C++ developers miss an easy and portable way of handling Unicode encoded strings. The original C++ Standard (known as C++98 or C++03) is Unicode agnostic. C++11 provides some support for Unicode on core language and library level: u8, u, and U character and string literals, char16_t and char32_t character types, u16string and u32string library classes, and codecvt support for conversions between Unicode encoding forms. In the meantime, developers use third party libraries like ICU, OS specific capabilities, or simply roll out their own solutions.
In order to easily handle UTF-8 encoded Unicode strings, I came up with a small, C++98 compatible generic library. For anybody used to work with STL algorithms and iterators, it should be easy and natural to use. The code is freely available for any purpose - check out the [license](./LICENSE). The library has been used a lot in the past ten years both in commercial and open-source projects and is considered feature-complete now. If you run into bugs or performance issues, please let me know and I'll do my best to address them.
The purpose of this article is not to offer an introduction to Unicode in general, and UTF-8 in particular. If you are not familiar with Unicode, be sure to check out [Unicode Home Page](http://www.unicode.org/) or some other source of information for Unicode. Also, it is not my aim to advocate the use of UTF-8 encoded strings in C++ programs; if you want to handle UTF-8 encoded strings from C++, I am sure you have good reasons for it.
## Examples of use
### Introductionary Sample
To illustrate the use of the library, let's start with a small but complete program that opens a file containing UTF-8 encoded text, reads it line by line, checks each line for invalid UTF-8 byte sequences, and converts it to UTF-16 encoding and back to UTF-8:
In the previous code sample, for each line we performed a detection of invalid UTF-8 sequences with `find_invalid`; the number of characters (more precisely - the number of Unicode code points, including the end of line and even BOM if there is one) in each line was determined with a use of `utf8::distance`; finally, we have converted each line to UTF-16 encoding with `utf8to16` and back to UTF-8 with `utf16to8`.
With a more modern compiler, the same operation would look like:
```cpp
u16string utf16line = utf8::utf8to16(line);
```
If `__cplusplus` macro points to a C++ 11 or later, the library exposes API that takes into
account C++ standard Unicode strings and move semantics. With an older compiler, it is still
possible to use the same functionality, just in a little less convenient way
In case you do not trust the `__cplusplus` macro or, for instance, do not want to include
the C++ 11 helper functions even with a modern compiler, define `UTF_CPP_CPLUSPLUS` macro
before including `utf8.h` and assign it a value for the standard you want to use - the values are the same as for the `__cplusplus` macro. This can be also useful with compilers that are conservative in setting the `__cplusplus` macro even if they have a good support for a recent standard edition - Microsoft's Visual C++ is one example.
Because the function `utf8::is_valid()` works with input iterators, we were able to pass an `istreambuf_iterator` to `it` and read the content of the file directly without loading it to the memory first.
Note that other functions that take input iterator arguments can be used in a similar way. For instance, to read the content of a UTF-8 encoded text file and convert the text to UTF-16, just do something like:
### Ensure that a string contains valid UTF-8 text
If we have some text that "probably" contains UTF-8 encoded text and we want to replace any invalid UTF-8 sequence with a replacement character, something like the following function may be used:
The function will replace any invalid UTF-8 sequence with a Unicode replacement character. There is an overloaded function that enables the caller to supply their own replacement character.
1. Generic: for better or worse, there are many C++ string classes out there, and the library should work with as many of them as possible.
2. Portable: the library should be portable both accross different platforms and compilers. The only non-portable code is a small section that declares unsigned integers of different sizes: three typedefs. They can be changed by the users of the library if they don't match their platform. The default setting should work for Windows (both 32 and 64 bit), and most 32 bit and 64 bit Unix derivatives. Support for post C++03 language features is included for modern compilers at API level only, so the library should work even with pretty old compilers.
3. Lightweight: follow the "pay only for what you use" guideline.
4. Unintrusive: avoid forcing any particular design or even programming style on the user. This is a library, not a framework.
