liburlparser 2.0.0__tar.gz → 2.1.0__tar.gz

{liburlparser-2.0.0 → liburlparser-2.1.0}/.github/workflows/orchestrator.yml

@@ -99,6 +99,9 @@ jobs:
       id-token: write
 
     steps:
+    - name: Checkout repository
+      uses: actions/checkout@v4
+
     - name: Download documentation artifact
       uses: actions/download-artifact@v4
       with:
@@ -111,6 +114,18 @@ jobs:
         name: cmake-coverage-report
         path: ./pages/coverage
 
+    # Benchmarks are NOT run in CI: the report is generated locally
+    # (cmake --build <dir> --target urlparser_benchmarks_report) and committed
+    # as benchmarks/results/report.html. CI only publishes it, as /benchmarks/.
+    - name: Add benchmark report
+      run: |
+        if [ ! -f benchmarks/results/report.html ]; then
+          echo "::error::benchmarks/results/report.html is missing - generate it locally and commit it"
+          exit 1
+        fi
+        mkdir -p pages/benchmarks
+        cp benchmarks/results/report.html pages/benchmarks/index.html
+
     - name: List pages directory
       run: find pages -maxdepth 2
 

{liburlparser-2.0.0 → liburlparser-2.1.0}/.github/workflows/wheels.yml

@@ -53,7 +53,7 @@ jobs:
     - uses: actions/checkout@v4
 
     - name: Build wheels
-      uses: pypa/cibuildwheel@v2.21.3
+      uses: pypa/cibuildwheel@v4.2.1
       env:
         MACOSX_DEPLOYMENT_TARGET: "10.13"
         CIBW_ARCHS_MACOS: universal2

{liburlparser-2.0.0 → liburlparser-2.1.0}/CMakeLists.txt

@@ -40,15 +40,7 @@ endif()
 
 find_package(Python3 REQUIRED COMPONENTS Interpreter)
 
-# --- httpp local patch --------------------------------------------------
-# Only treat this as "liburlparser is the scikit-build-core package being
-# installed" when SKBUILD_PROJECT_NAME actually IS liburlparser. Without
-# this guard, vendoring liburlparser as a subdirectory of another
-# scikit-build-core project (e.g. httpp) makes it wrongly think it's the
-# top-level package and try to build its own nanobind bindings, which
-# fails when nanobind isn't installed (it isn't a dependency of the
-# parent project at all). See src/third_party/README.md for details.
-if(DEFINED SKBUILD AND SKBUILD_PROJECT_NAME STREQUAL PROJECT_NAME)
+if(DEFINED SKBUILD)
     set(PYTHON_PROJECT_NAME "${SKBUILD_PROJECT_NAME}")
 elseif(BUILD_PYTHON)
     set(PYTHON_PROJECT_NAME "${CMAKE_BINARY_DIR}")
@@ -104,6 +96,44 @@ embed_file_as_string(
 )
 
