StringZilla 4.0!

.git-blame-ignore-revs

@@ -0,0 +1,40 @@
+6512f1d129aeddc8601c9df7332c135038914b68
+fc9e5d61e5fb1c5031f6f10920f6b50e2530de1e
+ad2af78f8651870727c5b39e1fea2eff26d71d2f
+49e8d9d240993bdf68715a9c87824a032752798d
+fc408fa0a0f2d947c610568bd7a5c4a60ecca443
+b835051c09a0ecfc420932de444f3c6839610764
+1ba7982559111d4fc9b58caa7bc7aa1c6e64257c
+5b55e19d1378c61da88309b30a38f9cf7c64bf79
+be4c63d926c8628451726863e4d14dbd1ea374dd
+8b401bd41e4bd9c29c8fad9a5b83d8232efa50c7
+295d49a38d66b08075357ac829ad66d80b5edab0
+2a1fcd113d217e3124f6501c38e93a318aca37f0
+2f7652141bd8dc3c2c38ab34321567bfcdb91d93
+9e3180019acffe5261f0a1713b4ea324dca79ea0
+45e57eefd796841cbd14ee7f75ec42b42b5bde0c
+66778d6b2b3aa0eed27e32fbdceef79b8c54eda5
+c357c3ea756523d3bcc8d8f25068ad08aef5456d
+9b1948b3771c21dd56954e5f43301ca8a0b8b1a9
+cbfe5c7ac6371047eae88621b092297474d0b82a
+085d2d3c8b99e0f90d320dd027040e554e410929
+3464cb428ae9a8721ab82a8c4bff214aa9ce6254
+5d0d2da422c7df96f9613ec843cd47c579a2edce
+89c46810c2f9bfafa31f8592339f9a1b45dcc245
+3f9c248fbf59add2246055462e8fc19dc9f1693b
+e23c35ff2c2d4ccb752f4ffbf9b6f39a1677b532
+7fdc58fd26e06c41052287d47a9c729c068a95ca
+10d829efcb8ed4cfa5f2db4050f8403184484423
+d74e5dca2e62eb0078cb2ebacc0dac2b8bb92d54
+1f60e6d7c81f0e285e594eb63fee6119e05a3e69
+a6768af38b40307fe66364403f141c285b3e164c
+08d0a20d35d3b29a44b9c8a826d53435c3ef839c
+9e9f2567d052d635722921a1d70ec63d69ec6669
+974ed78822dc0b519dd61bc1c4dc18d59fe4ad15
+b007ba571860e1d3737d1478c7f8d66ae1839e36
+14ba3bf3c43408438a7de9ad57118c747c1347b1
+9e577be71dcd2e20854bf55f08c54854b3e82989
+8cb0742b2d1b31b61fac5272f17017953c6677e6
+bd547453122e9f8565e5be15f137e7b0de37caca
+22e3d1e34d62d68c1e89df7c8bdc201faa18a9de
+ecb377541d0c706cf8997faff4f026b07e3f76f3

.github/workflows/prerelease.yml

@@ -275,7 +275,7 @@ jobs:
         # We can't run the produced builds, but we can make sure they exist
       - name: Test artifacts presense
         run: |
-          test -e build_artifacts/libstringzillite.so
+          test -e build_artifacts/libstringzilla_bare.so
           test -e build_artifacts/libstringzilla_shared.so
           test -e build_artifacts/stringzilla_test_cpp20
 

.github/workflows/release.yml

@@ -253,15 +253,15 @@ jobs:
 