#### Alternatives
In case you want to look into other means of working with UTF-8 strings from C++, here is the list of solutions I am aware of:
1. [ICU Library](http://icu.sourceforge.net/). It is very powerful, complete, feature-rich, mature, and widely used. Also big, intrusive, non-generic, and doesn't play well with the Standard Library. I definitelly recommend looking at ICU even if you don't plan to use it.
2. C++11 language and library features. Still far from complete, and not easy to use.
3. [Glib::ustring](http://www.gtkmm.org/gtkmm2/docs/tutorial/html/ch03s04.html). A class specifically made to work with UTF-8 strings, and also feel like `std::string`. If you prefer to have yet another string class in your code, it may be worth a look. Be aware of the licensing issues, though.
4. Platform dependent solutions: Windows and POSIX have functions to convert strings from one encoding to another. That is only a subset of what my library offers, but if that is all you need it may be good enough.
Note that `append` does not allocate any memory - it is the burden of the caller to make sure there is enough memory allocated for the operation. To make things more interesting, `append` can add anywhere between 1 and 4 octets to the sequence. In practice, you would most often want to use `std::back_inserter` to ensure that the necessary memory is allocated.
In case of an invalid code point, a `utf8::invalid_code_point` exception is thrown.
#### utf8::next
Available in version 1.0 and later.
Given the iterator to the beginning of the UTF-8 sequence, it returns the code point and moves the iterator to the next position.
`it`: a reference to an iterator pointing to the beginning of an UTF-8 encoded code point. After the function returns, it is incremented to point to the beginning of the next code point.
`end`: end of the UTF-8 sequence to be processed. If `it` gets equal to `end` during the extraction of a code point, an `utf8::not_enough_room` exception is thrown.
Return value: the 32 bit representation of the processed UTF-8 code point.
Example of use:
```cpp
char* twochars = "\xe6\x97\xa5\xd1\x88";
char* w = twochars;
int cp = next(w, twochars + 6);
assert (cp == 0x65e5);
assert (w == twochars + 3);
```
This function is typically used to iterate through a UTF-8 encoded string.
In case of an invalid UTF-8 seqence, a `utf8::invalid_utf8` exception is thrown.
#### utf8::peek_next
Available in version 2.1 and later.
Given the iterator to the beginning of the UTF-8 sequence, it returns the code point for the following sequence without changing the value of the iterator.
`it`: an iterator pointing to the beginning of an UTF-8 encoded code point.
`end`: end of the UTF-8 sequence to be processed. If `it` gets equal to `end` during the extraction of a code point, an `utf8::not_enough_room` exception is thrown.
Return value: the 32 bit representation of the processed UTF-8 code point.
Example of use:
```cpp
char* twochars = "\xe6\x97\xa5\xd1\x88";
char* w = twochars;
int cp = peek_next(w, twochars + 6);
assert (cp == 0x65e5);
assert (w == twochars);
```
In case of an invalid UTF-8 seqence, a `utf8::invalid_utf8` exception is thrown.
#### utf8::prior
Available in version 1.02 and later.
Given a reference to an iterator pointing to an octet in a UTF-8 sequence, it decreases the iterator until it hits the beginning of the previous UTF-8 encoded code point and returns the 32 bits representation of the code point.
`it`: a reference pointing to an octet within a UTF-8 encoded string. After the function returns, it is decremented to point to the beginning of the previous code point.
`start`: an iterator to the beginning of the sequence where the search for the beginning of a code point is performed. It is a safety measure to prevent passing the beginning of the string in the search for a UTF-8 lead octet.
Return value: the 32 bit representation of the previous code point.
Example of use:
```cpp
char* twochars = "\xe6\x97\xa5\xd1\x88";
unsigned char* w = twochars + 3;
int cp = prior (w, twochars);
assert (cp == 0x65e5);
assert (w == twochars);
```
This function has two purposes: one is two iterate backwards through a UTF-8 encoded string. Note that it is usually a better idea to iterate forward instead, since `utf8::next` is faster. The second purpose is to find a beginning of a UTF-8 sequence if we have a random position within a string. Note that in that case `utf8::prior` may not detect an invalid UTF-8 sequence in some scenarios: for instance if there are superfluous trail octets, it will just skip them.
