def nearest_nodes(df: cudf.DataFrame, nodes: cudf.DataFrame) -> cudf.DataFrame:
"""
Find nearest road node to point of interest
Uses `cuml` nearest neighbours for GPU accelerated nearest points.
It is assumed that all points use a planar coordinate system like BNG.
Parameters
----------
df : cudf.DataFrame
POI df containing coordinate information
nodes : cudf.DataFrame
Road nodes with coordinate information
Returns
-------
cudf.DataFrame:
Road nodes that are nearest neighbour to some POI
"""
df = df.dropna(subset=["easting", "northing"])
nbrs = NearestNeighbors(n_neighbors=1,
output_type="cudf",
algorithm="brute").fit(
nodes[["easting", "northing"]])
_, indices = nbrs.kneighbors(df[["easting", "northing"]])
df["node_id"] = nodes.iloc[indices]["node_id"].reset_index(drop=True)
return df
def apply_op(
self,
gdf: cudf.DataFrame,
columns_ctx: dict,
input_cols,
target_cols=["base"],
stats_context=None,
):
target_columns = self.get_columns(columns_ctx, input_cols, target_cols)
new_gdf = gdf.dropna(subset=target_columns or None)
new_gdf.reset_index(drop=True, inplace=True)
self.update_columns_ctx(columns_ctx, input_cols, new_gdf.columns,
target_columns)
return new_gdf
def transform(self, columns: ColumnNames,
gdf: cudf.DataFrame) -> cudf.DataFrame:
new_gdf = gdf.dropna(subset=columns or None)
new_gdf.reset_index(drop=True, inplace=True)
return new_gdf
