I want to understand the difference between char and wchar_t ? I understand that wchar_t uses more bytes but can I get a clear cut example to differentiate when I would use char vs wchar_t
Short anwser:
You should never use wchar_t in modern C++, except when interacting with OS-specific APIs (basically use wchar_t only to call Windows API functions).
Long answer:
Design of standard C++ library implies there is only one way to handle Unicode - by storing UTF-8 encoded strings in char arrays, as almost all functions exist only in char variants (think of std::exception::what).
In a C++ program you have two locales:
Standard C library locale set by std::setlocale
Standard C++ library locale set by std::locale::global
Unfortunately, none of them defines behavior of standard functions that open files (like std::fopen, std::fstream::open etc). Behavior differs between OSes:
Linux is encoding agnostic, so those function simply pass char string to underlying system call
On Windows char string is converted to wide string using user specific locale before system call is made
Everything usually works fine on Linux as everyone uses UTF-8 based locales so all user input and arguments passed to main functions will be UTF-8 encoded. But you might still need to switch current locales to UTF-8 variants explicitly as by default C++ program starts using default "C" locale. At this point, if you only care about Linux and don't need to support Windows, you can use char arrays and std::string assuming it is UTF-8 sequences and everything "just works".
Problems appear when you want to support Windows, as there you always have additional 3rd locale: the one set for the current user which can be configured somewhere in "Control Panel". The main issue is that this locale is never a unicode locale, so it is impossible to use functions like std::fopen(const char *) and std::fstream::open(const char *) to open a file using Unicode path. On Windows you will have to use custom wrappers that use non-standard Windows specific functions like _wfopen, std::fstream::open(const wchar_t *) on Windows. You can check Boost.Nowide (not yet included in Boost) to see how this can be done: http://cppcms.com/files/nowide/html/
With C++17 you can use std::filesystem::path to store file path in a portable way, but it is still broken on Windows:
Implicit constructor std::filesystem::path::path(const char *) uses user-specific locale on MSVC and there is no way to make it use UTF-8. Function std::filesystem::u8string should be used to construct path from UTF-8 string, but it is too easy to forget about this and use implicit constructor instead.
std::error_category::message(int) for both error categories returns error description using user-specific encoding.
So what we have on Windows is:
Standard library functions that open files are broken and should never be used.
Arguments passed to main(int, char**) are broken and should never be used.
WinAPI functions ending with *A and macros are broken and should never be used.
std::filesystem::path is partially broken and should never be used directly.
Error categories returned by std::generic_category and std::system_category are broken and should never be used.
If you need long term solution for a non-trivial project, I would recommend:
Using Boost.Nowide or implementing similar functionality directly - this fixes broken standard library.
Re-implementing standard error categories returned by std::generic_category and std::system_category so that they would always return UTF-8 encoded strings.
Wrapping std::filesystem::path so that new class would always use UTF-8 when converting path to string and string to path.
Wrapping all required functions from std::filesystem so that they would use your path wrapper and your error categories.
Unfortunately, this won't fix issues with other libraries that work with files, but many are broken anyway (do not support unicode).
You can check this link for further explanation: http://utf8everywhere.org/
Fundamentally, use wchar_t when the encoding has more symbols than a char can contain.
Background
The char type has enough capacity to hold any character (encoding) in the ASCII character set.
The issue is that many languages require more encodings than the ASCII accounts for. So, instead of 127 possible encodings, more are needed. Some languages have more than 256 possible encodings. A char type does not guarantee a range greater than 256. Thus a new data type is required.
The wchar_t, a.k.a. wide characters, provides more room for encodings.
Summary
Use char data type when the range of encodings is 256 or less, such as ASCII. Use wchar_t when you need the capacity for more than 256.
Prefer Unicode to handle large character sets (such as emojis).
Never use wchar_t.
When possible, use (some kind of array of) char, such as std::string, and ensure that it is encoded in UTF-8.
When you must interface with APIs that don't speak UTF-8, use char16_t or char32_t. Never use them otherwise; they provide only illusory advantages and encourage faulty code.