           cmake --build build_release --config Release
 
-          cp build_release/libstringzillite.so "stringzillite_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}.so"
-          mkdir -p "stringzillite_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/DEBIAN"
-          touch "stringzillite_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/DEBIAN/control"
-          mkdir -p "stringzillite_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/usr/local/lib"
-          mkdir "stringzillite_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/usr/local/include"
-          cp include/stringzilla/stringzilla.h "stringzillite_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/usr/local/include/"
-          cp build_release/libstringzillite.so "stringzillite_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/usr/local/lib/"
-          echo -e "Package: stringzilla\nVersion: ${{ steps.set_version.outputs.version }}\nMaintainer: Ash Vardanian\nArchitecture: ${{ matrix.arch }}\nDescription: SIMD-accelerated string search, sort, hashes, fingerprints, & edit distances" > "stringzillite_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/DEBIAN/control"
-          dpkg-deb --build "stringzillite_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}"
+          cp build_release/libstringzilla_bare.so "stringzilla_bare_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}.so"
+          mkdir -p "stringzilla_bare_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/DEBIAN"
+          touch "stringzilla_bare_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/DEBIAN/control"
+          mkdir -p "stringzilla_bare_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/usr/local/lib"
+          mkdir "stringzilla_bare_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/usr/local/include"
+          cp include/stringzilla/stringzilla.h "stringzilla_bare_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/usr/local/include/"
+          cp build_release/libstringzilla_bare.so "stringzilla_bare_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/usr/local/lib/"
+          echo -e "Package: stringzilla\nVersion: ${{ steps.set_version.outputs.version }}\nMaintainer: Ash Vardanian\nArchitecture: ${{ matrix.arch }}\nDescription: SIMD-accelerated string search, sort, hashes, fingerprints, & edit distances" > "stringzilla_bare_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}/DEBIAN/control"
+          dpkg-deb --build "stringzilla_bare_linux_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}"
 
       - name: Upload library
         uses: xresloader/upload-to-github-release@v1
@@ -314,14 +314,14 @@ jobs:
         run: |
           cmake -DCMAKE_BUILD_TYPE=Release -B build_release
           cmake --build build_release --config Release
-          tar -cvf "stringzillite_windows_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}.tar" "build_release/stringzillite.dll" "./include/stringzilla/stringzilla.h"
+          tar -cvf "stringzilla_bare_windows_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}.tar" "build_release/stringzilla_bare.dll" "./include/stringzilla/stringzilla.h"
 
       - name: Upload archive
         uses: xresloader/upload-to-github-release@v1
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         with:
-          file: "stringzillite_windows_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}.tar"
+          file: "stringzilla_bare_windows_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}.tar"
           update_latest_release: true
 
   create_macos_library:
@@ -349,7 +349,7 @@ jobs:
         run: |
           cmake -DCMAKE_BUILD_TYPE=Release -B build_release
           cmake --build build_release --config Release        
-          zip -r stringzillite_macos_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}.zip build_release/libstringzillite.dylib include/stringzilla/stringzilla.h
+          zip -r stringzilla_bare_macos_${{ matrix.arch }}_${{ steps.set_version.outputs.version }}.zip build_release/libstringzilla_bare.dylib include/stringzilla/stringzilla.h
 
       - name: Upload archive
         uses: xresloader/upload-to-github-release@v1

.vscode/settings.json

@@ -27,6 +27,7 @@
     "Baeza",
     "basicsize",
     "bigram",
+    "bigrams",
     "bioinformaticians",
     "bioinformatics",
     "Bitap",
@@ -38,6 +39,7 @@
     "cheminformatics",
     "cibuildwheel",
     "CONCAT",
+    "constexpr",
     "copydoc",
     "Corasick",
     "cptr",
@@ -50,6 +52,7 @@
     "getslice",
     "Giancarlo",
     "Gonnet",
+    "Haswell",
     "Heikki",
     "hexdigits",
     "Hirschberg's",
@@ -80,6 +83,7 @@
     "Merkle-Damgård",
     "Mersenne",
     "MODINIT",
+    "MSVC",
     "napi",
     "nargsf",
     "ndim",
@@ -102,6 +106,7 @@
     "readlines",
     "releasebuffer",
     "rfind",
+    "rfinds",
     "richcompare",
     "Ritchie",
     "rmatcher",
@@ -111,11 +116,13 @@
     "rsplits",
     "rstrip",
     "SIMD",
+    "Skylake",
     "splitlines",
     "ssize",
     "startswith",
     "STL",
     "stringzilla",
+    "stringzilla_bare",
     "Strs",
     "strzl",
     "substr",
@@ -129,6 +136,7 @@
     "unpoison",
     "usecases",
     "Vardanian",
+    "VBMI",
     "vectorcallfunc",
     "Wagner",
     "whitespaces",

