v5: More backends! More UTF8! #965

Workflow file for this run

.github/workflows/prerelease.yml at 097feb7

	name: "Pre-Release"

	on:
	push:
	branches: ["main*"]
	pull_request:
	branches: ["main*"]

	env:
	BUILD_TYPE: Release
	GH_TOKEN: ${{ secrets.SEMANTIC_RELEASE_TOKEN }}
	PYTHON_VERSION: 3.13
	SWIFT_VERSION: 6.2
	PYTHONUTF8: 1

	# Sets permissions of the GITHUB_TOKEN to allow deployment to GitHub Pages
	permissions:
	contents: read

	jobs:
	versioning:
	name: Update Version
	runs-on: ubuntu-24.04
	steps:
	- name: Checkout
	uses: actions/checkout@v6
	with:
	fetch-depth: 0
	fetch-tags: true
	persist-credentials: false
	- name: Run TinySemVer
	uses: ashvardanian/tinysemver@v3.0.1
	with:
	verbose: "true"
	version-file: "VERSION"
	update-version-in: \|
	Cargo.toml:^version = "(\d+\.\d+\.\d+)"
	package.json:"version": "(\d+\.\d+\.\d+)"
	CMakeLists.txt:VERSION (\d+\.\d+\.\d+)
	update-major-version-in: \|
	include/stringzilla/stringzilla.h:^#define STRINGZILLA_H_VERSION_MAJOR (\d+)
	update-minor-version-in: \|
	include/stringzilla/stringzilla.h:^#define STRINGZILLA_H_VERSION_MINOR (\d+)
	update-patch-version-in: \|
	include/stringzilla/stringzilla.h:^#define STRINGZILLA_H_VERSION_PATCH (\d+)
	dry-run: "true"

	test_ubuntu_gcc:
	name: Ubuntu (GCC 12) - ${{ matrix.os }}
	runs-on: ${{ matrix.os }}
	strategy:
	fail-fast: false
	matrix:
	os: [ubuntu-22.04, ubuntu-22.04-arm]
	env:
	CC: gcc-12
	CXX: g++-12

	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	- name: Log hardware capabilities
	run: \|
	echo "────────────────────────────────────────────────────────────────"
	echo "Hardware Capabilities"
	echo "────────────────────────────────────────────────────────────────"
	lscpu
	echo ""
	echo "CPU Info:"
	grep -E "model name\|flags" /proc/cpuinfo \| head -5 \|\| true
	echo "────────────────────────────────────────────────────────────────"

	# C/C++
	# If the compilation fails, we want to log the compilation commands in addition to
	# the standard output.
	- name: Install dependencies
	run: \|
	sudo apt update
	sudo apt install -y cmake build-essential gcc-12 g++-12

	- name: Build C/C++
	run: \|
	cmake -B build_artifacts \
	-DCMAKE_BUILD_TYPE=RelWithDebInfo \
	-DCMAKE_EXPORT_COMPILE_COMMANDS=1 \
	-DSTRINGZILLA_BUILD_BENCHMARK=1 \
	-DSTRINGZILLA_BUILD_TEST=1

	cmake --build build_artifacts --config RelWithDebInfo > build_artifacts/logs.txt 2>&1 \|\| {
	echo "Compilation failed. Here are the logs:"
	cat build_artifacts/logs.txt
	echo "The original compilation commands:"
	cat build_artifacts/compile_commands.json
	echo "GCC Version:"
	gcc-12 --version
	echo "G++ Version:"
	g++-12 --version
	exit 1
	}
	- name: Test C++
	run: build_artifacts/stringzilla_test_cpp20
	- name: Check stringzilla_bare is built correctly
	run: test -z "$(ldd build_artifacts/libstringzilla_bare.so \| grep '.so')"
	- name: Build shared library for Go binding
	run: \|
	cmake -B build_shared -DCMAKE_BUILD_TYPE=RelWithDebInfo -DSTRINGZILLA_BUILD_SHARED=1
	cmake --build build_shared --target stringzilla_shared --config RelWithDebInfo
	- name: Set up Go
	uses: actions/setup-go@v6
	with:
	go-version: ">=1.24"
	cache: false
	- name: Test Go binding
	working-directory: golang
	env:
	LD_LIBRARY_PATH: ${{ github.workspace }}/build_shared
	run: go test -v -tags no_sha256 # Skip SHA tests for GCC
	- name: Test on Real World Data
	run: \|
	build_artifacts/stringzilla_bench_memory_cpp20 ${DATASET_PATH} # for string copies and fills
	build_artifacts/stringzilla_bench_find_cpp20 ${DATASET_PATH} # for substring search
	build_artifacts/stringzilla_bench_token_cpp20 ${DATASET_PATH} # for hashing, equality comparisons, etc.
	build_artifacts/stringzilla_bench_sequence_cpp20 ${DATASET_PATH} # for sorting arrays of strings
	build_artifacts/stringzilla_bench_container_cpp20 ${DATASET_PATH} # for STL containers with string keys
	env:
	DATASET_PATH: ./README.md
	# Don't overload GitHub with our benchmarks.
	# The results in such an unstable environment will be meaningless anyway.
	if: 0

