core-labs/core
core-labs/core
1
0
Fork 0
Code Issues 1 Pull requests 1 Projects Releases Packages Wiki Activity

fix(algebra): close runtime versor applications in float64

Browse source
This commit is contained in:
Shay 2026-05-15 23:18:07 -07:00
parent fbb6570a7d
commit d09b1c8ad0
1 changed files with 14 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

30
algebra/versor.py
View file

@ -14,6 +14,7 @@ _CONSTRUCTION_RESIDUE_TOLERANCE = 1e-2
_NEAR_ZERO_TOLERANCE = 1e-12 _NEAR_ZERO_TOLERANCE = 1e-12
_DENSE_SEED_MIN_COMPONENTS = 8 _DENSE_SEED_MIN_COMPONENTS = 8
_SEED_BIVECTORS = (6, 7, 8, 10, 11, 13) _SEED_BIVECTORS = (6, 7, 8, 10, 11, 13)
_RUNTIME_FIELD_DTYPE = np.dtype(np.float64)
def _array_dtype(v: np.ndarray) -> np.dtype: def _array_dtype(v: np.ndarray) -> np.dtype:
@ -94,29 +95,26 @@ def construction_seed_versor(v: np.ndarray) -> np.ndarray:
return _seed_to_rotor(v, _array_dtype(v)) return _seed_to_rotor(v, _array_dtype(v))
def _close_applied_versor(v: np.ndarray, dtype: np.dtype) -> np.ndarray: def _close_applied_versor(v: np.ndarray) -> np.ndarray:
"""Close algebra-produced sandwich results for runtime field states. """Close runtime field sandwich results at float64 precision.
``versor_apply`` is the field-propagation API: callers expect the returned ``versor_apply`` is the runtime field-propagation API. Vocabulary entries
multivector to satisfy ``versor_condition(result)``. CGA point/null-vector may be stored compactly as float32, but live ``FieldState.F`` values are
preservation belongs behind an explicit geometry API, not this runtime judged against a strict ``versor_condition < 1e-6`` invariant. Closing and
manifold boundary. returning float64 avoids leaking float32 roundoff as false manifold drift.
""" """
arr = np.asarray(v, dtype=dtype) arr = np.asarray(v, dtype=_RUNTIME_FIELD_DTYPE)
try: try:
return unitize_versor(arr).astype(dtype) return unitize_versor(arr).astype(_RUNTIME_FIELD_DTYPE)
except ValueError: except ValueError:
return construction_seed_versor(arr).astype(dtype) return _seed_to_rotor(arr, _RUNTIME_FIELD_DTYPE).astype(_RUNTIME_FIELD_DTYPE)
def versor_apply(V: np.ndarray, F: np.ndarray) -> np.ndarray: def versor_apply(V: np.ndarray, F: np.ndarray) -> np.ndarray:
dtype = np.result_type(V, F) V = np.asarray(V, dtype=_RUNTIME_FIELD_DTYPE)
if dtype not in (np.dtype(np.float32), np.dtype(np.float64)): F = np.asarray(F, dtype=_RUNTIME_FIELD_DTYPE)
dtype = np.dtype(np.float32) applied = geometric_product(geometric_product(V, F), reverse(V)).astype(_RUNTIME_FIELD_DTYPE)
V = np.asarray(V, dtype=dtype) return _close_applied_versor(applied)
F = np.asarray(F, dtype=dtype)
applied = geometric_product(geometric_product(V, F), reverse(V)).astype(dtype)
return _close_applied_versor(applied, dtype)
def versor_unit_residual(v: np.ndarray, *, allow_negative: bool = False) -> float: def versor_unit_residual(v: np.ndarray, *, allow_negative: bool = False) -> float: