>> I feel like Prolog is sold by many as being able to solve predicates in both/many directions, with all the grandfather/ancestor/append examples, and my apparent frustration is in trying to use it that way in general and finding that doesn't work and that trying to produce working predicates that do work that way is seemingly impossible (without clpfd or cuts).

You're not wrong about that and it's not your fault, or any misunderstanding. Certainly Prolog has been "sold" this way, by some. I can't name those "some" because I'm not sure who they are exactly (although I have a few specific people in mind). I don't think it's something that's made a big issue of in the usually recommended textbooks (Clocksin and Mellish, Sterling and Shapiro, O'Keefe, Bratko) but I can't be sure.

Obviously, you can run some programs in multiple modes, in Prolog. You can't always write every program to be all-modes. Pretending that it's possible to do so is overselling the language. What's more it's trying to sell a nice feature, as a defining feature. The idea I think is that First-Order Logic predicates are relations and so should not have "inputs" and "outputs". But Prolog predicates are not FOL predicates. If they were, they'd be simultaneously capable of being run in all modes, and less expressive than they are now (because Prolog is higher-order, as I say in other comments, trying not to sound like a dangerous loon).

I think this is the age-old debate about purity and formal rigour in Prolog, that has never done the language any good. There has always been a part of the Prolog community who wanted Prolog to be some kind of pure and perfect jewel, that should remain untouched from the real world. There's always been another part, who wanted to treat Prolog as a programming language to do real work with. I get the feeling that the latter, more practically-minded, folks have slowly weeded themselves out and now the majority of those who remain are of the "purist" camp. Entirely coincidentally, the number of people picking up Prolog has also constantly dwindled.

Entirely coincidentally.

You can probably tell which camp I'm on. Actually, I'm not. I got a foot in each. I serve two masters and all that. I just want to know that things work, that's my schtick. Sometimes you need formal rigor, sometimes you just need to bodge some codes together. And use the cut :P

>> I'd be happy to engage in offline conversation about it, but I'd likely just be yelling at clouds. I am interested in discovering which logic languages can trivially provide an any-mode rle and in working up from there to larger problems. Clearly, from this thread, prolog+clpfd is one.

You're welcome to get in touch:

  eIA!patsantzis17CTHULHUimperialIA!acIA!uk

But first open vim and praise Cthulhu:

  s/IA!/./g
  s/CTHULHU/@/g