 
+# --- C++ core library (static) ---
+# Built in every mode (including scikit-build), so the Python extension
+# links the very same library that gets shipped to C++ consumers.
+# Install layout: the wheel carries include/, lib/ and cmake/ inside the
+# `liburlparser` package (nanobind/pybind11 pattern, see get_cmake_dir() in
+# the Python package); a plain `cmake --install` uses the usual prefix layout.
+if(DEFINED SKBUILD)
+    set(_urlparser_lib_dest   liburlparser/lib)
+    set(_urlparser_inc_dest   liburlparser/include)
+    set(_urlparser_cmake_dest liburlparser/cmake)
+else()
+    set(_urlparser_lib_dest   lib)
+    set(_urlparser_inc_dest   include)
+    set(_urlparser_cmake_dest lib/cmake/liburlparser)
+endif()
+
+add_library(urlparser STATIC ${SOURCES})
+# The public target is `urlparser::urlparser` - the same name whether the
+# library comes from FetchContent/add_subdirectory (this alias) or from an
+# installed package (find_package(liburlparser), exported below).
+add_library(urlparser::urlparser ALIAS urlparser)
+
+set_target_properties(urlparser PROPERTIES
+    POSITION_INDEPENDENT_CODE ON        # linked into the Python extension
+    CXX_VISIBILITY_PRESET hidden
+    VISIBILITY_INLINES_HIDDEN ON
+    EXPORT_NAME urlparser               # installed as urlparser::urlparser (see install(EXPORT))
+    OUTPUT_NAME urlparser               # liburlparser.a / urlparser.lib
+)
+target_include_directories(urlparser
+    PUBLIC
+        $<BUILD_INTERFACE:${PROJECT_SOURCE_DIR}/include>
+        $<INSTALL_INTERFACE:${_urlparser_inc_dest}>
+    PRIVATE
+        ${PROJECT_SOURCE_DIR}/src
+        ${GENERATED_HEADERS_DIR})
+
+
 if(DEFINED PYTHON_PROJECT_NAME)
     # Try to import all Python components potentially needed by nanobind
     #set(Python3_FIND_STRATEGY LOCATION)
@@ -140,10 +170,10 @@ if(DEFINED PYTHON_PROJECT_NAME)
             ${NB_ABI_ARGS}
 
             # Source code goes here
-            ${PROJECT_SOURCE_DIR}/src/bindings/python/binding_py.cpp ${SOURCES}
+            ${PROJECT_SOURCE_DIR}/src/bindings/python/binding_py.cpp
     )
 
-    target_include_directories(${NB_MODULE} PRIVATE include src ${GENERATED_HEADERS_DIR})
+    target_link_libraries(${NB_MODULE} PRIVATE urlparser)
 
     if(DEFINED SKBUILD)
         install(TARGETS ${NB_MODULE}
@@ -157,37 +187,53 @@ if(DEFINED PYTHON_PROJECT_NAME)
     endif()
 endif()
 
-if(DEFINED SKBUILD AND SKBUILD_PROJECT_NAME STREQUAL PROJECT_NAME)
-    # (see the httpp local-patch comment above) only early-return here when
-    # liburlparser itself is the package scikit-build-core is installing —
-    # not when it's vendored as a subdirectory of another project.
+# --- install: find_package(liburlparser CONFIG) -> urlparser::urlparser ---
+include(CMakePackageConfigHelpers)
+install(TARGETS urlparser EXPORT liburlparserTargets
+    ARCHIVE DESTINATION ${_urlparser_lib_dest}
+    LIBRARY DESTINATION ${_urlparser_lib_dest})
+install(DIRECTORY include/ DESTINATION ${_urlparser_inc_dest})
+install(EXPORT liburlparserTargets
+    FILE liburlparserTargets.cmake
+    NAMESPACE urlparser::
+    DESTINATION ${_urlparser_cmake_dest})
+
+# urlparser_LIB_PATH / urlparser_INCLUDE_PATH for consumers that want plain
+# paths. The exact artifact name (liburlparser.a, urlparser.lib, ...) is only
+# known to the generator, so the file is produced with file(GENERATE) - the
+# generator expression is resolved when the build system is generated, and the
+# result is installed next to the package config. Paths are written relative
+# to the installed file itself, so the package stays relocatable.
+file(RELATIVE_PATH _urlparser_rel_lib "/_/${_urlparser_cmake_dest}" "/_/${_urlparser_lib_dest}")
+file(RELATIVE_PATH _urlparser_rel_inc "/_/${_urlparser_cmake_dest}" "/_/${_urlparser_inc_dest}")
+string(REGEX REPLACE "/+$" "" _urlparser_rel_lib "${_urlparser_rel_lib}")
+string(REGEX REPLACE "/+$" "" _urlparser_rel_inc "${_urlparser_rel_inc}")
+file(GENERATE
+    OUTPUT ${PROJECT_BINARY_DIR}/liburlparserPaths.cmake
+    CONTENT
+"# Generated by liburlparser's CMakeLists.txt - do not edit.
+get_filename_component(urlparser_LIB_PATH
+    \"\${CMAKE_CURRENT_LIST_DIR}/${_urlparser_rel_lib}/$<TARGET_FILE_NAME:urlparser>\" ABSOLUTE)
+get_filename_component(urlparser_INCLUDE_PATH
+    \"\${CMAKE_CURRENT_LIST_DIR}/${_urlparser_rel_inc}\" ABSOLUTE)
+")
+
+write_basic_package_version_file(
+    ${PROJECT_BINARY_DIR}/liburlparserConfigVersion.cmake
+    VERSION ${PROJECT_VERSION}
+    COMPATIBILITY SameMajorVersion)
+install(FILES
+    ${PROJECT_SOURCE_DIR}/cmake/liburlparserConfig.cmake
+    ${PROJECT_BINARY_DIR}/liburlparserPaths.cmake
+    ${PROJECT_BINARY_DIR}/liburlparserConfigVersion.cmake
+    DESTINATION ${_urlparser_cmake_dest})
+
+if(DEFINED SKBUILD)
     RETURN()
 endif()
 
