I’m thinking about how to save/load game state in SWI-Prolog. Here is the code skeleton of my game: main.pl :- dynamic script_iterator/2. init_iterator(Goal, Iterator) :- reset(Goal,YE,Cont), ( Cont == 0 -> Iterator = done ; YE = yield(Element) -> Iterator = next(Element, Cont) …

[image] jan: These days you can do such things using virtual machines. Sounds a bit like an overkill. As much as it is overkill it is practical, works and often used as a starting or stop gap measure but once in place and working tends to be left as is. I asked a related question concerning …

[image] jan: A continuation may be enough, but note that it doesn’t capture the entire state. Notably choice points are not part of it. I think so too. Other game data are all first-order (no closures, low-level pointers, etc), they can be serialized easily. [image] jan: You could work a…

[image] chansey97: Do you mean to modify the delimited continuations’ source code? This is a bit of a challenge for me :smiling_face: (this may require more WAM knowledge?). No. The continuation is just a Prolog term. The only problem is that it contains references to clauses that are in th…

[image] jan: Or are you also going to implement the whole continuation stuff in the meta interpreter? Yes, that is what I mean. Just give the meta interpreter an additional argument for continuations (a stack of goals), see Schrijvers et al., 2013 ( preprint PDF ) section 3 for details.

[image] jan: Now <clause>(0x55e67cad9680), is a reference to the t2/0 clause. You can figure that out using nth_clause/3 or clause_property/2 . So, you turn that into e.g. '$CLAUSE'(t20, 1) to indicate the clause. Then you have a term you can save. Now at restore you find the clause reference and…

I just used compound_name_arguments/3 to compose/decompose continuations: ?- nth_clause(t2, 1, Reference), compound_name_arguments(ContInternal, '$cont$', [[], Reference, 9 | []]), Cont = call_continuation([ContInternal]), call(Cont). b Reference = <clause>(0000000000354290), ContInternal = $con…

[image] chansey97: Yes, that is what I mean. Just give the meta interpreter an additional argument for continuations (a stack of goals), see Schrijvers et al., 2013 ( preprint PDF ) section 3 for details. You’d have to try, but I fear it is way too slow. I don’t recall all the details, but m…

Thanks. Disturb you again, could you explain a little further (it is not related to the topic though): It knows the term created by shift/1 has no internal references, while the one you have does. What do you mean by “the term created by shift/1 has no internal references” ? P.s. I also success…

A Prolog term is one of the usual things (atom, int, …, compound, var). Internally, there is one more thing: references. You can see the need for this best after running A = B. Now A and B become the same variable. This is done by making A a reference to B. Binding the combined variable wi…

[image] jan: there is one more thing: references Thanks. Is the reference you mentioned related to e.g. <clause>(0000000000354290)? My understanding is that the term created by shift/1 includes an actual reference, i.e. <clause>(0000000000354290), instead of a logic variable Reference, but …

Is there a way to serialize continuation or program execution state?

Help!

EricGT June 23, 2022, 2:20pm 15

Is a reference really a blob? e.g. Blob types

Topic		Replies	Views
DCGs implemented with delimited continuations Help!	16	1785	August 23, 2022
Solving two consecutive dependent goals from command line Help!	10	1237	September 25, 2019
Dealing with state Discussion	32	1324	August 11, 2024
Save prolog execution state Resources discussion	3	696	October 31, 2020
Naming clarification about delimited continuation (reset/3 and shift/1) Discussion discussion , nice-to-know , continuations	8	1167	July 4, 2022

Is there a way to serialize continuation or program execution state?

Related topics