eyeprolog 1.5.62 → 1.5.65

package/README.md

@@ -69,14 +69,15 @@ printf 'human(socrates).\nmortal(X) :- human(X).\n' |
 - [Example decks](examples/deck/README.md) — explainable RDF/Prolog scenarios with reproducible roundtrips
 - [Introduction to EyeProlog](https://eyereasoner.github.io/eyeprolog/examples/deck/introduction-to-eyeprolog) — short presentation deck for first-time audiences
 - [Symbiotic Knowledge Graphs](https://eyereasoner.github.io/eyeprolog/examples/deck/symbiotic-knowledge-graphs) — RDF ↔ Prolog heatwave-response demo for human/AI/KG co-evolution
-- [rdf-prolog-roundtrip](https://github.com/eyereasoner/rdf-prolog-roundtrip) — standalone RDF 1.2 ↔ ISO Prolog bridge used by the RDF examples
+- [rdf-prolog-interchange](https://github.com/eyereasoner/rdf-prolog-interchange) — standalone RDF 1.2 ↔ ISO Prolog bridge used by the RDF examples
+- [RDF and Prolog: Two Standards-Based Legs](https://github.com/eyereasoner/rdf-prolog-interchange/blob/main/why-rdf-prolog.md) — why W3C RDF and ISO Prolog form the foundation
 - [ISO conformance review](test/conformance/ISO-COMPLIANCE.md) — supported Part 1 profile
 - [Latest Neumerkel conformity](test/conformance/NEUMERKEL-LATEST.md) — tracked result from the current live upstream inventory
 - [Conformance report](conformance-report.md) — generated executable conformance status, local corpus summary, and known deviations
 - [OpenRuleBench](openrulebench/README.md) — portable benchmark profile
 ## RDF, Prolog, and symbiotic knowledge graphs
 
-EyeProlog can sit behind an RDF knowledge graph without inventing a private graph representation. [`rdf-prolog-roundtrip`](https://github.com/eyereasoner/rdf-prolog-roundtrip) converts RDF 1.2 datasets to ordinary `rdf(Subject, Predicate, Object, Graph)` facts, EyeProlog applies portable rules, and ground `rdf/4` results can be converted back to RDF.
+EyeProlog can sit behind an RDF knowledge graph without inventing a private graph representation. [`rdf-prolog-interchange`](https://github.com/eyereasoner/rdf-prolog-interchange) converts RDF 1.2 datasets to ordinary `rdf(Subject, Predicate, Object, Graph)` facts, EyeProlog applies portable rules, and ground `rdf/4` results can be converted back to RDF.
 
 The checked [Symbiotic Knowledge Graphs example](examples/symbiotic-knowledge-graph.pl) uses named graphs and RDF 1.2 triple terms to distinguish trusted knowledge, AI-proposed statements, and human review. Its [wide-audience companion](https://eyereasoner.github.io/eyeprolog/examples/deck/symbiotic-knowledge-graphs) explains why this is a useful present-day software model for human/AI/KG co-evolution: RDF supplies shared semantic memory, Prolog supplies explicit deliberation, AI supplies new hypotheses, and people remain participants in meaning and judgment.
 

package/examples/cross-organization-data-sharing.pl

@@ -1,4 +1,4 @@
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/data/README.md

@@ -4,7 +4,7 @@ This directory contains source and materialized data that support runnable examp
 
 ## Symbiotic Knowledge Graphs
 
-The heatwave-response example uses [`rdf-prolog-roundtrip`](https://github.com/eyereasoner/rdf-prolog-roundtrip) as its RDF boundary:
+The heatwave-response example uses [`rdf-prolog-interchange`](https://github.com/eyereasoner/rdf-prolog-interchange) as its RDF boundary:
 
 ```text
 symbiotic-knowledge-graph-input.nq

package/examples/data/symbiotic-knowledge-graph-rules.pl

@@ -1,6 +1,6 @@
 % Symbiotic knowledge graph rules for EyeProlog.
 %
-% The source rdf/4 facts above are generated by rdf-prolog-roundtrip from
+% The source rdf/4 facts above are generated by rdf-prolog-interchange from
 % symbiotic-knowledge-graph-input.nq.  RDF stays the interchange and storage
 % model; these rules provide the explicit reasoning and governance layer.
 
@@ -317,10 +317,10 @@ knowledge_exchange_fact(prolog_to_rdf, materialized_ground_rdf4).
 
 pipeline_step(N, Description) :- pipeline_step_fact(N, Description).
 pipeline_step_fact(1, 'RDF 1.2 N-Quads: named graphs keep source and governance boundaries explicit').
-pipeline_step_fact(2, 'rdf-prolog-roundtrip: RDF becomes ordinary rdf/4 Prolog facts without a solver').
+pipeline_step_fact(2, 'rdf-prolog-interchange: RDF becomes ordinary rdf/4 Prolog facts without a solver').
 pipeline_step_fact(3, 'EyeProlog: ISO Prolog rules validate candidates, apply review policy, and derive actions').
 pipeline_step_fact(4, 'EyeProlog: result_rdf/4 materializes accepted knowledge and decisions as ground RDF-shaped facts').
-pipeline_step_fact(5, 'rdf-prolog-roundtrip: ground rdf/4 facts become RDF again for publication or federation').
+pipeline_step_fact(5, 'rdf-prolog-interchange: ground rdf/4 facts become RDF again for publication or federation').
 
 cognitive_parallel(Prolog, Human) :- cognitive_parallel_fact(Prolog, Human).
 cognitive_parallel_fact(fact, remembered_assertion).

package/examples/deck/introduction-to-eyeprolog.md

@@ -166,7 +166,7 @@ Boolean constraints are compact when the domain is “on/off”, “selected/not
 
 ## RDF and knowledge graphs
 
-EyeProlog pairs naturally with RDF through `rdf-prolog-roundtrip`:
+EyeProlog pairs naturally with RDF through `rdf-prolog-interchange`:
 
 ```text
 RDF dataset

package/examples/deck/symbiotic-knowledge-graphs.md

@@ -10,7 +10,7 @@ The demonstration does not claim that EyeProlog is a brain-computer interface. I
 
 The architecture uses two small projects with deliberately separate jobs:
 
-- [`rdf-prolog-roundtrip`](https://github.com/eyereasoner/rdf-prolog-roundtrip) moves RDF 1.2 datasets to and from ordinary `rdf/4` Prolog facts. It contains no Prolog solver.
+- [`rdf-prolog-interchange`](https://github.com/eyereasoner/rdf-prolog-interchange) moves RDF 1.2 datasets to and from ordinary `rdf/4` Prolog facts. It contains no Prolog solver.
 - [EyeProlog](https://github.com/eyereasoner/eyeprolog) reasons over those facts with explicit ISO Prolog rules.
 
 That separation is the central design choice. **RDF remains the shared interchange and publication layer; Prolog becomes the transparent reasoning layer.**
@@ -105,7 +105,7 @@ p2 confidencePercent 93
 
 The embedded triple is a **statement being talked about**. It is not yet an asserted operational fact.
 