CMakeLists.txt

@@ -1,4 +1,37 @@
-cmake_minimum_required(VERSION 3.1)
+# StringZilla CMakeLists.txt
+#
+# This file defines several library build & installation targets:
+#
+# - stringzilla_header: A header-only library with the StringZilla C and C++ headers.
+# - stringzilla_shared: A shared library with the StringZilla C and C++ headers and dynamic SIMD dispatch.
+# - stringzilla_bare: A shared library with the StringZilla headers, but without linking the standard C library.
+# 
+# Tests for different C++ standards:
+#
+#   - stringzilla_test_cpp11: C++11 baseline support.
+#   - stringzilla_test_cpp14: C++14 support with `std::less<std::string>`-like function objects.
+#   - stringzilla_test_cpp17: C++17 support with `std::string_view` compatibility.
+#   - stringzilla_test_cpp20: C++20 support with `<=>` operator and more `constexpr` features.
+#
+# Tests for different SIMD architectures:
+#
+#   - stringzilla_test_cpp20_serial: A test executable for serial execution.
+#   - stringzilla_test_cpp20_haswell: A test executable for AVX2.
+#   - stringzilla_test_cpp20_ice: A test executable for AVX-512.
+#   - stringzilla_test_cpp20_neon: A test executable for ARM Neon.
+#   - stringzilla_test_cpp20_sve: A test executable for ARM Scalable Vector Extension.
+#
+# Benchmarks:
+#
+#   - stringzilla_bench_search: A benchmark for substring search operations.
+#   - stringzilla_bench_similarity: A benchmark for similarity operations.
+#   - stringzilla_bench_sort: A benchmark for sorting operations.
+#   - stringzilla_bench_token: A benchmark for comparators and hash functions.
+#   - stringzilla_bench_container: A benchmark for STL containers powered by StringZilla.
+#   - stringzilla_bench_memory: A benchmark for LibC-style low-level memory operations.
+#
+# For higher-level language bindings separate build scripts are provided, native to each toolchain.
+cmake_minimum_required(VERSION 3.14 FATAL_ERROR)
 project(
   stringzilla
   VERSION 3.11.0
@@ -7,7 +40,7 @@ project(
   HOMEPAGE_URL "https://github.com/ashvardanian/stringzilla")
 
 set(CMAKE_C_STANDARD 99)
-set(CMAKE_CXX_STANDARD 17) # This gives many issues for msvc and clang-cl, especially if later on you set it to std-c++11 later on in the tests...
+set(CMAKE_CXX_STANDARD 11)
 
 set(CMAKE_C_EXTENSIONS OFF)
 set(CMAKE_CXX_EXTENSIONS OFF)
@@ -46,7 +79,7 @@ elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64|AARCH64|arm64|ARM64")
   message(STATUS "Platform: ARM")
 endif()
 
-# Determine if StringZilla is built as a subproject (using `add_subdirectory`)
+# Determine if StringZilla is built as a sub-project (using `add_subdirectory`)
 # or if it is the main project
 set(STRINGZILLA_IS_MAIN_PROJECT OFF)
 
@@ -78,19 +111,9 @@ endif()
 
 # Configuration
 include(GNUInstallDirs)
-set(STRINGZILLA_TARGET_NAME ${PROJECT_NAME})
 set(STRINGZILLA_INCLUDE_BUILD_DIR "${PROJECT_SOURCE_DIR}/include/")
 set(STRINGZILLA_INCLUDE_INSTALL_DIR "${CMAKE_INSTALL_INCLUDEDIR}")
 
-# Define our library
-add_library(${STRINGZILLA_TARGET_NAME} INTERFACE)
-add_library(${PROJECT_NAME}::${STRINGZILLA_TARGET_NAME} ALIAS ${STRINGZILLA_TARGET_NAME})
-
-target_include_directories(
-  ${STRINGZILLA_TARGET_NAME}
-  INTERFACE $<BUILD_INTERFACE:${STRINGZILLA_INCLUDE_BUILD_DIR}>
-  $<INSTALL_INTERFACE:include>)
-
 
 if(${CMAKE_VERSION} VERSION_EQUAL 3.13 OR ${CMAKE_VERSION} VERSION_GREATER 3.13)
   include(CTest)
@@ -99,7 +122,7 @@ endif()
 
 if (MSVC)
   # Remove /RTC* from MSVC debug flags by default (it will be added back in the set_compiler_flags function)
-  # Beacuse /RTC* cannot be used without the crt so it needs to be disabled for that specifc target
+  # Because /RTC* cannot be used without the crt so it needs to be disabled for that specific target
   string(REGEX REPLACE "/RTC[^ ]*" "" CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG}")
   string(REGEX REPLACE "/RTC[^ ]*" "" CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG}")
 endif()
@@ -109,7 +132,6 @@ function(set_compiler_flags target cpp_standard target_arch)
   get_target_property(target_type ${target} TYPE)
 
   target_include_directories(${target} PRIVATE scripts)
-  target_link_libraries(${target} PRIVATE ${STRINGZILLA_TARGET_NAME})
 
