/
githubmirror
/
nghttp2
Обзор
Документация
Войти
/
githubmirror
/
nghttp2
Код
Запросы
0
Пакеты
0
Релизы
0
Аналитика
Безопасность
master
src/util.h
1 396 строк
44 KB
Tatsuhiro Tsujikawa
src: Upper case hex integer literals
06 июн 2026, 13:29
06 июн 2026, 13:29
41004d2
Код
Авторство
О чём код?
/* * nghttp2 - HTTP/2 C Library * * Copyright (c) 2012 Tatsuhiro Tsujikawa * * Permission is hereby granted, free of charge, to any person obtaining * a copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including * without limitation the rights to use, copy, modify, merge, publish, * distribute, sublicense, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ #ifndef UTIL_H #define UTIL_H #include "nghttp2_config.h" #ifdef HAVE_UNISTD_H # include <unistd.h> #endif // defined(HAVE_UNISTD_H) #include <getopt.h> #ifdef HAVE_NETDB_H # include <netdb.h> #endif // defined(HAVE_NETDB_H) #ifdef __QNX__ # include <sys/time.h> #endif // defined(__QNX__) #include <cmath> #include <cstring> #include <cassert> #include <vector> #include <string> #include <algorithm> #include <memory> #include <chrono> #include <unordered_map> #include <random> #include <ranges> #include <bit> #include <expected> #ifdef HAVE_LIBEV # include <ev.h> #endif // defined(HAVE_LIBEV) #include "urlparse.h" #include "template.h" #include "network.h" #include "allocator.h" #include "errors.h" using namespace std::literals; namespace nghttp2 { inline constexpr auto NGHTTP2_H2_ALPN = "\x2h2"sv; inline constexpr auto NGHTTP2_H2 = "h2"sv; inline constexpr auto NGHTTP2_H1_1_ALPN = "\x8http/1.1"sv; inline constexpr auto NGHTTP2_H1_1 = "http/1.1"sv; namespace util { template <std::predicate<size_t> Pred> constexpr auto pred_tbl_gen256(Pred pred) { std::array<bool, 256> tbl; for (size_t i = 0; i < tbl.size(); ++i) { tbl[i] = pred(i); } return tbl; } constexpr auto alpha_pred(size_t i) noexcept { return ('A' <= i && i <= 'Z') || ('a' <= i && i <= 'z'); } inline constexpr auto is_alpha_tbl = pred_tbl_gen256(alpha_pred); constexpr bool is_alpha(char c) noexcept { return is_alpha_tbl[static_cast<uint8_t>(c)]; } constexpr auto digit_pred(size_t i) noexcept { return '0' <= i && i <= '9'; } inline constexpr auto is_digit_tbl = pred_tbl_gen256(digit_pred); constexpr bool is_digit(char c) noexcept { return is_digit_tbl[static_cast<uint8_t>(c)]; } constexpr auto hex_digit_pred(size_t i) noexcept { return digit_pred(i) || ('A' <= i && i <= 'F') || ('a' <= i && i <= 'f'); } inline constexpr auto is_hex_digit_tbl = pred_tbl_gen256(hex_digit_pred); constexpr bool is_hex_digit(char c) noexcept { return is_hex_digit_tbl[static_cast<uint8_t>(c)]; } // Returns true if a range [|first|, |last|) is hex string. template <std::forward_iterator I> constexpr bool is_hex_string(I first, I last) { return !(std::ranges::distance(first, last) & 1) && std::ranges::all_of(first, last, is_hex_digit); } // Returns true if |r| is hex string. template <std::ranges::forward_range R> requires(!std::is_array_v<std::remove_cvref_t<R>>) constexpr bool is_hex_string(R &&r) { return is_hex_string(std::ranges::begin(r), std::ranges::end(r)); } constexpr auto rfc3986_unreserved_chars_pred(size_t i) noexcept { switch (i) { case '-': case '.': case '_': case '~': return true; } return digit_pred(i) || alpha_pred(i); } inline constexpr auto in_rfc3986_unreserved_chars_tbl = pred_tbl_gen256(rfc3986_unreserved_chars_pred); constexpr bool in_rfc3986_unreserved_chars(char c) noexcept { return in_rfc3986_unreserved_chars_tbl[static_cast<uint8_t>(c)]; } constexpr auto rfc3986_sub_delims_pred(size_t i) noexcept { switch (i) { case '!': case '$': case '&': case '\'': case '(': case ')': case '*': case '+': case ',': case ';': case '=': return true; } return false; } inline constexpr auto in_rfc3986_sub_delims_tbl = pred_tbl_gen256(rfc3986_sub_delims_pred); constexpr bool in_rfc3986_sub_delims(char c) noexcept { return in_rfc3986_sub_delims_tbl[static_cast<uint8_t>(c)]; } constexpr auto token_pred(size_t i) noexcept { switch (i) { case '!': case '#': case '$': case '%': case '&': case '\'': case '*': case '+': case '-': case '.': case '^': case '_': case '`': case '|': case '~': return true; } return digit_pred(i) || alpha_pred(i); } inline constexpr auto in_token_tbl = pred_tbl_gen256(token_pred); // Returns true if |c| is in token (HTTP-p1, Section 3.2.6) constexpr bool in_token(char c) noexcept { return in_token_tbl[static_cast<uint8_t>(c)]; } constexpr auto attr_char_pred(size_t i) noexcept { switch (i) { case '*': case '\'': case '%': return false; } return token_pred(i); } inline constexpr auto in_attr_char_tbl = pred_tbl_gen256(attr_char_pred); constexpr bool in_attr_char(char c) noexcept { return in_attr_char_tbl[static_cast<uint8_t>(c)]; } inline constexpr auto hex_to_uint_tbl = [] { std::array<uint32_t, 256> tbl; std::ranges::fill(tbl, 256); for (char i = '0'; i <= '9'; ++i) { tbl[static_cast<uint8_t>(i)] = static_cast<uint32_t>(i - '0'); } for (char i = 'A'; i <= 'F'; ++i) { tbl[static_cast<uint8_t>(i)] = static_cast<uint32_t>(i - 'A' + 10); } for (char i = 'a'; i <= 'f'; ++i) { tbl[static_cast<uint8_t>(i)] = static_cast<uint32_t>(i - 'a' + 10); } return tbl; }(); // Returns integer corresponding to hex notation |c|. If // is_hex_digit(c) is false, it returns 256. constexpr uint32_t hex_to_uint(char c) noexcept { return hex_to_uint_tbl[static_cast<uint8_t>(c)]; } template <std::input_iterator I, std::weakly_incrementable O> requires(std::indirectly_copyable<I, O>) constexpr O percent_decode(I first, I last, O result) { using result_type = std::iter_value_t<O>; for (; first != last; ++first) { if (*first != '%') { *result++ = static_cast<result_type>(*first); continue; } auto dig1 = std::ranges::next(first, 1); if (dig1 == last || !is_hex_digit(*dig1)) { *result++ = static_cast<result_type>(*first); continue; } auto dig2 = std::ranges::next(dig1, 1); if (dig2 == last || !is_hex_digit(*dig2)) { *result++ = static_cast<result_type>(*first); continue; } *result++ = static_cast<result_type>((hex_to_uint(*dig1) << 4) | hex_to_uint(*dig2)); first = dig2; } return result; } template <std::forward_iterator I> constexpr std::string percent_decode(I first, I last) { std::string result; result.resize_and_overwrite( as_unsigned(std::ranges::distance(first, last)), [first = std::move(first), last = std::move(last)](auto p, auto len) mutable { return std::ranges::distance( p, percent_decode(std::move(first), std::move(last), p)); }); return result; } template <std::ranges::forward_range R> requires(!std::is_array_v<std::remove_cvref_t<R>>) constexpr std::string percent_decode(R &&r) { return percent_decode(std::ranges::begin(r), std::ranges::end(r)); } template <std::ranges::sized_range R> requires(!std::is_array_v<std::remove_cvref_t<R>>) std::string_view percent_decode(BlockAllocator &balloc, R &&r) { auto iov = make_byte_ref(balloc, std::ranges::size(r) + 1); auto p = percent_decode(std::ranges::begin(r), std::ranges::end(r), std::ranges::begin(iov)); *p = '\0'; return as_string_view(std::ranges::begin(iov), p); } // Quote a range [|first|, |last|) and stores the result in another // range, beginning at |result|. It returns an output iterator to the // element past the last element stored. Currently, this function // just replace '"' with '\"'. template <std::input_iterator I, std::weakly_incrementable