feat: add error handling for non-compliant GeoArrow structs in ddbs_open_dataset

Author: e-kotov

R/io_open_dataset.R

@@ -195,11 +195,11 @@ ddbs_open_dataset <- function(path,
       scan_query <- glue::glue("read_parquet('{path}'{p_args_str})")
 
       # Resolve geometry column
+      try_cols <- tryCatch({
+          DBI::dbGetQuery(conn, glue::glue("DESCRIBE SELECT * FROM {scan_query}"))
+      }, error = function(e) NULL)
+
       if (is.null(geom_col)) {
-         try_cols <- tryCatch({
-             DBI::dbGetQuery(conn, glue::glue("DESCRIBE SELECT * FROM {scan_query}"))
-         }, error = function(e) NULL)
-         
          if (!is.null(try_cols)) {
              possibles <- c("geometry", "geom", "wkb_geometry")
              found <- try_cols$column_name[try_cols$column_name %in% possibles]
@@ -208,6 +208,22 @@ ddbs_open_dataset <- function(path,
              geom_col <- NULL
          }
       }
+
+      # Intercept GeoArrow structs (native Arrow encoding) which DuckDB cannot parse
+      if (!is.null(geom_col) && !is.null(try_cols)) {
+          col_type <- try_cols$column_type[try_cols$column_name == geom_col]
+          if (length(col_type) > 0 && grepl("STRUCT", toupper(col_type[1]))) {
+              cli::cli_abort(c(
+                  "The geometry column {.val {geom_col}} is encoded as a native Arrow Struct, which is not supported by DuckDB's spatial extension.",
+                  "x" = "DuckDB requires standard GeoParquet 1.1 with WKB (Well-Known Binary) encoding.",
+                  "i" = "To fix this, please resave the file using:",
+                  " " = "  {.code duckspatial::ddbs_write_dataset(data, path)}",
+                  " " = "  # OR if using geoarrow:",
+                  " " = "  {.code data${geom_col} <- geoarrow::as_geoarrow_vctr(data${geom_col}, schema = geoarrow::geoarrow_wkb())}",
+                  " " = "  {.code arrow::write_parquet(data, path)}"
+              ))
+          }
+      }
 
       view_query <- create_temp_table(
         name = view_name,

tests/testthat/test-ddbs_open_dataset.R

@@ -265,3 +265,19 @@ test_that("ddbs_open_dataset handles missing file gracefully", {
     ignore.case = TRUE
   )
 })
+
+test_that("ddbs_open_dataset fails gracefully on non-compliant GeoArrow structs", {
+  skip_if_not_installed("arrow")
+  skip_if_not_installed("geoarrow")
+  
+  # Create a file with Arrow native encoding (NOT WKB) using direct write_parquet
+  # This triggers the GeoArrow native struct encoding
+  tmp_bad <- tempfile(fileext = ".parquet")
+  on.exit(unlink(tmp_bad))
+  arrow::write_parquet(countries_sf, tmp_bad)
+  
+  expect_error(
+    ddbs_open_dataset(tmp_bad),
+    "encoded as a native Arrow Struct"
+  )
+})