	- name: Test Parallel Algorithms on Real World Data
	run: \|
	build_artifacts/stringzillas_bench_similarities_cpp20 ${DATASET_PATH} # for edit distances and alignment scores
	build_artifacts/stringzillas_bench_fingerprints_cpp20 ${DATASET_PATH} # for multi-needle search in many strings
	env:
	DATASET_PATH: ./README.md
	# Don't overload GitHub with our benchmarks.
	# The results in such an unstable environment will be meaningless anyway.
	if: 0

	# Python
	- name: Set up Python ${{ env.PYTHON_VERSION }}
	uses: actions/setup-python@v6.2.0
	with:
	python-version: ${{ env.PYTHON_VERSION }}
	- name: Build Python
	run: \|
	python -m pip install --upgrade pip
	pip install pytest pytest-repeat numpy pyarrow
	python -m pip install .
	- name: Log Python capabilities
	run: \|
	python - << 'PY'
	import platform
	import stringzilla as sz
	print("platform:", platform.platform())
	print("machine:", platform.machine(), "processor:", platform.processor())
	print("stringzilla.__capabilities__:", sz.__capabilities__)
	PY
	- name: Test Python
	run: python -X faulthandler -m pytest test/ --ignore=test/stringzillas.py -s -vv --maxfail=1 --full-trace

	# JavaScript
	- name: Set up Node.js
	uses: actions/setup-node@v6
	with:
	node-version: 22
	- name: Build and test JavaScript
	run: npm install && npm test

	# Rust
	- name: Set up Rust
	uses: dtolnay/rust-toolchain@stable

	test_ubuntu_clang:
	name: Ubuntu (Clang 16) - ${{ matrix.os }}
	runs-on: ${{ matrix.os }}
	strategy:
	fail-fast: false
	matrix:
	os: [ubuntu-22.04, ubuntu-22.04-arm]
	env:
	CC: clang-16
	CXX: clang++-16

	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	- name: Log hardware capabilities
	run: \|
	echo "────────────────────────────────────────────────────────────────"
	echo "Hardware Capabilities"
	echo "────────────────────────────────────────────────────────────────"
	lscpu
	echo ""
	echo "CPU Info:"
	grep -E "model name\|flags" /proc/cpuinfo \| head -5 \|\| true
	echo "────────────────────────────────────────────────────────────────"

	# C/C++
	# Clang 16 isn't available from default repos on Ubuntu 22.04, so we have to install it manually
	- name: Install dependencies
	run: \|
	sudo apt update
	sudo apt install -y cmake build-essential libjemalloc-dev
	wget https://apt.llvm.org/llvm.sh
	chmod +x llvm.sh
	sudo ./llvm.sh 16

	- name: Build C/C++
	run: \|
	cmake -B build_artifacts \
	-DCMAKE_BUILD_TYPE=RelWithDebInfo \
	-DCMAKE_EXPORT_COMPILE_COMMANDS=1 \
	-DSTRINGZILLA_BUILD_BENCHMARK=1 \
	-DSTRINGZILLA_BUILD_TEST=1

	cmake --build build_artifacts --config RelWithDebInfo > build_artifacts/logs.txt 2>&1 \|\| {
	echo "Compilation failed. Here are the logs:"
	cat build_artifacts/logs.txt
	echo "The original compilation commands:"
	cat build_artifacts/compile_commands.json
	echo "Clang Version:"
	clang-16 --version
	echo "Clang++ Version:"
	clang++-16 --version
	exit 1
	}
	- name: Test C++
	run: build_artifacts/stringzilla_test_cpp20
	- name: Check stringzilla_bare is built correctly
	run: test -z "$(ldd build_artifacts/libstringzilla_bare.so \| grep '.so')"
	- name: Build shared library for Go binding
	run: \|
	cmake -B build_shared -DCMAKE_BUILD_TYPE=RelWithDebInfo -DSTRINGZILLA_BUILD_SHARED=1
	cmake --build build_shared --target stringzilla_shared --config RelWithDebInfo
	- name: Set up Go
	uses: actions/setup-go@v6
	with:
	go-version: ">=1.24"
	cache: false
	- name: Test Go binding
	working-directory: golang
	env:
	LD_LIBRARY_PATH: ${{ github.workspace }}/build_shared
	run: go test -v
	- name: Test on Real World Data
	run: \|
	build_artifacts/stringzilla_bench_find_cpp20 ${DATASET_PATH} # for substring search
	build_artifacts/stringzilla_bench_token_cpp20 ${DATASET_PATH} # for hashing, equality comparisons, etc.
	build_artifacts/stringzilla_bench_sequence_cpp20 ${DATASET_PATH} # for sorting arrays of strings
	build_artifacts/stringzilla_bench_container_cpp20 ${DATASET_PATH} # for STL containers with string keys
	env:
	DATASET_PATH: ./README.md
	# Don't overload GitHub with our benchmarks.
	# The results in such an unstable environment will be meaningless anyway.
	if: 0

