Thanks for clearing this up. I had encountered similar error messages and figured that it was due to an incomplete implementation, but I didn't know what the reasoning was. Perhaps, a comment could be added to Prelude.curry explaining the situation?
On Apr 19, 2013, at 8:21 AM, Michael Hanus <mh_at_informatik.uni-kiel.de> wrote:
> On 04/19/2013 12:34 PM, Wolfgang Jeltsch wrote:
>> I have a problem with using KiCS2. I have a source file that contains
>> the following code:
>>
>>> sublist :: [a] -> [a]
>>> sublist (pre ++ sub ++ post) = sub
>>
>> Now I load this into KiCS2 and enter the expression
>>
>> findall (\sub -> sub =:= sublist "Hallo!")
>>
>> at the prompt. KiCS2 answers with the message
>>
>> Main: external_ndho_C_try .
>>
>> What does this mean? Does this indicate a bug?
>
> The error message is not good and should be improved.
> Actually, the primitive encapsulation operator "try"
> which is used in the standard prelude to implement
> "findall" is not implemented in KiCS2.
>
> Since appropriate definitions of search operators
> is an ongoing topic for discussion (since the "naive"
> inclusion of findall in the first version of Curry),
> KiCS2 supports two more appropriate alternatives
> to encapsulate non-deterministic computations:
>
> 1. Strong encapsulation: This means that all potential
> non-determinism is encapsulated. Since this might
> result in dependencies on the evaluation strategy
> (if non-deterministic computations are defined "outside"
> and passed "inside" an encapsulation operator), this kind
> of encapsulation is only available as I/O operations.
> For instance, the library AllSolutions provides
>
> getAllSolutions :: (a -> Success) -> IO [a]
>
> so that you can compute all your values by
>
> getAllSolutions (\sub -> sub =:= sublist "Hallo!") >>= print
>
> There is also the library SearchTree which provides
> programmable search strategies (depth-first, breadth-first,
> iterative deepening).
>
> 2. Weak encapsulation: This means that only the non-determinism
> defined inside an encapsulation operator is encapsulated.
> Since this is independent on the evaluation strategy,
> these concept can be also used independent of the I/O monad.
> Conceptually, these operators are offered as "set functions"
> which compute the set of all results but do not encapsulate
> non-determinism in the actual arguments. KiCS2 provides this
> functionality by the library SetFunctions. For instance, you
> can compute the set of all your results by
>
> (set1 sublist "Hallo!")
>
> Since the result is a set, you can only map or fold the results
> (see operations in the library SetFunctions) or transform them
> into a list by sorting them (of course, this works only for
> finite search spaces).
>
> I hope this explanation helps. If you need more infos or pointers
> to the literature, please let me know.
>
> Best regards,
>
> Michael
>
> _______________________________________________
> curry mailing list
> curry_at_lists.RWTH-Aachen.DE
> http://MailMan.RWTH-Aachen.DE/mailman/listinfo/curry
Received on Fr Apr 19 2013 - 18:06:03 CEST