-
-
-add_library(urlparser STATIC ${SOURCES})
-add_library(url::base ALIAS urlparser)
-target_include_directories(urlparser PRIVATE ${GENERATED_HEADERS_DIR})
-
-
-target_include_directories(urlparser
-        PUBLIC
-            ${PROJECT_SOURCE_DIR}/include
-        PRIVATE
-            ${PROJECT_SOURCE_DIR}/src)
-
-
 add_executable(example examples/main.cpp)
-target_link_libraries(example PRIVATE url::base)
-
-install(TARGETS urlparser
-    LIBRARY DESTINATION lib
-    ARCHIVE DESTINATION lib
-    INCLUDES DESTINATION include
-)
-
-
+target_link_libraries(example PRIVATE urlparser::urlparser)
 
 enable_testing()
 find_package(Threads REQUIRED)

{liburlparser-2.0.0 → liburlparser-2.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: liburlparser
-Version: 2.0.0
+Version: 2.1.0
 Summary: Fastest Url parser in the world
 Author-Email: Mohammad Raziei <mohammadraziei1375@gmail.com>
 Classifier: Development Status :: 4 - Beta
@@ -39,7 +39,7 @@ Description-Content-Type: text/markdown
 [![forks - liburlparser](https://img.shields.io/github/forks/mohammadraziei/liburlparser?style=social)](https://github.com/mohammadraziei/liburlparser)
 
 [![PyPi](https://img.shields.io/pypi/v/liburlparser.svg)](https://pypi.org/project/liburlparser/)
-![Python](https://img.shields.io/badge/Python-3.8%20%7C%203.9%20%7C%203.10%20%7C%203.11-blue)
+![Python](https://img.shields.io/badge/Python-3.10%20%7C%203.11%20%7C%203.12-blue)
 ![Cpp](https://img.shields.io/badge/C++-17-blue)
 
 [![GitHub release](https://img.shields.io/github/release/mohammadraziei/liburlparser?include_prereleases=&sort=semver&color=purple)](https://github.com/mohammadraziei/liburlparser/releases/)
@@ -240,49 +240,102 @@ you can see all methods in python we can use in c++ very easily
 
 ## Installation
 
-### C++:
+### C++
+
+Requires a C++17 compiler and CMake >= 3.19. The library is a static library plus one header (`urlparser.h`),
+and its CMake target is **`urlparser::urlparser`** - the same name no matter how you get it:
+
+#### 1. FetchContent
 
