MultiPoint. Supported geometries also return
MultiPoint for every size, including zero and one. Sampling is
evaluated by native Spark and Sedona expressions and remains
distributed.
Signature
Parameters
int or array - like
required
Number of points to sample from each geometry. An array-like or
pandas-on-Spark
Series supplies one positional size per row.
Local sequences are materialized on the driver; use a distributed
Series for large per-row size vectors.str
default:"'uniform'"
Sampling method. Sedona currently supports only
"uniform".optional
default:"None"
Deprecated alias for
rng.optional
default:"None"
Any seed accepted by
numpy.random.default_rng. Fresh generators
initialized with the same seed produce reproducible results. Passing
an existing Generator or BitGenerator consumes one draw
immediately to derive the engine seed.Accepted for GeoPandas signature compatibility and ignored for
uniform sampling.
Returns
GeoSeries
Sampled points with the original index and CRS and the name
"sampled_points".Sedona and GeoPandas use different random number generators, so the
sampled coordinates are not expected to be identical for the same
seed. An integer seed is reused for every row, matching GeoPandas’s
behavior in which congruent geometries receive the same relative sample
positions. With
rng=None, no reproducibility guarantee is made.A stateful NumPy Generator or BitGenerator is reduced to one
engine seed and then varied deterministically by row position, so its
state advancement also differs. Validation of values in a distributed
size Series is lazy and errors are raised when the result is
evaluated.Sampling work and per-row intermediate memory grow with size. The
current line sampler materializes an array of sampled points and costs
roughly O(size * vertices) per geometry before collecting those
points into a MultiPoint. Unsupported nonempty geometry types produce
empty MultiPoints without GeoPandas’s per-row warning, avoiding an eager
type scan.Sedona follows GeoPandas 1.1.4 and later by always returning a
MultiPoint. GeoPandas 1.1.3 and earlier instead return a Point
for size=1 and an empty GeometryCollection for size=0, and
may collapse duplicate samples.