-After `rdf-prolog-roundtrip`, the same structure is an ordinary Prolog term:
+After `rdf-prolog-interchange`, the same structure is an ordinary Prolog term:
 
 ```prolog
 rdf(
@@ -246,10 +246,10 @@ That closeness makes it easier for a person to enter the machine's reasoning loo
 
 Without RDF, the Prolog example could be dismissed as a hand-written toy knowledge base.
 
-With `rdf-prolog-roundtrip`, the roles are clearer:
+With `rdf-prolog-interchange`, the roles are clearer:
 
 - RDF provides Web identifiers, graph boundaries, literals, RDF 1.2 triple terms, and interoperable publication;
-- `rdf-prolog-roundtrip` preserves those structures as ordinary ISO Prolog terms;
+- `rdf-prolog-interchange` preserves those structures as ordinary ISO Prolog terms;
 - EyeProlog applies explicit rules and explores alternatives;
 - materialized conclusions return to RDF.
 
@@ -279,7 +279,7 @@ A direct neural interface would change the input/output channel. It would not re
 
 ## Reproduce the roundtrip
 
-Assuming `rdf-prolog-roundtrip` and EyeProlog are available on `PATH`:
+Assuming `rdf-prolog-interchange` and EyeProlog are available on `PATH`:
 
 ```sh
 rdf-to-prolog \
@@ -327,5 +327,5 @@ The deepest reason Prolog fits this vision is not nostalgia or syntax. It is tha
 ## References
 
 - Ruben Taelman, [*Symbiotic Knowledge Graphs: A Vision for Semantic Brain-Computer Interfaces*](https://rubensworks.github.io/article-iswc2026-vision-symbiotic-knowledge-graphs/).
-- [`eyereasoner/rdf-prolog-roundtrip`](https://github.com/eyereasoner/rdf-prolog-roundtrip) — standalone RDF 1.2 ↔ ISO Prolog roundtripping toolkit.
+- [`eyereasoner/rdf-prolog-interchange`](https://github.com/eyereasoner/rdf-prolog-interchange) — standalone RDF 1.2 ↔ ISO Prolog roundtripping toolkit.
 - [`eyereasoner/eyeprolog`](https://github.com/eyereasoner/eyeprolog) — ISO Prolog reasoning and proof engine.

package/examples/dpv-odrl-purpose-mapping.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/explainable-ev-depot-configuration.pl

@@ -1,4 +1,4 @@
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/odrl-dpv-fpv-trust-flow.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/odrl-dpv-healthcare-risk-ranked.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/odrl-dpv-risk-ranked.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/odrl-policy-advanced.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/odrl-policy.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/operational-incident-response.pl

@@ -1,4 +1,4 @@
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/output/symbiotic-knowledge-graph.pl

@@ -29,10 +29,10 @@ knowledge_exchange(machine_to_human, recommendation(after_review, riverside, ope
 knowledge_exchange(rdf_to_prolog, ordinary_rdf4_facts).
 knowledge_exchange(prolog_to_rdf, materialized_ground_rdf4).
 pipeline_step(1, 'RDF 1.2 N-Quads: named graphs keep source and governance boundaries explicit').
-pipeline_step(2, 'rdf-prolog-roundtrip: RDF becomes ordinary rdf/4 Prolog facts without a solver').
+pipeline_step(2, 'rdf-prolog-interchange: RDF becomes ordinary rdf/4 Prolog facts without a solver').
 pipeline_step(3, 'EyeProlog: ISO Prolog rules validate candidates, apply review policy, and derive actions').
 pipeline_step(4, 'EyeProlog: result_rdf/4 materializes accepted knowledge and decisions as ground RDF-shaped facts').
-pipeline_step(5, 'rdf-prolog-roundtrip: ground rdf/4 facts become RDF again for publication or federation').
+pipeline_step(5, 'rdf-prolog-interchange: ground rdf/4 facts become RDF again for publication or federation').
 cognitive_parallel(fact, remembered_assertion).
 cognitive_parallel(rule, reusable_generalization).
 cognitive_parallel(query, explicit_question).

package/examples/rdf12-annotated-claims.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/rdf12-annotation.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/rdf12-directional-language.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/rdf12-nested-triple-term.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/rdf12-trig-graph-join.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/rdf12-trig-named-graph.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/rdf12-trig-triple-term.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/rdf12-triple-term.pl

@@ -1,7 +1,7 @@
-% Adapted from rdf-prolog-roundtrip.
+% Adapted from rdf-prolog-interchange.
 % This standalone program combines generated RDF input with its ISO Prolog rules.
 
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/sbom-vulnerability-response.pl

@@ -1,4 +1,4 @@
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/scientific-evidence-graph.pl

@@ -1,4 +1,4 @@
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 

package/examples/symbiotic-knowledge-graph.pl

@@ -1,4 +1,4 @@
-% Generated by rdf-prolog-roundtrip.
+% Generated by rdf-prolog-interchange.
 
 % RDF terms use rdf(Subject, Predicate, Object, Graph).
 
@@ -63,7 +63,7 @@ rdf(iri('https://example.org/city/review/r3'), iri('https://example.org/vocab/no
 % Rules
 % Symbiotic knowledge graph rules for EyeProlog.
 %
-% The source rdf/4 facts above are generated by rdf-prolog-roundtrip from
+% The source rdf/4 facts above are generated by rdf-prolog-interchange from
 % symbiotic-knowledge-graph-input.nq.  RDF stays the interchange and storage
 % model; these rules provide the explicit reasoning and governance layer.
 
@@ -380,10 +380,10 @@ knowledge_exchange_fact(prolog_to_rdf, materialized_ground_rdf4).
 
 pipeline_step(N, Description) :- pipeline_step_fact(N, Description).
 pipeline_step_fact(1, 'RDF 1.2 N-Quads: named graphs keep source and governance boundaries explicit').
-pipeline_step_fact(2, 'rdf-prolog-roundtrip: RDF becomes ordinary rdf/4 Prolog facts without a solver').
+pipeline_step_fact(2, 'rdf-prolog-interchange: RDF becomes ordinary rdf/4 Prolog facts without a solver').
 pipeline_step_fact(3, 'EyeProlog: ISO Prolog rules validate candidates, apply review policy, and derive actions').
 pipeline_step_fact(4, 'EyeProlog: result_rdf/4 materializes accepted knowledge and decisions as ground RDF-shaped facts').
-pipeline_step_fact(5, 'rdf-prolog-roundtrip: ground rdf/4 facts become RDF again for publication or federation').
+pipeline_step_fact(5, 'rdf-prolog-interchange: ground rdf/4 facts become RDF again for publication or federation').
 
 cognitive_parallel(Prolog, Human) :- cognitive_parallel_fact(Prolog, Human).
 cognitive_parallel_fact(fact, remembered_assertion).

package/package.json

@@ -3,7 +3,7 @@
   "publishConfig": {
     "access": "public"
   },
-  "version": "1.5.62",
+  "version": "1.5.65",
   "description": "EyeProlog turns facts and rules into answers and proofs.",
   "type": "module",
   "main": "./index.js",

package/src/lib/error.pl

@@ -143,12 +143,15 @@ call_with_error_context(Goal, Pair) :-
           ( copy_term(Pair, Element),
             throw(error(Error, [Element|Context])) )).
 
-%  The element must be a pair (issue #99). This cannot go through must_be/2:
-%  must_be/2 declares its own context with call_with_error_context/2, so
-%  checking with must_be/2 here would recurse. The reported context matches
-%  what must_be(pair, _) would have produced.
+%  The element must be a pair (issue #99). The test is inlined rather than
+%  delegated to must_be(pair, _) so that the wrapper stays independent of the
+%  rest of this module. The reported context matches what must_be(pair, _)
+%  produces.
+%  Written as indexed clauses rather than an if-then-else: this runs on every
+%  call_with_error_context/2 call, and an if-then-else costs several times what
+%  first-argument clause selection does.
+error__require_pair(Pair) :- var(Pair), !,
+    throw(error(instantiation_error, [predicate-must_be/2])).
+error__require_pair(_-_) :- !.
 error__require_pair(Pair) :-
-    ( var(Pair) -> throw(error(instantiation_error, [predicate-must_be/2]))
-    ; Pair = _-_ -> true
-    ; throw(error(type_error(pair, Pair), [predicate-must_be/2]))
-    ).
+    throw(error(type_error(pair, Pair), [predicate-must_be/2])).

package/src/lib/si.pl

@@ -8,6 +8,7 @@
     character_si/1,
     term_si/1,
     chars_si/1,
+    compare_si/3,
     dif_si/2,
     not_si/1,
     when_si/2
@@ -16,24 +17,24 @@
 :- meta_predicate(not_si(0)).
 :- meta_predicate(when_si(+, 0)).
 
