Точки
и сетка

Canvas: W × H Normalized coordinate system: x,y ∈ [0,1] Grid origin: (ox, oy) Grid steps: sx = usable_width / Gx sy = usable_height / Gy Node(i,j): Nx = ox + i × sx Ny = oy + j × sy Nearest node: i* = round((x - ox) / sx) j* = round((y - oy) / sy) Snap distance: d_snap = distance(P, N(i*,j*)) d_norm = d_snap / min(sx, sy) Controlled deviation: δx = (x_final - Nx) / sx δy = (y_final - Ny) / sy Recommended default: |δx|, |δy| ≤ 0.18 cell Hard snap: δ = 0 Soft snap: 0 < |δ| ≤ allowed_deviation

Snap decision: if role in [anchor, connector_node, measurement_marker]: mode = HARD elif role in [support, rhythm, micro_marker]: mode = SOFT elif role == focus: mode = CANDIDATE else: mode = SOFT Hard constraints may force a different mode.

Node occupancy score: occupancy_score = 1 - clamp(O_norm(node), 0, 1) If role == focus: prefer O_norm ≈ 0.10..0.55 If role == support: prefer O_norm ≈ 0.20..0.70 If role == micro_noise: allow O_norm ≈ 0.35..0.85

Compo 1.06 DOES NOT OWN: free_space_score hierarchy_score balance_score alignment_score eye_flow_score negative_space_score These are DEFERRED_INPUTS. They may be attached to GridCandidate later by their owners. Local candidate result: GridCandidate { nodeId, snappedPosition, deviation, components: { snap_distance_score, occupancy_fit, deviation_score, mechanicality_penalty }, grid_fit_score, deferredContext: null | ExternalMetricMap }

Priority order: 1. hard safety / protected content 2. semantic constraints 3. global composition constraints 4. hierarchy / balance / eye flow 5. grid alignment 6. local aesthetic variation 7. randomization Therefore: GRID MUST NEVER OVERRIDE A BAD COMPOSITION.

GRID FIT AGGREGATE active = { snap_distance_score, occupancy_fit, deviation_score } w = normalizeActiveWeights(stylePreset.gridWeights) grid_fit_score = w.snap * snap_distance_score + w.occupancy * occupancy_fit + w.deviation * deviation_score - w.mechanicality * mechanicality_penalty grid_fit_score = clamp(grid_fit_score, 0, 1) IMPORTANT ANTI-DOUBLE-COUNTING: Global solver may consume EITHER: A) grid_fit_score OR B) its component metrics but MUST NOT sum A + B together.

function proposePointGridCandidates(point, grid, scene, style): nodes = kNearestAllowedNodes(point.preferredPosition, grid, k=8) candidates = [] for node in nodes: validation = requestHardValidation(point, node, scene) if validation.status == REJECT: continue mode = resolveSnapMode(point.role, style) offsets = generateAllowedOffsets(node, mode, point.role, style) for offset in offsets: pos = node.position + offset snapScore = computeSnapDistanceScore(pos, node, grid) occFit = computeRoleOccupancyFit(point.role, node, scene.gridOccupancy) devScore = computeDeviationScore(offset, grid, point.role, style) gridFit = aggregateGridFit( snapScore, occFit, devScore, currentMechanicalityState(), style ) candidates.push({ nodeId: node.id, position: pos, snapMode: mode, gridFitScore: gridFit, components: {snapScore, occFit, devScore}, deferredContext: null }) if candidates is empty: return {status:"NO_GRID_CANDIDATE", fallback:"SOLVE_WITHOUT_GRID"} sortDescending(candidates, by=gridFitScore) return { status:"LOCAL_PROPOSAL", candidates, finalDecision:false } // NEXT STAGE, OUTSIDE COMPO 1.06: // owners attach hierarchy/balance/negative-space/eye-flow/etc. // global solver accepts, moves, re-ranks or rejects candidates.

{ "document_id": "Compo 1.06", "version": "2.0", "module": "point_grid_coupling", "finality": "LOCAL_PROPOSAL_ONLY", "coordinate_system": "normalized_0_1", "owns_metrics": [ "grid.snap_distance_score", "grid.node_occupancy", "grid.occupancy_fit", "grid.deviation_score", "grid.mechanicality_penalty", "grid.fit_score" ], "metric_groups": { "grid.fit_score": { "components": [ "grid.snap_distance_score", "grid.occupancy_fit", "grid.deviation_score", "grid.mechanicality_penalty" ], "aggregation_rule": "consume aggregate OR components, never both" } }, "grid": { "types": ["uniform_rect","column","modular","baseline","radial","custom_nodes"], "default_columns": {"value":6,"status":"HEURISTIC"}, "default_rows": {"value":8,"status":"HEURISTIC"}, "safe_margin_pct": {"value":[0.04,0.08],"status":"HEURISTIC"} }, "snap_modes": { "anchor": "hard", "connector_node": "hard", "focus": "candidate", "support": "soft", "rhythm": "soft", "noise": "soft" }, "deviation": { "focus_cell_fraction": {"value":[0.00,0.12],"status":"HEURISTIC"}, "support_cell_fraction": {"value":[0.00,0.18],"status":"HEURISTIC"}, "rhythm_cell_fraction": {"value":[0.00,0.14],"status":"HEURISTIC"}, "noise_cell_fraction": {"value":[0.00,0.28],"status":"HEURISTIC"}, "reject_if_over": {"value":0.30,"status":"HEURISTIC"} }, "deferred_inputs": [ "global.hierarchy_score", "global.balance_score", "global.eye_flow_score", "global.negative_space_score", "global.free_space_score", "global.alignment_score" ], "external_validators": [ "protected_content_validator", "Compo1.09.clearance_or_tangency_validator" ], "forbidden_local_redefinitions": [ "hierarchy_score", "balance_score", "eye_flow_score", "negative_space_score", "free_space_score", "alignment_score", "tangency_penalty", "collision_penalty" ], "override_policy": { "global_may_override_local_grid": true, "grid_may_override_hard_constraint": false, "random_may_override_grid": false } }