NewGraphEnvironment · NewGraphEnvironment · Sep 29, 2026 · Sep 28, 2026 · Sep 28, 2026 · Sep 28, 2026
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -1,6 +1,6 @@
 Package: drift
 Title: Detecting Riparian and Inland Floodplain Transitions
-Version: 0.17.0
+Version: 0.17.1
 Date: 2026-09-06
 Authors@R: c(
     person("Allan", "Irvine", , "al@newgraphenvironment.com", role = c("aut", "cre"),
@@ -53,6 +53,7 @@ Suggests:
     tidyterra,
     tmap
 Remotes:
+    appelmar/gdalcubes,
     bhaskarvk/leaflet.extras,
     NewGraphEnvironment/flooded,
     NewGraphEnvironment/gq

diff --git a/NEWS.md b/NEWS.md
@@ -1,3 +1,7 @@
+# drift 0.17.1
+
+- **gdalcubes installs from GitHub (#80).** gdalcubes was archived on CRAN on 2026-09-16. That broke the pkgdown workflow's dependency resolve, which had failed on every push since 2026-09-26, and it left new users unable to install what `dft_stac_fetch()` and `dft_stac_cube()` need. `appelmar/gdalcubes` is now in `Remotes:` with no suffix, and the README says to install it with `pak::pak("appelmar/gdalcubes")`. So does the prompt `dft_stac_fetch()`, `dft_stac_cube()` and `dft_index_expr()` raise when it is missing. rlang's own install offer means CRAN and would fail. Upstream master (0.7.5) has merged the `filter_geom` segfault fix from `NewGraphEnvironment/gdalcubes@newgraph`, so the fork is no longer needed. On 0.7.5 the offline suite passes (1007 tests), and so does the opt-in network end-to-end for the cube and fetch paths (109 and 165). That run also surfaced five network assertions that had listed the pre-#48 cache path since that change, so they failed on every opt-in run. They now use `cache_scheme_dir()`.
+
 # drift 0.17.0
 
 - **The article gains patch shape and location, and the corridor question comes back negative (#73).** [What a Land-Cover Change Figure Is Made Of](https://newgraphenvironment.github.io/drift/articles/temporal-composition.html) reported *how much* of a two-epoch comparison is temporally unstable and never raised the spatial axis. It now reports both grains. Body prose 731 -> 1469 words against a cap raised 1000 -> 1500; the cap is a variable now, because the literal appeared twice on one line and changing one would have left the message lying.

diff --git a/R/dft_index_expr.R b/R/dft_index_expr.R
@@ -50,7 +50,7 @@ dft_index_expr <- function(cube,
                            roles = NULL,
                            scale = NULL,
                            offset = NULL) {
-  rlang::check_installed("gdalcubes", reason = "to apply an index to a cube")
+  check_gdalcubes("to apply an index to a cube")
   cfg <- dft_stac_config(source)
   roles <- roles %||% cfg$roles
   scale <- scale %||% cfg$scale %||% 1

diff --git a/R/dft_stac_cube.R b/R/dft_stac_cube.R
@@ -162,7 +162,7 @@ dft_stac_cube <- function(aoi,
                           cache_dir = NULL,
                           force = FALSE,
                           sign_fn = rstac::sign_planetary_computer()) {
-  rlang::check_installed("gdalcubes", reason = "to fetch STAC cubes")
+  check_gdalcubes("to fetch STAC cubes")
 
   # gdalcubes worker processes for the read. drift never set this before v0.9.0,
   # so every fetch ran single-threaded — not a considered choice, just the

diff --git a/R/dft_stac_fetch.R b/R/dft_stac_fetch.R
@@ -98,7 +98,7 @@ dft_stac_fetch <- function(aoi,
                            cache_dir = NULL,
                            force = FALSE,
                            sign_fn = rstac::sign_planetary_computer()) {
-  rlang::check_installed("gdalcubes", reason = "to fetch STAC rasters")
+  check_gdalcubes("to fetch STAC rasters")
 