-atom_si(A) :- ( var(A) -> throw(error(instantiation_error, atom_si/1)) ; atom(A) ).
-integer_si(I) :- ( var(I) -> throw(error(instantiation_error, integer_si/1)) ; integer(I) ).
-atomic_si(A) :- ( var(A) -> throw(error(instantiation_error, atomic_si/1)) ; atomic(A) ).
+atom_si(A) :- ( var(A) -> throw(error(instantiation_error, [predicate-atom_si/1])) ; atom(A) ).
+integer_si(I) :- ( var(I) -> throw(error(instantiation_error, [predicate-integer_si/1])) ; integer(I) ).
+atomic_si(A) :- ( var(A) -> throw(error(instantiation_error, [predicate-atomic_si/1])) ; atomic(A) ).
 
 list_si(List) :- si__list(List).
+si__list(Term) :-
+    var(Term), !, throw(error(instantiation_error, [predicate-list_si/1])).
 si__list([]).
 si__list([_|Tail]) :- !, si__list(Tail).
-si__list(Term) :-
-    ( var(Term) -> throw(error(instantiation_error, list_si/1)) ; fail ).
 
 character_si(C) :-
-    ( var(C) -> throw(error(instantiation_error, character_si/1))
+    ( var(C) -> throw(error(instantiation_error, [predicate-character_si/1]))
     ; atom(C), atom_length(C, 1)
     ).
 
 term_si(Term) :-
     ( ground(Term) -> acyclic_term(Term)
-    ; throw(error(instantiation_error, term_si/1))
+    ; throw(error(instantiation_error, [predicate-term_si/1]))
     ).
 
 chars_si(Chars) :-
@@ -44,19 +45,56 @@ si__chars([C|Cs]) :- character_si(C), si__chars(Cs).
 
 dif_si(X, Y) :-
     X \== Y,
-    ( X \= Y -> true ; throw(error(instantiation_error, dif_si/2)) ).
+    ( X \= Y -> true ; throw(error(instantiation_error, [predicate-dif_si/2])) ).
 
 not_si(Goal) :- term_si(Goal), \+ Goal.
 
 when_si(Condition, Goal) :-
     ( si__condition(Condition) ->
         ( call(Condition) -> call(Goal)
-        ; throw(error(instantiation_error, when_si/2)) )
+        ; throw(error(instantiation_error, [predicate-when_si/2])) )
     ; throw(error(domain_error(when_condition_si, Condition), when_si/2))
     ).
 
-si__condition(Condition) :- var(Condition), !, throw(error(instantiation_error, when_si/2)).
+si__condition(Condition) :- var(Condition), !, throw(error(instantiation_error, [predicate-when_si/2])).
 si__condition(ground(_)).
 si__condition(nonvar(_)).
 si__condition((A,B)) :- si__condition(A), si__condition(B).
 si__condition((A;B)) :- si__condition(A), si__condition(B).
+
+%  compare_si(Order, A, B) holds when compare(Order, A, B) holds for *every*
+%  instance of A and B, and raises an instantiation_error otherwise. The point
+%  is to raise it as rarely as possible: an unbound variable only matters when
+%  it is what decides the ordering. So f(X) @< g(Y) is decided by the functor
+%  names, f(X,a) @< f(X,b) by the second arguments, and X == X gives (=), while
+%  f(a) vs f(Y) and X vs 1 are genuinely undecided.
+compare_si(Order, A, B) :-
+    (  A == B
+    -> Order = (=)
+    ;  si__order_decided(A, B)
+    -> compare(Order, A, B)
+    ;  throw(error(instantiation_error, [predicate-compare_si/3]))
+    ).
+
+%  Both arguments are known to be non-identical here.
+si__order_decided(A, B) :- ( var(A) ; var(B) ), !, fail.
+si__order_decided(A, B) :-
+    compound(A), compound(B), !,
+    functor(A, NameA, ArityA),
+    functor(B, NameB, ArityB),
+    (  ArityA =\= ArityB -> true
+    ;  NameA \== NameB -> true
+    ;  A =.. [_|ArgsA],
+       B =.. [_|ArgsB],
+       si__args_decided(ArgsA, ArgsB)
+    ).
+%  Otherwise both are non-variables and at least one is atomic, so either the
+%  type order or the values themselves settle it and no instantiation can
+%  change the outcome.
+si__order_decided(_, _).
+
+si__args_decided([A|As], [B|Bs]) :-
+    (  A == B
+    -> si__args_decided(As, Bs)
+    ;  si__order_decided(A, B)
+    ).

package/src/library-autoload-index.js

@@ -90,6 +90,7 @@ export const eyePrologLibraryAutoload = Object.freeze({
   "chars_utf8bytes/2": "charsio",
   "circuit/1": "clpz",
   "clpz_t/2": "clpz",
+  "compare_si/3": "si",
   "complement/2": "ugraphs",
   "compose/3": "ugraphs",
   "cond_t/3": "reif",

package/src/standard-library.js

@@ -178,7 +178,7 @@ export const eyePrologPortableLibraryIndicators = Object.freeze([
   ',/3', ';/3', '=/3', 'cond_t/3', 'dif/3', 'if_/3', 'memberd_t/3',
   'tfilter/3', 'tmember/2', 'tmember_t/3', 'tpartition/4',
   'atom_si/1', 'integer_si/1', 'atomic_si/1', 'list_si/1', 'character_si/1',
-  'term_si/1', 'chars_si/1', 'dif_si/2', 'not_si/1', 'when_si/2',
+  'term_si/1', 'chars_si/1', 'compare_si/3', 'dif_si/2', 'not_si/1', 'when_si/2',
   'matches/3', 'split/3', 'replace/4', 'lowercase/2', 'uppercase/2', 'trim/2',
   'number_string/2', 'atom_string/2', 'term_string/2', 'string_concat/3',
   'contains/2', 'matches/2', 'join/3', 'substring/4',

package/test/regression/cases-api.mjs

@@ -1331,7 +1331,7 @@ answer(ok) :-
         assertEqual(registry.get('setup_call_cleanup', 3), null, 'setup_call_cleanup/3 is absent from the ISO registry');
         assertEqual(Boolean(library.get('setup_call_cleanup', 3)), true, 'setup_call_cleanup/3 is an EyeProlog cleanup control');
         assertEqual(registeredNativeEyePrologLibraryNames().length, 122, 'public host-supported EyeProlog library count');
-        assertEqual(eyePrologPortableLibraryIndicators.length, 281, 'portable Prolog library count');
+        assertEqual(eyePrologPortableLibraryIndicators.length, 282, 'portable Prolog library count');
         assertEqual(eyePrologInteropLibraryIndicators.length, 221, 'cross-implementation interop profile count');
         assertEqual(eyePrologInteropLibraryModules.length, 27, 'common explicit library module profile count');
         assertEqual(eyePrologInteropAutoload['member/2'], 'lists', 'member/2 canonical autoload');
@@ -1349,7 +1349,7 @@ answer(ok) :-
         assertEqual(eyePrologNativeLibraryIndicators.length, 122, 'host-supported library count');
         assertEqual(eyePrologNativeLibraryIndicators.includes('call_nth/2'), true, 'call_nth/2 remains classified as host-supported');
         assertEqual(eyePrologNativeLibraryIndicators.includes('random/1'), true, 'stateful random/1 is classified as host-supported');
-        assertEqual(eyePrologLibraryIndicators.length, 403, 'complete non-ISO EyeProlog library surface');
+        assertEqual(eyePrologLibraryIndicators.length, 404, 'complete non-ISO EyeProlog library surface');
         assertEqual(registry.get('eyeprolog__call_nth', 2), null, 'private call_nth adapter is absent from ISO registry');
         assertEqual(Boolean(library.get('eyeprolog__call_nth', 2)), true, 'private call_nth adapter is registered for EyeProlog');
         assertEqual(Boolean(library.get('eyeprolog__call_residue_vars', 2)), true, 'private call_residue_vars adapter is registered for EyeProlog');