Note that there are plenty of cases where more than one char32_t is required to represent a single user-visible character. OTOH, using UTF-8 with char forces you to handle variable width very early.
Related
I want to understand the difference between char and wchar_t ? I understand that wchar_t uses more bytes but can I get a clear cut example to differentiate when I would use char vs wchar_t
Short anwser:
You should never use wchar_t in modern C++, except when interacting with OS-specific APIs (basically use wchar_t only to call Windows API functions).
Long answer:
Design of standard C++ library implies there is only one way to handle Unicode - by storing UTF-8 encoded strings in char arrays, as almost all functions exist only in char variants (think of std::exception::what).
In a C++ program you have two locales:
Standard C library locale set by std::setlocale
Standard C++ library locale set by std::locale::global
Unfortunately, none of them defines behavior of standard functions that open files (like std::fopen, std::fstream::open etc). Behavior differs between OSes:
Linux is encoding agnostic, so those function simply pass char string to underlying system call
On Windows char string is converted to wide string using user specific locale before system call is made
Everything usually works fine on Linux as everyone uses UTF-8 based locales so all user input and arguments passed to main functions will be UTF-8 encoded. But you might still need to switch current locales to UTF-8 variants explicitly as by default C++ program starts using default "C" locale. At this point, if you only care about Linux and don't need to support Windows, you can use char arrays and std::string assuming it is UTF-8 sequences and everything "just works".
Problems appear when you want to support Windows, as there you always have additional 3rd locale: the one set for the current user which can be configured somewhere in "Control Panel". The main issue is that this locale is never a unicode locale, so it is impossible to use functions like std::fopen(const char *) and std::fstream::open(const char *) to open a file using Unicode path. On Windows you will have to use custom wrappers that use non-standard Windows specific functions like _wfopen, std::fstream::open(const wchar_t *) on Windows. You can check Boost.Nowide (not yet included in Boost) to see how this can be done: http://cppcms.com/files/nowide/html/
With C++17 you can use std::filesystem::path to store file path in a portable way, but it is still broken on Windows:
Implicit constructor std::filesystem::path::path(const char *) uses user-specific locale on MSVC and there is no way to make it use UTF-8. Function std::filesystem::u8string should be used to construct path from UTF-8 string, but it is too easy to forget about this and use implicit constructor instead.
std::error_category::message(int) for both error categories returns error description using user-specific encoding.
So what we have on Windows is:
Standard library functions that open files are broken and should never be used.
Arguments passed to main(int, char**) are broken and should never be used.
WinAPI functions ending with *A and macros are broken and should never be used.
std::filesystem::path is partially broken and should never be used directly.
Error categories returned by std::generic_category and std::system_category are broken and should never be used.
If you need long term solution for a non-trivial project, I would recommend:
Using Boost.Nowide or implementing similar functionality directly - this fixes broken standard library.
Re-implementing standard error categories returned by std::generic_category and std::system_category so that they would always return UTF-8 encoded strings.
Wrapping std::filesystem::path so that new class would always use UTF-8 when converting path to string and string to path.
Wrapping all required functions from std::filesystem so that they would use your path wrapper and your error categories.
Unfortunately, this won't fix issues with other libraries that work with files, but many are broken anyway (do not support unicode).
You can check this link for further explanation: http://utf8everywhere.org/
Fundamentally, use wchar_t when the encoding has more symbols than a char can contain.
Background
The char type has enough capacity to hold any character (encoding) in the ASCII character set.
The issue is that many languages require more encodings than the ASCII accounts for. So, instead of 127 possible encodings, more are needed. Some languages have more than 256 possible encodings. A char type does not guarantee a range greater than 256. Thus a new data type is required.
The wchar_t, a.k.a. wide characters, provides more room for encodings.
Summary
Use char data type when the range of encodings is 256 or less, such as ASCII. Use wchar_t when you need the capacity for more than 256.
Prefer Unicode to handle large character sets (such as emojis).
Never use wchar_t.
When possible, use (some kind of array of) char, such as std::string, and ensure that it is encoded in UTF-8.
When you must interface with APIs that don't speak UTF-8, use char16_t or char32_t. Never use them otherwise; they provide only illusory advantages and encourage faulty code.
Note that there are plenty of cases where more than one char32_t is required to represent a single user-visible character. OTOH, using UTF-8 with char forces you to handle variable width very early.
So, i've been trying to do a bit of research of strings and wstrings as i need to understand how they work for a program i'm creating so I also looked into ASCII and unicode, and UTF-8 and UTF-16.
I believe i have an okay understanding of the concept of how these work, but what i'm still having trouble with is how they are actually stored in 'char's, 'string's, 'wchar_t's and 'wstring's.
So my questions are as follows:
Which character set and encoding is used for char and wchar_t? and are these types limited to using only these character sets / encoding?
If they are not limited to these character sets / encoding, how is it decided what character set / encoding is used for a particular char or wchar_t? is it automatically decided at compile for example or do we have to explicitly tell it what to use?
From my understanding UTF-8 uses 1 byte when using the first 128 code points in the set but can use more than 1 byte when using code point 128 and above. If so how is this stored? for example is it simply stored identically to ASCII if it only uses 1 byte? and how does the type (char or wchar_t or whatever) know how many bytes it is using?
Finally, if my understanding is correct I get why UTF-8 and UTF-16 are not compatible, eg. a string can't be used where a wstring is needed. But in a program that requires a wstring would it be better practice to write a conversion function from a string to a wstring and the use this when a wstring is required to make my code exclusively string-based or just use wstring where needed instead?
Thanks, and let me know if any of my questions are incorrectly worded or use the wrong terminology as i'm trying to get to grips with this as best as I can.
i'm working in C++ btw
They use whatever characterset and encoding you want. The types do not imply a specific characterset or encoding. They do not even imply characters - you could happily do math problems with them. Don't do that though, it's weird.
How do you output text? If it is to a console, the console decides which character is associated with each value. If it is some graphical toolkit, the toolkit decides. Consoles and toolkits tend to conform to standards, so there is a good chance they will be using unicode, nowadays. On older systems anything might happen.
UTF8 has the same values as ASCII for the range 0-127. Above that it gets a bit more complicated; this is explained here quite well: https://en.wikipedia.org/wiki/UTF-8#Description
wstring is a string made up of wchar_t, but sadly wchar_t is implemented differently on different platforms. For example, on Visual Studio it is 16 bits (and could be used to store UTF16), but on GCC it is 32 bits (and could thus be used to store unicode codepoints directly). You need to be aware of this if you want your code to be portable. Personally I chose to only store strings in UTF8, and convert only when needed.
Which character set and encoding is used for char and wchar_t? and are these types limited to using only these character sets / encoding?
This is not defined by the language standard. Each compiler will have to agree with the operating system on what character codes to use. We don't even know how many bits are used for char and wchar_t.
On some systems char is UTF-8, on others it is ASCII, or something else. On IBM mainframes it can be EBCDIC, a character encoding already in use before ASCII was defined.
If they are not limited to these character sets / encoding, how is it decided what character set / encoding is used for a particular char or wchar_t? is it automatically decided at compile for example or do we have to explicitly tell it what to use?
The compiler knows what is appropriate for each system.
From my understanding UTF-8 uses 1 byte when using the first 128 code points in the set but can use more than 1 byte when using code point 128 and above. If so how is this stored? for example is it simply stored identically to ASCII if it only uses 1 byte? and how does the type (char or wchar_t or whatever) know how many bytes it is using?
The first part of UTF-8 is identical to the corresponding ASCII codes, and stored as a single byte. Higher codes will use two or more bytes.
The char type itself just store bytes and doesn't know how many bytes we need to form a character. That's for someone else to decide.
The same thing for wchar_t, which is 16 bits on Windows but 32 bits on other systems, like Linux.
