Separate and merge .h/.hpp files in duckdb (#8566)

- Merge small .h files into vortex_duckdb.h
- Separate C++ code not used in bindgen to .hpp files
- Remove dead code from duckdb: duckdb_vector_to_string, Config,  
ClientContext, duckdb_vx_add_extension_option, DDBString creation.
- Remove rest duckdbfs code. Duckdbfs was partially removed in #8468 but
the Rust-side code survived.

Signed-off-by: Mikhail Kot <mikhail@spiraldb.com>

Author: myrrc

benchmarks/duckdb-bench/src/lib.rs

@@ -16,7 +16,6 @@ use vortex_bench::Format;
 use vortex_bench::IdempotentPath;
 use vortex_bench::generate_duckdb_registration_sql;
 use vortex_bench::runner::BenchmarkQueryResult;
-use vortex_duckdb::duckdb::Config;
 use vortex_duckdb::duckdb::Connection;
 use vortex_duckdb::duckdb::Database;
 use vortex_duckdb::duckdb::QueryResult;
@@ -75,30 +74,21 @@ impl DuckClient {
         path: Option<PathBuf>,
         threads: Option<usize>,
     ) -> Result<(Database, Connection)> {
-        let mut config = Config::new().vortex_expect("failed to create duckdb config");
-
-        // Set DuckDB thread count if specified
-        if let Some(thread_count) = threads {
-            config.set("threads", &format!("{}", thread_count))?;
-        }
-
         let db = match path {
-            Some(path) => Database::open_with_config(path, config),
-            None => Database::open_in_memory_with_config(config),
+            Some(path) => Database::open(path),
+            None => Database::open_in_memory(),
         }?;
 
         let connection = db.connect()?;
         vortex_duckdb::initialize(&db)?;
 
-        // Enable Parquet metadata cache for all benchmark runs.
-        //
+        if let Some(thread_count) = threads {
+            connection.query(&format!("SET threads = {thread_count}"))?;
+        }
+
         // `parquet_metadata_cache` is an extension-specific option that's
         // only available after the Parquet extension is loaded. The Parquet
         // extension is loaded after the connection is established.
-        //
-        // Passing the option to `open_with_config` before leads to
-        // "Invalid Input Error: The following options were not recognized:
-        // parquet_metadata_cache" when running DuckDB in debug mode.
         connection.query("SET parquet_metadata_cache = true")?;
 
         Ok((db, connection))

vortex-duckdb/build.rs

@@ -27,30 +27,19 @@ const DEFAULT_DUCKDB_VERSION: &str = "1.5.3";
 
 const BUILD_ARTIFACTS: [&str; 3] = ["libduckdb.dylib", "libduckdb.so", "libduckdb_static.a"];
 
-const SOURCE_FILES: [&str; 18] = [
-    "cpp/client_context.cpp",
-    "cpp/config.cpp",
+const SOURCE_FILES: [&str; 7] = [
+    "cpp/vortex_duckdb.cpp",
     "cpp/copy_function.cpp",
-    "cpp/data.cpp",
-    "cpp/data_chunk.cpp",
-    "cpp/error.cpp",
     "cpp/expr.cpp",
-    "cpp/file_system.cpp",
-    "cpp/logical_type.cpp",
     "cpp/optimizer.cpp",
-    "cpp/replacement_scan.cpp",
-    "cpp/reusable_dict.cpp",
-    "cpp/scalar_function.cpp",
     "cpp/table_filter.cpp",
     "cpp/table_function.cpp",
-    "cpp/value.cpp",
     "cpp/vector.cpp",
-    "cpp/vector_buffer.cpp",
 ];
 
 // Duckdb C API function we use.
 // This lowers codegen'd src/cpp.rs by four times.
-const DUCKDB_C_API_FUNCTIONS: [&str; 134] = [
+const DUCKDB_C_API_FUNCTIONS: [&str; 133] = [
     "duckdb_array_type_array_size",
     "duckdb_array_type_child_type",
     "duckdb_array_vector_get_child",
@@ -184,7 +173,16 @@ const DUCKDB_C_API_FUNCTIONS: [&str; 134] = [
     "duckdb_vector_reference_value",
     "duckdb_vector_reference_vector",
     "duckdb_vector_size",
-    "duckdb_vector_to_string",
+];
+
+const DUCKDB_C_API_HEADERS: [&str; 7] = [
+    "cpp/include/vortex_duckdb.h",
+    "cpp/include/expr.h",
+    "cpp/include/table_filter.h",
+    "cpp/include/vector.h",
+    "cpp/include/copy_function.h",
+    "cpp/include/table_function.h",
+    "cpp/include/optimizer.h",
 ];
 
 const DOWNLOAD_MAX_RETRIES: i32 = 3;
@@ -462,7 +460,7 @@ fn try_build_duckdb(
 /// Generate rust functions with bindgen from C sources.
 fn bindgen_c2rust(crate_dir: &Path, duckdb_include_dir: &Path) {
     let mut builder = bindgen::Builder::default()
-        .header("cpp/include/duckdb_vx.h")
+        .headers(DUCKDB_C_API_HEADERS)
         .override_abi(Abi::CUnwind, ".*")
         .raw_line("#![allow(dead_code)]")
         .raw_line("#![allow(non_camel_case_types)]")

vortex-duckdb/cpp/client_context.cpp

@@ -1,8 +1,9 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright the Vortex contributors
-#include "duckdb_vx/data.hpp"
-#include "duckdb_vx/error.hpp"
-#include "duckdb_vx/table_function.h"
+#include "data.hpp"
+#include "error.hpp"
+#include "vortex_duckdb.h"
+#include "table_function.h"
 #include "vortex.h"
 #include "duckdb/function/copy_function.hpp"
 #include "duckdb/main/capi/capi_internal.hpp"
@@ -11,8 +12,6 @@
 #include "duckdb/parser/parsed_data/create_copy_function_info.hpp"
 
 using namespace duckdb;
-using vortex::CData;
-using vortex::IntoErrString;
 
 struct CopyBindData final : TableFunctionData {
     CopyBindData(unique_ptr<CData> ffi_data) : ffi_data(std::move(ffi_data)) {