function

I would put 2 last. We require things to be more indented to indicate
that they "belong". If we allowed things to "belong" at the same
indentation we'd write thing like:

foo x = ...
where
  y = ...

As for 1 vs 3, if H'98 had always made it an error then I would not
argue especially strongly that it should be allowed rather than continue
to be declared a layout error. That said, if the proposal was to make it
meaningful then I would not object since it would not break existing
programs, would not be visually ambiguous in most cases and would allow
a more compact layout (using less horizontal space).

Duncan

_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc

Reply via email to