|
1 | | -#' GDAL VRT utilities for R |
| 1 | +#' @title vrtility: GDAL VRT utilities for R |
| 2 | +#' |
| 3 | +#' @description vrtility is a package for building raster (primarily remote sensing) |
| 4 | +#' processing pipelines. It makes use of GDAL's VRT (virtual raster format) |
| 5 | +#' capabilities for efficient processing of large raster datasets. This |
| 6 | +#' package's primary focus is on the use of GDAL VRT pixel functions. These |
| 7 | +#' pixel functions (currently implemented with python) are used to apply |
| 8 | +#' cloud masks and summarise pixel values (e.g. median) from multiple images |
| 9 | +#' (i.e create a composite image). We hope to add C++ or expression based |
| 10 | +#' pixel functions in time. |
| 11 | +#' @section Spatial helpers: |
| 12 | +#' \describe{ |
| 13 | +#' \item{\code{\link{bbox_to_projected}}}{Convert a long/lat bounding box to a projected bounding box} |
| 14 | +#' \item{\code{\link{to_wkt}}}{Get a wkt string from a numeric epsg code or character such as a proj4 string} |
| 15 | +#' } |
| 16 | +#' @section STAC helpers: |
| 17 | +#' \describe{ |
| 18 | +#' \item{\code{\link{stac_query}}}{Query a STAC catalog} |
| 19 | +#' \item{\code{\link{sentinel2_stac_query}}}{Query a STAC catalog for Sentinel-2 data} |
| 20 | +#' \item{\code{\link{sign_planetary_computer}}}{Sign a URL for the Planetary Computer} |
| 21 | +#' } |
| 22 | +#' @section VRT utilities: |
| 23 | +#' \describe{ |
| 24 | +#' \item{\code{\link{vrt_collect}}}{Create a vrt_collection object} |
| 25 | +#' \item{\code{\link{vrt_set_maskfun}}}{Set a mask function for a vrt object} |
| 26 | +#' \item{\code{\link{vrt_stack}}}{Create a vrt_stack object} |
| 27 | +#' \item{\code{\link{vrt_set_pixelfun}}}{Set a pixel function for a vrt stack object} |
| 28 | +#' \item{\code{\link{vrt_warp}}}{Warp a vrt_x object to a warped vrt} |
| 29 | +#' \item{\code{\link{vrt_compute}}}{Compute a vrt pipeline (using GDAL)} |
| 30 | +#' } |
| 31 | +#' @section VRT pixel functions: |
| 32 | +#' \describe{ |
| 33 | +#' \item{\code{\link{vrt_set_pixelfun}}}{Set a pixel function for a vrt stack object} |
| 34 | +#' \item{\code{\link{bitmask_numba}}}{A pixel function to apply a bitmask} |
| 35 | +#' \item{\code{\link{bitmask_numpy}}}{A pixel function to compute the median} |
| 36 | +#' \item{\code{\link{median_numba}}}{A pixel function to compute the median} |
| 37 | +#' \item{\code{\link{median_numpy}}}{A pixel function to compute the median} |
| 38 | +#' } |
| 39 | +#' @section python environment helpers: |
| 40 | +#' \describe{ |
| 41 | +#' \item{\code{\link{vrtility_py_require}}}{Require a python package} |
| 42 | +#' \item{\code{\link{set_py_env_vals}}}{Set python environment variables as options} |
| 43 | +#' } |
| 44 | +#' @section VRT helpers: |
| 45 | +#' \describe{ |
| 46 | +#' \item{\code{\link{save_vrt}}}{Save a VRT object to a file} |
| 47 | +#' \item{\code{\link{vrt_schema}}}{(data object)The official GDAL VRT schema as a character object} |
| 48 | +#' } |
| 49 | +#' |
2 | 50 | #' |
3 | | -#' THE GDAL VRT (Virtual Format) is an extremely powerful tool for combining |
4 | | -#' and manipulating raster data. This package provides a set of utilities |
5 | | -#' for working with VRTs in R. It is primarily focussed on harnessing the |
6 | | -#' power of VRT pixel functions to carry out complex raster operations. This |
7 | | -#' is achieved through the use of python and in particular the numba library |
8 | | -#' which allows for the creation of fast, compiled functions that can be |
9 | | -#' used in the VRT. All very much a work in progress... |
10 | 51 | #' |
11 | 52 | #' @docType package |
12 | 53 | #' @aliases vrtility-package |
|
15 | 56 |
|
16 | 57 |
|
17 | 58 | .onLoad <- function(libname, pkgname) { |
18 | | - python_init_checks() |