package/test/regression/cases-documentation-sync.mjs

@@ -61,15 +61,15 @@ export function documentationSyncCases() {
           encoding: 'utf8',
         });
         assertEqual(result.status, 0, 'exit status');
-        assertIncludes(result.stdout, 'predicate reference is up to date (532 predicates)', 'stdout');
+        assertIncludes(result.stdout, 'predicate reference is up to date (533 predicates)', 'stdout');
         assertEqual(result.stderr, '', 'stderr');
 
         const book = fs.readFileSync(path.join(packageRoot, 'the-art-of-eyeprolog.md'), 'utf8');
         const section = between(book, '<!-- eyeprolog-predicate-reference:start -->', '<!-- eyeprolog-predicate-reference:end -->');
         assertEqual(section.split('\n').some((line) => line.trimStart().startsWith('|')), false, 'predicate reference avoids wide table markup');
-        assertEqual((section.match(/^- \*\*`/gm) ?? []).length, 532, 'one reference entry per predicate');
-        assertEqual((section.match(/<a id="predicate-reference-\d{4}"><\/a>/g) ?? []).length, 532, 'one explicit anchor per predicate');
-        assertEqual((section.match(/\]\(#predicate-reference-\d{4}\)/g) ?? []).length, 532, 'one direct index link per predicate');
+        assertEqual((section.match(/^- \*\*`/gm) ?? []).length, 533, 'one reference entry per predicate');
+        assertEqual((section.match(/<a id="predicate-reference-\d{4}"><\/a>/g) ?? []).length, 533, 'one explicit anchor per predicate');
+        assertEqual((section.match(/\]\(#predicate-reference-\d{4}\)/g) ?? []).length, 533, 'one direct index link per predicate');
         assertNotIncludes(section, '[Symbols](#predicate-reference-symbols)', 'predicate index does not rely on renderer-generated group anchors');
 
         const chapter = between(book, '## 39. Predicate reference', '## 40. Running EyeProlog: command line and corpus');

