Check it out: gdalraster

The gdalraster package provides a comprehensive interface to the GDAL library for R. It handles raster I/O, vector I/O, the virtual filesystem, coordinate transformation, and a large collection of GDAL utilities — all through a direct, low-level API that doesn’t impose a high-level spatial data model.

Raster I/O

The core raster interface is GDALRaster, an Rcpp module class that wraps a GDAL dataset handle:

library(gdalraster)

f <- system.file("extdata/storm_lake.lcp", package = "gdalraster")
ds <- new(GDALRaster, f, TRUE)

ds$info()
ds$dim()       # c(ncol, nrow, nbands)
ds$res()       # c(xres, yres)
ds$bbox()      # c(xmin, ymin, xmax, ymax)
ds$getProjection()

## read a band
data <- ds$read(band = 1,
                xoff = 0, yoff = 0,
                xsize = ds$getRasterXSize(),
                ysize = ds$getRasterYSize(),
                out_xsize = ds$getRasterXSize(),
                out_ysize = ds$getRasterYSize())

ds$close()

The read() method maps directly to GDAL’s RasterIO — you specify the source window and the output size. This gives full control over decimation, oversampling, and partial reads. No intermediate abstraction, no automatic resampling decisions you didn’t ask for.

Vector I/O

gdalraster provides vector read/write through OGR:

## list layers in a data source
dsn <- system.file("extdata/ynp_fires_1984_2022.gpkg", package = "gdalraster")
ogr_ds_layer_names(dsn)

## feature count and extent
ogr_ds_layer_count(dsn)
ogr_layer_field_names(dsn, "mtbs_perims")

## read features as a data frame with WKB geometry
lyr <- new(GDALVector, dsn, "mtbs_perims", read_only = TRUE)
feat <- lyr$fetch(-1)  # all features
lyr$close()

The geometry comes back as WKB — directly compatible with wk. You can pass it to wk::as_wkt(), geos::as_geos_geometry(), or sf::st_as_sfc() as needed. gdalraster reads the data; what you do with the geometry is up to you and whatever packages you choose.

Virtual filesystem

GDAL’s /vsicurl/, /vsizip/, /vsimem/ and other virtual filesystem handlers are all available:

## read a remote file without downloading
vsi_read_dir("/vsicurl/https://geodata.ucdavis.edu/gadm/gadm4.1/gpkg/")

## open a GeoPackage over HTTP
dsn <- "/vsicurl/https://geodata.ucdavis.edu/gadm/gadm4.1/gpkg/gadm41_AUS.gpkg"
ogr_ds_layer_names(dsn)

## read inside a remote zip
vsi_read_dir("/vsizip//vsicurl/https://github.com/OSGeo/gdal/raw/refs/heads/master/autotest/gcore/data/byte.tif.zip")
#[1] "byte.tif"

This is essential for cloud-native workflows. The same dataset string that works locally works over HTTP, S3, or Azure — GDAL handles the transport.

The warper

warp() provides access to GDAL’s warper API for raster reprojection and resampling:

## warp to a new projection
warp(
  src_filename = "input.tif",
  dst_filename = "output.tif",
  t_srs = "EPSG:3577",
  tr = c(100, 100),           # target resolution
  r = "bilinear"              # resampling method
)

This is the same warper that gdalwarp uses on the command line. For building VRTs, mosaics, and complex raster pipelines, gdalraster provides buildVRT(), translate(), and rasterFromRaster().

VRT construction

One of the most powerful features is programmatic VRT (Virtual Raster) construction:

## create a VRT from multiple source files
buildVRT(
  vrt_filename = "mosaic.vrt",
  input_rasters = c("tile1.tif", "tile2.tif", "tile3.tif")
)

## add sources to a VRT programmatically
ds <- new(GDALRaster, "output.vrt", FALSE)
ds$addSimpleSource(band = 1, src_filename = "source.tif", ...)
ds$close()

VRTs are GDAL’s answer to lazy raster operations — the XML file describes the operation but nothing is computed until you read pixels.

Configuration and utilities

gdalraster exposes GDAL configuration options:

## set GDAL config for performance tuning
set_config_option("GDAL_CACHEMAX", "512")
set_config_option("GDAL_HTTP_MULTIRANGE", "YES")
set_config_option("GDAL_NUM_THREADS", "ALL_CPUS")

## check what's set
get_config_option("GDAL_CACHEMAX")

## GDAL/PROJ/GEOS version info
gdal_version()

Plus a full set of geometry utilities (g_buffer(), g_intersection(), etc.) that use GEOS through GDAL’s OGR layer, and coordinate transformation via transform_xy().

Install from CRAN:

install.packages("gdalraster")

Links:

Check it out series

This is part of a series of posts for foundational spatial support in R. These posts include: