gdalBindings: wraps around GDAL C library using Rcpp for reading and writing to Raster

Authors: Caio Hamamura, Carlos Alberto Silva

The gdalBindings package wraps around GDAL Raster and Band classes for reading and writing directly from RasterBlock in R semantic [[]] and familiar syntax for accessing RasterBand and reading/writing to blocks.

It also provides a formulaCalculate() function for allowing to perform multiband calculations using formula, you only need to provide a named list of RasterBands and the formula will evaluate against the provided names from the list while updating the updateBand argument. It by reading the RasterBlocks in an efficient way.

The package was developed mainly because the other solutions available don't work with the RasterBlock which is the most common way to read/save and operate over raster datasets. This ensures you will be able to read/write data in small chunks which fits in the memory and works with very efficient continuous file pointers.

This package allows you to access lower level GDAL dataset, so you can efficiently implement your own algorithms and calculations for operating over raster, without sacrificing too much performance. It also allows you to work with extremely large datasets, by accessing only specific parts of it in an efficient way and block by block.

Using the package

Installation

#The CRAN version:
install.packages("gdalBindings")

# The development version:
#install.packages("remotes")
remotes::install_github('caiohamamura/gdalBindings-r')

Creating a new raster dataset

raster_path <- file.path(tempdir(), "output.tif")

ul_lat <- -15
ul_lon <- -45
lr_lat <- -25
lr_lon <- -35

res <- c(0.01, -0.01)
datatype <- gdalBindings::GDALDataType$GDT_Int32
nbands <- 2
projstring <- "EPSG:4326"
nodata <- -1
co <- c("TILED=YES", "BLOCKXSIZE=512", "BLOCKYSIZE=512", "COMPRESSION=LZW")

# Create a new raster dataset
ds <- createDataset(
  raster_path = raster_path,
  nbands = nbands,
  datatype = datatype,
  projstring = projstring,
  lr_lat = lr_lat,
  ul_lat = ul_lat,
  ul_lon = ul_lon,
  lr_lon = lr_lon,
  res = res,
  nodata = nodata,
  co = co
)

Opening an existing dataset

# For read only
raster <- GDALOpen("path/for/file.tif")

# For read and write
rasterWritable <- GDALOpen("path/for/file.tif", readonly = FALSE)

Getting information

Currently the GDALDataset only have GetRasterXSize() and GetRasterYSize() for retrieving information.

# Get raster dimensions
width <- ds$GetRasterXSize()
height <- ds$GetRasterYSize()

Accessing bands, reading and writing

The functions for accessing bands, reading and writing blocks are very R-like friendly to R users, you only need to index it with double brackets [[]] as you are already used to.

Before reading and writing the pixels values you need to first choose a band from for the dataset (1-index based):

# Get the band 1 from ds
band <- ds[[1]]

Now you can read the blocks from the band retrieved using the indexing [[]] operator, you can both write or read data as you are used to with other types of data.

# Set some dummy values
band[[0, 0]] <- 1:(512 * 512)

# Read the beggining of the block
head(band[[0, 0]])

#
# [1] 1 2 3 4 5 6
#

Note that both writing and reading works with 1-D vector, that is the way GDAL operates, and it is kept that way for performance reasons withing this package.

Accessing band information

Band has additional information on how the data is stored with the functions:

1. GetBlockXSize and GetBlockYSize: so you can see how large tiles were saved.
2. GetNoDataValue: to know what value is considered nodata within the raster.
3. GetRasterDataType: information on which type of number (byte, integer, float, etc.).

# Get the data type for the raster
datatype <- ds$GetRasterDataType()

Raster algebra with gdalBindings

Raster algebra operations is done using the function formulaCalculate. It accepts three parameters:

1. formula: a character formula, in which you can use any legal operations within R formula's notation.
2. data: a list of named bands, the names given within the list will be used when evaluating the formula.
3. updateBand: the band where the result will be written on

# Update band 2
updateBand = ds[[2]]

# Calculate double - 10, we will call the band "x" by naming the list index
formulaCalculate(
  formula = ~x * 2 - 10,
  data = list(x = band),
  updateBand = updateBand
)

You could use as well bands from different datasets in the data parameter, just make sure they are all opened and then don't forget to close them in the end.

Close the dataset after finishing to ensure data is saved and the file is unlocked

ds$Close()

Reporting Issues

Please report any issue regarding the gdalBindings package to Dr. Hamamura (hamamura.caio@ifsp.edu.br)

Citing gdalBindings-R

Hamamura, C. gdalBindings-R: wraps around GDAL C library using Rcpp for reading and writing to Raster, version 0.1.33, accessed on May 28 2024, available at: https://github.com/caiohamamura/gdalBindings-R