`it` will typically point to the beginning of a code point, and `start` will point to the beginning of the string to ensure we don't go backwards too far. `it` is decreased until it points to a lead UTF-8 octet, and then the UTF-8 sequence beginning with that octet is decoded to a 32 bit representation and returned.
In case `start` is reached before a UTF-8 lead octet is hit, or if an invalid UTF-8 sequence is started by the lead octet, an `invalid_utf8` exception is thrown.
In case `start` equals `it`, a `not_enough_room` exception is thrown.
#### utf8::advance
Available in version 1.0 and later.
Advances an iterator by the specified number of code points within an UTF-8 sequence.
void advance (octet_iterator& it, distance_type n, octet_iterator end);
```
`octet_iterator`: an input iterator.
`distance_type`: an integral type convertible to `octet_iterator`'s difference type.
`it`: a reference to an iterator pointing to the beginning of an UTF-8 encoded code point. After the function returns, it is incremented to point to the nth following code point.
`n`: number of code points `it` should be advanced. A negative value means decrement.
`end`: limit of the UTF-8 sequence to be processed. If `n` is positive and `it` gets equal to `end` during the extraction of a code point, an `utf8::not_enough_room` exception is thrown. If `n` is negative and `it` reaches `end` while `it` points t a trail byte of a UTF-8 sequence, a `utf8::invalid_code_point` exception is thrown.
`first`: an iterator to a beginning of a UTF-8 encoded code point.
`last`: an iterator to a "post-end" of the last UTF-8 encoded code point in the sequence we are trying to determine the length. It can be the beginning of a new code point, or not.
Return value the distance between the iterators, in code points.
This function is used to find the length (in code points) of a UTF-8 encoded string. The reason it is called _distance_, rather than, say, _length_ is mainly because developers are used that _length_ is an O(1) function. Computing the length of an UTF-8 string is a linear operation, and it looked better to model it after `std::distance` algorithm.
In case of an invalid UTF-8 seqence, a `utf8::invalid_utf8` exception is thrown. If `last` does not point to the past-of-end of a UTF-8 seqence, a `utf8::not_enough_room` exception is thrown.
`start`: an iterator pointing to the beginning of the UTF-8 encoded string to convert. <br/>`end`: an iterator pointing to pass-the-end of the UTF-8 encoded string to convert.
`result`: an output iterator to the place in the UTF-16 string where to append the result of conversion.
Return value: An iterator pointing to the place after the appended UTF-16 string.
In case of an invalid UTF-8 seqence, a `utf8::invalid_utf8` exception is thrown. If `end` does not point to the past-of-end of a UTF-8 seqence, a `utf8::not_enough_room` exception is thrown.
In case of an invalid UTF-8 seqence, a `utf8::invalid_utf8` exception is thrown. If `end` does not point to the past-of-end of a UTF-8 seqence, a `utf8::not_enough_room` exception is thrown.
Available in version 3.0 and later. Requires a C++ 11 compliant compiler.
Detects an invalid sequence within a UTF-8 string.
```cpp
std::size_t find_invalid(const std::string& s);
```
`s`: a UTF-8 encoded string.
Return value: the index of the first invalid octet in the UTF-8 string. In case none were found, equals `std::string::npos`.
Example of use:
```cpp
string utf_invalid = "\xe6\x97\xa5\xd1\x88\xfa";
auto invalid = find_invalid(utf_invalid);
assert (invalid == 5);
```
This function is typically used to make sure a UTF-8 string is valid before processing it with other functions. It is especially important to call it if before doing any of the _unchecked_ operations on it.
#### utf8::find_invalid
Available in version 3.2 and later. Requires a C++ 17 compliant compiler.
Detects an invalid sequence within a UTF-8 string.
```cpp
std::size_t find_invalid(std::string_view s);
```
`s`: a UTF-8 encoded string.
Return value: the index of the first invalid octet in the UTF-8 string. In case none were found, equals `std::string_view::npos`.
This function is typically used to make sure a UTF-8 string is valid before processing it with other functions. It is especially important to call it if before doing any of the _unchecked_ operations on it.