	# Python
	- name: Set up Python ${{ env.PYTHON_VERSION }}
	uses: actions/setup-python@v6.2.0
	with:
	python-version: ${{ env.PYTHON_VERSION }}
	- name: Build Python
	run: \|
	python -m pip install --upgrade pip
	pip install pytest pytest-repeat numpy pyarrow
	python -m pip install .
	- name: Log Python capabilities
	run: \|
	python - << 'PY'
	import platform
	import stringzilla as sz
	print("platform:", platform.platform())
	print("machine:", platform.machine(), "processor:", platform.processor())
	print("stringzilla.__capabilities__:", sz.__capabilities__)
	PY
	- name: Test Python
	run: python -X faulthandler -m pytest test/ --ignore=test/stringzillas.py -s -vv --maxfail=1 --full-trace

	# Rust
	- name: Set up Rust
	uses: dtolnay/rust-toolchain@stable

	# Swift
	# Fails due to: https://github.com/swift-actions/setup-swift/issues/591
	# - name: Set up Swift ${{ env.SWIFT_VERSION }}
	# uses: swift-actions/setup-swift@v2.4.0
	# with:
	# swift-version: ${{ env.SWIFT_VERSION }}
	# - name: Build Swift
	# run: swift build -c release --static-swift-stdlib
	# - name: Test Swift
	# run: swift test -c release

	test_ubuntu_cpus:
	name: Ubuntu (StringZillas-CPUs)
	runs-on: ubuntu-22.04
	env:
	CC: gcc-12
	CXX: g++-12

	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	- name: Log hardware capabilities
	run: \|
	echo "────────────────────────────────────────────────────────────────"
	echo "Hardware Capabilities"
	echo "────────────────────────────────────────────────────────────────"
	lscpu
	echo ""
	echo "CPU Info:"
	grep -E "model name\|flags" /proc/cpuinfo \| head -5 \|\| true
	echo "────────────────────────────────────────────────────────────────"

	# C/C++
	- name: Install dependencies
	run: \|
	sudo apt update
	sudo apt install -y cmake build-essential gcc-12 g++-12

	- name: Build C/C++
	run: \|
	cmake -B build_artifacts \
	-DCMAKE_BUILD_TYPE=RelWithDebInfo \
	-DCMAKE_EXPORT_COMPILE_COMMANDS=1 \
	-DSTRINGZILLA_BUILD_BENCHMARK=1 \
	-DSTRINGZILLA_BUILD_TEST=1

	cmake --build build_artifacts --config RelWithDebInfo > build_artifacts/logs.txt 2>&1 \|\| {
	echo "Compilation failed. Here are the logs:"
	cat build_artifacts/logs.txt
	echo "The original compilation commands:"
	cat build_artifacts/compile_commands.json
	echo "GCC Version:"
	gcc-12 --version
	echo "G++ Version:"
	g++-12 --version
	exit 1
	}
	- name: Test C++
	run: build_artifacts/stringzilla_test_cpp20
	- name: Check stringzilla_bare is built correctly
	run: test -z "$(ldd build_artifacts/libstringzilla_bare.so \| grep '.so')"

	# Python StringZillas-CPUs
	- name: Set up Python ${{ env.PYTHON_VERSION }}
	uses: actions/setup-python@v6.2.0
	with:
	python-version: ${{ env.PYTHON_VERSION }}
	# Ensure modern build backend is available for editable installs
	- name: Build Python StringZillas-CPUs
	run: \|
	python -m pip install --upgrade pip
	pip install "setuptools>=64" wheel
	pip install pytest pytest-repeat numpy pyarrow affine-gaps
	SZ_TARGET=stringzilla pip install -e . --force-reinstall --no-build-isolation
	SZ_TARGET=stringzillas-cpus pip install -e . --force-reinstall --no-build-isolation
	- name: Log Python capabilities
	run: \|
	python - << 'PY'
	import platform
	import stringzilla as sz
	import stringzillas as szs
	print("platform:", platform.platform())
	print("machine:", platform.machine(), "processor:", platform.processor())
	print("stringzilla.__capabilities__:", sz.__capabilities__)
	print("stringzillas.__capabilities__:", szs.__capabilities__)
	PY
	- name: Test Python StringZillas-CPUs
	run: python -X faulthandler -m pytest test/stringzillas.py test/doctests.py -s -vv --maxfail=1 --full-trace

	test_ubuntu_cuda:
	name: Ubuntu (StringZillas-CUDA)
	runs-on: ubuntu-22.04
	env:
	CC: gcc-12
	CXX: g++-12

	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	# Install CUDA Toolkit
	- name: Install CUDA Toolkit
	uses: Jimver/cuda-toolkit@v0.2.35
	with:
	cuda: "12.9.1"
	method: "network"

	- name: Log hardware capabilities
	run: \|
	echo "────────────────────────────────────────────────────────────────"
	echo "Hardware Capabilities"
	echo "────────────────────────────────────────────────────────────────"
	lscpu
	echo ""
	echo "CPU Info:"
	grep -E "model name\|flags" /proc/cpuinfo \| head -5 \|\| true
	echo ""
	echo "GPU Info:"
	nvcc --version 2>/dev/null \|\| echo "nvcc not yet installed"
	nvidia-smi 2>/dev/null \|\| echo "nvidia-smi not yet available"
	echo "────────────────────────────────────────────────────────────────"

