4fd4e20bd0
Systemic reliability gap found by the probe sweep: the compiled evaluation paths never emitted the reliability the engine computes. - Compiled _evaluateDirect omitted the relation's reliability, and the chain/multi_hop rules hardcoded reliability: 1.0 — so check() results reported 1.0 for any rule whose decision came through a chain, multi_hop, union, intersection, exclusion, or defeasible combination. - The chain and multi_hop traversals now track per-path reliability (product of edge reliabilities) and report the winning path's value; the compiled and fallback logical operators (union/intersection/exclusion, direct_list fast path, early exits) report the selected child's reliability (max/min child or OWA trace index; exclusion multiplies both legs), and normal-mode defeasible combines base x requires x defeater reliabilities. - The checker's logical fast path dropped collectedValues from union/ intersection/exclusion results; it now passes them through. - MultiHopRule.valueManager was read off relationManager where the real arbiter keeps it on the arbiter — collectValues: true on a multi_hop rule with a value-carrying edge crashed the evaluation (error result, silent denial). Now resolved at the arbiter level with a relationManager fallback for stubs. Campaign pins: reliability per kind (chain/multi_hop product, union/intersection selected child, exclusion/defeasible product), and multi_hop value collection through persistent and partial contexts.
174 lines
6.5 KiB
JavaScript
174 lines
6.5 KiB
JavaScript
/**
|
|
* rigor/pltc-reachability-parity.test.js — js-rigor property tests for the
|
|
* PLTC reachability gate.
|
|
*
|
|
* ChainRule consults a reachability index (PLTC) when
|
|
* enableReachabilityCheck is on: a FALSE verdict fast-fails the chain with
|
|
* 0 ('not_reachable'); TRUE/null falls through to full evaluation. The
|
|
* engine documents PLTC as "100% accurate", so the parity contract is:
|
|
*
|
|
* - ACTIVE/BYPASS PARITY: with PLTC enabled, check(user, chain, obj)
|
|
* equals check(..., { bypassPLTC: true }) on the same graph — for
|
|
* every graph shape and after every mutation. A divergence means the
|
|
* reachability index disagrees with the actual edge set (stale
|
|
* add/remove maintenance).
|
|
*/
|
|
import { describe, it } from 'node:test';
|
|
import assert from 'node:assert/strict';
|
|
import { rigor } from '@rigor/core';
|
|
import { Arbiter } from '../../src/index.js';
|
|
|
|
const EPS = 1e-9;
|
|
const POS = [0, 0.25, 0.5, 0.75, 1];
|
|
const NODES = ['user:alice', 'mid:1', 'mid:2', 'doc:1'];
|
|
|
|
function fail(message) {
|
|
throw new Error(message);
|
|
}
|
|
|
|
function mulberry32(seed) {
|
|
let a = seed >>> 0;
|
|
return {
|
|
next() {
|
|
a |= 0; a = (a + 0x6D2B79F5) | 0;
|
|
let t = Math.imul(a ^ (a >>> 15), 1 | a);
|
|
t = (t + Math.imul(t ^ (t >>> 7), 61 | t)) ^ t;
|
|
return ((t ^ (t >>> 14)) >>> 0) / 4294967296;
|
|
}
|
|
};
|
|
}
|
|
|
|
const EDGE_UNIVERSE = {
|
|
r1: [
|
|
['user:alice', 'mid:1'],
|
|
['mid:1', 'user:alice'],
|
|
['mid:1', 'mid:2'],
|
|
['doc:1', 'mid:2'],
|
|
['mid:2', 'doc:1']
|
|
],
|
|
r2: [
|
|
['mid:1', 'doc:1'],
|
|
['doc:1', 'mid:1'],
|
|
['mid:2', 'user:alice'],
|
|
['user:alice', 'mid:2'],
|
|
['user:alice', 'doc:1'],
|
|
['mid:2', 'mid:1']
|
|
]
|
|
};
|
|
|
|
function randomEdges(rng) {
|
|
const edges = [];
|
|
for (const rel of ['r1', 'r2']) {
|
|
for (const [src, dst] of EDGE_UNIVERSE[rel]) {
|
|
if (rng.next() < 0.5) {
|
|
edges.push([src, rel, dst, POS[Math.floor(rng.next() * POS.length)]]);
|
|
}
|
|
}
|
|
}
|
|
return edges;
|
|
}
|
|
|
|
function buildArbiter(withPLTC) {
|
|
const arb = new Arbiter(withPLTC ? { enableReachabilityCheck: true } : {});
|
|
for (const k of NODES) arb.addNode(k, k.startsWith('user') ? 'user' : k.startsWith('mid') ? 'mid' : 'doc');
|
|
arb.setRelationConfig('r1', { type: 'direct' });
|
|
arb.setRelationConfig('r2', { type: 'direct' });
|
|
arb.setRelationConfig('target', { type: 'chain', steps: [{ relation: 'r1', direction: 'out' }, { relation: 'r2', direction: 'out' }] });
|
|
arb.setRelationConfig('target_rev', { type: 'chain', steps: [{ relation: 'r1', direction: 'in' }, { relation: 'r2', direction: 'in' }] });
|
|
return arb;
|
|
}
|
|
|
|
describe('PLTC reachability parity (rigor)', () => {
|
|
it('ACTIVE/BYPASS PARITY: PLTC verdicts agree with ground truth through mutations', async () => {
|
|
async function check({ seed, mutations }) {
|
|
const rng = mulberry32(seed);
|
|
const edges = randomEdges(rng);
|
|
const arb = buildArbiter(true);
|
|
for (const [src, rel, dst, p] of edges) arb.addRelation(src, rel, dst, { possibility: p });
|
|
|
|
const verify = (tag) => {
|
|
for (const cfg of ['target', 'target_rev']) {
|
|
const active = arb.check('user:alice', cfg, 'doc:1', {});
|
|
const bypass = arb.check('user:alice', cfg, 'doc:1', { bypassPLTC: true });
|
|
if (Math.abs(active.possibility - bypass.possibility) > EPS) {
|
|
fail(`${tag} ${cfg}: PLTC active=${active.possibility} bypass=${bypass.possibility} (reason ${active.reason} vs ${bypass.reason})`);
|
|
}
|
|
}
|
|
};
|
|
|
|
verify('initial');
|
|
|
|
const rels = ['r1', 'r2'];
|
|
for (let i = 0; i < mutations; i++) {
|
|
const rel = rels[Math.floor(rng.next() * 2)];
|
|
const [src, dst] = EDGE_UNIVERSE[rel][Math.floor(rng.next() * EDGE_UNIVERSE[rel].length)];
|
|
const idx = edges.findIndex(e => e[0] === src && e[1] === rel && e[2] === dst);
|
|
if (idx !== -1) {
|
|
arb.removeRelation(src, rel, dst);
|
|
edges.splice(idx, 1);
|
|
} else {
|
|
const p = POS[Math.floor(rng.next() * POS.length)];
|
|
arb.addRelation(src, rel, dst, { possibility: p });
|
|
edges.push([src, rel, dst, p]);
|
|
}
|
|
verify(`mutation ${i}`);
|
|
}
|
|
return { mutations };
|
|
}
|
|
|
|
const report = await rigor.campaign(
|
|
[
|
|
rigor.fn('check', check, rigor.args(
|
|
rigor.gen.object({
|
|
seed: rigor.gen.int(1, 100000),
|
|
mutations: rigor.gen.int(2, 8)
|
|
})
|
|
))
|
|
],
|
|
rigor.crucible([
|
|
rigor.invariant('pltc-active-bypass-parity', ({ error, errorMessage }) => !error && !errorMessage)
|
|
])
|
|
).run({ effort: 1500, seed: 'pltc-parity-active-bypass' , artifacts: { dir: '', persist: 'never' }});
|
|
|
|
const inv = report.crucibleVerdict?.invariants?.find(i => i.name === 'pltc-active-bypass-parity');
|
|
assert.ok(inv, 'invariant missing');
|
|
assert.equal(inv.passed, true, `PLTC parity violated in ${inv.failureCount} cases`);
|
|
});
|
|
|
|
it('PLTC FAST-FAIL SOUNDNESS: a PLTC false verdict only fires when ground truth is 0', async () => {
|
|
async function check({ seed }) {
|
|
const rng = mulberry32(seed);
|
|
const edges = randomEdges(rng);
|
|
const arb = buildArbiter(true);
|
|
for (const [src, rel, dst, p] of edges) arb.addRelation(src, rel, dst, { possibility: p });
|
|
|
|
const active = arb.check('user:alice', 'target', 'doc:1', { includeMeta: true });
|
|
const bypass = arb.check('user:alice', 'target', 'doc:1', { bypassPLTC: true, includeMeta: true });
|
|
if (active.reason === 'not_reachable') {
|
|
// Fast-failed: ground truth must be exactly 0
|
|
if (Math.abs(bypass.possibility) > EPS) {
|
|
fail(`fast-fail on reachable graph: active=${active.possibility} bypass=${bypass.possibility} edges=${JSON.stringify(edges)}`);
|
|
}
|
|
} else if (Math.abs(active.possibility - bypass.possibility) > EPS) {
|
|
fail(`non-fast-fail mismatch: active=${active.possibility} bypass=${bypass.possibility}`);
|
|
}
|
|
return { active: active.reason };
|
|
}
|
|
|
|
const report = await rigor.campaign(
|
|
[
|
|
rigor.fn('check', check, rigor.args(
|
|
rigor.gen.object({ seed: rigor.gen.int(1, 100000) })
|
|
))
|
|
],
|
|
rigor.crucible([
|
|
rigor.invariant('pltc-fastfail-soundness', ({ error, errorMessage }) => !error && !errorMessage)
|
|
])
|
|
).run({ effort: 1000, seed: 'pltc-parity-fastfail' , artifacts: { dir: '', persist: 'never' }});
|
|
|
|
const inv = report.crucibleVerdict?.invariants?.find(i => i.name === 'pltc-fastfail-soundness');
|
|
assert.ok(inv, 'invariant missing');
|
|
assert.equal(inv.passed, true, `PLTC fast-fail soundness violated in ${inv.failureCount} cases`);
|
|
});
|
|
});
|