[Timber] Bindings

[Johan Nordlander]

Well, this is not a hard problem in the technical sense, it's just  
"hard" to choose between the properties we want.

The core of the problem is whether a binding

   x = exp

should be seen as an extreme form of a pattern binding

   pat = exp

(letting pat be the trivial pattern x), or an extreme function binding

   f pat11 .. pat1n | qual1 = exp1
   ..
   f patj1 .. patjn | qualj = expj

(letting f = x and n = 0).

Timber has so far followed Haskell by taking the secound route,  
because it enables the use of guards as in

   msg | null params = "Hello anybody!"
       | otherwise   = "Hello " ++ concat params

(Recall that bindings can occur in local scopes as well, where  
variables like "params" might be statically unknown.)

But of course, this example could just as well be written

   msg = if null params then "Hello anybody!" else "Hello " ++ concat  
params

Which perhaps is not a high price to pay in order to forbid the  
obviously erroneous variant

   msg = "Hello anybody!"
   msg = "Hello everybody!"

And come to think of it: there should be nothing to stop us from  
switching to the first route above *and* extending the syntax of  
pattern bindings to

   pat | qual = exp

This would meet both demands, I believe.

In any case, I see this as yet another example of the need for a  
thorough overhaul of Timber's concrete syntax!

-- Johan

> Hi
>
> On May 3, 2010, at 2:54 AM, Andrey Kruglyak wrote:
>
>> Oh, when two name bindings follow one another, this is probably  
>> seen as an extreme form of pattern matching, just as when we define  
>> a function on more than one line.
>
> Yes, I think this could be seen as an extreme for overlapping  
> patterns. At least Haskell makes an attempt warning the programmer  
> of such patterns, I do not know if it is left out in timberc  
> deliberately or if it is to be considered a bug. Johan once  
> mentioned this as a hard problem to solve in general, but at least  
> some obvious overlapping patterns should be brought to the  
> programmers attention.
>
> /Per
>
>
>
>
>
>>
>> On the other hand, maybe the compiler should detect that there are  
>> no arguments and consider this an error? And what is the correct  
>> behavior for:
>>
>> A x = z
>> B x = z
>>
>> (assuming z is a value of the data type D = A Int | B Int)? Should  
>> the first line cause an exception if z is B 4, or should  the  
>> second line succeed?
>>
>> Andrey
>> Sent from my iPhone
>>
>> On May 3, 2010, at 1:44, Andrey Kruglyak <kruglyak at mac.com> wrote:
>>
>>> Hi!
>>>
>>> This must be a bug (and one that should be fixed ASAP), not only  
>>> because it's not how we define bindings (a name can only be bound  
>>> once), but also because if we separate the two definitions of  
>>> "temp" we do get a compile error "duplicate variable":
>>>
>>> import POSIX
>>>
>>> root :: RootType
>>> root w = do
>>> temp = "12"
>>> env = new posix w
>>> temp = "abc"
>>> env.stdout.write (show temp)
>>>
>>> I hope that Johan can fix it soon.
>>>
>>> Regards,
>>> Andrey
>>>
>>> On May 2, 2010, at 23:41, Per Lindgren wrote:
>>>
>>>> Hi
>>>>
>>>> Consider:
>>>> -------------
>>>> module Err1 where
>>>>
>>>> import POSIX
>>>>
>>>> root :: RootType
>>>> root w = do
>>>>     env = new posix w
>>>>     temp = "12"
>>>>     temp = "abc"
>>>>     env.stdout.write (show temp)
>>>>
>>>> What is the defined behavior?
>>>>
>>>> (The result using timberc is no compiler error and the output  
>>>> "12" when executed. If that is what's intended, then we might be  
>>>> in for some confused programmers. )
>>>>
>>>> /Per
>>>>
>>>>
>>>> _______________________________________________
>>>> Timber mailing list
>>>> Timber at haskell.org
>>>> http://www.haskell.org/mailman/listinfo/timber
>>>
>>> ------------------------------------------------------------------------
>>> Andrey Kruglyak
>>> PhD student
>>>
>>> http://www.andreykruglyak.com
>>>
>>> Dept. of Computer Science and Electrical Engineering
>>> Lulea University of Technology, Lulea, Sweden
>>> T +46 920 49 23 57 (office)
>>> T +46 706 16 76 03 (mobile)
>>> Office A2314 (A-huset)
>>>
>>> _______________________________________________
>>> Timber mailing list
>>> Timber at haskell.org
>>> http://www.haskell.org/mailman/listinfo/timber
>> _______________________________________________
>> Timber mailing list
>> Timber at haskell.org
>> http://www.haskell.org/mailman/listinfo/timber
>
> _______________________________________________
> Timber mailing list
> Timber at haskell.org
> http://www.haskell.org/mailman/listinfo/timber
>