setof/3 incorrect

[UWN]

Item#42

[UWN]

This is important.

[infradig]

I don't doubt it. It seems to be common problem though, from a common source.

…

On Tue, Jul 4, 2023 at 4:12 PM UWN ***@***.***> wrote: This *is* important. — Reply to this email directly, view it on GitHub <#233 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFNKSEVWJ7ZXZ622IVBB2OLXOOX6XANCNFSM6AAAAAAZ3JEL6M> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[UWN]

Common? Apart from Scryer, I do not see another system with this problem.

[infradig]

B Prolog. CxProlog. Tau Prolog.

…

On Wed, Jul 5, 2023 at 12:32 AM UWN ***@***.***> wrote: Common? Apart from Scryer, I do not see another system with this problem. — Reply to this email directly, view it on GitHub <#233 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFNKSETSJ3RN3ZKLELXTJV3XOQSRHANCNFSM6AAAAAAZ3JEL6M> . You are receiving this because you commented.Message ID: ***@***.***>

[UWN]

Thank you, I'm getting sloppy... . But at least I corrected my test cases.

[UWN]

Tau works with setof/3, but has a bug with handling underscore variables. tau-prolog/tau-prolog#354

[infradig]

I've been looking into this. Trealla like a number of systems uses a version of ROK's aggregates.pl, which I thought may be faulty here. But I looked at Ciao which behaves correctly, and it does the same (but uses an older version, I believe, that does assert/retract temporarily). Replacing Trealla's code with that code and... it's still wrong 😢. That said, I think the fault is in replace_instance_/5, possibly to do with arg/3.

…

On Sun, Jul 2, 2023 at 5:44 PM UWN ***@***.***> wrote: Item#42 <http://www.complang.tuwien.ac.at/ulrich/trealla-prolog/Item#42> — Reply to this email directly, view it on GitHub <#233>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFNKSER6PSSQXQTRRMHJVB3XOERGHANCNFSM6AAAAAAZ3JEL6M> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[infradig]

I write the above mainly as a mental note to myself.

…

On Sun, Jul 2, 2023 at 5:44 PM UWN ***@***.***> wrote: Item#42 <http://www.complang.tuwien.ac.at/ulrich/trealla-prolog/Item#42> — Reply to this email directly, view it on GitHub <#233>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFNKSER6PSSQXQTRRMHJVB3XOERGHANCNFSM6AAAAAAZ3JEL6M> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[pmoura]

I added those two tests to the latest Logtalk git version: LogtalkDotOrg/logtalk3@cc74f95. I bit tricky to adapt given that order of solutions returned by bagof/3 is undefined per the ISO Prolog Core standard (not to mention having to use bagof/3 to collect all the solutions to the test goal so that variables are preserved).

[infradig]

I have a fix, indeed was arg/3. I thank RO'K for the clue.

…

On Sun, Jul 2, 2023 at 5:44 PM UWN ***@***.***> wrote: Item#42 <http://www.complang.tuwien.ac.at/ulrich/trealla-prolog/Item#42> — Reply to this email directly, view it on GitHub <#233>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFNKSER6PSSQXQTRRMHJVB3XOERGHANCNFSM6AAAAAAZ3JEL6M> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[UWN]

Now it's

?- L=[_,_,_], setof(t,member(E,L),_).
   L = [_A,_,_], E = _, unexpected, unsound.

These anonymous underscores in answers are all odd. Sometimes they appear, and sometimes they are named. It would be ideal to have no anonymous underscores in answers, similar to SICStus, Scryer.

[infradig]

Yeah, it's coming. But not until tomorrow.

…

On Wed, 5 July 2023, 21:09 UWN, ***@***.***> wrote: Now it's ?- L=[_,_,_], setof(t,member(E,L),_). L = [_A,_,_], E = _, unexpected, unsound. These anonymous underscores in answers are all odd. Sometimes they appear, and sometimes they are named. It would be ideal to have no anonymous underscores in answers, similar to SICStus, Scryer. — Reply to this email directly, view it on GitHub <#233 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFNKSEWLW4RPAMHRYM76T3TXOVDODANCNFSM6AAAAAAZ3JEL6M> . You are receiving this because you commented.Message ID: ***@***.***>

[pmoura]

