From 52d4375b10d920a59f1309c272a2e525feb1c25d Mon Sep 17 00:00:00 2001 From: Roland Reichwein Date: Sat, 1 Jan 2022 20:25:34 +0100 Subject: Separated out headers files; optimizations; type traits; better naming --- include/unicode/utf.h | 448 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 448 insertions(+) create mode 100644 include/unicode/utf.h (limited to 'include/unicode/utf.h') diff --git a/include/unicode/utf.h b/include/unicode/utf.h new file mode 100644 index 0000000..dd504a7 --- /dev/null +++ b/include/unicode/utf.h @@ -0,0 +1,448 @@ +#pragma once + +#include +#include +#include + +namespace unicode::detail { + + using namespace std::string_literals; + + template + inline bool is_utf8_leading_byte(value_type byte) noexcept + { + static_assert(sequence_length <= 4); + + if constexpr(sequence_length == 1) { + return !(byte & 0x80); + } else { + return (byte & static_cast(0xFF << (7 - sequence_length))) == static_cast(0xFF << (8 - sequence_length)); + } + } + + template + inline bool is_utf8_followup_byte(value_type b) noexcept + { + return (b & 0b11000000) == 0b10000000; + } + + template + inline bool is_utf8_sequence(value_type byte0, Tbytes... bytes) noexcept + { + constexpr auto sequence_length{sizeof...(Tbytes) + 1}; + + static_assert(sequence_length <= 4, "UTF-8 sequences of 1 through 4 code units are supported"); + + return is_utf8_leading_byte(byte0) && + (... && is_utf8_followup_byte(bytes)); // left fold for linear evaluation from left to right + } + + template = true> + inline bool validate_utf(const std::basic_string& s) + { + int i{}; + auto size{s.size()}; + while (i < size) { + if (is_utf8_sequence(s[i])) { + i++; + } else if ((i < size - 1) && is_utf8_sequence(s[i], s[i + 1])) { + i += 2; + } else if ((i < size - 2) && is_utf8_sequence(s[i], s[i + 1], s[i + 2])) { + if (((s[i] & 0xF) == 0xD) && ((s[i + 1] & 0x20) == 0x20)) + return false; // Reserved for UTF-16 surrogates: 0xD800..0xDFFF + i += 3; + } else if ((i < size - 3) && is_utf8_sequence(s[i], s[i + 1], s[i + 2], s[i + 3])) { + if ((((s[i] & 7) << 2) | ((s[i + 1] >> 4) & 3)) >= 0x11) + return false; // Unicode too big above 0x10FFFF + i += 4; + } else { + return false; + } + } + return true; + } + + template + inline bool is_utf16_sequence(value_type word0, Twords... words) noexcept + { + constexpr auto sequence_length{sizeof...(Twords) + 1}; + + static_assert(sequence_length <= 2, "UTF-16 sequences of only 1 or 2 code units are supported"); + + if constexpr(sequence_length == 1) { + return is_valid_unicode(word0); + } else { + char16_t unit0 {static_cast(word0)}; + char16_t unit1 {static_cast((words, ...))}; + return (unit0 & 0xFC00) == 0xD800 && (unit1 & 0xFC00) == 0xDC00; + } + } + + template = true> + inline bool validate_utf(const std::basic_string& s) + { + int i{}; + auto size{s.size()}; + while (i < size) { + if (is_utf16_sequence(s[i])) { + i++; + } else if ((i < size - 1) && is_utf16_sequence(s[i], s[i + 1])) { + i += 2; + } else { + return false; + } + } + return true; + } + + template = true> + inline bool validate_utf(const std::basic_string& s) + { + for (auto i: s) + if (!is_valid_unicode(i)) + return false; + return true; + } + + template + inline char32_t decode_utf8_leading_byte(value_type b) noexcept + { + return static_cast(b & (0b1111111 >> sequence_length)) << ((sequence_length - 1) * 6); + } + + template + inline char32_t decode_utf8_followup_byte(value_type b) noexcept + { + return static_cast(b & 0b00111111); + } + + template + inline char32_t decode_utf8_followup_byte(value_type b, Targs... bytes) noexcept + { + return decode_utf8_followup_byte(b) << (6 * sizeof...(Targs)) | decode_utf8_followup_byte(bytes...); + } + + template + inline char32_t decode_utf8_sequence(value_type b, Targs... bytes) noexcept + { + size_t constexpr sequence_length{sizeof...