This function is typically used to make sure a UTF-8 string is valid before processing it with other functions. It is especially important to call it if before doing any of the _unchecked_ operations on it.
`is_valid` is a shorthand for `find_invalid(start, end) == end;`. You may want to use it to make sure that a byte seqence is a valid UTF-8 string without the need to know where it fails if it is not valid.
`replace_invalid` does not perform in-place replacement of invalid sequences. Rather, it produces a copy of the original string with the invalid sequences replaced with a replacement marker. Therefore, `out` must not be in the `[start, end]` range.
The typical use of this function is to check the first three bytes of a file. If they form the UTF-8 BOM, we want to skip them before processing the actual UTF-8 encoded text.
#### utf8::starts_with_bom
Available in version 3.2 and later. Requires a C++ 17 compliant compiler.
Checks whether a string starts with a UTF-8 byte order mark (BOM)
The typical use of this function is to check the first three bytes of a file. If they form the UTF-8 BOM, we want to skip them before processing the actual UTF-8 encoded text.
The typical use of this function is to check the first three bytes of a file. If they form the UTF-8 BOM, we want to skip them before processing the actual UTF-8 encoded text.
### Types From utf8 Namespace
#### utf8::exception
Available in version 2.3 and later.
Base class for the exceptions thrown by UTF CPP library functions.
```cpp
class exception : public std::exception {};
```
Example of use:
```cpp
try {
code_that_uses_utf_cpp_library();
}
catch(const utf8::exception& utfcpp_ex) {
cerr <<utfcpp_ex.what();
}
```
#### utf8::invalid_code_point
Available in version 1.0 and later.
Thrown by UTF8 CPP functions such as `advance` and `next` if an UTF-8 sequence represents and invalid code point.
```cpp
class invalid_code_point : public exception {
public:
uint32_t code_point() const;
};
```
Member function `code_point()` can be used to determine the invalid code point that caused the exception to be thrown.
#### utf8::invalid_utf8
Available in version 1.0 and later.
Thrown by UTF8 CPP functions such as `next` and `prior` if an invalid UTF-8 sequence is detected during decoding.
```cpp
class invalid_utf8 : public exception {
public:
uint8_t utf8_octet() const;
};
```
Member function `utf8_octet()` can be used to determine the beginning of the byte sequence that caused the exception to be thrown.
#### utf8::invalid_utf16
Available in version 1.0 and later.
Thrown by UTF8 CPP function `utf16to8` if an invalid UTF-16 sequence is detected during decoding.
```cpp
class invalid_utf16 : public exception {
public:
uint16_t utf16_word() const;
};
```
Member function `utf16_word()` can be used to determine the UTF-16 code unit that caused the exception to be thrown.
#### utf8::not_enough_room
Available in version 1.0 and later.
Thrown by UTF8 CPP functions such as `next` if the end of the decoded UTF-8 sequence was reached before the code point was decoded.
```cpp
class not_enough_room : public exception {};
```
#### utf8::iterator
Available in version 2.0 and later.
Adapts the underlying octet iterator to iterate over the sequence of code points, rather than raw octets.
`explicit iterator (const octet_iterator& octet_it, const octet_iterator& range_start, const octet_iterator& range_end);` a constructor that initializes the underlying octet_iterator with octet_it and sets the range in which the iterator is considered valid.
The purpose of `utf8::iterator` adapter is to enable easy iteration as well as the use of STL algorithms with UTF-8 encoded strings. Increment and decrement operators are implemented in terms of `utf8::next()` and `utf8::prior()` functions.
Note that `utf8::iterator` adapter is a checked iterator. It operates on the range specified in the constructor; any attempt to go out of that range will result in an exception. Even the comparison operators require both iterator object to be constructed against the same range - otherwise an exception is thrown. Typically, the range will be determined by sequence container functions `begin` and `end`, i.e.:
```cpp
std::string s = "example";
utf8::iterator i (s.begin(), s.begin(), s.end());
```
### Functions From utf8::unchecked Namespace
#### utf8::unchecked::append
Available in version 1.0 and later.
Encodes a 32 bit code point as a UTF-8 sequence of octets and appends the sequence to a UTF-8 string.