package/test/regression/cases-regression.mjs

@@ -4528,6 +4528,82 @@ answer(Result) :- countdown(2048, Result), Result = 2048.
         if (elapsed >= 20000) throw new Error(`20k must_be/2 calls took ${elapsed}ms`);
       },
     },
+    {
+      // The pair check runs on every call_with_error_context/2 call, so it is
+      // written as indexed clauses. An if-then-else costs several times what
+      // clause selection does, and this guard used to dominate the wrapper.
+      name: 'call_with_error_context/2 pair check stays off the hot path',
+      run: () => {
+        const input = '%% goal: answer(X)\np(X) :- integer(X).\n' +
+          'loop(0) :- !.\nloop(N) :- call_with_error_context(p(N), c-1), M is N - 1, loop(M).\n' +
+          'answer(ok) :- loop(20000).\n';
+        const started = Date.now();
+        const result = runCli(['-'], { input });
+        const elapsed = Date.now() - started;
+        assertEqual(result.status, 0, `exit status; stderr=${result.stderr}`);
+        assertIncludes(result.stdout, 'answer(ok)', 'loop completes');
+        if (elapsed >= 20000) throw new Error(`20k wrapped calls took ${elapsed}ms`);
+      },
+    },
+    {
+      name: 'sufficient-instantiation list checks reject unknown tails without binding elements',
+      run: () => {
+        const input = ':- use_module(library(si)).\n%% goal: answer(X)\n' +
+          'answer(ok) :- ' +
+          'catch(list_si(X), error(instantiation_error,_), C1 = caught), C1 == caught, var(X), ' +
+          'catch(list_si([a|T]), error(instantiation_error,_), C2 = caught), C2 == caught, var(T), ' +
+          'catch(chars_si(Cs), error(instantiation_error,_), C3 = caught), C3 == caught, var(Cs), ' +
+          'catch(chars_si([a|Tail]), error(instantiation_error,_), C4 = caught), C4 == caught, var(Tail), ' +
+          'list_si([]), list_si([Element]), var(Element), chars_si([a,b]), ' +
+          '\\+ list_si([a|bad]), \\+ chars_si([ab]).\n';
+        const result = runCli(['-'], { input });
+        assertEqual(result.status, 0, `exit status; stderr=${result.stderr}`);
+        assertIncludes(result.stdout, 'answer(ok)', 'list checks preserve sufficient instantiation');
+      },
+    },
+    {
+      // Issue #100: compare_si/3 decides the standard order only when no
+      // instantiation could change it, and raises instantiation_error as
+      // rarely as possible.
+      name: 'compare_si/3 decides the order only when instantiation cannot change it',
+      run: () => {
+        const decided = [
+          ['compare_si(O, 1, a)', '<'],
+          ['compare_si(O, X, X)', '='],
+          ['compare_si(O, f(X), g(Y))', '<'],
+          ['compare_si(O, f(X,a), f(X,b))', '<'],
+          ['compare_si(O, f(a,b), f(a,b,c))', '<'],
+          ['compare_si(O, [a|X], [b|Y])', '<'],
+          ['compare_si(O, 1, 1.0)', '>'],
+        ];
+        for (const [goal, order] of decided) {
+          const input = ':- use_module(library(si)).\n%% goal: answer(X)\n' +
+            `answer(O) :- ${goal}.\n`;
+          const result = runCli(['-'], { input });
+          assertEqual(result.status, 0, `${goal} status; stderr=${result.stderr}`);
+          assertIncludes(result.stdout, `answer(${order})`, `${goal} order`);
+        }
+        for (const goal of ['compare_si(O, X, 1)', 'compare_si(O, f(a), f(Y))',
+          'compare_si(O, X, Y)', 'compare_si(O, [a|X], [a|Y])']) {
+          const input = ':- use_module(library(si)).\n%% goal: answer(X)\n' +
+            `answer(E) :- catch(${goal}, error(E,_), true).\n`;
+          const result = runCli(['-'], { input });
+          assertEqual(result.status, 0, `${goal} status; stderr=${result.stderr}`);
+          assertIncludes(result.stdout, 'answer(instantiation_error)', `${goal} is undecided`);
+        }
+      },
+    },
+    {
+      // compare_si/3 must not bind anything while deciding.
+      name: 'compare_si/3 leaves its arguments unbound',
+      run: () => {
+        const input = ':- use_module(library(si)).\n%% goal: answer(X)\n' +
+          'answer(ok) :- X = f(Y), compare_si(<, X, g(1)), var(Y).\n';
+        const result = runCli(['-'], { input });
+        assertEqual(result.status, 0, `exit status; stderr=${result.stderr}`);
+        assertIncludes(result.stdout, 'answer(ok)', 'no bindings made');
+      },
+    },
     {
       // Issue #99: the context element must be a pair, as in Scryer and
       // Trealla, and predicate contexts use the predicate-F/A convention.