m-peko/bitflags

Single-header header-only C++11 / C++14 / C++17 library for easily managing set of auto-generated type-safe flags.

Single-header header-only C++11 / C++14 / C++17 library for easily managing set of auto-generated type-safe flags.

Quick start

#include <bitflags/bitflags.hpp>

BEGIN_BITFLAGS(Flags)
    FLAG(none)
    FLAG(flag_a)
    FLAG(flag_b)
    FLAG(flag_c)
END_BITFLAGS(Flags)

int main() {
    Flags flags = Flags::flag_a | Flags::flag_b;

    if (flags & Flags::flag_a) {
        std::cout << "flag_a is set" << std::endl;
    } else {
        std::cout << "flag_a is not set" << std::endl;
    }

    flags.toggle(Flags::flag_a);

    // ...

    return 0;
}

Table of Contents

• Motivation
• Getting Started
  • Raw flags vs flags
  • How to Declare Set of Flags?
  • Important Notice For C++11
  • Bits and Names
  • Bitwise Operators
  • Is Specific Flag Set?
  • All or Empty
  • Set and Remove Specific Flag
  • Toggle Flags
  • Clear Flags
• Benchmark
• Building Tests
• Compiler Compatibility
• Contributing
• License
• Support

Motivation

Inspiration for writing this bitflags library I got from the homonymous Rust's crate that you may find here.

Some people may ask: Why not just use enum class and assign binary literals?

Following example presents using enum class with std::uint8_t as underlying type (so that max number of flags is 9):

enum class Flags : std::uint8_t {
    none   = 0b0000,
    flag_a = 0b0001,
    flag_b = 0b0010,
    flag_c = 0b0010
};

Nothing is wrong with above usage, right?

Oh... you might not have noticed that both Flags::flag_b and Flags::flag_c have the same value assigned! That's the exact same mistake that developers make in everyday work... This mistake happened with std::uint8_t as underlying type but imagine how often would be mistakes if the underlying value is std::uint32_t or even std::uint64_t.

You got the point?

Assigning binary literals to flags is manual and error-prone process, especially if there is lots of flags and if developers change them over time.

So, why should I use bitflags library?

bitflags library provides you decently safe way of specifying your flags with 2 core features:

• auto - generated flag values
• auto - detected underlying type

Getting Started

bitflags is a single-header header-only C++11 / C++14 / C++17 library for easily managing set of auto-generated type-safe flags.

Raw flags vs flags?

bitflags library provides you with 2 flag types:

• raw_flag without its string representation (sort of lightweight version)
• flag with its string representation

How to Declare Set of Flags?

In order to declare a set of auto-generated flags, there are few helper macros that hide kind of "ugly" declaration syntax and provide auto-generated value for each flag.

Macros for creating set of raw flags, i.e. flags without string representation:

• BEGIN_RAW_BITFLAGS(NAME)

  • NAME - name of set of raw flags

• RAW_FLAG(NAME)

  • NAME - name of specific raw flag

• END_RAW_BITFLAGS(NAME)

  • NAME - name of set of raw flags

Macros for creating set of ordinary flags, i.e. flags with string representation:

• BEGIN_BITFLAGS(NAME)

  • NAME - name of set of flags

• FLAG(NAME)

  • NAME - name of specific flag

• END_BITFLAGS(NAME)

  • NAME - name of set of flags

Following snippet shows the use case of the above macros:

BEGIN_BITFLAGS(Flags)
    FLAG(none)
    FLAG(flag_a)
    FLAG(flag_b)
    FLAG(flag_c)
END_BITFLAGS()

and is translated into:

template <typename T>
struct FlagsImpl {
    using flag = bf::internal::flag<FlagsImpl, T>
    static constexpr flag none{ 0b0000, "none" };
    static constexpr flag flag_a{ 0b0001, "flag_a" };
    static constexpr flag flag_b{ 0b0010, "flag_b" };
    static constexpr flag flag_c{ 0b0100, "flag_c" };
};

using Flags = bf::bitflags<
    FlagsImpl<
        bf::bitflags<
            FlagsImpl<std::uint8_t>
        >::underlying_type
    >
>;

Usage is basically the same for raw_flags.

Important Notice For C++11

In case you are using C++11, you need to define your flags separately like:

DEFINE_FLAG(Flags, none)
DEFINE_FLAG(Flags, flag_a)
DEFINE_FLAG(Flags, flag_b)
DEFINE_FLAG(Flags, flag_c)

This is because C++11 requires static class members to have an out-of-class definition.

Bits and Names

While bits are part of both raw_flag and flag, names are part of flag type only.

Once the flags are specified, it is possible to get bits representing each flag as well as string representation of each flag:

BEGIN_RAW_BITFLAGS(RawFlags)
    RAW_FLAG(none)
    RAW_FLAG(flag_a)
    RAW_FLAG(flag_b)
END_RAW_BITFLAGS(RawFlags)

std::cout << RawFlags::flag_a.bits << std::endl;

and

BEGIN_BITFLAGS(Flags)
    FLAG(none)
    FLAG(flag_a)
    FLAG(flag_b)
END_BITFLAGS(Flags)

std::cout << Flags::flag_a.bits << " - " << Flags::flag_a.name << std::endl;

Bitwise Operators

The following binary operators are implemented for the generated flags:

• NOT (~) operator
• AND (&) operator
• OR (|) operator
• XOR (^) operator

Is Specific Flag Set?

