Skip to contents

Extract raster values at points.

extract_points.Rd

Samples a raster at points via gdalraster::pixel_extract(), which reads only the blocks containing points. Accepts anything that function accepts (a path or a GDALRaster, passed through untouched) plus a garry LazyRaster/LazyDataset that is already materialised – i.e. one whose graph is a bare source over local files, as materialise() returns.

Usage

extract_points(raster, xy, bands = NULL, interp = NULL, ...)

Arguments

raster

A LazyRaster, LazyDataset, raster path, or GDALRaster object.

xy

Points: a wk::xy() vector (its CRS supplies xy_srs), or a two-column matrix/data frame as gdalraster expects.

bands

Bands to extract (default all).

interp

Interpolation: NULL/"nearest" (default), "bilinear", "cubic", "cubicspline".

...

Passed to gdalraster::pixel_extract() (krnl_dim, xy_srs, max_ram, as_data_frame).

Value

As gdalraster::pixel_extract(): a matrix, or a data frame with as_data_frame = TRUE.

Details

An unmaterialised pipeline is an error, not a silent cube write. Getting pixels onto disk costs real time and space, and the cube is almost always wanted again (for the predict pass, say), so the caller should own that step:

cube <- materialise(pipeline)     # explicit, reusable
vals <- extract_points(cube, pts)

xy may be a wk::xy() point vector, in which case its CRS supplies xy_srs and GDAL reprojects as needed; a matrix or data frame behaves exactly as in gdalraster.

See also

materialise(), collect()

Examples

if (FALSE) { # \dontrun{
pts <- wk::xy(c(512300, 514800), c(4600100, 4601900), crs = "EPSG:32632")
cube <- materialise(composite)
extract_points(cube, pts, interp = "bilinear")
extract_points("composite.tif", pts)   # a path works too
} # }