Skip to content

DuffsDevice/tiny-utf8

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

TINY UTF8 Art 4.4

Build Status  Licence  Donation

DESCRIPTION

Tiny-utf8 is a library for extremely easy integration of Unicode into an arbitrary C++11 project. The library consists solely of the class utf8_string, which acts as a drop-in replacement for std::string. Its implementation is successfully in the middle between small memory footprint and fast access. All functionality of std::string is therefore replaced by the corresponding codepoint-based UTF-32 version - translating every access to UTF-8 under the hood.

CHANGES BETWEEN Version 4.4 and 4.3

  • tiny-utf8 used to only work with byte-index-based iterator types. The set of iterator types has now been completed with codepoint-based versions and
  • the default has been changed. That means (c)(r)begin/end now return codepoint-based iterators, while raw_(c)(r)begin/end now return byte-based iterators.
  • The upside with byte-based iterators is: they are usually quicker than code-point-based iterators. The downside is: They get invalidated very quickly. Example: str.erase( std::remove( str.begin() , str.end() , U'W' ) , str.end() ) will work, but str.erase( std::remove(str.raw_begin(),str.raw_end(), U'W' ) ,str.raw_end()) will not (at least not always). The reason is: after the call to std::remove, the size of the string data might have changed and the second call to str.raw_end() might have yielded a now-invalidated iterator.

FEATURES

  • Drop-in replacement for std::string
  • Lightweight and self-contained (~5K SLOC)
  • Very fast, i.e. highly optimized decoder, encoder and traversal routines
  • Advanced Memory Layout, i.e. Random Access is
    • O(1) for ASCII-only strings (!) and
    • O(#Codepoints ∉ ASCII) for the average case.
    • O(n) for strings with a high amount of non-ASCII code points (>25%)
  • Small String Optimization (SSO) for strings up to an UTF8-encoded length of sizeof(utf8_string)! That is, including the trailing \0
  • Growth in Constant Time (Amortized)
  • On-the-fly Conversion between UTF32 and UTF8
  • size() returns the size of the data in bytes, length() returns the number of codepoints contained.
  • Codepoint Range of 0x0 - 0xFFFFFFFF, i.e. 1-7 Code Units/Bytes per Codepoint (Note: This is more than specified by UTF8, but until now otherwise considered out of scope)
  • Complete support for embedded zeros (Note: all methods taking const char*/const char32_t* also have an overload for const char (&)[N]/const char32_t (&)[N], allowing correct interpretation of string literals with embedded zeros)
  • Single Header File
  • Straightforward C++11 Design
  • Possibility to prepend the UTF8 BOM (Byte Order Mark) to any string when converting it to an std::string
  • Supports raw (Byte-based) access for occasions where Speed is needed
  • Supports shrink_to_fit()
  • Malformed UTF8 sequences will lead to defined behaviour

THE PURPOSE OF TINY-UTF8

Back when I decided to write a UTF8 solution for C++, I knew I wanted a drop-in replacement for std::string. At the time mostly because I found it neat to have one and felt C++ always lacked accessible support for UTF8. Since then, several years have passed and the situation has not improved much. That said, things currently look like they are about to improve - but that doesn't say much, eh?

The opinion shared by many "experienced Unicode programmers" (e.g. published on UTF-8 Everywhere) is that "non-experienced" programmers both under and overestimate the need for Unicode- and encoding-specific treatment: This need is...

  1. overestimated, because many times we really should care less about codepoint/grapheme borders within string data;
  2. underestimated, because if we really want to "support" unicode, we need to think about normalizations, visual character comparisons, reserved codepoint values, illegal code unit sequences and so on and so forth.

Unicode is not rocket science but nonetheless hard to get right. Tiny-utf8 does not intend to be an enterprise solution like ICU for C++. The goal of tiny-utf8 is to

  • bridge as many gaps to "supporting Unicode" as possible by 'just' replacing std::string with a custom class which means to
  • provide you with a Codepoint Abstraction Layer that takes care of the Run-Length Encoding, without you noticing.

Tiny-utf8 aims to be the simple-and-dependable groundwork which you build Unicode infrastructure upon. And, if 1) C++2xyz should happen to make your Unicode life easier than tiny-utf8 or 2) you decide to go enterprise, you have not wasted much time replacing std::string with tiny_utf8::string either. That's what makes tiny-utf8 so agreeable.

WHAT TINY-UTF8 IS NOT AIMED AT

  • Conversion between ISO encodings and UTF8
  • Interfacing with UTF16
  • Visible character comparison ('ch' vs. 'c'+'h')
  • Codepoint Normalization
  • Correction of invalid Code Unit sequences
  • Detection of Grapheme Clusters

Note: ANSI suppport was dropped in Version 2.0 in favor of execution speed.

EXAMPLE

#include <iostream>
#include <algorithm>
#include <tinyutf8/tinyutf8.h>
using namespace std;

int main()
{
    tiny_utf8::string str = u8"!🌍 olleH";
    for_each( str.rbegin() , str.rend() , []( char32_t codepoint ){
      cout << codepoint;
    } );
    return 0;
}

EXCEPTION BEHAVIOR

  • Tiny-utf8 should automatically detect, whether your build system allows the use of exceptions or not. This is done by checking for the feature test macro __cpp_exceptions.
  • If you would like tiny-utf8 to be noexcept anyway, #define the macro TINY_UTF8_NOEXCEPT.
  • If you would like tiny-utf8 to use a different exception strategy, #define the macro TINY_UTF8_THROW( location , failing_predicate ). For using assertions, you would write #define TINY_UTF8_THROW( _ , pred ) assert( pred ).
  • Hint: If exceptions are disabled, TINY_UTF8_THROW( ... ) is automatically defined as void(). This works well, because all uses of TINY_UTF8_THROW are immediately followed by a ; as well as a proper return statement with a fallback value. That also means, TINY_UTF8_THROW can safely be a NO-OP.

BACKWARDS-COMPATIBILITY

CHANGES BETWEEN Version 4.3 and 4.2

  • Class tiny_utf8::basic_utf8_string has been renamed to basic_string, which better resembles its drop-in-capabilities for std::string.

CHANGES BETWEEN Version 4.1 and 4.0

  • tinyutf8.h has been moved into the folder include/tinyutf8/ in order to mimic the structuring of many other C++-based open source projects.

CHANGES BETWEEN Version 4.0 and 3.2.4

  • Class utf8_string is now defined inside namespace tiny_utf8. If you want the old declaration in the global namespace, #define TINY_UTF8_GLOBAL_NAMESPACE
  • Support for C++20: Use class tiny_utf8::u8string, which uses char8_t as underlying data type (instead of char)

CHANGES BETWEEN Version 4.0 and Version 3.2

  • If you would like to stay compatible with 3.2.* and have utf8_string defined in the global namespace, #define the macro TINY_UTF8_GLOBAL_NAMESPACE.

BUGS

If you encounter any bugs, please file a bug report through the "Issues" tab. I'll try to answer it soon!

THANK YOU

  • @iainchesworth
  • @vadim-berman
  • @MattHarrington
  • @evanmoran
  • @bakerstu
  • @revel8n
  • @githubuser0xFFFF
  • @marekfoltyn
  • @Megaxela
  • @vfiksdal
  • @maddouri
  • @Abdullah-AlAttar
  • @s9w

for taking your time to improve tiny-utf8.

Cheers, Jakob