libpEpDatatypes/src/nfc.hh

// This file is under GNU General Public License 3.0
// see LICENSE.txt

#ifndef LIBPEPDATATYPES_NFC_HH
#define LIBPEPDATATYPES_NFC_HH

#include <string_view>
#include <string>
#include <stdexcept>
#include <iosfwd>
#include <boost/operators.hpp>
#include <pEp/identity_list.h>

namespace pEp {

enum class IsNFC
{
	No=0,    // contains a character that cannot occur in NFC
	Maybe=1, // contains a character that is only allowed in certain positions in NFC
	Yes=2    // contains no invalid or partially valid character
};

std::ostream& operator<<(std::ostream& o, IsNFC is_nfc);


class illegal_utf : public std::runtime_error
{
public:
	illegal_utf(   std::string_view, unsigned position, const std::string& reason);
	illegal_utf(std::u16string_view, unsigned position, const std::string& reason);
	explicit illegal_utf(const std::string& message);
};


/// Common class template to define the same functions for all 3 Unicode Transfer Formats.
template<class CharT>
class UTF
{
public:
	/// parses a sequence of input code units into one Unicode code point and updates the input iterator c.
	/// \todo change to iterator templates?
	static
	uint32_t parse(const CharT*& c, const CharT* end);
	
	/// generates a UTF sequence from a given Unicode code point.
	template<class OutIter>
	static
	void generate(const char32_t c, OutIter& out);
	

	/// returns whether the sequence starts with IsNFC==Yes char
	static
	bool is_safe_NFC_start(std::basic_string_view<CharT> s);

	/// returns No or Maybe, if at least one character with NFC_Quickcheck class is "No" or "Maybe"
	/// might throw illegal_utf exception
	static
	IsNFC isNFC_quick_check(std::basic_string_view<CharT> s);

	/// runs first quick check and a deep test if quick check returns "Maybe".
	static
	bool isNFC(std::basic_string_view<CharT> s);

	/// returns true if the sequence is valid UTF-8
	bool isUtf(const CharT* begin, const CharT* end);

	/// converts a C++ string (in UTF-8/-16) into NFC form
	static
	std::basic_string<CharT> toNFC(std::basic_string_view<CharT> s);

	/// calculates the number of "code units" in the target Unicode Transfer Format.
	static
	size_t utf_length(std::u32string_view s);

	/// generates a whole u32string at once
	static
	std::basic_string<CharT> generate(const std::u32string& s);

	/// creates an NFD u32string from UTF-8/UTF-16 input string s
	static
	std::u32string fromUtf_decompose(std::basic_string_view<CharT> s);
	
	
	/// class holding a NFC-conform Unicode string.
	/// content is mostly read-only, because arbitrary modifications might destroy NFC conformacy.
	class nfc_string : public boost::totally_ordered2<nfc_string, std::basic_string_view<CharT>>
	{
	public:
		typedef std::basic_string<CharT>      String;
		typedef std::basic_string_view<CharT> StringView;
		
		/// only const_reference is supported.
		typedef typename String::const_reference const_reference;
		typedef typename String::const_pointer   const_pointer;
		
		/// only forward iterator. Does a backward_iterator make sense in UTF-encoded strings?
		typedef typename String::const_iterator  const_iterator;
		
		static const size_t npos = String::npos;
		
		
		explicit nfc_string(StringView  src);
		explicit nfc_string(String &&   src);
		
		/// construct from a NUL-terminated src
		explicit nfc_string(const CharT* src)
		: nfc_string{ StringView{src} }
		{}
		
		nfc_string(const CharT* src, size_t length)
		: nfc_string{ StringView{src, length} }
		{}
		
		nfc_string(const nfc_string&  src) = default;
		nfc_string(      nfc_string&& src) = default;
		
		nfc_string& operator=(const nfc_string&  src) = default;
		nfc_string& operator=(      nfc_string&& src) = default;
		
		nfc_string& assign(StringView src);
		nfc_string& assign(String &&  src);
		
		
		/// read-only: shares representation
		operator const String&() const noexcept { return s; }
		
		const String& get() const noexcept { return s;}
		
		/// read write: copy content
		operator String() const { return s; }
		
		const CharT* c_str() const noexcept { return s.c_str(); }
		const CharT* data()  const noexcept { return s.data();  }
		std::size_t  size()  const noexcept { return s.size();  }
		bool         empty() const noexcept { return s.empty(); }
		
		std::size_t capacity() const noexcept { return s.capacity(); }
		void reserve(std::size_t new_capacity) { s.reserve(new_capacity); }
		void shrink_to_fit() { s.shrink_to_fit(); }
		
		const_reference operator[](std::size_t ofs) const noexcept { return s[ofs]; }
		const_reference at(std::size_t ofs) const { return s.at(ofs); }
		const_reference front() const noexcept { return s.front(); }
		const_reference back()  const noexcept { return s.back(); }
		operator StringView()   const noexcept { return StringView{s}; }
		
		const_iterator  begin() const noexcept { return s.cbegin(); }
		const_iterator cbegin() const noexcept { return s.cbegin(); }  /// r/o access only
		const_iterator  end()   const noexcept { return s.cend(); }
		const_iterator cend()   const noexcept { return s.cend(); }    /// r/o access only
		
		void clear() { s.clear(); }
		
		/// I am lazy and delegate all the 10 different insert() overloads directly to s.
		template<typename... Args>
		nfc_string& insert(Args&& ...args)
		{
			s.insert( std::forward<Args>(args)... );
			normalize();
			return *this;
		}
		