-#### build steps:
+```cmake
+include(FetchContent)
+FetchContent_Declare(
+  liburlparser
+  GIT_REPOSITORY https://github.com/mohammadraziei/liburlparser.git
+  GIT_TAG        master
+)
+FetchContent_MakeAvailable(liburlparser)
+
+target_link_libraries(my_app PRIVATE urlparser::urlparser)
+```
+
+#### 2. From the pip package
+
+The Python wheel also ships the compiled C++ library, its header and a CMake
+package config (same idea as `python -m nanobind --cmake_dir`), so a C++
+project can use a plain `pip install liburlparser` without building anything:
+
+```cmake
+find_package(Python3 REQUIRED COMPONENTS Interpreter)
+execute_process(
+    COMMAND ${Python3_EXECUTABLE} -m liburlparser --cmake-dir
+    OUTPUT_VARIABLE liburlparser_DIR
+    OUTPUT_STRIP_TRAILING_WHITESPACE
+    COMMAND_ERROR_IS_FATAL ANY)
+
+find_package(liburlparser CONFIG REQUIRED)
+target_link_libraries(my_app PRIVATE urlparser::urlparser)
+```
+
+For build systems that prefer plain paths over the target, `find_package` also sets:
+
+| Variable                 | Value                                                              |
+| ------------------------ | ------------------------------------------------------------------ |
+| `urlparser_LIB_PATH`     | full path of the compiled library (`liburlparser.a`, `urlparser.lib`, ...) |
+| `urlparser_INCLUDE_PATH` | directory containing `urlparser.h`                                 |
+
+The same locations are available from the command line and from Python:
 
 ```sh
-# 1. Clone the repository with submodules (recursive)
-# --recursive is essential to download third-party libs like googletest
-git clone --recursive https://github.com/mohammadraziei/liburlparser
-cd liburlparser
+python -m liburlparser --cmake-dir     # directory with liburlparserConfig.cmake
+python -m liburlparser --include-dir   # directory with urlparser.h
+python -m liburlparser --lib-dir       # directory with the compiled library
+```
+
+```python
+import liburlparser
+liburlparser.get_cmake_dir(), liburlparser.get_include_dir(), liburlparser.get_lib_dir()
+```
 
-# 2. Configure the project
-# -B build tells CMake to create a 'build' directory and generate files there
-cmake -B build
+A complete example project is in
+[examples/find_liburlparser_via_python](examples/find_liburlparser_via_python).
 
-# 3. Build the project
-# --build abstracts the underlying build tool (Make, Ninja, MSBuild, etc.)
-cmake --build build --config Debug
+#### 3. Build and install from source
 
-# 4. Run examples
-# Note: On Windows, the path might be ./build/Debug/example.exe
+```sh
+git clone https://github.com/mohammadraziei/liburlparser
+cd liburlparser
+
+# configure and build (-B creates the 'build' directory)
+cmake -B build -DCMAKE_BUILD_TYPE=Release
+cmake --build build
+
+# run the example (on Windows with multi-config generators: ./build/Release/example.exe)
 ./build/example
 
-# 5. Make install
-# --install handles the installation process
-sudo cmake --install build
+# install the library, header and CMake package (add --prefix <dir> to choose where)
+cmake --install build
+```
+
+Then, in your project (add `-DCMAKE_PREFIX_PATH=<prefix>` if you used a custom prefix):
+
+```cmake
+find_package(liburlparser CONFIG REQUIRED)
+target_link_libraries(my_app PRIVATE urlparser::urlparser)
 ```
 
 ### Python and Command Line:
 
-Be aware that it required `python>=3.8`
+Be aware that it requires `python>=3.10`
 