   # Set output directory for single-configuration generators (like Make)
   set_target_properties(${target} PROPERTIES
@@ -245,6 +267,7 @@ endfunction()
 function(define_launcher exec_name source cpp_standard target_arch)
   add_executable(${exec_name} ${source})
   set_compiler_flags(${exec_name} ${cpp_standard} "${target_arch}")
+  target_link_libraries(${exec_name} PRIVATE stringzilla_header)
   add_test(NAME ${exec_name} COMMAND ${exec_name})
 endfunction()
 
@@ -270,25 +293,36 @@ if(${STRINGZILLA_BUILD_TEST})
   if(SZ_PLATFORM_X86)
     # x86 specific backends
     if (MSVC)
-      define_launcher(stringzilla_test_cpp20_x86_serial scripts/test.cpp 20 "AVX")
-      define_launcher(stringzilla_test_cpp20_x86_avx2 scripts/test.cpp 20 "AVX2")
-      define_launcher(stringzilla_test_cpp20_x86_avx512 scripts/test.cpp 20 "AVX512")
+      define_launcher(stringzilla_test_cpp20_serial scripts/test.cpp 20 "AVX")
+      define_launcher(stringzilla_test_cpp20_haswell scripts/test.cpp 20 "AVX2")
+      define_launcher(stringzilla_test_cpp20_ice scripts/test.cpp 20 "AVX512")
     else()
-      define_launcher(stringzilla_test_cpp20_x86_serial scripts/test.cpp 20 "ivybridge")
-      define_launcher(stringzilla_test_cpp20_x86_avx2 scripts/test.cpp 20 "haswell")
-      define_launcher(stringzilla_test_cpp20_x86_avx512 scripts/test.cpp 20 "sapphirerapids")
+      define_launcher(stringzilla_test_cpp20_serial scripts/test.cpp 20 "ivybridge")
+      define_launcher(stringzilla_test_cpp20_haswell scripts/test.cpp 20 "haswell")
+      define_launcher(stringzilla_test_cpp20_ice scripts/test.cpp 20 "sapphirerapids")
     endif()
   elseif(SZ_PLATFORM_ARM)
     # ARM specific backends
-    define_launcher(stringzilla_test_cpp20_arm_serial scripts/test.cpp 20 "armv8-a")
-    define_launcher(stringzilla_test_cpp20_arm_neon scripts/test.cpp 20 "armv8-a+simd")
+    define_launcher(stringzilla_test_cpp20_serial scripts/test.cpp 20 "armv8-a")
+    define_launcher(stringzilla_test_cpp20_neon scripts/test.cpp 20 "armv8-a+simd")
+    define_launcher(stringzilla_test_cpp20_sve scripts/test.cpp 20 "armv8.2-a+sve")
   endif()
 endif()
 
+# Define our libraries, first the header-only version
+add_library(stringzilla_header INTERFACE)
+add_library(${PROJECT_NAME}::stringzilla_header ALIAS stringzilla_header)
+target_include_directories(
+  stringzilla_header
+  INTERFACE $<BUILD_INTERFACE:${STRINGZILLA_INCLUDE_BUILD_DIR}>
+  $<INSTALL_INTERFACE:include>)
+
+
 if(${STRINGZILLA_BUILD_SHARED})
 
   function(define_shared target)
     add_library(${target} SHARED c/lib.c)
+    add_library(${PROJECT_NAME}::${target} ALIAS ${target})
 
     set_target_properties(${target} PROPERTIES
       VERSION ${PROJECT_VERSION}
@@ -303,18 +337,20 @@ if(${STRINGZILLA_BUILD_SHARED})
       endif()
 
       target_compile_definitions(${target} PRIVATE
-        "SZ_USE_X86_AVX512=1"
-        "SZ_USE_X86_AVX2=1"
-        "SZ_USE_ARM_NEON=0"
-        "SZ_USE_ARM_SVE=0")
+        "SZ_USE_HASWELL=1"
+        "SZ_USE_SKYLAKE=1"
+        "SZ_USE_ICE=1"
+        "SZ_USE_NEON=0"
+        "SZ_USE_SVE=0")
     elseif(SZ_PLATFORM_ARM)
       set_compiler_flags(${target} "" "armv8-a")
 