In case we want to check whether specific flag is set or not, we have 2 options:

1. use AND operator

BEGIN_BITFLAGS(Flags)
    FLAG(none)
    FLAG(flag_a)
    FLAG(flag_b)
END_BITFLAGS(Flags)

Flags flags = Flags::flag_a;

std::cout << static_cast<bool>(flags & Flags::flag_a) << std::endl; // true
std::cout << static_cast<bool>(flags & Flags::flag_b) << std::endl; // false

2. use contains member function

BEGIN_BITFLAGS(Flags)
    FLAG(none)
    FLAG(flag_a)
    FLAG(flag_b)
    FLAG(flag_c)
END_BITFLAGS(Flags)

Flags flags_1 = Flags::flag_a;

std::cout << flags_1.contains(Flags::flag_a) << std::endl; // true
std::cout << flags_1.contains(Flags::flag_b) << std::endl; // false

Flags flags_2 = Flags::flag_a | Flags::flag_b;

std::cout << flags_2.contains(Flags::flag_a, Flags::flag_b) << std::endl; // true
std::cout << flags_2.contains(Flags::flag_a, Flags::flag_c) << std::endl; // false

All or Empty

Following member functions are available for setting all the flags or setting no flag:

1. all / is_all

BEGIN_BITFLAGS(Flags)
    FLAG(none)
    FLAG(flag_a)
    FLAG(flag_b)
END_BITFLAGS(Flags)

Flags flags = Flags::all();

std::cout << flags.is_all() << std::endl;                // true
std::cout << flags.contains(Flags::flag_a) << std::endl; // true
std::cout << flags.contains(Flags::flag_b) << std::endl; // true
std::cout << flags.is_empty() << std::endl;              // false

2. empty / is_empty

BEGIN_BITFLAGS(Flags)
    FLAG(none)
    FLAG(flag_a)
    FLAG(flag_b)
END_BITFLAGS(Flags)

Flags flags = Flags::empty();

std::cout << flags.is_all() << std::endl;                // false
std::cout << flags.contains(Flags::flag_a) << std::endl; // false
std::cout << flags.contains(Flags::flag_b) << std::endl; // false
std::cout << flags.is_empty() << std::endl;              // true

Set and Remove Specific Flag

Not only that one can set and remove specific flag by using bitwise operators, but there are also special member functions set and remove that have the same purpose.

BEGIN_BITFLAGS(Flags)
    FLAG(none)
    FLAG(flag_a)
    FLAG(flag_b)
END_BITFLAGS(Flags)

Flags flags = Flags::empty();

std::cout << flags.contains(Flags::flag_a) << std::endl; // false
std::cout << flags.contains(Flags::flag_b) << std::endl; // false

flags.set(Flags::flag_a);
flags.set(Flags::flag_b);

std::cout << flags.contains(Flags::flag_a) << std::endl; // true
std::cout << flags.contains(Flags::flag_b) << std::endl; // true

flags.remove(Flags::flag_a);

std::cout << flags.contains(Flags::flag_a) << std::endl; // false
std::cout << flags.contains(Flags::flag_b) << std::endl; // true

Toggle Flags

It is possible to toggle specific flag, i.e. if the flag is not already set, it will be set. On the other hand, if the flag is already set, it will be unset.

BEGIN_BITFLAGS(Flags)
    FLAG(none)
    FLAG(flag_a)
    FLAG(flag_b)
END_BITFLAGS(Flags)

Flags flags = Flags::flag_a;

std::cout << flags.contains(Flags::flag_a) << std::endl; // true
flags.toggle(Flags::flag_a);
std::cout << flags.contains(Flags::flag_a) << std::endl; // false

std::cout << flags.contains(Flags::flag_b) << std::endl; // false
flags.toggle(Flags::flag_b);
std::cout << flags.contains(Flags::flag_b) << std::endl; // true

Clear Flags

In order to clear all the flags currently set, one can use clear member function.

BEGIN_BITFLAGS(Flags)
    FLAG(none)
    FLAG(flag_a)
    FLAG(flag_b)
    FLAG(flag_c)
END_BITFLAGS(Flags)

Flags flags = Flags::flag_a | Flags::flag_b;

std::cout << flags.contains(Flags::flag_a) << std::endl; // true
std::cout << flags.contains(Flags::flag_b) << std::endl; // true
std::cout << flags.contains(Flags::flag_c) << std::endl; // false

flags.clear();

std::cout << flags.contains(Flags::flag_a) << std::endl; // false
std::cout << flags.contains(Flags::flag_b) << std::endl; // false
std::cout << flags.contains(Flags::flag_c) << std::endl; // false

Benchmark

As you can see from the following chart, using raw_flags is as fast as using std::bitset. However, using ordinary flags (i.e. flags with string representation) is a bit slower (as it is expected because of additional feature of having string representation).

If you want to run the benchmark yourself, you can use plot.py script like:

$ python3 plot.py --benchmarks-dir <benchmark-json-dir>

Building Tests

$ git clone https://github.com/m-peko/bitflags
$ cd bitflags

$ # create the build directory
$ mkdir build
$ cd build

$ # configure the project
$ cmake -DBITFLAGS_BUILD_TESTS=ON ..

$ # compile
$ make

$ # compile tests
$ make tests

$ # run tests
$ make test

Compiler Compatibility

• Clang/LLVM >= 5
• MSVC++ >= 14.11 / Visual Studio >= 2017
• GCC >= 7.3

There are no 3^rd party dependencies.

Contributing

Feel free to contribute.

If you find that any of the tests fail, please create a ticket in the issue tracker indicating the following information:

• platform
• architecture
• library version
• minimal reproducible example

License

The project is available under the MIT license.

Support

If you like the work bitflags library is doing, please consider supporting it: