chore: polish ArrowDeviceArray (#8023)

## Changes

- Generate Arrow C Device ABI bindings from a vendored Arrow reference
header
- Fixed VarBin/Utf8 Arrow export layout by reporting the correct
`n_buffers = 3`
- Fixed Bool export to ensure buffers are resident on the CUDA device
- Fixed struct child `ArrowArray` lifetime/release handling
- Replaced unsupported-export panics with recoverable errors
- Added schema normalization so exported schemas match the physical
Arrow Device layout (`Utf8View`/`BinaryView` → `Utf8`/`Binary`)

## Notes

This PR is scoped to Arrow Device array export. The intent is to build
cuDF support in a follow up on top of this.

---------

Signed-off-by: Alexander Droste <alexander.droste@protonmail.com>

Author: 0ax1

.github/workflows/ci.yml

@@ -270,6 +270,7 @@ jobs:
         run: |
           git ls-files vortex-cuda vortex-cxx vortex-duckdb vortex-ffi \
             | grep -E '\.(cpp|hpp|cu|cuh|h)$' \
+            | grep -v 'arrow/reference/arrow_c_device\.h$' \
             | grep -v 'kernels/src/bit_unpack_.*\.cu$' \
             | grep -v 'kernels/src/bit_unpack_.*_lanes\.cuh$' \
             | xargs clang-format --dry-run --Werror --style=file

Cargo.lock

@@ -46,3 +46,8 @@ path = ["vortex-cuda/kernels/src/bit_unpack_*"]
 precedence = "override"
 SPDX-FileCopyrightText = "Copyright the Vortex contributors"
 SPDX-License-Identifier = "Apache-2.0"
+
+[[annotations]]
+path = "vortex-cuda/src/arrow/reference/arrow_c_device.h"
+SPDX-FileCopyrightText = "2016-2025 Copyright The Apache Software Foundation"
+SPDX-License-Identifier = "Apache-2.0"

REUSE.toml

@@ -24,6 +24,7 @@ unstable_encodings = ["vortex/unstable_encodings"]
 
 [dependencies]
 arc-swap = { workspace = true }
+arrow-schema = { workspace = true, features = ["ffi"] }
 async-trait = { workspace = true }
 bytes = { workspace = true }
 cudarc = { workspace = true, features = ["f16"] }

vortex-cuda/Cargo.toml

@@ -56,6 +56,7 @@ fn main() {
     generate_unpack::<u64>(&kernels_src, 16).expect("Failed to generate unpack for u64");
 
     let out_dir = PathBuf::from(env::var("OUT_DIR").expect("OUT_DIR not set"));
+    generate_arrow_device_array_bindings(Path::new(&manifest_dir), &out_dir);
     generate_dynamic_dispatch_bindings(&kernels_src, &out_dir);
     generate_patches_bindings(&kernels_src, &out_dir);
 
@@ -196,6 +197,30 @@ fn nvcc_compile_ptx(
     Ok(())
 }
 
+/// Generate bindings for the vendored Arrow C Device ABI header.
+fn generate_arrow_device_array_bindings(manifest_dir: &Path, out_dir: &Path) {
+    let header = manifest_dir.join("src/arrow/reference/arrow_c_device.h");
+    println!("cargo:rerun-if-changed={}", header.display());
+
+    let bindings = bindgen::Builder::default()
+        .header(header.to_string_lossy())
+        .allowlist_type("ArrowArray")
+        .allowlist_type("ArrowDeviceArray")
+        .allowlist_type("ArrowDeviceType")
+        .allowlist_var("ARROW_DEVICE_.*")
+        // ArrowArray/ArrowDeviceArray own producer state through release/private_data.
+        // Shallow copies must use Arrow C move semantics, not Rust Copy/Clone.
+        .derive_copy(false)
+        .derive_debug(true)
+        .layout_tests(false)
+        .generate()
+        .expect("Failed to generate Arrow C Device bindings");
+
+    bindings
+        .write_to_file(out_dir.join("arrow_c_abi.rs"))
+        .expect("Failed to write arrow_c_abi.rs");
+}
+
 /// Generate bindings for the dynamic dispatch shared header.
 fn generate_dynamic_dispatch_bindings(kernels_src: &Path, out_dir: &Path) {
     let header = kernels_src.join("dynamic_dispatch.h");