   # Normalize tile_size ONCE so the path gate (is.null) and the cache key derive
   # from the same snapped scalar. When tiling, tune GDAL for the many extra
@@ -863,6 +863,26 @@ cache_hit_ok <- function(cache_file, label) {
 }
 
 
+#' Require gdalcubes, naming where to get it
+#'
+#' gdalcubes was archived on CRAN on 2026-09-16 (#80). `rlang::check_installed()`
+#' offers to install a missing package by bare name, which now means CRAN and
+#' fails, and it never reads drift's `Remotes:`. So this aborts with the GitHub
+#' install instead, for anyone who meets it before the README.
+#' @noRd
+check_gdalcubes <- function(reason) {
+  if (gdalcubes_available()) return(invisible(TRUE))
+  cli::cli_abort(c(
+    "The {.pkg gdalcubes} package is required {reason}.",
+    "i" = "It was archived on CRAN on 2026-09-16. Install it from GitHub:",
+    " " = "{.code pak::pak(\"appelmar/gdalcubes\")}"
+  ))
+}
+
+#' Whether gdalcubes can be loaded (a seam, so tests can simulate its absence)
+#' @noRd
+gdalcubes_available <- function() requireNamespace("gdalcubes", quietly = TRUE)
+
 #' Auto-detect UTM EPSG code from sf geometry
 #' @noRd
 auto_utm_epsg <- function(x) {

diff --git a/README.md b/README.md
@@ -15,6 +15,14 @@ Monitoring riparian vegetation loss matters for fish habitat, water quality, and
 pak::pak("NewGraphEnvironment/drift")
 ```
 
+The STAC fetch and cube functions (`dft_stac_fetch()`, `dft_stac_cube()`) also need
+[gdalcubes](https://github.com/appelmar/gdalcubes), which was archived on CRAN on
+2026-09-16. Install it from GitHub until it is back:
+
+```r
+pak::pak("appelmar/gdalcubes")
+```
+
 ## Quick start
 
 ```r

diff --git a/inst/notes/gdalcubes-pc-gotchas.md b/inst/notes/gdalcubes-pc-gotchas.md
@@ -10,6 +10,12 @@ Provenance, because it is now mixed: the #30-era bullets were verified on
 (`8bad203`), which is 0.7.4 plus the `filter_geom` segfault fix — with terra
 1.9.34. Each bullet says which.
 
+**Where gdalcubes comes from now (#80, 2026-09-28).** gdalcubes was archived on
+CRAN on 2026-09-16. drift lists `appelmar/gdalcubes` in `Remotes:` with no suffix,
+and upstream master (0.7.5, `ed68331`) has merged the `filter_geom` fix (their
+PR #111), so the `NewGraphEnvironment/gdalcubes@newgraph` fork is no longer needed.
+Do not install the fork over 0.7.5.
+
 - **`gdalcubes::filter_geom()` is not worth using — now for measured reasons, not
   because it crashes (#47).** The original defect was a segfault in the compute
   worker (`gc_exec_worker`, `address 0x120`) or, intermittently, a silent all-NA

diff --git a/tests/testthat/test-dft_stac_cube.R b/tests/testthat/test-dft_stac_cube.R
@@ -401,8 +401,8 @@ test_that("dft_stac_cube fetches an index stack end-to-end", {
   # and coverage is not one surviving pixel. Loose on purpose: cloud masking
   # legitimately removes a lot, so this guards the degenerate case, not quality.
   expect_lt(mean(is.na(vals[inpoly, , drop = FALSE])), 0.9)
-  # second call hits the cache (one cube_<key>.tif under the source dir)
-  expect_length(list.files(file.path(cache, "sentinel-2-l2a"),
+  # the call cached exactly one cube_<key>.tif under the source dir (#48 scheme)
+  expect_length(list.files(drift:::cache_scheme_dir(cache, "sentinel-2-l2a"),
                            pattern = "^cube_.*\\.tif$"), 1)
 })
 
