Support Anonymous variables in GetLink BindLink

[linas]

Currently, the pattern matcher requires that all things to be searched for have to be given a name, by using a VariableNode or a GlobNode. But this is not really needed: it could be done anonymously, thus making the patterns more compact and a bit easier to read. For example, a current search is of the form:

(Get
   (TypedVariable (Variable "X") (Type 'ConceptNode))
   (Present
      (Evaluation (Predicate "P")
         (List (Concept "foo") (Variable "X")))))

It could be done anonlymously:

(Get
   (Present
      (Evaluation (Predicate "P")
         (List (Concept "foo") (Signature (Type 'ConceptNode))))))

with the SignatureLink indicating what to look for.

Implementing this should be pretty easy: scan the clauses for SignatureLinks, and under the covers, create a corresponding type VariableNode with some UUID variable name. Do NOT put the variable into the main atomspace; but when the search is to be run, poke it into the transient atomspace, and do the search. ... and pretty much that's it ... everything else runs as before.

This is somewhat related to enhancment request #2599 which scans for variable declarations not in the VariableList ... this is effectively doing the same thing, only it scans for SignatureLinks, and treats them like variables... Huh.

[ngeiswei]

That's nice.

How would you represent queries involving repeated occurrences of the same variable though, like

(Get
  (TypedVariable (Variable "$X") (Type 'ConceptNode))
  (Present
    (Inheritance (Variable "$X") (Concept "A"))
    (Inheritance (Variable "$X") (Concept "B"))))

?

[ngeiswei]

I guess one would need to introduce adequate combinators, as in combinatory logic.

[linas]

How would you represent

Sometimes, things do need to be named. In order to say "I want this thing to be the same as that thing", you need to point: either with a finger, or a variable.

combinators

I'm doubtful, having seen them in lambda calculus. But I don't know how one might write graphs in this fashion, maybe something interesting happens.

[linas]

Anyway, this is maybe part of an effort to sheaf-ify everything. In case its not clear, plugging things into variables is just a special case of connecting together sheaf sections. (I started writing a PDF about this and then got bored.)