scnlib

scnlib is a modern C++ library for scanning values. Think of it as a more C++-y scanf, or the inverse of {fmt} / std::format.

The code lives over at GitHub, at https://github.com/eliaskosunen/scnlib.

#include <scn/scan.h>
#include <print> // for std::print, C++23

int main() {
    auto result = scn::scan<int, double>("[42, 3.14]", "[{}, {}]");
    if (result) {
        auto [first, second] = result->values();
        std::println("first: {}, second: {}", first, second);
        // Prints "first: 42, second: 3.14\n"
    }
}

About this documentation

This documentation is for the version 2.0 of the library. For version 1.1, see https://v1.scnlib.dev/.

An introductory guide to the library can be found at Guide. Instructions for migrating to v2.0 from v1.1 can be found at Migration Guide v1.1 -> v2.0.

The API documentation is organized into modules, that can be found under Modules, behind the link at the top of the page. It can be searched directly using the search function in the navbar, or by pressing the TAB key. The most important modules are probably:

Installation

scnlib uses CMake for building. If you're already using CMake for your project, integration is easy with find_package.

$ mkdir build
$ cd build
$ cmake ..
$ make -j
$ make install

# Find scnlib package
find_package(scn CONFIG REQUIRED)

# Target which you'd like to use scnlib
add_executable(my_program ...)
target_link_libraries(my_program scn::scn)

Another option would be usage through CMake's FetchContent module.

FetchContent_Declare(
        scn
        GIT_REPOSITORY  https://github.com/eliaskosunen/scnlib
        GIT_TAG         v2.0.0
        GIT_SHALLOW     TRUE
)
FetchContent_MakeAvailable(scn)

# ...

target_link_libraries(my_program scn::scn)

Dependencies

scnlib internally depends on fast_float and simdutf.

By default, the CMake machinery automatically fetches, builds, and links these libraries through FetchContent. These libraries are only used in the implementation, and they are not visible to the users of the library.

Alternatively, by setting the CMake options SCN_USE_EXTERNAL_FAST_FLOAT or SCN_USE_EXTERNAL_SIMDUTF to ON, these libraries are searched for using find_package. Use these options, if you already have these libraries installed.

If your standard library doesn't have an available C++20 <ranges> implementation, a single-header version of NanoRange is also bundled with the library, inside the directory include/scn/external.

The tests and benchmarks described below depend on GTest and Google Benchmark, respectively. These libraries are also fetched with FetchContent, if necessary.

Tests and benchmarks

To build and run the tests and benchmarks for scnlib, clone the repository, and build it with CMake.

Building and running tests:

$ mkdir build
$ cd build
$ cmake -DCMAKE_BUILD_TYPE=Debug ..
$ make -j
$ ctest -j

Building and running the runtime performance benchmarks:

# create build directory like above
$ cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INTERPROCEDURAL_OPTIMIZATION=ON -DSCN_NATIVE_ARCH=ON ..
$ make -j

# disable CPU frequency scaling
$ sudo cpupower frequency-set --governor performance

# run the benchmark of your choosing
$ ./benchmark/runtime/integer/scn_int_bench

# re-enable CPU frequency scaling
$ sudo cpupower frequency-set --governor powersave

Without CMake

As mentioned above, the implementation of scnlib depends on fast_float and simdutf. If you're not using CMake, you'll need to download and build these libraries yourself.

Headers for scnlib can be found from the include/ directory, and source files from the src/ directory.

Building and linking the library:

$ mkdir build
$ cd build
$ c++ -c -I../include/ ../src/*.cpp -Ipath-to-fast-float -Ipath-to-simdutf
$ ar rcs libscn.a *.o

libscn.a can then be linked, as usual, with your project. Don't forget to also include fast_float and simdutf.

# in your project
$ c++ ... -Lpath-to-scn/build -lscn -Lpath-to-fast-float -lfast_float -Lpath-to-simdutf -lsimdutf

Note, that scnlib requires at least C++17, so –std=c++17 (or equivalent, or newer) may need to be included in the build flags.

License

The library is open source, licensed under the Apache License, version 2.0.
Copyright (c) 2017 Elias Kosunen
For further details, see the LICENSE file in the repository.