 #### Installation
 
 ###### from source
 
-To install from source, you must ensure all submodules are cloned:
-
 ```bash
-# 1. Clone recursively to get third-party dependencies
-git clone --recursive https://github.com/mohammadraziei/liburlparser
+git clone https://github.com/mohammadraziei/liburlparser
 cd liburlparser
 
-# 2. Install using pip
 pip install .
 
 # Optional: If you want to see build logs
@@ -314,6 +367,18 @@ pip install ./liburlparser
 
 ### Performance
 
+Up-to-date, reproducible results for both C++ and Python (compared against other libraries)
+are on the **[Benchmarks page](https://mohammadraziei.github.io/liburlparser/benchmarks/)**.
+The report is generated locally and published with the documentation:
+
+```sh
+cmake -S benchmarks -B build-bench -DCMAKE_BUILD_TYPE=Release
+cmake --build build-bench --target urlparser_benchmarks_report
+# -> benchmarks/results/report.html
+```
+
+The two tables below are the original measurements from March 2022.
+
 #### Extract From Host
 
 Tests were run on a file containing 10 million random domains from various top-level domains (Mar. 13rd 2022)
@@ -354,5 +419,6 @@ Project Link:
 
 - [https://github.com/mohammadraziei/liburlparser](https://github.com/mohammadraziei/liburlparser)
 - [https://pypi.org/project/liburlparser](https://pypi.org/project/liburlparser)
+- [Documentation](https://mohammadraziei.github.io/liburlparser/) and [Benchmarks](https://mohammadraziei.github.io/liburlparser/benchmarks/)
 
 [license-shield]: https://img.shields.io/github/license/othneildrew/Best-README-Template.svg?style=flat-square

{liburlparser-2.0.0 → liburlparser-2.1.0}/README.md

@@ -15,7 +15,7 @@
 [![forks - liburlparser](https://img.shields.io/github/forks/mohammadraziei/liburlparser?style=social)](https://github.com/mohammadraziei/liburlparser)
 
 [![PyPi](https://img.shields.io/pypi/v/liburlparser.svg)](https://pypi.org/project/liburlparser/)
-![Python](https://img.shields.io/badge/Python-3.8%20%7C%203.9%20%7C%203.10%20%7C%203.11-blue)
+![Python](https://img.shields.io/badge/Python-3.10%20%7C%203.11%20%7C%203.12-blue)
 ![Cpp](https://img.shields.io/badge/C++-17-blue)
 
 [![GitHub release](https://img.shields.io/github/release/mohammadraziei/liburlparser?include_prereleases=&sort=semver&color=purple)](https://github.com/mohammadraziei/liburlparser/releases/)
@@ -216,49 +216,102 @@ you can see all methods in python we can use in c++ very easily
 
 ## Installation
 
-### C++:
+### C++
+
+Requires a C++17 compiler and CMake >= 3.19. The library is a static library plus one header (`urlparser.h`),
+and its CMake target is **`urlparser::urlparser`** - the same name no matter how you get it:
+
+#### 1. FetchContent
 
-#### build steps:
+```cmake
+include(FetchContent)
+FetchContent_Declare(
+  liburlparser
+  GIT_REPOSITORY https://github.com/mohammadraziei/liburlparser.git
+  GIT_TAG        master
+)
+FetchContent_MakeAvailable(liburlparser)
+
+target_link_libraries(my_app PRIVATE urlparser::urlparser)
+```
+
+#### 2. From the pip package
+
+The Python wheel also ships the compiled C++ library, its header and a CMake
+package config (same idea as `python -m nanobind --cmake_dir`), so a C++
+project can use a plain `pip install liburlparser` without building anything:
+
+```cmake
+find_package(Python3 REQUIRED COMPONENTS Interpreter)
+execute_process(
+    COMMAND ${Python3_EXECUTABLE} -m liburlparser --cmake-dir
+    OUTPUT_VARIABLE liburlparser_DIR
+    OUTPUT_STRIP_TRAILING_WHITESPACE
+    COMMAND_ERROR_IS_FATAL ANY)
+
+find_package(liburlparser CONFIG REQUIRED)
+target_link_libraries(my_app PRIVATE urlparser::urlparser)
+```
+
+For build systems that prefer plain paths over the target, `find_package` also sets:
+
+| Variable                 | Value                                                              |
+| ------------------------ | ------------------------------------------------------------------ |
+| `urlparser_LIB_PATH`     | full path of the compiled library (`liburlparser.a`, `urlparser.lib`, ...) |
+| `urlparser_INCLUDE_PATH` | directory containing `urlparser.h`                                 |
+
+The same locations are available from the command line and from Python:
 
 ```sh
-# 1. Clone the repository with submodules (recursive)
-# --recursive is essential to download third-party libs like googletest
-git clone --recursive https://github.com/mohammadraziei/liburlparser
-cd liburlparser
+python -m liburlparser --cmake-dir     # directory with liburlparserConfig.cmake
+python -m liburlparser --include-dir   # directory with urlparser.h
+python -m liburlparser --lib-dir       # directory with the compiled library
+```
+
+```python
+import liburlparser
+liburlparser.get_cmake_dir(), liburlparser.get_include_dir(), liburlparser.get_lib_dir()
+```
 
-# 2. Configure the project
-# -B build tells CMake to create a 'build' directory and generate files there
-cmake -B build
+A complete example project is in
+[examples/find_liburlparser_via_python](examples/find_liburlparser_via_python).
 