		/// delegates all erase() overloads to s.
		template<typename... Args>
		nfc_string& erase(Args&& ...args)
		{
			s.erase( std::forward<Args>(args)... );
			normalize();
			return *this;
		}
		
		nfc_string& push_back(CharT c);
		
		/// delegates all 9 append() overloads to s.
		template<typename... Args>
		nfc_string& append(Args&& ...args)
		{
			s.append( std::forward<Args>(args)... );
			normalize();
			return *this;
		}
		
		nfc_string& operator+=(StringView s);
		
		/// optimization possible to avoid re-normalization in most cases.
		nfc_string& operator+=(const nfc_string& s);
		
		/// optimization possible to avoid re-normalization in most cases.
		nfc_string& operator+=(CharT c) { push_back(c); return *this; }
		
		/// delegates all 9 compare() overloads to s
		template<typename... Args>
		int compare(Args&& ...args) const
		{
			return s.compare( std::forward<Args>(args)... );
		}
		
		/// stolen from C++20
		bool starts_with(StringView s) const noexcept;
		
		/// stolen from C++20
		bool ends_with(StringView s) const noexcept;
		
		/// delegates all 5 find() overloads to s
		template<typename... Args>
		std::size_t find(Args&& ...args) const
		{
			return s.find( std::forward<Args>(args)... );
		}
		
		/// might throw illegal_utf, if a multi-char sequence is clipped.
		nfc_string substr(std::size_t pos=0, std::size_t count=npos) const;
		
	private:
		std::basic_string<CharT> s;
		
		/// (re-)normalize the content string s.
		void normalize();
	};
};

/// can be more efficient than the operator+() below.
template<class CharT>
typename
UTF<CharT>::nfc_string operator+(
		typename UTF<CharT>::nfc_string left,
		const typename UTF<CharT>::nfc_string& right);

template<class CharT, class T>
inline
typename
UTF<CharT>::nfc_string operator+(typename UTF<CharT>::nfc_string left, const T& right)
{
	return left+=right;
}

template<class CharT, class T>
inline
typename
UTF<CharT>::nfc_string operator+(typename UTF<CharT>::nfc_string&& left, const T& right)
{
	return left+=right;
}


template<class CharT, class T>
inline
typename
UTF<CharT>::nfc_string operator+(const T& left, const typename UTF<CharT>::nfc_string& right)
{
	typename UTF<CharT>::nfc_string left_s{left};
	return left_s+=right;
}

template<class CharT>
inline
bool operator<(const typename UTF<CharT>::nfc_string& left, std::basic_string_view<CharT> right)
{
	return left<right;
}

template<class CharT>
inline
bool operator==(const typename UTF<CharT>::nfc_string& left, std::basic_string_view<CharT> right)
{
	return left==right;
}


/// convenient alias names:
using UTF8 = UTF<char>;
using UTF16 = UTF<char16_t>;

using nfc_string = UTF8::nfc_string;
using nfc_u16string = UTF16::nfc_string;

// throws illegal_utf8 exception if s is not valid UTF-8
void assert_utf8(std::string_view s);


// convert NFD to NFC
std::u32string createNFC(std::u32string nfd_string);

/*
// return No or Maybe, if at least one character with NFC_Quickcheck class is "No" or "Maybe"
// might throw illegal_utf exception
template<class CharT>
IsNFC isNFC_quick_check(std::basic_string_view<CharT> s);

// runs first quick check and a deep test if quick check returns "Maybe".
template<class CharT>
bool isNFC(std::basic_string_view<CharT> s);

// returns true if the sequence is valid UTF-8
bool isUtf8(const char* begin, const char* end);

// converts a C++ string (in UTF-8) into NFC form
// s is ''moved'' to the return value if possible so no copy is done here.
template<class CharT>
std::basic_string<CharT> toNFC(std::basic_string_view<CharT> s);
*/

// creates a UTF-8-encoded NFC string from s
std::string toNFC_8(std::u16string_view s);

// convenience functions to avoid ::strdup(pEp::toNFC<char>(text).c_str());
// and unecessary temporary std::string etc.
char* strdup_NFC(std::string_view s);

pEp_identity *identity_dup_NFC(const ::pEp_identity* value);
::identity_list* identity_list_dup_NFC(const ::identity_list* value);


} // end of namespace pEp

#endif  // LIBPEPDATATYPES_NFC_HH