	# C/C++
	- name: Install dependencies
	run: \|
	sudo apt update
	sudo apt install -y cmake build-essential gcc-12 g++-12

	- name: Build C/C++ with CUDA
	run: \|
	# Note: Using Release build type instead of RelWithDebInfo
	# Optimized debugging not supported in CUDA and results in PTXAS fatal errors
	cmake -B build_artifacts \
	-DCMAKE_BUILD_TYPE=Release \
	-DCMAKE_EXPORT_COMPILE_COMMANDS=1 \
	-DSTRINGZILLA_BUILD_BENCHMARK=1 \
	-DSTRINGZILLA_BUILD_TEST=1

	cmake --build build_artifacts --config Release > build_artifacts/logs.txt 2>&1 \|\| {
	echo "Compilation failed. Here are the logs:"
	cat build_artifacts/logs.txt
	echo "The original compilation commands:"
	cat build_artifacts/compile_commands.json
	echo "CUDA Version:"
	nvcc --version
	echo "GCC Version:"
	gcc-12 --version
	echo "G++ Version:"
	g++-12 --version
	exit 1
	}
	- name: Test C++
	run: build_artifacts/stringzilla_test_cpp20
	- name: Check stringzilla_bare is built correctly
	run: test -z "$(ldd build_artifacts/libstringzilla_bare.so \| grep '.so')"

	# Python StringZillas-CUDA
	- name: Set up Python ${{ env.PYTHON_VERSION }}
	uses: actions/setup-python@v6.2.0
	with:
	python-version: ${{ env.PYTHON_VERSION }}
	# Ensure modern build backend is available for editable installs
	- name: Build Python StringZillas-CUDA
	run: \|
	python -m pip install --upgrade pip
	pip install "setuptools>=64" wheel
	pip install pytest pytest-repeat numpy pyarrow wheel affine-gaps
	SZ_TARGET=stringzilla pip install -e . --force-reinstall --no-build-isolation
	SZ_TARGET=stringzillas-cuda pip install -e . --force-reinstall --no-build-isolation
	- name: Log Python capabilities
	run: \|
	python - << 'PY'
	import platform
	import stringzilla as sz
	import stringzillas as szs
	print("platform:", platform.platform())
	print("machine:", platform.machine(), "processor:", platform.processor())
	print("stringzilla.__capabilities__:", sz.__capabilities__)
	print("stringzillas.__capabilities__:", szs.__capabilities__)
	PY
	- name: Test Python StringZillas-CUDA
	run: python -X faulthandler -m pytest test/stringzillas.py test/doctests.py -s -vv --maxfail=1 --full-trace

	# Temporary workaround to run Swift tests on Linux
	# Based on: https://github.com/swift-actions/setup-swift/issues/591#issuecomment-1685710678
	test_ubuntu_swift:
	name: Swift on Linux
	runs-on: ubuntu-22.04
	container: swift:latest
	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	- name: Log hardware capabilities
	run: \|
	echo "────────────────────────────────────────────────────────────────"
	echo "Hardware Capabilities"
	echo "────────────────────────────────────────────────────────────────"
	lscpu \|\| true
	echo ""
	echo "CPU Info:"
	grep -E "model name\|flags" /proc/cpuinfo \| head -5 \|\| true
	echo "────────────────────────────────────────────────────────────────"

	- name: Test Swift
	run: swift test

	test_ubuntu_dotnet:
	name: C# on Linux
	runs-on: ubuntu-24.04
	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	- name: Install build tools
	run: \|
	sudo apt update
	sudo apt install -y cmake build-essential

	- name: Build shared library
	run: \|
	cmake -B build_shared -DCMAKE_BUILD_TYPE=Release -DSTRINGZILLA_BUILD_SHARED=1
	cmake --build build_shared --target stringzilla_shared --config Release

	# The runtime resolver probes runtimes/{rid}/native first; the shared target emits
	# libstringzilla_shared.so, so copy it under the name the binding looks up.
	- name: Stage native library
	run: \|
	mkdir -p csharp/runtimes/linux-x64/native
	cp -L build_shared/libstringzilla_shared.so csharp/runtimes/linux-x64/native/libstringzilla.so

	- name: Set up .NET
	uses: actions/setup-dotnet@v5
	with:
	dotnet-version: 8.0.x

	- name: Check formatting
	run: \|
	dotnet format StringZilla.csproj --verify-no-changes
	dotnet format csharp/StringZillaTests.csproj --verify-no-changes

	- name: Test C#
	run: dotnet test csharp/StringZillaTests.csproj -c Release --nologo

	test_ubuntu_java:
	name: Java on Linux
	runs-on: ubuntu-24.04
	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	- name: Install build tools
	run: \|
	sudo apt update
	sudo apt install -y cmake build-essential