@@ -441,7 +441,7 @@ test_that("dft_stac_cube tiled read reproduces the untiled cube over the AOI", {
   expect_equal(terra::nlyr(tiled), terra::nlyr(untiled))   # same monthly axis
   expect_false(anyNA(terra::time(tiled)))                  # time set per layer
   # tiled and untiled each cache one cube_<key>.tif, keyed apart (2 files total)
-  expect_length(list.files(file.path(cache, "sentinel-2-l2a"),
+  expect_length(list.files(drift:::cache_scheme_dir(cache, "sentinel-2-l2a"),
                            pattern = "^cube_.*\\.tif$"), 2)
 
   # the efficiency claim: for this diagonal reach the tiled read streams fewer

diff --git a/tests/testthat/test-dft_stac_fetch.R b/tests/testthat/test-dft_stac_fetch.R
@@ -19,6 +19,22 @@ test_that("auto_utm_epsg handles prime meridian", {
   expect_equal(epsg, "EPSG:32631")
 })
 
+test_that("a missing gdalcubes names the GitHub install, not only CRAN", {
+  # gdalcubes is archived on CRAN (#80), so a bare install offer fails; the
+  # message has to say where it lives now. The tests below skip whenever
+  # gdalcubes is installed, so this simulates its absence through the seam.
+  testthat::local_mocked_bindings(gdalcubes_available = function() FALSE)
+  expect_error(drift:::check_gdalcubes("to fetch STAC rasters"),
+               "appelmar/gdalcubes", fixed = TRUE)
+  expect_error(drift:::check_gdalcubes("to fetch STAC rasters"),
+               "to fetch STAC rasters", fixed = TRUE)
+  aoi <- sf::st_read(system.file("extdata", "example_aoi.gpkg", package = "drift"),
+                     quiet = TRUE)
+  expect_error(dft_stac_fetch(aoi, source = "io-lulc", years = 2020),
+               "appelmar/gdalcubes", fixed = TRUE)
+  expect_error(dft_stac_cube(aoi), "appelmar/gdalcubes", fixed = TRUE)
+})
+
 test_that("dft_stac_fetch requires gdalcubes", {
   skip_if(requireNamespace("gdalcubes", quietly = TRUE),
           "gdalcubes is installed, can't test missing-package path")
@@ -616,7 +632,7 @@ test_that("dft_stac_fetch tiled result matches untiled over the AOI", {
       cfg$stac_url, cfg$collection, cfg$asset, tile_size = NULL
     )
   )
-  expect_length(list.files(file.path(cache, "io-lulc"),
+  expect_length(list.files(drift:::cache_scheme_dir(cache, "io-lulc"),
                            pattern = paste0("^2020_", attr(untiled_list, "cache_key"),
                                             "\\.nc$")), 1)
   # small tile_size relative to the AOI bbox → several tiles, most bbox-only
@@ -629,9 +645,9 @@ test_that("dft_stac_fetch tiled result matches untiled over the AOI", {
   expect_s4_class(tiled, "SpatRaster")
   expect_equal(terra::nlyr(tiled), 1L)
   # extension routing: untiled caches a gdalcubes .nc, tiled a terra .tif
-  expect_length(list.files(file.path(cache, "io-lulc"),
+  expect_length(list.files(drift:::cache_scheme_dir(cache, "io-lulc"),
                            pattern = "^2020_.*\\.nc$"), 1)
-  expect_length(list.files(file.path(cache, "io-lulc"),
+  expect_length(list.files(drift:::cache_scheme_dir(cache, "io-lulc"),
                            pattern = "^2020_.*\\.tif$"), 1)
   # tiled == untiled over their common in-AOI cells: tiling changes only which
   # bbox pixels are streamed, not the classification. Put the tiled mosaic onto