chore(flow-php/arrow-ext): bump arrow/parquet to 58.3 and ext-php-rs to 0.15.15

.github/workflows/job-arrow-extension.yml

@@ -58,6 +58,11 @@ jobs:
         run: |
           brew install llvm
           echo "LIBCLANG_PATH=$(brew --prefix llvm)/lib" >> "$GITHUB_ENV"
+          # Pin the clang executable to the one shipped with the same llvm as
+          # libclang. Otherwise clang-sys (used by ext-php-rs's bindgen) falls
+          # back to `xcodebuild -find clang` and mixes Xcode's clang headers
+          # (e.g. arm_neon.h) with brew's newer libclang, which fails to compile.
+          echo "CLANG_PATH=$(brew --prefix llvm)/bin/clang" >> "$GITHUB_ENV"
 
       - name: Set LIBCLANG_PATH (Ubuntu)
         if: runner.os == 'Linux'
@@ -147,6 +152,9 @@ jobs:
         run: |
           brew install autoconf automake libtool llvm
           echo "LIBCLANG_PATH=$(brew --prefix llvm)/lib" >> "$GITHUB_ENV"
+          # See the build job: keep clang and libclang from the same llvm so
+          # bindgen does not mix brew's libclang with Xcode's clang headers.
+          echo "CLANG_PATH=$(brew --prefix llvm)/bin/clang" >> "$GITHUB_ENV"
 
       - name: Set LIBCLANG_PATH (Ubuntu)
         if: runner.os == 'Linux'
@@ -180,7 +188,7 @@ jobs:
         run: |
           sudo pie repository:remove packagist.org
           sudo pie repository:add path ${{ github.workspace }}/src/extension/arrow-ext
-          sudo env "PATH=$PATH" "LIBCLANG_PATH=$LIBCLANG_PATH" "RUSTUP_TOOLCHAIN=$RUSTUP_TOOLCHAIN" "CARGO_HOME=$CARGO_HOME" "RUSTUP_HOME=$RUSTUP_HOME" pie install flow-php/arrow-ext:0.0.9999@dev
+          sudo env "PATH=$PATH" "LIBCLANG_PATH=$LIBCLANG_PATH" ${CLANG_PATH:+"CLANG_PATH=$CLANG_PATH"} "RUSTUP_TOOLCHAIN=$RUSTUP_TOOLCHAIN" "CARGO_HOME=$CARGO_HOME" "RUSTUP_HOME=$RUSTUP_HOME" pie install flow-php/arrow-ext:0.0.9999@dev
 
       - name: Verify extension is loaded
         run: |

documentation/contributing/c.md

@@ -87,10 +87,10 @@ nix-shell --arg with-pg-query-ext false --arg with-c true --run "cd src/extensio
 The extension is only half of the story. A given `libpg_query` version pins a specific PostgreSQL
 **grammar**, and that grammar is shared across three places that must always agree:
 
-| Place | What it is | Used by |
-|-------|------------|---------|
-| `src/extension/pg-query-ext/vendor/libpg_query` (Makefile `PG_VERSION`) | the C library compiled into the extension | CI (`pie` build) |
-| `.nix/pkgs/php-pg-query-ext/package.nix` (`libpg_query` `version`/`rev`/`hash`) | the C library compiled into the **nix dev shell** | local `nix-shell` |
+| Place                                                                                                                           | What it is                                                                | Used by                           |
+|---------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------|-----------------------------------|
+| `src/extension/pg-query-ext/vendor/libpg_query` (Makefile `PG_VERSION`)                                                         | the C library compiled into the extension                                 | CI (`pie` build)                  |
+| `.nix/pkgs/php-pg-query-ext/package.nix` (`libpg_query` `version`/`rev`/`hash`)                                                 | the C library compiled into the **nix dev shell**                         | local `nix-shell`                 |
 | `src/lib/postgresql/resources/proto/pg_query.proto` + the generated stubs in `src/lib/postgresql/src/Flow/PostgreSql/Protobuf/` | the protobuf schema the PHP side serializes/deserializes parse trees with | the `flow-php/postgresql` library |
 
 The PHP library round-trips parse trees as protobuf (`pg_query_parse_protobuf` → mutate → `pg_query_deparse`).

documentation/contributing/rust.md

@@ -3,7 +3,7 @@
 [TOC]
 
 This document describes how to develop the `arrow-ext` PHP extension, which is written in Rust
-using the [ext-php-rs](https://github.com/nicedragons/ext-php-rs) framework.
+using the [ext-php-rs](https://github.com/extphprs/ext-php-rs) framework.
 
