r-lyeh-archived/sole

:doughnut: Sole is a lightweight C++11 library to generate universally unique identificators (UUID), both v1 and v4.

sole's Issues

Add method to check if string is a valid guid

I want to check if a guid string is valid. Currently I've been doing sole::rebuild(guidString) but maybe there is a more optimised way.
An extra method like sole::isValid would come in handy.

Tag release v1.0.2

Can we tag a release? :)

-Wunused-value compiler warning

Hey great library. One minor issue (which I have locally fixed) is that on gcc version:

gcc (GCC) 4.9.2 20150212 (Red Hat 4.9.2-6)

I had to manually disable a -Wunused-value compiler warning. So for those who require warning-free compilations, you would have to edit the sole.cpp file to disable this warning:

#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wunused-value"
$linux({
    struct ifreq ifr;

    int s = socket(PF_INET, SOCK_DGRAM, 0);
    if (s == -1) return ("cannot open socket"), false;

    std::strcpy(ifr.ifr_name, "eth0");
    int rc = ioctl(s, SIOCGIFHWADDR, &ifr);
    close(s);
    if (rc < 0) return ("cannot get MAC address"), false;
    struct sockaddr* sa = reinterpret_cast<struct sockaddr*>(&ifr.ifr_addr);
    _node.resize( sizeof(sa->sa_data) );
    std::memcpy(_node.data(), sa->sa_data, _node.size() );
    return true;
})
#pragma GCC diagnostic pop

Not sure if you care to fix this or not, but just letting you know. Thanks!

time to create a new release?

@r-lyeh-archived can you please tag the new release so that I can start using it in my project without needing to use older 1.0.2 tag or the master branch?

Build fails when building on GCC mac.

Had some errors while using GCC 5 on a Mac.
To fix this we included <sys/types.h> in the sole.hpp file.

Warning: possible misuse of comma operator here

While compiling my code on XCode, I get this warning.

sole.hpp:724:12: warning: possible misuse of comma operator here [-Wcomma]
                u.ab = ab, u.cd = cd;
                         ^
sole.hpp:724:3: note: cast expression to void to silence warning
                u.ab = ab, u.cd = cd;
                ^~~~~~~~~
                static_cast<void>( )
1 warning generated.

uuid4 does not seem to be thread safe.

uuid4() constructor uses a static variable : static std::random_device rd;
I don't think that random_device is guaranteed to be thread safe.
If that's the case, a potential fix could be to add the thread_local qualifier to the variable declaration.

uuid4 seems unsafe base on C++ standards

I have found the following content on std::random_device:

std::random_device is a uniformly-distributed integer random number generator that produces non-deterministic random numbers.
std::random_device may be implemented in terms of an implementation-defined pseudo-random number engine if a non-deterministic source (e.g. a hardware device) is not available to the implementation. In this case each std::random_device object may generate the same number sequence.

I am writing a cross platform game and wanting a good UUID, and the fact that these could come out as the same number on some devices (I am thinking Android in particular) is a bit terrifying. Please let me know if I am wrong and there is a better assurance that this couldn't happen.

socket leak

I used sole in my program to generate uuid , and found a sock leak in this line code;
https://github.com/r-lyeh-archived/sole/blob/master/sole.hpp#L604

It could be fixed by using this line code.
if (std::strcmp("lo", ifap->ifa_name) == 0) { close(s); continue;}

Thanks for your work.

clang warning on exit-time declaration

This code at around line 367:
inline std::string uuid::base62() const {
static const std::string base62 =
"0123456789"
"ABCDEFGHIJKLMNOPQRSTUVWXYZ"
"abcdefghijklmnopqrstuvwxyz";
return rebase( ab, base62 ) + "-" + rebase( cd, base62 );
}

and this code at around line 737:
static const std::string base62 =
"0123456789"
"ABCDEFGHIJKLMNOPQRSTUVWXYZ"
"abcdefghijklmnopqrstuvwxyz";

shows clang warning as below:

warning: declaration requires an exit-time destructor [-Wexit-time-destructors]
static const std::string base62 =
^