Finally, if my understanding is correct I get why UTF-8 and UTF-16 are not compatible, eg. a string can't be used where a wstring is needed. But in a program that requires a wstring would it be better practice to write a conversion function from a string to a wstring and the use this when a wstring is required to make my code exclusively string-based or just use wstring where needed instead?
You will likely have to convert. Unfortunately the conversion needed will be different for different systems, as character sizes and encodings vary.
In later C++ standards you have new types char16_t and char32_t, with the string types u16string and u32string. Those have known sizes and encodings.
Everything about used encoding is implementation defined. Check your compiler documentation. It depends on default locale, encoding of source file and OS console settings.
Types like string, wstring, operations on them and C facilities, like strcmp/wstrcmp expect fixed-width encodings. So the would not work properly with variable width ones like UTF8 or UTF16 (but will work with, e.g., UCS-2). If you want to store variable-width encoded strings, you need to be careful and not use fixed-width operations on it. C-string do have some functions for manipulation of such strings in standard library .You can use classes from codecvt header to convert between different encodings for C++ strings.
I would avoid wstring and use C++11 exact width character string: std::u16string or std::u32string
As an example here is some info on how windows uses these types/encodings.
char stores ASCII values (with code pages for non-ASCII values)
wchar_t stores UTF-16, note this means that some unicode characters will use 2 wchar_t's
If you call a system function, e.g. puts then the header file will actually pick either puts or _putws depending on how you've set things up (i.e. if you are using unicode).
So on windows there is no direct support for UTF-8, which means that if you use char to store UTF-8 encoded strings you have to covert them to UTF-16 and call the corresponding UTF-16 system functions.
I have a question:
Some libraries use WCHAR as the text parameter and others use CHAR (as UTF-8): I need to know when to use WCHAR or CHAR when I write my own library.
Use char and treat it as UTF-8. There are a great many reasons for this; this website summarises it much better than I can:
http://utf8everywhere.org/
It recommends converting from wchar_t to char (UTF-16 to UTF-8) as soon as you receive it from any library, and converting back when you need to pass strings to it. So to answer your question, always use char except at the point that an API requires you to pass or receive wchar_t.
WCHAR (or wchar_t on Visual C++ compiler) is used for Unicode UTF-16 strings.
This is the "native" string encoding used by Win32 APIs.
CHAR (or char) can be used for several other string formats: ANSI, MBCS, UTF-8.
Since UTF-16 is the native encoding of Win32 APIs, you may want to use WCHAR (and better a proper string class based on it, like std::wstring) at the Win32 API boundary, inside your app.
And you can use UTF-8 (so, CHAR/char and std::string) to exchange your Unicode text outside your application boundary. For example: UTF-8 is widely used on the Internet, and when you exchange UTF-8 text between different platforms you don't have the problem of endianness (instead with UTF-16 you have to consider both the UTF-16BE big-endian and the UTF-16LE little-endian cases).
You can convert between UTF-16 and UTF-8 using the WideCharToMultiByte() and MultiByteToWideChar() Win32 APIs. These are pure-C APIs, and these can be conveniently wrapped in C++ code, using string classes instead of raw character pointers, and exceptions instead of raw error codes. You can find an example of that here.
The right question is not which type to use, but what should be your contract with your library users. Both char and wchar_t can mean more than one thing.
The right answer to me, is use char and consider everything utf-8 encoded, as utf8everywhere.org suggests. This will also make it easier to write cross-platform libraries.
Make sure you make correct use of strings though. Some APIs like fopen(), would accept a char* string and treat it differently (not as UTF-8) when compiled on Windows. If Unicode is important to you (and it probably is, when you are dealing with strings), be sure to handle your strings correctly. A good example can be seen in boost::locale. I also recommend using boost::nowide on Windows to get strings handled correctly inside your library.
In Windows we stick to WCHARS. std::wstring. Mainly because if you don't you end up having to convert because calling Windows functions.
I have a feeling that trying to use utf8 internally simply because of http://utf8everywhere.org/ is gonna bite us in the bum later on down the line.