-# 3. Build the project
-# --build abstracts the underlying build tool (Make, Ninja, MSBuild, etc.)
-cmake --build build --config Debug
+#### 3. Build and install from source
 
-# 4. Run examples
-# Note: On Windows, the path might be ./build/Debug/example.exe
+```sh
+git clone https://github.com/mohammadraziei/liburlparser
+cd liburlparser
+
+# configure and build (-B creates the 'build' directory)
+cmake -B build -DCMAKE_BUILD_TYPE=Release
+cmake --build build
+
+# run the example (on Windows with multi-config generators: ./build/Release/example.exe)
 ./build/example
 
-# 5. Make install
-# --install handles the installation process
-sudo cmake --install build
+# install the library, header and CMake package (add --prefix <dir> to choose where)
+cmake --install build
+```
+
+Then, in your project (add `-DCMAKE_PREFIX_PATH=<prefix>` if you used a custom prefix):
+
+```cmake
+find_package(liburlparser CONFIG REQUIRED)
+target_link_libraries(my_app PRIVATE urlparser::urlparser)
 ```
 
 ### Python and Command Line:
 
-Be aware that it required `python>=3.8`
+Be aware that it requires `python>=3.10`
 
 #### Installation
 
 ###### from source
 
-To install from source, you must ensure all submodules are cloned:
-
 ```bash
-# 1. Clone recursively to get third-party dependencies
-git clone --recursive https://github.com/mohammadraziei/liburlparser
+git clone https://github.com/mohammadraziei/liburlparser
 cd liburlparser
 
-# 2. Install using pip
 pip install .
 
 # Optional: If you want to see build logs
@@ -290,6 +343,18 @@ pip install ./liburlparser
 
 ### Performance
 
+Up-to-date, reproducible results for both C++ and Python (compared against other libraries)
+are on the **[Benchmarks page](https://mohammadraziei.github.io/liburlparser/benchmarks/)**.
+The report is generated locally and published with the documentation:
+
+```sh
+cmake -S benchmarks -B build-bench -DCMAKE_BUILD_TYPE=Release
+cmake --build build-bench --target urlparser_benchmarks_report
+# -> benchmarks/results/report.html
+```
+
+The two tables below are the original measurements from March 2022.
+
 #### Extract From Host
 
 Tests were run on a file containing 10 million random domains from various top-level domains (Mar. 13rd 2022)
@@ -330,5 +395,6 @@ Project Link:
 
 - [https://github.com/mohammadraziei/liburlparser](https://github.com/mohammadraziei/liburlparser)
 - [https://pypi.org/project/liburlparser](https://pypi.org/project/liburlparser)
+- [Documentation](https://mohammadraziei.github.io/liburlparser/) and [Benchmarks](https://mohammadraziei.github.io/liburlparser/benchmarks/)
 
 [license-shield]: https://img.shields.io/github/license/othneildrew/Best-README-Template.svg?style=flat-square