$ logtalk_tester -p trealla
% Batch testing started @ 2023-07-05 23:05:17
%         Logtalk version: 3.68.0-b01
%         Trealla Prolog version: 2.21.16
%
% Documents/Logtalk/logtalk3/tests/prolog/predicates/setof_3
%         40 tests: 2 skipped, 38 passed, 0 failed (0 flaky)
%         completed tests from object tests in 1 second
%         clause coverage n/a
%
% Skipped tests
Documents/Logtalk/logtalk3/tests/prolog/predicates/setof_3/tests.lgt - eddbali_setof_3_25 @ tests
Documents/Logtalk/logtalk3/tests/prolog/predicates/setof_3/tests.lgt - lgt_setof_3_36 @ tests
%
% 1 test sets: 1 completed, 0 skipped, 0 broken, 0 timedout, 0 crashed
% 40 tests: 2 skipped, 38 passed, 0 failed (0 flaky)
%
% Batch testing ended @ 2023-07-05 23:05:19

[UWN]

?- L=[_,_,_], setof(t,member(E,L),_).
   L = [_A,_B,_C], E = _A
;  L = [_A,_B,_C], E = _B
;  L = [_A,_B,_C], E = _C.

It's all perfect, but why and when do you write these intermediary variables? In other words, why not the same as in

?- length(L,3), setof(t,member(E,L),_).
   L = [E,_A,_B]
;  L = [_A,E,_B]
;  L = [_A,_B,E].

(Could be, that you treat the underscores as kind-of-named variables)

[infradig]

It must be the order the variables are bound.

…

On Thu, 6 July 2023, 16:27 UWN, ***@***.***> wrote: ?- L=[_,_,_], setof(t,member(E,L),_). L = [_A,_B,_C], E = _A ; L = [_A,_B,_C], E = _B ; L = [_A,_B,_C], E = _C. It's all perfect, but why and when do you write these intermediary variables? In other words, why not the same as in ?- length(L,3), setof(t,member(E,L),_). L = [E,_A,_B] ; L = [_A,E,_B] ; L = [_A,_B,E]. (Could be, that you treat the underscores as kind-of-named variables) — Reply to this email directly, view it on GitHub <#233 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFNKSEVKNNGERRDGE6YB7NDXOZLF5ANCNFSM6AAAAAAZ3JEL6M> . You are receiving this because you commented.Message ID: ***@***.***>

[UWN]

?- length(L,3), setof(t,member(E,L),_),L=[_,_,_].
   L = [E,_A,_B]
;  L = [_A,E,_B]
;  L = [_A,_B,E].
?- length(L,3), L=[_,_,_],setof(t,member(E,L),_).
   L = [_A,_B,_C], E = _A
;  L = [_A,_B,_C], E = _B
;  L = [_A,_B,_C], E = _C.

What do you mean by the order the variables are bound?

[infradig]

Yes, one would think these first two would be the same, if not the third as well.. /trealla (devel) $ tpl Trealla Prolog (c) Infradig 2020-2023, v2.21.16-dirty ?- length(L,3), setof(t,member(E,L),_). L = [E,_A,_B] ; L = [_A,E,_B] ; L = [_A,_B,E]. ?- L=[_,_,_], setof(t,member(E,L),_). L = [_A,_B,_C], E = _A ; L = [_A,_B,_C], E = _B ; L = [_A,_B,_C], E = _C. ?- L=[A,B,C], setof(t,member(E,L),_). L = [A,B,C], E = A ; L = [A,B,C], E = B ; L = [A,B,C], E = C. ?-

…

On Thu, Jul 6, 2023 at 5:27 PM UWN ***@***.***> wrote: ?- length(L,3), setof(t,member(E,L),_),L=[_,_,_]. L = [E,_A,_B] ; L = [_A,E,_B] ; L = [_A,_B,E]. ?- length(L,3), L=[_,_,_],setof(t,member(E,L),_). L = [_A,_B,_C], E = _A ; L = [_A,_B,_C], E = _B ; L = [_A,_B,_C], E = _C. What do you mean by *the order the variables are bound*? — Reply to this email directly, view it on GitHub <#233 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFNKSET24ZYRXNBFJZAWBQ3XOZSELANCNFSM6AAAAAAZ3JEL6M> . You are receiving this because you commented.Message ID: ***@***.***>

[UWN]

In the third case, there are three times two named variables which are unified. So in this situation, one has to state that E = A or A = E &ct. So it is inevitable to have two answer substitutions. But not for the first two cases. (The reason this interests me so much is my suspicion that this differences is somehow induced by the internal representation. So some things that should not be visible, become visible. After all, we have still #216 open.