It is best recommended that, when developing a Windows application, resort to TCHARs. The good thing about TCHARs is that they can be either regular chars or wchars, depending whether the unicode setting is set or not. Once you resort to TCHARs, you make sure that all string manipulations that you use also start with the _t prefix (e.g. _tcslen for length of string). That way you will know that your code will work both in Unicode and ASCII environments.
I'm currently taking care of an application that uses std::string and char for string operations - which is fine on linux, since Linux is agnostic to Unicode (or so it seems; I don't really know, so please correct me if I'm telling stories here). This current style naturally leads to this kind of function/class declarations:
std::string doSomethingFunkyWith(const std::string& thisdata)
{
/* .... */
}
However, if thisdata contains unicode characters, it will be displayed wrongly on windows, since std::string can't hold unicode characters on Windows.
So I thought up this concept:
namespace MyApplication {
#ifdef UNICODE
typedef std::wstring string_type;
typedef wchar_t char_type;
#else
typedef std::string string_type;
typedef char char_type;
#endif
/* ... */
string_type doSomethingFunkyWith(const string_type& thisdata)
{
/* ... */
}
}
Is this a good concept to go with to support Unicode on windows?
My current toolchain consists of gcc/clang on Linux, and wine+MinGW for Windows support (crosstesting also happens via wine), if that matters.
Multiplatform issues comes from the fact that there are many encodings, and a wrong encoding pick will lead to encóding Ãssues. Once you tackle that problem, you should be able to use std::wstring on all your program.
The usual workflow is:
raw_input_data = read_raw_data()
input_encoding = "???" // What is your file or terminal encoding?
unicode_data = convert_to_unicode(raw_input_data, input_encoding)
// Do something with the unicode_data, store in some var, etc.
output_encoding = "???" // Is your terminal output encoding the same as your input?
raw_output_data = convert_from_unicode(unicode_data, output_encoding)
print_raw_data(raw_data)
Most Unicode issues comes from wrongly detecting the values of input_encoding and output_encoding. On a modern Linux distribution this is usually UTF-8. On Windows YMMV.
Standard C++ don't know about encodings, you should use some library like ICU to do the conversion.
How you store a string within your application is entirely up to you -- after all, nobody would know as long as the strings stay within your application. The problem starts when you try to read or write strings from the outside world (console, files, sockets etc.) and this is where the OS matters.
Linux isn't exactly "agnostic" to Unicode -- it does recognize Unicode but the standard library functions assume UTF-8 encoding, so Unicode strings fit into standard char arrays. Windows, on the other hand, uses UTF-16 encoding, so you need a wchar_t array to represent 16-bit characters.
The typedefs you proposed should work fine, but keep in mind that this alone doesn't make your code portable. As an example, if you want to store text in files in a portable manner, you should choose one encoding and stick to it across all platforms -- this could require converting between encodings on certain platforms.
Linux does support Unicode, it simply uses UTF-8. Probably a better way to make your system portable would be to make use of International Components for Unicode and treat all std::string objects as containing UTF-8 characters, and convert them to UTF-16 as needed when invoking Windows functions. It almost always makes sense to use UTF-8 over UTF-16, as UTF-8 uses less space for some of the most commonly used characters (e.g. English*) and more space for less frequent characters, whereas UTF-16 wastes space equally for all characters, no matter how frequently they are used.
While you can use your typedefs, this will mean that you have to write two copies of every single function that has to deal with strings. I think it would be more efficient to simply do all internal computations in UTF-8 and simply translate that to/from UTF-16 if necessary when inputting/outputting as needed.
*For HTML, XML, and JSON that use English as part of the encoding (e.g. "<html>, <body>, etc.) regardless of the language of the values, this can still be a win for foreign languages.
The problem for Linux and using Unicode is that all the IO and most system functions use UTF-8 and the wide character type is 32 bit. Then there is interfacing to Java and other programs which requires UTF-16.
As a suggestion for Unicode support, see the OpenRTL library at http://code.google.com/p/openrtl which supports all UTF-8, UTF-16 and UTF-32 on windows, Linux, Osx and Ios. The Unicode support is not just the character types, but also Unicode collation, normalization, case folding, title casing and about 64 different Unicode character properties per full unsigned 32 bit character.
The OpenRTL code is ready now to support char8_t, char16_t and char32_t for the new C++ standards as well, allthough the same character types are supported using macros for existing C and C++ compilers. I think for Unicode and strings processing that it might be what you want for your library.
The point is that if you use OpenRTL, you can build the system using the OpenRTL "char_t" type. This supports the notion that your entire library can be built in either UTF8, UTF16 or UTF32 mode, even on Linux, because OpenRTL is already handling all the interfacing to a lot of system functions like files and io stuff. It has its own print_f functions for example.
By default the char_t is mapping to the wide character type. So on windows it is 32 bit and on Linux it is 32 bit. But you can make it also make it 8 bit everywhere for example. Also it has the support to do fast UTF decoding inside loops using macros.
So instead of ifdeffing between wchar_t and char, you can build everything using char_t and OpenRTL takes care of the rest.
At my company we have a cross platform(Linux & Windows) library that contains our own extension of the STL std::string, this class provides all sort of functionality on top of the string; split, format, to/from base64, etc. Recently we were given the requirement of making this string unicode "friendly" basically it needs to support characters from Chinese, Japanese, Arabic, etc. After initial research this seems fine on the Linux side since every thing is inherently UTF-8, however I am having trouble with the Windows side; is there a trick to getting the STL std::string to work as UTF-8 on windows? Is it even possible? Is there a better way? Ideally we would keep ourselves based on the std::string since that is what the string class is based on in Linux.
Thank you,
There are several misconceptions in your question.
Neither C++ nor the STL deal with encodings.
std::string is essentially a string of bytes, not characters. So you should have no problem stuffing UTF-8 encoded Unicode into it. However, keep in mind that all string functions also work on bytes, so myString.length() will give you the number of bytes, not the number of characters.
Linux is not inherently UTF-8. Most distributions nowadays default to UTF-8, but it should not be relied upon.
Yes - by being more aware of locales and encodings.
Windows has two function calls for everything that requires text, a FoobarA() and a FoobarW(). The *W() functions take UTF-16 encoded strings, the *A() takes strings in the current codepage. However, Windows doesn't support a UTF-8 code page, so you can't directly use it in that sense with the *A() functions, nor would you want to depend on that being set by users. If you want "Unicode" in Windows, use the Unicode-capable (*W) functions. There are tutorials out there, Googling "Unicode Windows tutorial" should get you some.
If you are storing UTF-8 data in a std::string, then before you pass it off to Windows, convert it to UTF-16 (Windows provides functions for doing such), and then pass it to Windows.
Many of these problems arise from C/C++ being generally encoding-agnostic. char isn't really a character, it's just an integral type. Even using char arrays to store UTF-8 data can get you into trouble if you need to access individual code units, as char's signed-ness is left undefined by the standards. A statement like str[x] < 0x80 to check for multiple-byte characters can quickly introduce a bug. (That statement is always true if char is signed.) A UTF-8 code unit is an unsigned integral type with a range of 0-255. That maps to the C type of uint8_t exactly, although unsigned char works as well. Ideally then, I'd make a UTF-8 string an array of uint8_ts, but due to old APIs, this is rarely done.
