Add cond link

[kasimebrahim]

As per opencog/asmoses#46

[ngeiswei]

Suggested change

	DEFINE_LINK_FACTORY(CondLink, COND_LINK)
	DEFINE_LINK_FACTORY(CondLink, COND_LINK)

[linas]

What Nil says. Your editor seems to forget to inert a newline.

[linas]

Looks very nice! I don't really have any complaints; I don't have any major complaints; all my remarks are about style, rather than design. Fix them up, and I have no problems with a merge.

[linas]

Instead of this check, and the checks below, it would be better to check if the even atoms are evaluatable, and the odd ones are executable. nameserver().isA(type, EVALUATABLE_LINK) and nameserver.isA(type, EXECUTABLE_LINK) The EVALUATABLE_LINK type already exists, and "does the right thing"; the EXECUTABLE_LINK does not yet exist) (I started adding it a few weeks ago, but did not finish. Perhaps you can add it, in a separate pull req?)

[linas]

This is not wrong, but there is now a new and better way of doing this: call exps[i]->execute(scratch).

This generic execute() method is brand new, introduced just a week ago, and it is meant to resolve multiple issues with calling the instantiator directly.

[kasimebrahim]

Hi @linas, In case of eager execution the expressions in the CondLink might have been already executed right? such as an ArithmeticLink might have been already reduced to a NumberNode and calling execute on that NumberNode should throw the "Not Executable" exception in Atom.h? or am I missing something?

[linas]

Yes, for eager evaluation, the arguments will already have been reduced. There should not be any exceptions as a result, as the reduction is (meant to be) performed only once, not twice.

Lay evaluation would be better; it almost works; there's some open issue saying "must finish this work. Anyway, toe CondLink code looks OK to me.

[linas]

What Nil says. Your editor seems to forget to inert a newline.

[noskill]

I suggest adding one test for EvaluationLink with grounded predicate

[kasimebrahim]

Hi @noskill my scheme knowledge is very limited, will this suffice

(define (grounded_cond1 ) (cog-new-stv 0 0))

(define (grounded_cond2 ) (cog-new-stv 1 1))

(define grounded-cond
    (CondLink
	(Evaluation
	     (GroundedPredicate "scm:grounded_cond1")
	     (List ))
	(NumberNode 1)

	(Evaluation
	     (GroundedPredicate "scm:grounded_cond2")
	     (List ))
	(NumberNode 2)))

[linas]

looks like it was added

[noskill]

Yes, looks good, thanks

[linas]

Thanks! This looks good. Sorry for all the earlier confusion about whether this was needed or not.

[noskill]

Is it the same as?:

else if false: 
     (PlusLink (Number 1)(Number 2))

[kasimebrahim]

yep

[noskill]

Is there an example of common usecase:

if (expression) 
     some code
else
      some other code

?

[kasimebrahim]

do you mean like this?

(CondLink
  |expression|                  ; EvaluatableLink
  |some-code|                  ; if |expression| evaluates to true
  |some-other-code|          ; else
)

https://wiki.opencog.org/w/CondLink

[noskill]

Yes