CHR: constraints allow hyps, not considered in the clique error check

[FissoreD]

The grammar has been extended to support the new behavior

A test has been added

[gares]

Please define a type with names so that there is no risk of confusing the hyps with the rest. Also, I think ?- is a better symbol than |- since it is the same.

[gares]

There is something I don't get. CHR.new_clique should now have a different type, a clique should be two sets on symbols, not the union of the two sets. And the code filtering the context should use only that set of predicate names.

[gares]

Also, I expect to see some code making the union of all the context filters (for the same clique).

[gares]

elpi/src/runtime.ml

Lines 3417 to 3422 in af3bb10

	match CHR.clique_of (head_of g) !chrules with
	| Some clique -> (* real constraint *)
	(* XXX head_of is weak because no clausify ? XXX *)
	delay_goal ~filter_ctx:(fun { hsrc = x } -> C.Set.mem (head_of x) clique)
	~depth prog ~goal:g (gid[@trace]) ~on:keys
	| None -> delay_goal ~depth prog ~goal:g (gid[@trace]) ~on:keys

[gares]

clique_of should also return the context filter, I guess

[FissoreD]

Yeah, I'm going in that direction:
type clique = {ctx_filter: Constants.Set.t; clique: Constants.Set.t} [@@deriving show]

[FissoreD]

I'm however a bit lost on how to correctly manage the ctx_filter

[gares]

please rebase and update the README.md