       target_compile_definitions(${target} PRIVATE
-        "SZ_USE_X86_AVX512=0"
-        "SZ_USE_X86_AVX2=0"
-        "SZ_USE_ARM_NEON=1"
-        "SZ_USE_ARM_SVE=1")
+        "SZ_USE_HASWELL=0"
+        "SZ_USE_SKYLAKE=0"
+        "SZ_USE_ICE=0"
+        "SZ_USE_NEON=1"
+        "SZ_USE_SVE=1")
     endif()
 
     if (MSVC)
@@ -333,16 +369,16 @@ if(${STRINGZILLA_BUILD_SHARED})
   target_compile_definitions(stringzilla_shared PRIVATE "SZ_OVERRIDE_LIBC=1")
 
   # Try compiling a version without linking the LibC
-  define_shared(stringzillite)
-  target_compile_definitions(stringzillite PRIVATE "SZ_AVOID_LIBC=1")
-  target_compile_definitions(stringzillite PRIVATE "SZ_OVERRIDE_LIBC=1")
+  define_shared(stringzilla_bare)
+  target_compile_definitions(stringzilla_bare PRIVATE "SZ_AVOID_LIBC=1")
+  target_compile_definitions(stringzilla_bare PRIVATE "SZ_OVERRIDE_LIBC=1")
 
-  # Avoid built-ins on MSVC and other compilers, as that will cause compileration errors
-  target_compile_options(stringzillite PRIVATE
+  # Avoid built-ins on MSVC and other compilers, as that will cause compilation errors
+  target_compile_options(stringzilla_bare PRIVATE
     "$<$<CXX_COMPILER_ID:GNU,Clang>:-fno-builtin;-nostdlib>"
     "$<$<CXX_COMPILER_ID:MSVC>:/Oi-;/GS->")
-  target_link_options(stringzillite PRIVATE "$<$<CXX_COMPILER_ID:GNU,Clang>:-nostdlib>")
-  target_link_options(stringzillite PRIVATE "$<$<CXX_COMPILER_ID:MSVC>:/NODEFAULTLIB>")
+  target_link_options(stringzilla_bare PRIVATE "$<$<CXX_COMPILER_ID:GNU,Clang>:-nostdlib>")
+  target_link_options(stringzilla_bare PRIVATE "$<$<CXX_COMPILER_ID:MSVC>:/NODEFAULTLIB>")
 
 
 endif()
@@ -360,7 +396,7 @@ if(STRINGZILLA_INSTALL)
     RESOURCE
     RUNTIME)
   install(
-    TARGETS stringzillite
+    TARGETS stringzilla_bare
     ARCHIVE
     BUNDLE
     FRAMEWORK

CONTRIBUTING.md

@@ -116,7 +116,7 @@ Replacing the default compiler is not recommended, as it may break the system, b
 
 ```bash
 brew install llvm
-cmake -D CMAKE_BUILD_TYPE=Release -D SIMSIMD_BUILD_TESTS=1 \
+cmake -D CMAKE_BUILD_TYPE=Release -D STRINGZILLA_BUILD_TEST=1 \
     -D CMAKE_C_COMPILER="$(brew --prefix llvm)/bin/clang" \
     -D CMAKE_CXX_COMPILER="$(brew --prefix llvm)/bin/clang++" \
     -B build_release
@@ -131,8 +131,8 @@ Using modern syntax, this is how you build and run the test suite:
 cmake -D STRINGZILLA_BUILD_TEST=1 -D CMAKE_BUILD_TYPE=Debug -B build_debug
 cmake --build build_debug --config Debug          # Which will produce the following targets:
 build_debug/stringzilla_test_cpp20                # Unit test for the entire library compiled for current hardware
-build_debug/stringzilla_test_cpp20_x86_serial     # x86 variant compiled for IvyBridge - last arch. before AVX2
-build_debug/stringzilla_test_cpp20_arm_serial     # Arm variant compiled without Neon
+build_debug/stringzilla_test_cpp20_serial     # x86 variant compiled for IvyBridge - last arch. before AVX2
+build_debug/stringzilla_test_cpp20_serial     # Arm variant compiled without Neon
 ```
 
 To use CppCheck for static analysis make sure to export the compilation commands.