The IDE I used is "Xcode Version 7.3.1"

Github url in the comments

May I suggest adding your Github URL in your library header comments? That way it is easier to track where the library comes from :)

error: call to 'clock_gettime' is ambiguous

In sole.hpp you have:

    $lelse( $belse( // if not linux, if not bsd... valid for apple/win32
        inline int clock_gettime( int /*clk_id*/, struct timespec* t ) {
            struct timeval now;
            int rv = gettimeofday(&now, NULL);
            if( rv ) return rv;
            t->tv_sec  = now.tv_sec;
            t->tv_nsec = now.tv_usec * 1000;
            return 0;
        }
    ))

It looks like the intention is that you define clock_gettime on all versions of all non-Linux non-BSD operating systems. That's a bit of a problem on macOS because although OS X 10.11 and earlier did not have clock_gettime, macOS 10.12 and later do.

EternalTerminal includes sole and building EternalTerminal 6.0.13 on macOS 10.13 works. I'm not sure how it works since you have a definition of clock_gettime and the operating system also has one, but somehow it works.

EternalTerminal also uses getline which didn't exist until Mac OS X 10.7, so the build failed on Mac OS X 10.6. In MacPorts we have a library of functions that can be included on older systems to provide compatibility implementations of things like getline and clock_gettime. But when I do that for EternalTerminal, the build fails because the implementation of clock_gettime in the compatibility library conflicts with the implementation in sole.

https://trac.macports.org/ticket/60779

In file included from .../build/cotire/et-lib_CXX_prefix.cxx:4:
In file included from .../EternalTerminal-6.0.13/src/base/Headers.hpp:87:
.../EternalTerminal-6.0.13/src/base/sole.hpp:455:9: error: call to 'clock_gettime' is ambiguous
        clock_gettime(0 /*CLOCK_REALTIME*/, &tp);
        ^~~~~~~~~~~~~
/opt/local/include/LegacySupport/time.h:39:12: note: candidate function
extern int clock_gettime( clockid_t clk_id, struct timespec *ts );
           ^
.../EternalTerminal-6.0.13/src/base/sole.hpp:439:20: note: candidate function
        inline int clock_gettime( int /*clk_id*/, struct timespec* t ) {
                   ^
.../EternalTerminal-6.0.13/src/base/sole.hpp:225:18: note: expanded from macro '$belse'
#define $belse   $yes
                 ^
1 error generated.

Full build log: https://build.macports.org/builders/ports-10.6_x86_64-builder/builds/54980/steps/install-port/logs/stdio

I wonder if there is a way that sole could only define its clock_gettime if the OS (or the MacPorts legacy support library) doesn't already define it.

Compile error in C++17

if use command line option /std:c++latest (c++17) in Visual Studio 2017,
compile error occurred about std::unary_function.

so i just remove the code public std::unary_function < ... > and it works.
like this

namespace std {
    template<>
    class hash< sole::uuid >/* public unary_function< sole::uuid, size_t > */ {
    public:
        // hash functor: hash uuid to size_t value by pseudorandomizing transform
        size_t operator()( const sole::uuid &uuid ) const {
            if( sizeof(size_t) > 4 ) {
                return size_t( uuid.ab ^ uuid.cd );
            } else {
                uint64_t hash64 = uuid.ab ^ uuid.cd;
                return size_t( uint32_t( hash64 >> 32 ) ^ uint32_t( hash64 ) );
            }
        }
    };
}

is it right?

Support forward declaration

I'm having trouble doing forward declaration. Is it currently possible? Otherwise, can it be supported?

I use a json library which has "Json::Value". They have a special import "json-forwards.h" that people can use. You could do something similar.

Tag a release

If possible you could tag a release

error: type of thread-local variable has non-trivial destruction

#include "sole.hpp" produces the following compile error on Linux. Sole version is 1.0.4.

