vibespatial.spatial.query_types¶
Attributes¶
Classes¶
Device-resident spatial join index pairs (Phase 2 zero-copy overlay). |
|
ADR-0042: Typed contract for low-level spatial kernel index output. |
Module Contents¶
- vibespatial.spatial.query_types.SUPPORTED_GEOM_TYPES¶
- class vibespatial.spatial.query_types.SpatialQueryExecution¶
- requested: vibespatial.runtime.ExecutionMode¶
- selected: vibespatial.runtime.ExecutionMode¶
- implementation: str¶
- reason: str¶
- class vibespatial.spatial.query_types.RegularGridPointIndex¶
- origin_x: float¶
- origin_y: float¶
- cell_width: float¶
- cell_height: float¶
- cols: int¶
- rows: int¶
- size: int¶
- class vibespatial.spatial.query_types.DeviceSpatialJoinResult¶
Device-resident spatial join index pairs (Phase 2 zero-copy overlay).
Holds left and right index arrays as CuPy int32 device arrays, eliminating the D->H->D round-trip when both sides of an overlay have owned (device-resident) geometry backing.
Use
to_host()when host-side numpy arrays are needed (e.g., for pandas attribute assembly or Shapely fallback paths).- d_left_idx: Any¶
- d_right_idx: Any¶
- to_host() tuple[numpy.ndarray, numpy.ndarray]¶
Copy index arrays to host as numpy int32 arrays.
- property size: int¶
Number of index pairs.
- class vibespatial.spatial.query_types.SpatialJoinIndices¶
ADR-0042: Typed contract for low-level spatial kernel index output.
Device-native workflows may carry richer result objects, but the low-level spatial-query seam still uses integer index arrays. This frozen dataclass enforces that dtype invariant at construction time.
- left: numpy.ndarray¶
- right: numpy.ndarray¶
- classmethod from_raw(left, right) SpatialJoinIndices¶
Validate and coerce raw arrays into the canonical index form.