Some people have recommended wchar_t, claiming it to be "A Unicode character type" or something like that. Again, here the standard is just as agnostic as before, as C is meant to work anywhere, and anywhere might not be using Unicode. Thus, wchar_t is no more Unicode than char. The standard states:
which is an integer type whose range of values can represent distinct codes for all members of the largest extended character set specified among the supported locales
In Linux, a wchat_t represents a UTF-32 code unit / code point. It is thus 4 bytes. However, in Windows, it's a UTF-16 code unit, and is only 2 bytes. (Which, I would have said does not conform to the above, since 2-bytes cannot represent all of Unicode, but that's the way it works.) This size difference, and difference in data encoding, clearly puts a strain on portability. The Unicode standard itself recommends against wchar_t if you need portability. (§5.2)
The end lesson: I find it easiest to store all my data in some well-declared format. (Typically UTF-8, usually in std::string's, but I'd really like something better.) The important thing here is not the UTF-8 part, but rather, I know that my strings are UTF-8. If I'm passing them to some other API, I must also know that that API expects UTF-8 strings. If it doesn't, then I must convert them. (Thus, if I speak to Window's API, I must convert strings to UTF-16 first.) A UTF-8 text string is an "orange", and a "latin1" text string is an "apple". A char array that doesn't know what encoding it is in is a recipe for disaster.
Putting UTF-8 code points into an std::string should be fine regardless of platform. The problem on Windows is that almost nothing else expects or works with UTF-8 -- it expects and works with UTF-16 instead. You can switch to an std::wstring which will store UTF-16 (at least on most Windows compilers) or you can write other routines that will accept UTF-8 (probably by converting to UTF-16, and then passing through to the OS).
Have you looked at std::wstring? It's a version of std::basic_string for wchar_t rather than the char that std::string uses.
No, there is no way to make Windows treat "narrow" strings as UTF-8.
Here is what works best for me in this situation (cross-platform application that has Windows and Linux builds).
Use std::string in cross-platform portion of the code. Assume that it always contains UTF-8 strings.
In Windows portion of the code, use "wide" versions of Windows API explicitly, i.e. write e.g. CreateFileW instead of CreateFile. This allows to avoid dependency on build system configuration.
In the platfrom abstraction layer, convert between UTF-8 and UTF-16 where needed (MultiByteToWideChar/WideCharToMultiByte).
Other approaches that I tried but don't like much:
typedef std::basic_string<TCHAR> tstring; then use tstring in the business code. Wrappers/overloads can be made to streamline conversion between std::string and std::tstring, but it still adds a lot of pain.
Use std::wstring everywhere. Does not help much since wchar_t is 16 bit on Windows, so you either have to restrict yourself to BMP or go to a lot of complications to make the code dealing with Unicode cross-platform. In the latter case, all benefits over UTF-8 evaporate.
Use ATL/WTL/MFC CString in the platfrom-specific portion; use std::string in cross-platfrom portion. This is actually a variant of what I recommend above. CString is in many aspects superior to std::string (in my opinion). But it introduces an additional dependency and thus not always acceptable or convenient.
If you want to avoid headache, don't use the STL string types at all. C++ knows nothing about Unicode or encodings, so to be portable, it's better to use a library that is tailored for Unicode support, e.g. the ICU library. ICU uses UTF-16 strings by default, so no conversion is required, and supports conversions to many other important encodings like UTF-8. Also try to use cross-platform libraries like Boost.Filesystem for things like path manipulations (boost::wpath). Avoid std::string and std::fstream.
In the Windows API and C runtime library, char* parameters are interpreted as being encoded in the "ANSI" code page. The problem is that UTF-8 isn't supported as an ANSI code page, which I find incredibly annoying.
I'm in a similar situation, being in the middle of porting software from Windows to Linux while also making it Unicode-aware. The approach we've taken for this is:
Use UTF-8 as the default encoding for strings.
In Windows-specific code, always call the "W" version of functions, converting string arguments between UTF-8 and UTF-16 as necessary.
This is also the approach Poco has taken.
It really platform dependant, Unicode is headache. Depends on which compiler you use. For older ones from MS (VS2010 or older), you would need use API described in MSDN
for VS2015
std::string _old = u8"D:\\Folder\\This \xe2\x80\x93 by ABC.txt"s;
according to their docs. I can't check that one.
for mingw, gcc, etc.
std::string _old = u8"D:\\Folder\\This \xe2\x80\x93 by ABC.txt";
std::cout << _old.data();
output contains proper file name...
You should consider using QString and QByteArray, it has good unicode support