In file included from /home/conan/w/prod/BuildSingleReference/conan-center-index/recipes/sole/all/test_package/test_package.cpp:2:
/home/conan/w/prod/BuildSingleReference/.conan/data/sole/1.0.4/_/_/package/5ab84d6acfe1f23c4fae0ab88f26e3a396351ac9/include/sole.hpp:668:78: error: type of thread-local variable has non-trivial destruction
        static $msvc(__declspec(thread)) $melse(__thread) std::random_device rd;
                                                                             ^
/home/conan/w/prod/BuildSingleReference/.conan/data/sole/1.0.4/_/_/package/5ab84d6acfe1f23c4fae0ab88f26e3a396351ac9/include/sole.hpp:668:78: note: use 'thread_local' to allow this
/home/conan/w/prod/BuildSingleReference/.conan/data/sole/1.0.4/_/_/package/5ab84d6acfe1f23c4fae0ab88f26e3a396351ac9/include/sole.hpp:669:99: error: initializer for thread-local variable must be a constant expression
        static $msvc(__declspec(thread)) $melse(__thread) std::uniform_int_distribution<uint64_t> dist(0, (uint64_t)(~0));
                                                                                                  ^~~~~~~~~~~~~~~~~~~~~~~
/home/conan/w/prod/BuildSingleReference/.conan/data/sole/1.0.4/_/_/package/5ab84d6acfe1f23c4fae0ab88f26e3a396351ac9/include/sole.hpp:669:99: note: use 'thread_local' to allow this
2 errors generated.

Compiler setup:

CC and CXX: clang, clang++ 
Found clang 11.1
clang>=8, using the major as version
Default settings
	os=Linux
	os_build=Linux
	arch=x86_64
	arch_build=x86_64
	compiler=clang
	compiler.version=11
	compiler.libcxx=libstdc++
	build_type=Release

Test file:

#include <iostream>
#include "sole.hpp"

int main() {
    sole::uuid u0 = sole::uuid0(), u1 = sole::uuid1(), u4 = sole::uuid4();

    std::cout << "uuid v0 string : " << u0 << std::endl;
    std::cout << "uuid v0 base62 : " << u0.base62() << std::endl;
    std::cout << "uuid v0 pretty : " << u0.pretty() << std::endl << std::endl;

    std::cout << "uuid v1 string : " << u1 << std::endl;
    std::cout << "uuid v1 base62 : " << u1.base62() << std::endl;
    std::cout << "uuid v1 pretty : " << u1.pretty() << std::endl << std::endl;

    std::cout << "uuid v4 string : " << u4 << std::endl;
    std::cout << "uuid v4 base62 : " << u4.base62() << std::endl;
    std::cout << "uuid v4 pretty : " << u4.pretty() << std::endl << std::endl;

    u1 = sole::rebuild("F81D4FAE-7DEC-11D0-A765-00A0C91E6BF6");
    std::cout << "uuid v1 rebuilt: " << u1 << " -> " << u1.pretty()
              << std::endl;

    u4 = sole::rebuild("GITheR4tLlg-BagIW20DGja");
    std::cout << "uuid v4 rebuilt: " << u4 << " -> " << u4.pretty()
              << std::endl;
}

Related: conan-io/conan-center-index#11275

msys2 compile warinig

compile waring on msys2

/sole.hpp:138:0: error: ignoring #pragma comment  [-Werror=unknown-pragmas]
 #   pragma comment(lib,"iphlpapi.lib")

Better performance for uuid4

I was doing some performance profiling and it seems that sole is the slowest bit in our code.

I notice the code is just doing:

std::random_device rd;
std::uniform_int_distribution<uint64_t> dist(0, (uint64_t)(~0));

The performance was better when I added static:

static std::random_device rd;
static std::uniform_int_distribution<uint64_t> dist(0, (uint64_t)(~0));

Not sure if there is any side effects but seems to work fine for us.

need support for android platform

Call to member function "transform" is ambiguous

Hey, thanks for this library, it's been useful.
I tried pulling the latest version and I'm running into this, here.
Candidate functions: both inline transforms.

Using LLVM 6.1 to compile for iOS.