This is a faster but less safe version of `utf8::append`. It does not check for validity of the supplied code point, and may produce an invalid UTF-8 sequence.
#### utf8::unchecked::next
Available in version 1.0 and later.
Given the iterator to the beginning of a UTF-8 sequence, it returns the code point and moves the iterator to the next position.
```cpp
template <typenameoctet_iterator>
uint32_t next(octet_iterator& it);
```
`it`: a reference to an iterator pointing to the beginning of an UTF-8 encoded code point. After the function returns, it is incremented to point to the beginning of the next code point.
Return value: the 32 bit representation of the processed UTF-8 code point.
Example of use:
```cpp
char* twochars = "\xe6\x97\xa5\xd1\x88";
char* w = twochars;
int cp = unchecked::next(w);
assert (cp == 0x65e5);
assert (w == twochars + 3);
```
This is a faster but less safe version of `utf8::next`. It does not check for validity of the supplied UTF-8 sequence.
#### utf8::unchecked::peek_next
Available in version 2.1 and later.
Given the iterator to the beginning of a UTF-8 sequence, it returns the code point.
```cpp
template <typenameoctet_iterator>
uint32_t peek_next(octet_iterator it);
```
`it`: an iterator pointing to the beginning of an UTF-8 encoded code point.
Return value: the 32 bit representation of the processed UTF-8 code point.
Example of use:
```cpp
char* twochars = "\xe6\x97\xa5\xd1\x88";
char* w = twochars;
int cp = unchecked::peek_next(w);
assert (cp == 0x65e5);
assert (w == twochars);
```
This is a faster but less safe version of `utf8::peek_next`. It does not check for validity of the supplied UTF-8 sequence.
#### utf8::unchecked::prior
Available in version 1.02 and later.
Given a reference to an iterator pointing to an octet in a UTF-8 seqence, it decreases the iterator until it hits the beginning of the previous UTF-8 encoded code point and returns the 32 bits representation of the code point.
```cpp
template <typenameoctet_iterator>
uint32_t prior(octet_iterator& it);
```
`it`: a reference pointing to an octet within a UTF-8 encoded string. After the function returns, it is decremented to point to the beginning of the previous code point.
Return value: the 32 bit representation of the previous code point.
Example of use:
```cpp
char* twochars = "\xe6\x97\xa5\xd1\x88";
char* w = twochars + 3;
int cp = unchecked::prior (w);
assert (cp == 0x65e5);
assert (w == twochars);
```
This is a faster but less safe version of `utf8::prior`. It does not check for validity of the supplied UTF-8 sequence and offers no boundary checking.
#### utf8::unchecked::advance
Available in version 1.0 and later.
Advances an iterator by the specified number of code points within an UTF-8 sequence.
`it`: a reference to an iterator pointing to the beginning of an UTF-8 encoded code point. After the function returns, it is incremented to point to the nth following code point.
`n`: number of code points `it` should be advanced. A negative value means decrement.
This is a faster but less safe version of `utf8::advance`. It does not check for validity of the supplied UTF-8 sequence and offers no boundary checking.
#### utf8::unchecked::distance
Available in version 1.0 and later.
Given the iterators to two UTF-8 encoded code points in a seqence, returns the number of code points between them.
`first`: an iterator to a beginning of a UTF-8 encoded code point.
`last`: an iterator to a "post-end" of the last UTF-8 encoded code point in the sequence we are trying to determine the length. It can be the beginning of a new code point, or not.
Return value: the distance between the iterators, in code points.
`start`: an iterator pointing to the beginning of the UTF-8 encoded string to convert. <br/>`end`: an iterator pointing to pass-the-end of the UTF-8 encoded string to convert.
`result`: an output iterator to the place in the UTF-16 string where to append the result of conversion.
Return value: An iterator pointing to the place after the appended UTF-16 string.
`replace_invalid` does not perform in-place replacement of invalid sequences. Rather, it produces a copy of the original string with the invalid sequences replaced with a replacement marker. Therefore, `out` must not be in the `[start, end]` range.
Unlike `utf8::replace_invalid`, this function does not verify validity of the replacement marker.