(Targs) + 1}; + + static_assert(sequence_length <= 4); + + if constexpr (sequence_length == 1) + return b; + else + return decode_utf8_leading_byte(b) | decode_utf8_followup_byte(bytes...); + } + + template> + struct utf_iterator + { + static_assert(sizeof(T) == 1 || sizeof(T) == 2 || sizeof(T) == 4); + + typedef T value_type; + typedef char32_t internal_type; + typedef char32_t& reference; + typedef char32_t* pointer; + typedef size_t difference_type; + typedef std::input_iterator_tag iterator_category; + typedef Container string_type; + + utf_iterator(const typename string_type::const_iterator& cbegin, const typename string_type::const_iterator& cend): + iterator(cbegin), end_iterator(cend) + { + } + + utf_iterator(const utf_iterator& other) = default; + utf_iterator& operator=(const utf_iterator& other) = default; + + inline size_t remaining_code_units() const noexcept + { + return std::distance(iterator, end_iterator); + } + + template + inline value_type get_code_unit() const noexcept + { + if constexpr (std::is_same_v>) { + // std::list doesn't support it + n + auto it{iterator}; + std::advance(it, index); + return *it; + } else { + return *(iterator + index); + } + } + + template + inline internal_type calculate_utf8_value(Tbytes... bytes) + { + size_t constexpr sequence_length{sizeof...(Tbytes)}; + static_assert(sequence_length >= 1 && sequence_length <= 4); + + if constexpr(sequence_length > 1) { + if (remaining_code_units() < sequence_length) + throw std::invalid_argument("Bad input: Not enough bytes left for decoding UTF-8 sequence"); + } + + if (is_utf8_sequence(bytes...)) { + std::advance(iterator, sequence_length); + internal_type result{decode_utf8_sequence(bytes...)}; + if (!unicode::is_valid_unicode(result)) + throw std::invalid_argument("Invalid Unicode character: "s + std::to_string(static_cast(result))); + return result; + } else { + if constexpr(sequence_length <= 3) // template recursion break condition: UTF-8 has 1..4 code units + return calculate_utf8_value(bytes..., static_cast(get_code_unit())); + else + throw std::invalid_argument("Bad UTF-8 input: Invalid 4 byte sequence"); + } + } + + template = true> + inline internal_type calculate_value() + { + return calculate_utf8_value(static_cast(get_code_unit<0>())); + } + + template = true> + inline internal_type calculate_value() + { + char16_t unit0 {static_cast(get_code_unit<0>())}; + + if (is_valid_unicode(unit0)) { // 1 unit (BMP Basic Multilingual Plane) + std::advance(iterator, 1); + return unit0; + } else { + if (remaining_code_units() < 2) + throw std::invalid_argument("Bad input: Continuation of first UTF-16 unit missing"); + + char16_t unit1 {static_cast(get_code_unit<1>())}; + if ((unit0 & 0xFC00) != 0xD800 || (unit1 & 0xFC00) != 0xDC00) + throw std::invalid_argument("Bad input: 2 malformed UTF-16 surrogates"); + + std::advance(iterator, 2); + return (static_cast(unit0 & 0x03FF) << 10 | (unit1 & 0x03FF)) + 0x10000; + } + } + + template = true> + inline internal_type calculate_value() + { + internal_type result {static_cast(get_code_unit<0>())}; + + if (!unicode::is_valid_unicode(result)) + throw std::invalid_argument("Invalid Unicode character: "s + std::to_string(static_cast(result))); + + std::advance(iterator, 1); + + return result; + } + + // pre-increment + utf_iterator& operator++() + { + return *this; + } + + bool operator!=(const utf_iterator& other) const + { + return std::distance(iterator, end_iterator) != std::distance(other.iterator, other.end_iterator); + } + + internal_type operator*() + { + return calculate_value(); + } + + utf_iterator& operator+=(size_t distance) + { + std::advance(iterator, distance); + return *this; + } + + size_t operator-(const utf_iterator& other) const + { + return iterator - other.iterator; + } + + private: + typename string_type::const_iterator iterator; + typename string_type::const_iterator end_iterator; + }; + + // n is number of UTF-8 bytes in sequence + template + inline To utf8_byte0_of(const From& value) + { + return (value >> 6 * (n - 1)) | (0xFF << (8 - n)); + } + + // n is index of 6-bit groups, counting from bit 0 + template + inline To utf8_trailing_byte(const From& value) + { + return ((value >> n * 6) & 0b111111) | 0b10000000; + } + + // calculate UTF-8 sequence byte for m >= 2 bytes sequences (i.e. non-ASCII) + // assume value to be valid Unicode value for given byte position + template + inline To utf8_byte_n_of_m(const From& value) + { + if constexpr (n == 0) + return utf8_byte0_of(value); + else + return utf8_trailing_byte(value); + } + + template> + struct utf_back_insert_iterator + { + static_assert(sizeof(T) == 1 || sizeof(T) == 2 || sizeof(T) == 4); + + typedef T value_type; + typedef char32_t internal_type; + typedef Container string_type; + typedef utf_back_insert_iterator& reference; + typedef utf_back_insert_iterator* pointer; + typedef size_t difference_type; + typedef std::output_iterator_tag iterator_category; + + utf_back_insert_iterator(string_type& s): s(s) {} + + utf_back_insert_iterator& operator=(const utf_back_insert_iterator& other) + { + if (std::addressof(other.s) != std::addressof(s)) + throw std::runtime_error("utf_back_insert_iterator assignment operator actually called! Iterator should not be assigned to."); + + return *this; + } + + // no-op + reference operator++() + { + return *this; + } + + // support *x = value, together with operator=() + reference operator*() + { + return *this; + } + + template + inline void append(Args&&... args) + { + if constexpr (std::is_same_v>) { + s.append({args...}); + } else { + (s.emplace_back(args), ...); + } + } + + template = true> + inline void append_utf(const internal_type& value) + { + using Y = internal_type; + if (value < 0x80) { // 1 byte + append(static_cast(value)); + } else if (value < 0x800) { // 2 bytes + append(utf8_byte_n_of_m<0,2,Y,X>(value), utf8_byte_n_of_m<1,2,Y,X>(value)); + } else if (value < 0x10000) { // 3 bytes + append(utf8_byte_n_of_m<0,3,Y,X>(value), utf8_byte_n_of_m<1,3,Y,X>(value), utf8_byte_n_of_m<2,3,Y,X>(value)); + } else { // 4 bytes + // expect value to be already valid Unicode values (checked in input iterator) + append(utf8_byte_n_of_m<0,4,Y,X>(value), utf8_byte_n_of_m<1,4,Y,X>(value), utf8_byte_n_of_m<2,4,Y,X>(value), utf8_byte_n_of_m<3,4,Y,X>(value)); + } + } + + template = true> + inline void append_utf(const internal_type& value) + { + if (value <= 0xFFFF) { // expect value to be already valid Unicode values (checked in input iterator) + append(static_cast(value)); + } else { + internal_type value_reduced{value - 0x10000}; + append(static_cast((value_reduced >> 10) + 0xD800), static_cast((value_reduced & 0x3FF) + 0xDC00)); + } + } + + template = true> + inline void append_utf(const internal_type& value) + { + // expect value to be already valid Unicode values (checked in input iterator) + append(static_cast(value)); + } + + reference operator=(const internal_type& value) + { + append_utf(value); + return *this; + } + + private: + typename utf_back_insert_iterator::string_type& s; + }; + +} // namespace unicode::detail + +namespace unicode { + + // Encoding for convert() and UTF-* + template + struct UTF + { + typedef typename OutputIt::value_type value_type; + typedef typename InputIt::string_type string_type; + + static InputIt begin(const typename InputIt::string_type& s) + { + return InputIt{s.cbegin(), s.cend()}; + } + + static InputIt end(const typename InputIt::string_type& s) + { + return InputIt{s.cend(), s.cend()}; + } + + static OutputIt back_inserter(typename OutputIt::string_type& s) + { + return OutputIt(s); + } + }; + + // Encoding for convert() + typedef UTF, utf_back_insert_iterator> UTF_8; + typedef UTF, utf_back_insert_iterator> UTF_16; + typedef UTF, utf_back_insert_iterator> UTF_32; + + // Helper to get correct Encoding from char type, e.g. Encoding::type or Encoding_t + template + struct Encoding + { + }; + + template<> + struct Encoding + { + typedef UTF_8 type; + }; + + template<> + struct Encoding + { + typedef UTF_16 type; + }; + + template<> + struct Encoding + { + typedef UTF_32 type; + }; + + template + using Encoding_t = typename Encoding::type; + +} // namespace unicode + -- cgit v1.2.3