	- name: Build shared library
	run: \|
	cmake -B build_shared -DCMAKE_BUILD_TYPE=Release -DSTRINGZILLA_BUILD_SHARED=1
	cmake --build build_shared --target stringzilla_shared --config Release

	# NativeLoader extracts /native/{os}-{arch}/libstringzilla.so from the classpath at runtime;
	# Maven copies src/main/resources into target/classes before the tests run.
	- name: Stage native library
	run: \|
	mkdir -p java/src/main/resources/native/linux-x86_64
	cp -L build_shared/libstringzilla_shared.so java/src/main/resources/native/linux-x86_64/libstringzilla.so

	# JDK 23: finalized Foreign Function & Memory API (since 22) and Palantir Java Format (needs <= 23).
	- name: Set up JDK
	uses: actions/setup-java@v5
	with:
	distribution: temurin
	java-version: "23"
	cache: maven

	- name: Check formatting
	run: mvn -B spotless:check

	- name: Test Java
	run: mvn -B test

	# Cross-compile the static test executables once per target architecture and publish them as artifacts. The
	# actual execution under QEMU happens in `test_cross_qemu`, which fans the vector-length-agnostic RISC-V
	# binary across several VLENs WITHOUT recompiling it.
	test_cross_linux:
	name: Cross Compile (${{ matrix.arch }} → ${{ matrix.target_arch }})
	runs-on: ${{ matrix.os }}
	strategy:
	fail-fast: false
	matrix:
	include:
	- os: ubuntu-24.04
	arch: x86-64
	target_arch: aarch64
	target: aarch64-linux-gnu
	target_processor: arm64
	cflags: ""
	file_id: aarch64 # substring `file` prints for this ELF
	- os: ubuntu-24.04-arm
	arch: aarch64
	target_arch: x86-64
	target: x86_64-linux-gnu
	target_processor: amd64
	cflags: ""
	file_id: x86-64
	- os: ubuntu-24.04
	arch: x86-64
	target_arch: riscv64
	target: riscv64-linux-gnu
	target_processor: riscv64
	cflags: "-march=rv64gcv"
	file_id: RISC-V # `file` prints "UCB RISC-V", not "riscv64"

	env:
	CC: clang-20
	CXX: clang++-20
	AR: llvm-ar-20
	NM: llvm-nm-20
	RANLIB: llvm-ranlib-20

	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	- name: Log hardware capabilities
	run: \|
	echo "────────────────────────────────────────────────────────────────"
	echo "Cross-Compilation: ${{ matrix.arch }} → ${{ matrix.target_arch }}"
	echo "────────────────────────────────────────────────────────────────"
	lscpu
	echo ""
	echo "CPU Info:"
	grep -E "model name\|flags" /proc/cpuinfo \| head -5 \|\| true
	echo "────────────────────────────────────────────────────────────────"

	- name: Install cross-compilation toolchain
	run: \|
	sudo apt-get update
	sudo apt-get install -y make build-essential crossbuild-essential-${{ matrix.target_processor }}
	wget https://apt.llvm.org/llvm.sh
	chmod +x llvm.sh
	sudo ./llvm.sh 20

	- name: Build C/C++ for ${{ matrix.target_arch }}
	run: \|
	# Static-link the test/benchmark executables so QEMU user-mode can run them without juggling the
	# cross dynamic loader + sysroot library paths. `clang` drives the GNU linker directly and does not
	# add `-lstdc++` on its own, so we point it at the cross libstdc++ and pull it in explicitly.
	gcc_libdir=$(dirname "$(find /usr/lib/gcc-cross/${{ matrix.target }} -name 'libstdc++.a' 2>/dev/null \| head -1)")
	cmake -B build_artifacts \
	-DCMAKE_BUILD_TYPE=RelWithDebInfo \
	-DCMAKE_EXPORT_COMPILE_COMMANDS=1 \
	-DCMAKE_C_COMPILER_TARGET=${{ matrix.target }} \
	-DCMAKE_CXX_COMPILER_TARGET=${{ matrix.target }} \
	-DCMAKE_SYSTEM_NAME=Linux \
	-DCMAKE_SYSTEM_PROCESSOR=${{ matrix.target_processor }} \
	-DCMAKE_C_FLAGS="${{ matrix.cflags }}" \
	-DCMAKE_CXX_FLAGS="${{ matrix.cflags }}" \
	-DCMAKE_EXE_LINKER_FLAGS="-static -L${gcc_libdir} -lstdc++ -lm" \
	-DSTRINGZILLA_BUILD_SHARED=1 \
	-DSTRINGZILLA_BUILD_BENCHMARK=1 \
	-DSTRINGZILLA_BUILD_TEST=1

	cmake --build build_artifacts --config RelWithDebInfo

	- name: Verify ${{ matrix.target_arch }} binaries
	run: \|
	test -e build_artifacts/libstringzilla_bare.so
	test -e build_artifacts/libstringzilla_shared.so
	test -e build_artifacts/stringzilla_test_cpp20
	file build_artifacts/stringzilla_test_cpp20 \| grep -q "${{ matrix.file_id }}"
	echo "✓ ${{ matrix.target_arch }} binaries built successfully"

	- name: Upload ${{ matrix.target_arch }} test binaries
	uses: actions/upload-artifact@v4
	with:
	name: cross-tests-${{ matrix.target_arch }}
	path: build_artifacts/stringzilla_test_
	if-no-files-found: error
	retention-days: 1

	# Run the cross-compiled (static) test binaries under QEMU user-mode. Build-and-verify alone never exercised
	# the SIMD kernels. Each row is its own pass/fail status; RVV/SVE are vector-length-agnostic, so the RISC-V
	# binary built once above is swept across VLEN 128/256/512 here (`-cpu rv64,v=true,vlen=N`) with no rebuild —
	# a single width under-tests VLA code.
	test_cross_qemu:
	name: QEMU Tests (${{ matrix.target_arch }}${{ matrix.variant }})
	needs: test_cross_linux
	runs-on: ubuntu-24.04
	strategy:
	fail-fast: false
	matrix:
	include:
	# ARM is vector-length-agnostic under SVE just like RISC-V, so the one binary is run NEON-only
	# (`sve=off`, else dynamic dispatch always prefers SVE2 and NEON is never exercised) plus SVE at
	# 128/256/512-bit (`sve-max-vq` = VL/128).
	- target_arch: aarch64
	qemu: qemu-aarch64-static
	qemu_cpu: "max,sve=off"
	variant: " (NEON)"
	- target_arch: aarch64
	qemu: qemu-aarch64-static
	qemu_cpu: "max,sve-max-vq=1"
	variant: " @ sve=128"
	- target_arch: aarch64
	qemu: qemu-aarch64-static
	qemu_cpu: "max,sve-max-vq=2"
	variant: " @ sve=256"
	- target_arch: aarch64
	qemu: qemu-aarch64-static
	qemu_cpu: "max,sve-max-vq=4"
	variant: " @ sve=512"
	- target_arch: x86-64
	qemu: qemu-x86_64-static
	qemu_cpu: max
	variant: ""
	- target_arch: riscv64
	qemu: qemu-riscv64-static
	qemu_cpu: "rv64,v=true,vlen=128"
	variant: " @ vlen=128"
	- target_arch: riscv64
	qemu: qemu-riscv64-static
	qemu_cpu: "rv64,v=true,vlen=256"
	variant: " @ vlen=256"
	- target_arch: riscv64
	qemu: qemu-riscv64-static
	qemu_cpu: "rv64,v=true,vlen=512"
	variant: " @ vlen=512"

	steps:
	- name: Install QEMU user-mode
	run: \|
	sudo apt-get update
	sudo apt-get install -y qemu-user-static

	- name: Download ${{ matrix.target_arch }} test binaries
	uses: actions/download-artifact@v4
	with:
	name: cross-tests-${{ matrix.target_arch }}
	path: tests

	- name: Run tests under QEMU (${{ matrix.target_arch }}${{ matrix.variant }})
	run: \|
	set -euo pipefail
	chmod +x tests/*
	for test_binary in $(find tests -type f -name 'stringzilla_test_' \| sort); do
	echo "▶ ${{ matrix.qemu }} -cpu ${{ matrix.qemu_cpu }} ${test_binary}"
	${{ matrix.qemu }} -cpu "${{ matrix.qemu_cpu }}" "${test_binary}"
	done

	test_wasm:
	name: WebAssembly (wasm32-wasip1)
	runs-on: ubuntu-24.04
	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	- name: Install WASI SDK
	run: \|
	WASI_VERSION_FULL=25.0
	curl -sSL "https://github.com/WebAssembly/wasi-sdk/releases/download/wasi-sdk-25/wasi-sdk-${WASI_VERSION_FULL}-x86_64-linux.tar.gz" \
	\| tar xz -C "$HOME"
	echo "WASI_SDK_PATH=$HOME/wasi-sdk-${WASI_VERSION_FULL}-x86_64-linux" >> "$GITHUB_ENV"

	- name: Install Wasmtime
	run: \|
	curl https://wasmtime.dev/install.sh -sSf \| bash
	echo "$HOME/.wasmtime/bin" >> "$GITHUB_PATH"

	- name: Set up Rust
	uses: dtolnay/rust-toolchain@stable
	with:
	targets: wasm32-wasip1

	# The C kernels need the WASI SDK clang + sysroot; `build.rs` passes `-msimd128` (+ probed
	# `-mrelaxed-simd`). Tests run under Wasmtime via the `.cargo/config.toml` runner, single-threaded
	# (wasm32-wasip1 has no threads).
	- name: Build & test (relaxed-SIMD + SIMD128)
	env:
	CC_wasm32_wasip1: ${{ env.WASI_SDK_PATH }}/bin/clang
	AR_wasm32_wasip1: ${{ env.WASI_SDK_PATH }}/bin/llvm-ar
	CFLAGS_wasm32_wasip1: --sysroot=${{ env.WASI_SDK_PATH }}/share/wasi-sysroot
	run: \|
	cargo build --target wasm32-wasip1
	cargo test --release --target wasm32-wasip1 -- --test-threads=1

	- name: Build & test (baseline SIMD128, relaxed off)
	env:
	CC_wasm32_wasip1: ${{ env.WASI_SDK_PATH }}/bin/clang
	AR_wasm32_wasip1: ${{ env.WASI_SDK_PATH }}/bin/llvm-ar
	CFLAGS_wasm32_wasip1: --sysroot=${{ env.WASI_SDK_PATH }}/share/wasi-sysroot
	SZ_USE_V128RELAXED: "0"
	run: cargo test --release --target wasm32-wasip1 -- --test-threads=1

	test_macos:
	name: MacOS
	runs-on: macos-15

	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	- name: Log hardware capabilities
	run: \|
	echo "────────────────────────────────────────────────────────────────"
	echo "Hardware Capabilities"
	echo "────────────────────────────────────────────────────────────────"
	sysctl -a \| grep machdep.cpu \|\| true
	echo ""
	echo "System Info:"
	system_profiler SPHardwareDataType \|\| true
	echo "────────────────────────────────────────────────────────────────"

	# C/C++
	- name: Install dependencies
	run: \|
	brew update
	brew reinstall cmake
	- name: Build C/C++
	run: \|
	cmake -B build_artifacts \
	-DCMAKE_BUILD_TYPE=RelWithDebInfo \
	-DCMAKE_EXPORT_COMPILE_COMMANDS=1 \
	-DSTRINGZILLA_BUILD_BENCHMARK=1 \
	-DSTRINGZILLA_BUILD_TEST=1
	cmake --build build_artifacts --config RelWithDebInfo
	- name: Test C++
	run: build_artifacts/stringzilla_test_cpp17
	- name: Test on Real World Data
	run: \|
	build_artifacts/stringzilla_bench_find_cpp20 ${DATASET_PATH} # for substring search
	build_artifacts/stringzilla_bench_token_cpp20 ${DATASET_PATH} # for hashing, equality comparisons, etc.
	build_artifacts/stringzilla_bench_sequence_cpp20 ${DATASET_PATH} # for sorting arrays of strings
	build_artifacts/stringzilla_bench_container_cpp20 ${DATASET_PATH} # for STL containers with string keys
	env:
	DATASET_PATH: ./README.md
	# Don't overload GitHub with our benchmarks.
	# The results in such an unstable environment will be meaningless anyway.
	if: 0

	# Python
	- name: Set up Python ${{ env.PYTHON_VERSION }}
	uses: actions/setup-python@v6.2.0
	with:
	python-version: ${{ env.PYTHON_VERSION }}
	- name: Build Python
	run: \|
	python -m pip install --upgrade pip
	pip install pytest pytest-repeat numpy pyarrow
	python -m pip install .
	env:
	MACOSX_DEPLOYMENT_TARGET: "11.0"
	- name: Test Python
	run: python -X faulthandler -m pytest test/ --ignore=test/stringzillas.py -s -vv --maxfail=1 --full-trace

	# Swift
	- name: Set up Swift ${{ env.SWIFT_VERSION }}
	uses: swift-actions/setup-swift@v2.4.0
	with:
	swift-version: ${{ env.SWIFT_VERSION }}
	- name: Build Swift
	run: swift build
	- name: Test Swift
	run: swift test

	# Rust
	- name: Set up Rust
	uses: dtolnay/rust-toolchain@stable

	test_windows:
	name: Windows - ${{ matrix.os }}
	runs-on: ${{ matrix.os }}
	strategy:
	fail-fast: false
	matrix:
	include:
	- os: windows-2022
	arch: x64
	- os: windows-11-arm
	arch: arm64
	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive
	- uses: ilammy/msvc-dev-cmd@v1
	with:
	arch: ${{ matrix.arch }}

	- name: Log hardware capabilities
	shell: powershell
	run: \|
	Write-Host "────────────────────────────────────────────────────────────────"
	Write-Host "Hardware Capabilities"
	Write-Host "────────────────────────────────────────────────────────────────"
	Get-CimInstance -ClassName Win32_Processor \| Select-Object Name, NumberOfCores, NumberOfLogicalProcessors, MaxClockSpeed \| Format-List
	Write-Host "────────────────────────────────────────────────────────────────"

	- name: Build C/C++
	shell: cmd
	run: \|
	cmake -GNinja -B build_artifacts ^
	-DCMAKE_BUILD_TYPE=RelWithDebInfo ^
	-DCMAKE_EXPORT_COMPILE_COMMANDS=1 ^
	-DSTRINGZILLA_BUILD_BENCHMARK=1 ^
	-DSTRINGZILLA_BUILD_TEST=1