 ## Overview
 

src/extension/arrow-ext/Cargo.toml

@@ -9,7 +9,7 @@ crate-type = ["cdylib"]
 
 [dependencies]
 ext-php-rs = "0.15"
-parquet = { version = "58", default-features = false, features = ["arrow", "arrow_canonical_extension_types", "snap", "flate2", "flate2-zlib-rs", "zstd", "lz4", "brotli"] }
+parquet = { version = "58", default-features = false, features = ["arrow", "arrow_canonical_extension_types", "snap", "flate2", "flate2-zlib-rs", "zstd", "lz4", "brotli", "simdutf8"] }
 arrow-array = "58"
 arrow-buffer = "58"
 arrow-schema = { version = "58", features = ["canonical_extension_types"] }

src/extension/arrow-ext/build.rs

@@ -16,7 +16,8 @@ fn main() {
 
     println!("cargo:rustc-env=ARROW_VERSION={version}");
 
-    let arrow_version = resolve_dep_version("arrow-schema").unwrap_or_else(|| "unknown".to_string());
+    let arrow_version =
+        resolve_dep_version("arrow-schema").unwrap_or_else(|| "unknown".to_string());
     let parquet_version = resolve_dep_version("parquet").unwrap_or_else(|| "unknown".to_string());
 
     println!("cargo:rustc-env=ARROW_LIB_VERSION={arrow_version}");
@@ -32,7 +33,11 @@ fn resolve_dep_version(crate_name: &str) -> Option<String> {
             for line in chunk.lines() {
                 let line = line.trim();
                 if line.starts_with("version = ") {
-                    return Some(line.trim_start_matches("version = ").trim_matches('"').to_string());
+                    return Some(
+                        line.trim_start_matches("version = ")
+                            .trim_matches('"')
+                            .to_string(),
+                    );
                 }
             }
         }

src/extension/arrow-ext/examples/generate_fixtures.rs

@@ -107,7 +107,7 @@ fn generate_all_flat_types() {
     let col_uint32 = UInt32Array::from(vec![Some(0u32), Some(4294967295u32), None]);
     let col_uint64 = UInt64Array::from(vec![Some(0u64), Some(u64::MAX), None]);
     let col_float = Float32Array::from(vec![Some(1.5f32), Some(-2.5f32), None]);
-    let col_double = Float64Array::from(vec![Some(3.14159f64), Some(-2.71828f64), None]);
+    let col_double = Float64Array::from(vec![Some(1.23456f64), Some(-7.89012f64), None]);
     let col_string = StringArray::from(vec![Some("hello"), Some("world"), None]);
 
     let col_binary: BinaryArray = vec![
@@ -317,10 +317,7 @@ fn generate_deeply_nested() {
     let scores_builder = ListBuilder::new(Int64Builder::new());
     let struct_builder = StructBuilder::new(
         struct_fields,
-        vec![
-            Box::new(StringBuilder::new()),
-            Box::new(scores_builder),
-        ],
+        vec![Box::new(StringBuilder::new()), Box::new(scores_builder)],
     );
     let mut groups_builder = ListBuilder::new(struct_builder);
 

src/extension/arrow-ext/src/lib.rs

@@ -21,6 +21,10 @@ pub extern "C" fn php_module_info(_module: *mut ModuleEntry) {
     info_table_end!();
 }
 
+/// # Safety
+///
+/// Invoked by the PHP/Zend engine during module startup. Must only be called by
+/// the engine through the registered startup hook, never directly.
 pub unsafe extern "C" fn module_startup(_type: i32, _module_number: i32) -> i32 {
     if let Err(e) = stream::output_stream::register() {
         eprintln!("arrow: failed to register Flow\\Arrow\\OutputStream: {e}");

src/extension/arrow-ext/src/parquet/exception.rs

@@ -17,8 +17,7 @@ pub fn register() -> Result<()> {
 }
 
 fn parquet_exception_ce() -> &'static ClassEntry {
-    ClassEntry::try_find("Flow\\Arrow\\Parquet\\Exception")
-        .unwrap_or_else(|| ce::exception())
+    ClassEntry::try_find("Flow\\Arrow\\Parquet\\Exception").unwrap_or_else(|| ce::exception())
 }
 
 pub fn parquet_exception(message: impl Into<String>) -> PhpException {