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. A column of points becomes 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. Converting once up front settles the type for everything downstream.

crs is stored verbatim in the array's metadata and is never interpreted, so nothing here reprojects. Leaving it NULL keeps whatever the input carried, which for a bare binary column is no CRS at all.

Mixed geometry types are left as a geometry array, since no narrower type fits.

See also

Other cast: ga_as_point(), ga_cast_geometry(), 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)