liburlparser-2.1.0/benchmarks/CMakeLists.txt

@@ -0,0 +1,148 @@
+cmake_minimum_required(VERSION 3.19)
+
+# ==============================================================================
+# liburlparser Benchmarks — fully standalone CMake project.
+#
+# This has NO relationship to the repository root — not even for liburlparser
+# itself. Every language folder fetches liburlparser the exact same way it
+# fetches any competitor: from https://github.com/mohammadraziei/liburlparser.git
+# (C++, via FetchContent) or `pip install git+https://github.com/mohammadraziei/
+# liburlparser.git` (Python). liburlparser is never a special case here — it's
+# benchmarked exactly like ada or tldextract: one more entry in the same
+# results table. (Mirrors the same structure as ctoon's own benchmarks/.)
+#
+# Build it on its own:
+#
+#   cd benchmarks
+#   cmake -S . -B build-bench -DCMAKE_BUILD_TYPE=Release
+#   cmake --build build-bench --target urlparser_benchmarks
+#
+# Layout: one folder per language (cpp, python), each with its own
+# CMakeLists.txt that fetches its own dependencies and defines an
+# urlparser_benchmarks_<lang> target. A language is skipped with a warning
+# only when its toolchain isn't found.
+# ==============================================================================
+
+project(UrlParserBenchmarks LANGUAGES CXX)
+
+set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_STANDARD_REQUIRED ON)
+
+if(NOT CMAKE_BUILD_TYPE)
+    set(CMAKE_BUILD_TYPE Release)
+endif()
+if(CMAKE_BUILD_TYPE STREQUAL "Debug")
+    message(WARNING "Benchmarks are being built in Debug mode — numbers will "
+                     "not reflect real performance. Pass -DCMAKE_BUILD_TYPE=Release.")
+endif()
+
+set(URLPARSER_BENCH_SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR})
+set(URLPARSER_BENCH_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR})
+set(URLPARSER_GITHUB_URL "https://github.com/mohammadraziei/liburlparser.git")
+
+# Two different "results" directories, two different lifetimes:
+#  - SCRATCH_DIR: each language's own raw JSON (cpp.json, python.json) -
+#    disposable build output, lives in the build dir, never committed.
+#  - RESULTS_DIR: the one artifact meant to survive - the standalone
+#    report.html that report/ renders from those JSONs - lives in the
+#    source tree so it can be committed and viewed on GitHub directly,
+#    with no build step required to see the last-generated numbers.
+set(URLPARSER_BENCH_RESULTS_SCRATCH_DIR ${URLPARSER_BENCH_BINARY_DIR}/results)
+file(MAKE_DIRECTORY ${URLPARSER_BENCH_RESULTS_SCRATCH_DIR})
+
+set(URLPARSER_BENCH_RESULTS_DIR "${CMAKE_CURRENT_SOURCE_DIR}/results"
+    CACHE PATH "Where the final standalone report.html is written -- the one artifact meant to be committed")
+file(MAKE_DIRECTORY ${URLPARSER_BENCH_RESULTS_DIR})
+
+# ── Toolchain detection ─────────────────────────────────────────────────
+# A language is skipped (with a warning) only when its toolchain itself
+# isn't found. Everything else (venvs, pip installs, FetchContent) happens
+# inside that language's own CMakeLists.txt once we know the toolchain
+# exists. Needed here too, now, since corpus generation below runs Python.
+find_package(Python3 COMPONENTS Interpreter)
+
+# ── Corpus: fetch a real-world domain list, derive a URL list from it ─────
+# 10,000 real, currently-registered domains (Majestic Million-derived),
+# covering a wide range of TLDs/ccTLDs and subdomain shapes — not synthetic
+# data generated just for this benchmark.
+#
+# Shared, single location for the downloaded/derived corpus - NOT per
+# build directory (${URLPARSER_BENCH_BINARY_DIR} changes for every fresh
+# `cmake -B <dir>`, which used to mean a fresh multi-hundred-KB download
+# every time). Every build reuses this one copy; .gitignore'd, never
+# committed. Standalone (non-CMake) tools under benchmarks/ read from here
+# too - this is the single source of truth for the corpus, nothing
