radiosity optable → matriximport fullseye as fs; fs.ledger.form_factor_matrix(patches, n_quad=3, symmetric=True, far_ratio=8.0, memory_budget_mb=256.0) (実装を直接呼ぶなら import radiosity3d; radiosity3d.form_factor_matrix(patches, n_quad=3, symmetric=True, far_ratio=8.0, memory_budget_mb=256.0)、台帳から引くなら opsoptics.get("form_factor_matrix"))All patch-to-patch view factors F_ij of a radiosity_patches set (occlusion by the scene’s rays).
Each pair is :func:form_factor_polygons (n_quad × n_quad Gauss points on the
receiver, Lambert’s contour sum to the sender) restricted to pairs that face each
other; when the patches came from a scene, each Gauss point tests one ray to the
sender’s centre against the scene. Pairs farther apart than far_ratio times the
larger patch’s diagonal use the receiver’s centroid alone (the sender is still
exact); far_ratio=None integrates every pair fully. symmetric=True (default) enforces
reciprocity exactly by averaging A_i F_ij and A_j F_ji (the raw values agree
to the quadrature accuracy) — the progressive solver’s F_ij A_i / A_j is then
exactly F_ji.
Candidate pairs use the former, faster all-at-once calculation while its largest
temporary fits memory_budget_mb, and 256-patch blocks otherwise. The default
256 MB keeps the former path for a 384-patch closed box but blocks 4,096 patches,
whose candidate expression alone needs 1.00 GiB. On the 2026-10-21 development
machine, whole-call CPU time / tracemalloc peak were 0.234 s / 115 MiB
(original), 0.219 s / 60 MiB (default), and 0.234 s / 60 MiB (forced blocks) for
384 patches; the isolated 4,096-patch candidate case measured 0.547 s / 1,152 MiB,
1.500 s / 195 MiB, and 1.516 s / 195 MiB respectively. Pass 0 to always block,
or a sufficiently large value to never block.
Returns the (N, N) matrix (matrix), zero diagonal. In a closed enclosure
each row sums to 1 (to the quadrature accuracy); an opening shows as a row sum < 1.
Raises ValueError: an empty or unmarked table (not radiosity_patches);
n_quad not an integer in [1, 16]; far_ratio below 1; or a negative or
non-finite memory_budget_mb.
optics の全 op は入力を検証してから計算する(黙って通さない):
_mm / _um / _deg / _mrad。mm と µm の取り違えは crash ではなく「もっともらしく間違った答え」なので、名前で防ぐ。大きさから単位を推測する処理は一切しない。ValueError — float('50') は成功してしまうため、未パースの設定値が長さとして通り抜ける(実測: thin_lens('50', '200') がもっともらしい 66.667 mm を返していた)。bool も True == 1 の暗黙昇格として拒否。ValueError(実数枠のみ。虚部の無言切り捨て・マスク剥がしを拒否)。NaN/Inf は全入力で ValueError。depth_of_field の過焦点距離以遠の far_mm = inf(それが過焦点距離の定義)と gaussian_beam のウエストでの wavefront_radius_mm = inf(平面波面の曲率半径)。どちらも有限の相棒(far_is_infinite / curvature_per_mm)を併せて返す。それ以外の無言 NaN/Inf は内部で検出して ValueError —「float64 が溢れた」と「答えが無限大」は別の主張なので、後者の顔で前者を返さない。optics.MAX_GRID(4096)、供給された場/PSF/開口は optics.MAX_FIELD_ELEMENTS(2^24)、ABCD 素子列は optics.MAX_SYSTEM_ELEMENTS(1024)、Zernike は MAX_ZERNIKE_TERMS(512)/ MAX_ZERNIKE_ORDER(40)/ MAX_ZERNIKE_BASIS(2^25)。小さな引数から巨大な内部確保が起きる経路(実測: n_max=40 × 4096² で 108 GB)を fail-closed で塞ぐ。| 物理的に不可能な状態も拒否: 偏光度 > 1 の Stokes ベクトル、負の透過率、負の強度、n- | m | が奇数などの不正な Zernike 添字。 |
py -3.11 examples/poc_pocket_interreflection.pymatrix を入力に取れる)abcd_trace · emva_filter_distortion · mueller_apply · mueller_checks · equirect_rotate · pixel_shift_capture · pixel_shift_merge · pixel_shift_motion_mask
radiosity)form_factor_parallel_rects · form_factor_perpendicular_rects · form_factor_polygons · radiosity_patches · radiosity_progressive_3d · render_optscene_gi
Provenance: radiosity3d.py — OPTICS operator registry. この per-op ノートは tools/opdocs.py md が自動生成(手編集しない)。
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.