Tuple constructors cannot be used as arguments to lambdas

[jasagredo]

Other constructors can be provided in the lambda arguments as follows:

?- call(\g(A,B)^A^true, g(a,b), X).
   X = a.

However trying to do so with tuples doesn't seem to work:

?- call(\A-B^A^true, a-b, X).
   error(existence_error(procedure,(-)/4),(-)/4).

not even in canonical form or wrapping the argument in parentheses:

?- call(\(A-B)^A^true, a-b, X).
   false.
?- call(\-(A,B)^A^true, a-b, X).
   false.
?- call(\-(A,B)^A^true, -(a,b), X).
   false.

Is this intentional? It is possible to just de-construct the tuple inside the lambda of course.

And as a side note perhaps related, I find the fact that you can't wrap arguments in parenthesis surprising, perhaps it is also intentional?

?- call(\(g(A,B))^A^true, g(a,b), X). % wrapping the complex argument results in false
   false.
?- call(\g(A,B)^(A)^true, g(a,b), X). % I can wrap simple variables
   X = a.

[UWN]

?- call(\A-B^A^true, a-b, X).
   error(existence_error(procedure,(-)/4),(-)/4).
?- call(\ (A-B)^A^true, a-b, X).
   X = a.

not even in canonical form or wrapping the argument in parentheses

The cases you give are not canonical form. What you probably mean was functional notation just for \ alone. That would be

?- call(\( (A-B)^A^true ), a-b, X).
   X = a.

[jasagredo]

I meant "canonical form of the tuple", but yeah, that space you wrote there seems to be the key I was missing, the correct one:

?- write_canonical(\ (A-B)^A^true).
\(^(-(_A,_B),^(_A,true)))   true.

\ --> ^ --> A-B
      |
      +---> ^ --> A
            |
            +---> true

Here we parse \(A-B) as the first argument to ^:

?- write_canonical(\(A-B)^A^true).
^(\(-(_A,_B)),^(_A,true))   true.

^ --> \ --> A-B
|
+---> ^ --> A
      |
      +---> true

And this one is just garbage, we parse \A and then B^A^true, in the two components of a tuple:

?- write_canonical(\A-B^A^true).
-(\(_A),^(_B,^(_A,true)))   true.

- --> \ --> A
|
+---> ^ --> B
      |
      +---> ^ --> A
            |
            +---> true

Thanks for the quick response!

[UWN]

-(\(_594635),^(_594637,^(_594635,true)))

These variables are very ugly. I thought that #2163 fixed this. What is worse is that if you just enter a new scryer I get:

?- write_canonical(\A-B^A^true).
-(\(_39),^(_41,^(_39,true)))   true.

and thus more readable names, but after several interactions this monster-names reappear. Can you narrow this down?

[jasagredo]

I manually replaced the big numbers by _A and _B in the other code snippets to make it readable, but I forgot about that last one.

[UWN]

What I meant was that these big underscore-numbers are an indication for something stateful within Scryer (vulgo a bug). And that we need to narrow this down such that Mark can look at it.

[UWN]

The \ is responsible for renaming. It renames all variables in its argument.

The ^ is responsible for parameter passing.