	cmake --build build_artifacts --config RelWithDebInfo > build_artifacts/logs.txt 2>&1 \|\| (
	echo "Compilation failed. Here are the logs:"
	type build_artifacts\logs.txt
	echo "The original compilation commands:"
	type build_artifacts\compile_commands.json
	exit 1
	)
	- name: Test C++
	run: .\build_artifacts\stringzilla_test_cpp20.exe
	- name: Check stringzilla_bare is built correctly
	shell: cmd
	run: dumpbin /nologo /dependents .\build_artifacts\stringzilla_bare.dll \| findstr /r "[a-zA-Z0-9]\.dll" \| findstr /r /v "KERNEL32.dll stringzilla" && exit /b 1 \|\| exit /b 0
	- name: Test on Real World Data
	run: \|
	.\build_artifacts\stringzilla_bench_find_cpp20.exe ${DATASET_PATH} # for substring search
	.\build_artifacts\stringzilla_bench_token_cpp20.exe ${DATASET_PATH} # for hashing, equality comparisons, etc.
	.\build_artifacts\stringzilla_bench_sequence_cpp20.exe ${DATASET_PATH} # for sorting arrays of strings
	.\build_artifacts\stringzilla_bench_container_cpp20.exe ${DATASET_PATH} # for STL containers with string keys
	env:
	DATASET_PATH: ./README.md
	# Don't overload GitHub with our benchmarks.
	# The results in such an unstable environment will be meaningless anyway.
	if: 0

	# Python
	- name: Set up Python ${{ env.PYTHON_VERSION }}
	uses: actions/setup-python@v6.2.0
	with:
	python-version: ${{ env.PYTHON_VERSION }}
	- name: Build Python
	run: \|
	python -m pip install --upgrade pip
	pip install pytest pytest-repeat numpy
	python -m pip install .
	- name: Test Python
	run: python -X faulthandler -m pytest test/ --ignore=test/stringzillas.py -s -vv --maxfail=1 --full-trace

	test_alpine:
	name: Alpine Linux
	runs-on: ubuntu-24.04
	container:
	image: alpine:latest
	options: --privileged # If needed for certain Docker operations
	steps:
	# bash and curl are required by dtolnay/rust-toolchain
	- name: Install Git for checkout
	run: \|
	apk add --no-cache git openssh ca-certificates bash curl
	git --version
	- name: Log Alpine and Git versions
	run: \|
	echo "Alpine: $(cat /etc/alpine-release 2>/dev/null \|\| echo 'unknown')"
	git --version \|\| echo "git not installed"
	- uses: actions/checkout@v6
	with:
	submodules: recursive

	- name: Log hardware capabilities
	run: \|
	echo "────────────────────────────────────────────────────────────────"
	echo "Hardware Capabilities"
	echo "────────────────────────────────────────────────────────────────"
	lscpu \|\| true
	echo ""
	echo "CPU Info:"
	grep -E "model name\|flags" /proc/cpuinfo \| head -5 \|\| true
	echo "────────────────────────────────────────────────────────────────"

	# C/C++
	- name: Build C/C++
	run: \|
	apk add --update make cmake g++ gcc
	cmake -B build_artifacts \
	-DCMAKE_BUILD_TYPE=RelWithDebInfo \
	-DCMAKE_EXPORT_COMPILE_COMMANDS=1 \
	-DSTRINGZILLA_BUILD_BENCHMARK=1 \
	-DSTRINGZILLA_BUILD_TEST=1
	cmake --build build_artifacts --config RelWithDebInfo
	- name: Test C++
	run: build_artifacts/stringzilla_test_cpp20
	- name: Check stringzilla_bare is built correctly
	run: test -z "$(ldd build_artifacts/libstringzilla_bare.so \| grep '.so' \| grep '=>')"

	# Python
	- name: Build Python
	run: \|
	apk add python3 py3-pip python3-dev
	pip install --break-system-packages pytest pytest-repeat
	pip install --break-system-packages .
	- name: Test Python
	run: pytest test/ --ignore=test/stringzillas.py -s -x

	# Rust
	- name: Set up Rust
	uses: dtolnay/rust-toolchain@stable

	build_wheels:
	name: Build Python ${{ matrix.python-version }} for ${{ matrix.os }}
	runs-on: ${{ matrix.os }}
	needs:
	[
	test_ubuntu_gcc,
	test_ubuntu_clang,
	test_macos,
	test_windows,
	test_cross_linux,
	]
	strategy:
	fail-fast: false
	matrix:
	os: [ubuntu-24.04, macos-15, windows-2022]
	python-version: ["310", "311", "312", "313", "314"]
	steps:
	- uses: actions/checkout@v6
	with:
	submodules: recursive
	- name: Set up Python
	uses: actions/setup-python@v6.2.0
	with:
	python-version: 3.x

	# We only need QEMU for Linux builds
	- name: Setup QEMU
	if: matrix.os == 'ubuntu-24.04'
	uses: docker/setup-qemu-action@v4
	- name: Install cibuildwheel
	run: python -m pip install cibuildwheel==3.4.1
	- name: Build wheels
	run: cibuildwheel --output-dir wheelhouse
	env:
	CIBW_BUILD: cp${{ matrix.python-version }}-*
	CIBW_ENVIRONMENT_LINUX: SZ_IS_QEMU_=1 # When emulating this will disable some tricky SIMD instructions

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

v5: More backends! More UTF8! #965

Workflow file

v5: More backends! More UTF8! #965

Uh oh!

Workflow file for this run