Basic CPP bindings (#19)

Nothing fancy yet, but it compiles and links to the c++ code, and allows
two functions to be called.

Note that this requires a change to C++ code -
fast-pack/FastPFOR#125 (we could have worked
around it, but this is much simpler)

Author: nyurik

.github/workflows/ci.yml

@@ -16,6 +16,7 @@ jobs:
       - uses: taiki-e/install-action@v2
         with: { tool: just }
       - uses: actions/checkout@v4
+        with: { submodules: recursive }
 #
 # Uncomment this and below once the crate has been published to crates.io
 #
@@ -39,6 +40,7 @@ jobs:
       - uses: taiki-e/install-action@v2
         with: { tool: just }
       - uses: actions/checkout@v4
+        with: { submodules: recursive }
       - uses: Swatinem/rust-cache@v2
         if: github.event_name != 'release' && github.event_name != 'workflow_dispatch'
       - name: Read crate metadata

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "cpp"]
+	path = cpp
+	url = https://github.com/nyurik/FastPFor-cpp.git

Cargo.toml

@@ -10,9 +10,20 @@ keywords = ["fastpfor", "compression"]
 categories = ["compression"]
 rust-version = "1.83.0"
 
+[features]
+# We probably want to build without the C++ bindings by default.
+# Keeping it on for now to simplify development.
+default = ["cpp"]
+cpp = ["dep:cmake", "dep:cxx", "dep:cxx-build"]
+
 [dependencies]
 rand = "0.8.5"
 thiserror = "2.0.7"
+cxx = { version = "1.0.136", optional = true }
+
+[build-dependencies]
+cmake = { version = "0.1.52", optional = true }
+cxx-build = { version = "1.0.136", optional = true }
 
 [lints.rust]
 unused_assignments = "allow"
@@ -21,4 +32,4 @@ dead_code = "allow"
 [lints.clippy]
 unnecessary_cast = "allow"
 ptr_arg = "allow"
-needless_range_loop="allow"
+needless_range_loop = "allow"

build.rs

@@ -0,0 +1,28 @@
+use std::path::Path;
+
+use cmake::Config;
+
+fn main() {
+    if !Path::new("cpp/CMakeLists.txt").exists() {
+        panic!("FastPFOR submodule not initialized. Run `git submodule update --init`.");
+    }
+
+    // Compile FastPFOR using CMake
+    let lib_path = Config::new("cpp").build().join("lib");
+    let lib_path = lib_path.to_str().unwrap();
+    println!("cargo:rerun-if-changed=cpp");
+
+    // Compile the bridge
+    cxx_build::bridge("src/ffi_wrapper.rs")
+        .include("cpp/headers")
+        .include("src/bridge")
+        .std("c++14")
+        .compile("fastpfor_bridge");
+    println!("cargo:rerun-if-changed=src/ffi.rs");
+    println!("cargo:rerun-if-changed=src/bridge/fastpfor_bridge.cc");
+    println!("cargo:rerun-if-changed=src/bridge/fastpfor_bridge.h");
+
+    // Link the FastPFOR library - must be done after the bridge is compiled
+    println!("cargo:rustc-link-search=native={lib_path}");
+    println!("cargo:rustc-link-lib=static=FastPFOR");
+}

cpp

@@ -34,8 +34,8 @@ check:
     cargo check --all-targets --workspace
 
 # Default build
-build:
-    cargo build --all-targets --workspace
+build *ARGS:
+    cargo build --all-targets --workspace {{ARGS}}
 
 # Run all tests
 test *ARGS: build

justfile

@@ -0,0 +1,73 @@
+#pragma once
+
+#include <memory>
+#include <rust/cxx.h>
+#include "codecfactory.h"
+#include "codecs.h"
+
+// Adding things to the same namespace as lib to keep things simpler
+namespace FastPForLib {
+
+// Instantiate Coder Factory
+inline std::unique_ptr<CODECFactory> new_codec_factory() {
+    return std::make_unique<CODECFactory>();
+}
+
+// Get a list of codec names
+inline std::unique_ptr<std::vector<std::string>> codec_factory_all_names(
+        const CODECFactory& factory
+) {
+    return std::make_unique<std::vector<std::string>>(factory.allNames());
+}
+
+// Get codec by name from a factory
+inline std::shared_ptr<IntegerCODEC> codec_factory_get_from_name(
+        const CODECFactory& factory,
+        rust::Str name
+) {
+    return factory.getFromName(std::string(name));
+}
+
+// Encode helpers
+inline size_t codec_encode32(
+        const std::shared_ptr<IntegerCODEC>& codec,
+        const rust::Slice<const uint32_t> in,
+        rust::Slice<uint32_t> out
+) {
+    size_t outSize = out.size();
+    codec->encodeArray(in.data(), in.size(), out.data(), outSize);
+    return outSize;
+}
+
+inline size_t codec_encode64(
+        const std::shared_ptr<IntegerCODEC>& codec,
+        const rust::Slice<const uint64_t> in,
+        rust::Slice<uint32_t> out
+) {
+    size_t outSize = out.size();
+    codec->encodeArray(in.data(), in.size(), out.data(), outSize);
+    return outSize;
+}
+
+// Decode helper: returns consumed input elements
+inline size_t codec_decode32(
+        const std::shared_ptr<IntegerCODEC>& codec,
+        const rust::Slice<const uint32_t> in,
+        rust::Slice<uint32_t> out
+) {
+    size_t outSize = out.size();
+    codec->decodeArray(in.data(), in.size(), out.data(), outSize);
+    return outSize;
+}
+
+inline size_t codec_decode64(
+        const std::shared_ptr<IntegerCODEC>& codec,
+        const rust::Slice<const uint32_t> in,
+        rust::Slice<uint64_t> out
+) {
+    size_t outSize = out.size();
+    codec->decodeArray(in.data(), in.size(), out.data(), outSize);
+    return outSize;
+}
+
+} // namespace FastPForLib