Trying tpartition/4

[triska]

Hi all,

today I saw a discussion about an unexpected answer of partition/4 that prompted me to try the given example with tpartition/4 from library(reif):

?- tpartition(=([X]), [1,2,[3],4,[5]], As, Bs).
   X = 3, As = [[3]], Bs = [1,2,4,[5]]
;  X = 5, As = [[5]], Bs = [1,2,[3],4]
;  As = [], Bs = [1,2,[3],4,[5]], dif:dif([X],[3]), dif:dif([X],[5]).

It is immediately clear that this is an immense improvement over partition/4, already from the two additional answers we see. From them, the partitioning is also easier to understand, and in particular that As = [[3],[5]] can impossibly arise in any solution, because the X in [X] cannot be both 3 and 5.

The predicates from library(reif) are really worth a look, they address long-standing incompleteness issues of many earlier library predicates such as include/3, exclude/3 and partition/4 which are noteworthy for their incorrectness when used with non-ground conditions.

All the best,
Markus

[aarroyoc]

Is there a video about library(reif) on your YouTube channel? Maybe it's a good idea to do one about that!

[bakaq]

I think this would be very useful for the community. The reif paper is very good, but it lacks practical examples for a bunch of the predicates. I had to learn how most of it works by trial and error. It's very nice now that I know it works, but it took a lot to get there. And I still have no idea of the use case of cond_t/3 for example. One thing that I especially had some difficulty is how to define reified versions of a predicate, which is important in a lot of use cases of library(reif).

[UWN]

The largest collection of examples is on SO.