+# downloads or derives it a second way anywhere else.
+#
+# The actual download + derivation logic lives in corpus/generate_corpus.py,
+# not here: string/URL construction is a poor fit for CMake's DSL, and the
+# derived urls.txt is round-tripped through ada_url's WHATWG-spec parser
+# there (the same URL library this project's own benchmarks compare
+# against) rather than trusted as hand-built strings.
+set(URLPARSER_BENCH_CORPUS_DIR ${CMAKE_CURRENT_SOURCE_DIR}/corpus)
+set(URLPARSER_BENCH_DOMAINS_FILE ${URLPARSER_BENCH_CORPUS_DIR}/domains.txt)
+set(URLPARSER_BENCH_URLS_FILE ${URLPARSER_BENCH_CORPUS_DIR}/urls.txt)
+
+if(Python3_Interpreter_FOUND AND (NOT EXISTS ${URLPARSER_BENCH_DOMAINS_FILE} OR NOT EXISTS ${URLPARSER_BENCH_URLS_FILE}))
+    # ada-url is a ~2MB prebuilt wheel with zero dependencies of its own -
+    # installed straight into the system/found Python3, no dedicated venv
+    # for it. (The separate, heavier venv in python/CMakeLists.txt -
+    # nanobind, scikit-build-core, tldextract, ... - is for actually
+    # *running* the Python benchmark later; unrelated to this one-off
+    # corpus-generation step and not worth sharing across the two.)
+    execute_process(COMMAND ${Python3_EXECUTABLE} -c "import ada_url" RESULT_VARIABLE _ada_url_present)
+    if(NOT _ada_url_present EQUAL 0)
+        execute_process(COMMAND ${Python3_EXECUTABLE} -m pip install -q --disable-pip-version-check --user ada-url
+                         RESULT_VARIABLE _pip_rc)
+        if(NOT _pip_rc EQUAL 0)
+            # PEP 668 ("externally-managed-environment") refuses a plain
+            # --user install on modern Debian/Ubuntu system Pythons -
+            # --break-system-packages is the documented escape hatch for
+            # exactly this: one small, dependency-free utility package,
+            # not a project dependency going into the system environment.
+            execute_process(
+                COMMAND ${Python3_EXECUTABLE} -m pip install -q --disable-pip-version-check --user
+                        --break-system-packages ada-url
+                RESULT_VARIABLE _pip_rc)
+        endif()
+        if(NOT _pip_rc EQUAL 0)
+            message(FATAL_ERROR "Failed to install ada-url (needed by benchmarks/corpus/generate_corpus.py)")
+        endif()
+    endif()
+
+    execute_process(
+        COMMAND ${Python3_EXECUTABLE} ${URLPARSER_BENCH_CORPUS_DIR}/generate_corpus.py
+        RESULT_VARIABLE _corpus_rc
+    )
+    if(NOT _corpus_rc EQUAL 0)
+        message(FATAL_ERROR "Failed to generate benchmark corpus")
+    endif()
+elseif(NOT Python3_Interpreter_FOUND AND (NOT EXISTS ${URLPARSER_BENCH_DOMAINS_FILE} OR NOT EXISTS ${URLPARSER_BENCH_URLS_FILE}))
+    message(FATAL_ERROR "Python3 is required to generate the benchmark corpus (benchmarks/corpus/generate_corpus.py)")
+endif()
+
+file(STRINGS ${URLPARSER_BENCH_DOMAINS_FILE} URLPARSER_BENCH_DOMAINS)
+list(LENGTH URLPARSER_BENCH_DOMAINS URLPARSER_BENCH_DOMAIN_COUNT)
+message(STATUS "liburlparser benchmarks: domain corpus has ${URLPARSER_BENCH_DOMAIN_COUNT} domains")
+
+add_custom_target(urlparser_benchmarks
+    COMMENT "Running all available liburlparser language benchmarks"
+)
+
+
+# C++ always runs: this project itself requires a CXX compiler to configure
+# at all, so there's nothing to conditionally detect.
+add_subdirectory(cpp)
+
+if(Python3_Interpreter_FOUND)
+    add_subdirectory(python)
+
+    # The report needs at least the Python venv (for jinja2) to render -
+    # C++ is not a hard requirement for report generation itself, but this
+    # project always has a CXX compiler available (see above), so in
+    # practice both results are always present in the report.
+    add_subdirectory(report)
+    add_dependencies(urlparser_benchmarks urlparser_benchmarks_report)
+else()
+    message(WARNING "Python3 not found — Python benchmark and HTML report skipped.")
+endif()