Use PresentLink for verbatim patterns

[ngeiswei]

I wonder if we could use PresentLink to avoid most quotations.

For instance, currently

(cog-execute! (Get (Present (Evaluation (GroundedPredicate "dummy") (Variable "$X")))))

raises an exception because it attempts to run the grounded predicate.

Wouldn't it be cool if it were instead equivalent to

(cog-execute! (Get (Quote (Evaluation (GroundedPredicate "dummy") (Unquote (Variable "$X"))))))

?

[linas]

Knee-jerk reaction is that it's a good idea.

Be forewarned that the code the code that implements AbsentLink is very old, predating most of the newer theory for the pattern matcher. There's another bug that shujing opened that pointed out that NotLink PresentLink is broken, and etc. I tried to fix this and making everything consistent, efficient, and kept breaking half-a-dozen unit tests no matter how I rearranged the code. Then I ran out of time.

So, sure, what you propose sounds like a pretty good idea, at the moment. Not sure just how easy or hard it would be to hack it. Making all the booleans work correctly with present/absent won't be easy.

[linas]

Here is my half-finished attempt to cleanup/modernize absentlink https://github.com/linas/atomspace/tree/absent-link-rework

[linas]

See also #1596

[ngeiswei]

Cool! I'll have a look at your old branch see if I can take a stab at it.

[linas]

Old branch: https://github.com/linas/atomspace/tree/absent-link-rework

[linas]

I think this is fixed in pull req #2172 -- a minimal proof that it works is in PresentLinkUTest

[ngeiswei]

Excellent, will test it in more depth as I port my code to use PresentLink over the next few weeks.

[linas]

OK, please close this issue as soon as you see fit; if you find problems, please open a new issue.

[ngeiswei]

@leungmanhin @amebel @Necr0x0Der @vsbogd and others, I give you the green-light to use PresentLink in your patterns. It makes patterns more readable, quotation less needed and helps the forward chainer to select rule premises.

[leungmanhin]

Awesome! I'll try it out

[linas]

@ngeiswei -- next step: use FormulaLink instead of GroundedPredicate "scm:foo" for the arithmetic formulas. I think it will make the PLN formulas more readable. Faster too, as it will avoid a C++-to-guile transition.