| 59 | + vrtility_py_require(c("numpy", "numba")) |
19 | 60 | cache_init_checks() |
20 | 61 | gdal_init_opts() |
21 | 62 | } |
22 | | - |
23 | | -python_init_checks <- function() { |
24 | | - vrtility_python <- Sys.getenv("VRTILITY_PYTHON", unset = NA) |
25 | | - |
26 | | - if (!is.na(vrtility_python)) { |
27 | | - reticulate::use_virtualenv(vrtility_python, required = TRUE) |
28 | | - set_py_env_vals(vrtility_python) |
29 | | - } else { |
30 | | - if (reticulate::virtualenv_exists("vrtilitypy")) { |
31 | | - reticulate::use_virtualenv("vrtilitypy", required = TRUE) |
32 | | - set_py_env_vals("vrtilitypy") |
33 | | - } else { |
34 | | - cli::cli_inform( |
35 | | - c( |
36 | | - "!" = "Cannot locate the {cli::style_bold('VRTILITY_PYTHON')} environment", |
37 | | - "i" = "Run {cli::code_highlight('`build_vrtility_python()`')} to install it" |
38 | | - ), |
39 | | - class = "packageStartupMessage" |
40 | | - ) |
41 | | - } |
42 | | - } |
43 | | -} |
44 | | - |
45 | | -cache_init_checks <- function() { |
46 | | - op <- options() |
47 | | - op_vrtility <- list( |
48 | | - vrt.cache = tempdir() |
49 | | - ) |
50 | | - |
51 | | - toset <- !(names(op_vrtility) %in% names(op)) |
52 | | - if (any(toset)) options(op_vrtility[toset]) |
53 | | -} |
54 | | - |
55 | | -gdal_init_opts <- function() { |
56 | | - gdal_config_opts <- c( |
57 | | - # Existing good settings |
58 | | - GDAL_VRT_ENABLE_PYTHON = "YES", |
59 | | - VSI_CACHE = "TRUE", |
60 | | - GDAL_CACHEMAX = "50%", |
61 | | - |
62 | | - # Optimize VSI caching for COGs |
63 | | - VSI_CACHE_SIZE = "25000000", # Increase cache size |
64 | | - # CPL_VSIL_CURL_CHUNK_SIZE = "262144", |
65 | | - # CPL_VSIL_CURL_CACHE_SIZE = "33554432", |
66 | | - |
67 | | - # Optimize HTTP connections |
68 | | - GDAL_HTTP_MAX_RETRY = "5", |
69 | | - GDAL_HTTP_RETRY_DELAY = "3", |
70 | | - GDAL_HTTP_MULTIPLEX = "YES", |
71 | | - CPL_VSIL_CURL_ALLOWED_EXTENSIONS = ".tif .TIF", # NEW: Allow both case variants |
72 | | - CPL_VSIL_CURL_USE_HEAD = "NO", # NEW: Skip HEAD requests |
73 | | - |
74 | | - # Existing settings |
75 | | - GDAL_DISABLE_READDIR_ON_OPEN = "EMPTY_DIR", |
76 | | - GDAL_MAX_DATASET_POOL_SIZE = "500", |
77 | | - GDAL_INGESTED_BYTES_AT_OPEN = "32000", |
78 | | - GDAL_HTTP_VERSION = "2", |
79 | | - GDAL_HTTP_MERGE_CONSECUTIVE_RANGES = "YES", |
80 | | - GDAL_NUM_THREADS = "ALL_CPUS" |
81 | | - ) |
82 | | - |
83 | | - gdal_warp_opts <- c( |
84 | | - # Resampling |
85 | | - "-r", |
86 | | - "near", |
87 | | - |
88 | | - # Compression options |
89 | | - "-co", |
90 | | - "COMPRESS=LZW", # CHANGED: Often faster than DEFLATE for many datasets |
91 | | - "-co", |
92 | | - "PREDICTOR=2", |
93 | | - "-co", |
94 | | - "NUM_THREADS=ALL_CPUS", |
95 | | - "-co", |
96 | | - "BIGTIFF=YES", |
97 | | - |
98 | | - # Tiling options - optimized for COGs |
99 | | - "-co", |
100 | | - "TILED=YES", |
101 | | - "-co", |
102 | | - "BLOCKXSIZE=256", # CHANGED: 256 often better aligns with internal COG structure |
103 | | - "-co", |
104 | | - "BLOCKYSIZE=256", # CHANGED: 256 often better aligns with internal COG structure |
105 | | - "-co", |
106 | | - "COPY_SRC_OVERVIEWS=YES", # NEW: Copy source overviews when available |
107 | | - |
108 | | - # Memory and threading options |
109 | | - "-wm", |
110 | | - "50%", |
111 | | - "-multi", |
112 | | - # "-wo", |
113 | | - # "NUM_THREADS=ALL_CPUS", |
114 | | - |
115 | | - # Overview options |
116 | | - "-ovr", |
117 | | - "AUTO" # NEW: Use overviews when available |
118 | | - ) |
119 | | - |
120 | | - op <- options() |
121 | | - op_gdal <- list( |
122 | | - vrt.gdal.config.options = gdal_config_opts, |
123 | | - vrt.gdal.warp.options = gdal_warp_opts |
124 | | - ) |
125 | | - |
126 | | - toset <- !(names(op_gdal) %in% names(op)) |
127 | | - if (any(toset)) options(op_gdal[toset]) |
128 | | -} |
0 commit comments