"""Parity and provenance tests for the shared VAL/NUL proposal kernel.""" from __future__ import annotations from dataclasses import FrozenInstanceError import networkx as nx import pytest from tnfr.alias import get_attr from tnfr.constants.aliases import ALIAS_DNFR, ALIAS_EPI, ALIAS_THETA, ALIAS_VF from tnfr.errors import TNFRValueError from tnfr.operators import apply_glyph from tnfr.operators._scale_operator_kernel import propose_scale_operator from tnfr.operators.metrics_structural import contraction_metrics from tnfr.types import Glyph, real_scalar_epi def _graph( *, glyph: Glyph, epi: float, vf: float = 2.0, dnfr: float = 0.3, factor: float, edge_aware: bool = True, epi_min: float = -1.0, epi_max: float = 1.0, clip_mode: str = "hard", ) -> nx.Graph: key = "VAL_scale" if glyph is Glyph.VAL else "NUL_scale" graph = nx.Graph( GLYPH_FACTORS={key: factor}, EDGE_AWARE_ENABLED=edge_aware, EPI_MIN=epi_min, EPI_MAX=epi_max, CLIP_MODE=clip_mode, ) graph.add_node( "target", **{ ALIAS_EPI[0]: epi, ALIAS_VF[0]: vf, ALIAS_DNFR[0]: dnfr, ALIAS_THETA[0]: 0.0, "glyph_history": ["AL"], }, ) return graph @pytest.mark.parametrize( ( "glyph", "epi", "factor", "edge_aware", "epi_min", "epi_max", "clip_mode", ), [ (Glyph.VAL, 0.4, 1.2, True, -1.0, 1.0, "hard"), (Glyph.VAL, 0.9, 2.0, True, -1.0, 1.0, "hard"), (Glyph.VAL, -0.6, 2.0, True, -0.7, 1.2, "hard"), (Glyph.VAL, 0.95, 1.02, True, -1.0, 1.0, "soft"), (Glyph.NUL, 0.4, 0.5, True, -1.0, 1.0, "hard"), (Glyph.NUL, 0.999, 0.99, True, -1.0, 1.0, "soft"), (Glyph.NUL, -0.4, 0.75, True, -0.8, 1.0, "hard"), (Glyph.VAL, 0.4, 1.2, False, -1.0, 1.0, "hard"), (Glyph.NUL, 0.4, 0.5, False, -1.0, 1.0, "hard"), ], ) def test_pure_proposal_matches_direct_glyph_binary64_state( glyph, epi, factor, edge_aware, epi_min, epi_max, clip_mode, ) -> None: graph = _graph( glyph=glyph, epi=epi, factor=factor, edge_aware=edge_aware, epi_min=epi_min, epi_max=epi_max, clip_mode=clip_mode, ) proposal = propose_scale_operator( glyph=glyph, factor=factor, vf_before=2.0, dnfr_before=0.3 if glyph is Glyph.NUL else None, configured_densification_factor=(1.0 / factor if glyph is Glyph.NUL else None), edge_aware_enabled=edge_aware, epi_before=epi if edge_aware else None, epi_min=epi_min, epi_max=epi_max, epsilon=1e-12, clip_mode=clip_mode, ) apply_glyph(graph, "target", glyph) assert float(get_attr(graph.nodes["target"], ALIAS_VF)) == proposal.vf_after observed_epi = real_scalar_epi(get_attr(graph.nodes["target"], ALIAS_EPI)) if proposal.write_epi: assert observed_epi == proposal.epi_after else: assert observed_epi == epi if glyph is Glyph.NUL: assert float(get_attr(graph.nodes["target"], ALIAS_DNFR)) == ( proposal.dnfr_after ) event = graph.graph["nul_densification_log"][-1] assert event["node"] == "target" assert event["dnfr_before"] == proposal.dnfr_before assert event["dnfr_after"] == proposal.dnfr_after assert event["densification_factor"] == proposal.densification_factor assert event["binary64_inverse_product_residual"] == ( proposal.binary64_inverse_product_residual ) if proposal.edge_aware_adapted: event = graph.graph["edge_aware_interventions"][-1] assert event["node"] == "target" assert event["scale_effective"] == proposal.effective_epi_scale def test_disabled_edge_branch_does_not_consume_boundary_inputs() -> None: proposal = propose_scale_operator( glyph=Glyph.VAL, factor=1.1, vf_before=2.0, edge_aware_enabled=False, epi_before=object(), epi_min="invalid", epi_max=None, epsilon="invalid", clip_mode=object(), ) assert proposal.vf_after == pytest.approx(2.2) assert proposal.write_epi is False assert proposal.epi_after is None def test_pure_kernel_rejects_boolean_epi_like_public_scalar_boundary() -> None: with pytest.raises(TNFRValueError, match="not boolean"): propose_scale_operator( glyph=Glyph.VAL, factor=1.1, vf_before=2.0, edge_aware_enabled=True, epi_before=True, ) def test_nul_proposal_exposes_nodal_drive_coefficient_identity() -> None: proposal = propose_scale_operator( glyph=Glyph.NUL, factor=0.37, vf_before=2.5, dnfr_before=-0.8, configured_densification_factor=1.0 / 0.37, edge_aware_enabled=False, ) assert proposal.densification_factor is not None assert proposal.dnfr_after is not None assert proposal.binary64_inverse_product_residual == ( proposal.requested_scale * proposal.densification_factor - 1.0 ) assert proposal.vf_after * proposal.dnfr_after == pytest.approx( proposal.vf_before * proposal.dnfr_before ) def test_scale_proposal_is_frozen() -> None: proposal = propose_scale_operator( glyph=Glyph.VAL, factor=1.1, vf_before=2.0, edge_aware_enabled=False, ) with pytest.raises(FrozenInstanceError): proposal.vf_after = 3.0 def test_contraction_metrics_select_matching_node_event_and_magnitude() -> None: graph = nx.path_graph(("first", "second")) graph.nodes["first"].update( { ALIAS_EPI[0]: 0.4, ALIAS_VF[0]: 0.5, ALIAS_DNFR[0]: -0.6, } ) graph.nodes["second"].update( { ALIAS_EPI[0]: 0.2, ALIAS_VF[0]: 0.5, ALIAS_DNFR[0]: 0.5, } ) graph.graph["nul_densification_log"] = [ { "node": "first", "dnfr_before": -0.3, "dnfr_after": -0.6, "densification_factor": 2.0, }, { "node": "second", "dnfr_before": 0.25, "dnfr_after": 0.5, "densification_factor": 2.0, }, ] metrics = contraction_metrics(graph, "first", 1.0, 0.8) assert metrics["dnfr_before"] == -0.3 assert metrics["dnfr_increase"] == pytest.approx(0.3) assert metrics["dnfr_signed_change"] == pytest.approx(-0.3) assert metrics["dnfr_densified"] is True def test_zero_pressure_contraction_is_not_reported_as_densification() -> None: graph = nx.Graph() graph.add_node( "target", **{ ALIAS_EPI[0]: 0.4, ALIAS_VF[0]: 0.5, ALIAS_DNFR[0]: 0.0, }, ) graph.graph["nul_densification_log"] = [ { "node": "target", "dnfr_before": 0.0, "dnfr_after": 0.0, "densification_factor": 2.0, } ] metrics = contraction_metrics(graph, "target", 1.0, 0.8) assert metrics["dnfr_densified"] is False assert metrics["dnfr_increase"] == 0.0 assert metrics["densification_ratio"] == 1.0