Read well known binary as a GeoArrow array

Parses a binary column of WKB into the narrowest GeoArrow type that fits every geometry in it. ga_from_wkb() turns a column of points into a point array.

Usage

ga_from_wkb(x, crs = NULL)

Arguments

x

a binary array of WKB, or any GeoArrow array

crs

a coordinate reference system to record, or NULL to keep the one x already has

Value

a GeoArrow array of the narrowest type that fits, the same length as x

Details

GeoParquet writers, and plenty of plain Parquet ones, store geometry as a bare binary column with no GeoArrow extension metadata on it. Most geoarrowrs functions read that directly, but the ones that need a specific geometry type, such as ga_dist_euclidean_pairwise(), cannot tell what is in it. If you convert once up front, you settle the type for everything downstream.

ga_from_wkb() stores crs verbatim in the array's metadata and never interprets it, so it never reprojects. If you leave it NULL, ga_from_wkb() keeps whatever the input carried, which for a bare binary column is no CRS at all.

ga_from_wkb() leaves mixed geometry types as a geometry array, since no narrower type fits.

See also

Other cast: ga_as_point(), ga_cast_geometry(), ga_collection_extract(), ga_downcast_geometry(), ga_explode(), ga_flatten()

Examples

sfc <- sf::st_sfc(sf::st_point(c(0, 0)), sf::st_point(c(3, 4)))
wkb <- nanoarrow::as_nanoarrow_array(
  arrow::Array$create(unclass(wk::as_wkb(sfc)), type = arrow::binary())
)

pts <- ga_from_wkb(wkb, crs = "EPSG:4326")
sf::st_as_sfc(geoarrow::as_geoarrow_vctr(pts))
#> Geometry set for 2 features 
#> Geometry type: POINT
#> Dimension:     XY
#> Bounding box:  xmin: 0 ymin: 0 xmax: 3 ymax: 4
#> Geodetic CRS:  WGS 84
#> POINT (0 0)
#> POINT (3 4)