O> requires(std::indirectly_copyable<I, O>) constexpr O quote_string(I first, I last, O result) noexcept { for (; first != last; ++first) { if (*first == '"') { *result++ = '\\'; *result++ = '"'; } else { *result++ = static_cast<std::iter_value_t<O>>(*first); } } return result; } template <std::ranges::input_range R, std::weakly_incrementable O> requires(std::indirectly_copyable<std::ranges::iterator_t<R>, O> && !std::is_array_v<std::remove_cvref_t<R>>) constexpr O quote_string(R &&r, O result) { return quote_string(std::ranges::begin(r), std::ranges::end(r), std::move(result)); } template <std::ranges::sized_range R> requires(!std::is_array_v<std::remove_cvref_t<R>>) std::string_view quote_string(BlockAllocator &balloc, R &&r) { auto cnt = std::ranges::count(r, '"'); if (cnt == 0) { return make_string_ref(balloc, std::forward<R>(r)); } auto iov = make_byte_ref(balloc, std::ranges::size(r) + static_cast<size_t>(cnt) + 1); auto p = quote_string(std::forward<R>(r), std::ranges::begin(iov)); *p = '\0'; return as_string_view(std::ranges::begin(iov), p); } // Returns the number of bytes written by quote_string with the same // |r| parameter. The return value does not include a terminal NUL // byte. template <std::ranges::input_range R> requires(!std::is_array_v<std::remove_cvref_t<R>>) constexpr size_t quote_stringlen(R &&r) { size_t n = 0; for (auto c : r) { if (c == '"') { n += 2; } else { ++n; } } return n; } inline constexpr char LOWER_XDIGITS[] = "0123456789abcdef"; template <std::weakly_incrementable O> requires(std::indirectly_writable<O, char>) constexpr O format_hex_uint8(uint8_t b, O result) { #ifdef __GNUC__ # pragma GCC diagnostic push # pragma GCC diagnostic ignored "-Wsign-conversion" #endif // __GNUC__ *result++ = LOWER_XDIGITS[b >> 4]; *result++ = LOWER_XDIGITS[b & 0xF]; #ifdef __GNUC__ # pragma GCC diagnostic pop #endif // __GNUC__ return result; } // Converts a range [|first|, |last|) in hex format, and stores the // result in another range, beginning at |result|. It returns an // output iterator to the element past the last element stored. template <std::input_iterator I, std::weakly_incrementable O> requires(std::indirectly_writable<O, char> && sizeof(std::iter_value_t<I>) == sizeof(uint8_t)) constexpr O format_hex(I first, I last, O result) { for (; first != last; ++first) { result = format_hex_uint8(static_cast<uint8_t>(*first), result); } return result; } // Converts |R| in hex format, and stores the result in another range, // beginning at |result|. It returns an output iterator to the // element past the last element stored. template <std::ranges::input_range R, std::weakly_incrementable O> requires(std::indirectly_writable<O, char> && !std::is_array_v<std::remove_cvref_t<R>> && sizeof(std::ranges::range_value_t<R>) == sizeof(uint8_t)) constexpr O format_hex(R &&r, O result) { return format_hex(std::ranges::begin(r), std::ranges::end(r), std::move(result)); } // Converts |R| in hex format, and stores the result in a buffer // allocated by |balloc|. It returns std::string_view that is backed by the // allocated buffer. The returned string is NULL terminated. template <std::ranges::sized_range R> requires(!std::is_array_v<std::remove_cvref_t<R>> && sizeof(std::ranges::range_value_t<R>) == sizeof(uint8_t)) std::string_view format_hex(BlockAllocator &balloc, R &&r) { auto iov = make_byte_ref(balloc, std::ranges::size(r) * 2 + 1); auto p = format_hex(std::forward<R>(r), std::ranges::begin(iov)); *p = '\0'; return as_string_view(std::ranges::begin(iov), p); } // Converts |R| in hex format, and returns the result. template <std::ranges::forward_range R> requires(!std::is_array_v<std::remove_cvref_t<R>> && sizeof(std::ranges::range_value_t<R>) == sizeof(uint8_t)) constexpr std::string format_hex(R &&r) { std::string res; res.resize_and_overwrite(as_unsigned(std::ranges::distance(r) * 2), [&r](auto p, auto len) { return std::ranges::distance(p, format_hex(r, p)); }); return res; } template <std::unsigned_integral T, std::weakly_incrementable O> requires(std::indirectly_writable<O, char>) constexpr O format_hex(T n, O result) { if constexpr (sizeof(n) == 1) { return format_hex_uint8(n, result); } if constexpr (std::endian::native == std::endian::little) { auto end = reinterpret_cast<uint8_t *>(&n); auto p = end + sizeof(n); for (; p != end; --p) { result = format_hex_uint8(*(p - 1), result); } } else { auto p = reinterpret_cast<uint8_t *>(&n); auto end = p + sizeof(n); for (; p != end; ++p) { result = format_hex_uint8(*p, result); } } return result; } inline constexpr char UPPER_XDIGITS[] = "0123456789ABCDEF"; template <std::weakly_incrementable O> requires(std::indirectly_writable<O, char>) constexpr O format_upper_hex_uint8(uint8_t b, O result) { #ifdef __GNUC__ # pragma GCC diagnostic push # pragma GCC diagnostic ignored "-Wsign-conversion" #endif // __GNUC__ *result++ = UPPER_XDIGITS[b >> 4]; *result++ = UPPER_XDIGITS[b & 0xF]; #ifdef __GNUC__ # pragma GCC diagnostic pop #endif // __GNUC__ return result; } // decode_hex decodes hex string in a range [|first|, |last|), and // stores the result in another range, beginning at |result|. It // returns an output iterator to the element past the last element // stored. This function assumes a range [|first|, |last|) is hex // string, that is is_hex_string(|first|, |last|) == true. template <std::input_iterator I, std::weakly_incrementable O> requires(std::indirectly_writable<O, uint8_t>) constexpr O decode_hex(I first, I last, O result) { for (; first != last; first = std::ranges::next(first, 2)) { *result++ = static_cast<std::iter_value_t<O>>( (hex_to_uint(*first) << 4) | hex_to_uint(*std::ranges::next(first, 1))); } return result; } // decode_hex decodes hex string |r|, and stores the result in another // range, beginning at |result|. It returns an output iterator to the // element past the last element stored. This function assumes |r| is // hex string, that is is_hex_string(r) == true. template <std::ranges::input_range R, std::weakly_incrementable O> requires(std::indirectly_writable<O, uint8_t> && !std::is_array_v<std::remove_cvref_t<R>>) constexpr O decode_hex(R &&r, O result) { return decode_hex(std::ranges::begin(r), std::ranges::end(r), std::move(result)); } // decode_hex decodes hex string in a range [|first|, |last|), returns // the decoded byte string, which is not NULL terminated. This // function assumes a range [|first|, |last|) is hex string, that is // is_hex_string(|first|, |last|) == true. template <std::forward_iterator I> std::span<const uint8_t> decode_hex(BlockAllocator &balloc, I first, I last) { auto iov = make_byte_ref(balloc, as_unsigned(std::ranges::distance(first, last) / 2)); auto p = decode_hex(std::move(first), std::move(last), std::ranges::begin(iov)); return {std::ranges::begin(iov), p}; } // decode_hex decodes hex string |r|, returns the decoded byte string, // which is not NULL terminated. This function assumes |r| is hex // string, that is is_hex_string(r) == true. template <std::ranges::forward_range R> requires(!std::is_array_v<std::remove_cvref_t<R>>) std::span<const uint8_t> decode_hex(BlockAllocator &balloc, R &&r) { return decode_hex(balloc, std::ranges::begin(r), std::ranges::end(r)); } // Percent encode a range [|first|, |last|) if a character is not in // token or '%'. template <std::input_iterator I, std::weakly_incrementable O> requires(std::indirectly_copyable<I, O>) constexpr O percent_encode_token(I first, I last, O result) noexcept { using result_type = std::iter_value_t<O>; for (; first != last; ++first) { auto c = static_cast<uint8_t>(*first); if (c != '%' && in_token(static_cast<char>(c))) { *result++ = static_cast<result_type>(c); continue; } *result++ = '%'; result = format_upper_hex_uint8(c, result); } return result; } template <std::ranges::input_range R, std::weakly_incrementable O> requires(std::indirectly_copyable<std::ranges::iterator_t<R>, O> && !std::is_array_v<std::remove_cvref_t<R>>) constexpr O percent_encode_token(R &&r, O result) { return percent_encode_token(std::ranges::begin(r), std::ranges::end(r), std::move(result)); } // Returns the number of bytes written by percent_encode_token with // the same |r| parameter. The return value does not include a // terminal NUL byte. template <std::ranges::input_range R> requires(!std::is_array_v<std::remove_cvref_t<R>>) constexpr size_t percent_encode_tokenlen(R &&r) noexcept { size_t n = 0; for (auto c : r) { if (c != '%' && in_token(c)) { ++n; continue; } // percent-encoded character '%ff' n += 3; } return n; } std::expected<time_t, Error> parse_http_date(std::string_view s); inline constexpr auto upcase_tbl = [] { std::array<char, 256> tbl; for (size_t i = 0; i < 256; ++i) { if ('a' <= i && i <= 'z') { tbl[i] = static_cast<char>(i - 'a' + 'A'); } else { tbl[i] = static_cast<char>(i); } } return tbl; }(); constexpr char upcase(char c) noexcept { return upcase_tbl[static_cast<uint8_t>(c)]; } inline constexpr auto lowcase_tbl = [] { std::array<char, 256> tbl; for (size_t i = 0; i < 256; ++i) { if ('A' <= i && i <= 'Z') { tbl[i] = static_cast<char>(i - 'A' + 'a'); } else { tbl[i] = static_cast<char>(i); } } return tbl; }(); constexpr char lowcase(char c) noexcept { return lowcase_tbl[static_cast<uint8_t>(c)]; } template <std::ranges::forward_range R1, std::ranges::forward_range R2> constexpr bool starts_with(R1 &&s, R2 &&prefix) { auto prefixlen = std::ranges::distance(prefix); return std::ranges::distance(s) >= prefixlen && std::ranges::equal(std::views::take(std::forward<R1>(s), prefixlen), std::forward<R2>(prefix)); } struct CaseCmp { constexpr bool operator()(char lhs, char rhs) const noexcept { return lowcase(lhs) == lowcase(rhs); } }; template <std::ranges::forward_range R1, std::ranges::forward_range R2> constexpr bool istarts_with(R1 &&s, R2 &&prefix) { auto prefixlen = std::ranges::distance(prefix); return std::ranges::distance(s) >= prefixlen && std::ranges::equal(std::views::take(std::forward<R1>(s), prefixlen), std::forward<R2>(prefix), CaseCmp()); } template <std::ranges::forward_range R1, std::ranges::forward_range R2> constexpr bool ends_with(R1 &&s, R2 &&suffix) { auto slen = std::ranges::distance(s); auto suffixlen = std::ranges::distance(suffix); return slen >= suffixlen && std::ranges::equal( std::views::drop(std::forward<R1>(s), slen - suffixlen), std::forward<R2>(suffix)); } template <std::ranges::forward_range R1, std::ranges::forward_range R2> constexpr bool iends_with(R1 &&s, R2 &&suffix) { auto slen = std::ranges::distance(s); auto suffixlen = std::ranges::distance(suffix); return slen >= suffixlen && std::ranges::equal( std::views::drop(std::forward<R1>(s), slen - suffixlen), std::forward<R2>(suffix), CaseCmp()); } template <std::ranges::input_range R1, std::ranges::input_range R2> constexpr bool strieq(R1 &&a, R2 &&b) { return std::ranges::equal(std::forward<R1>(a), std::forward<R2>(b), CaseCmp()); } template <std::ranges::input_range R1, std::ranges::input_range R2> constexpr bool streq(R1 &&a, R2 &&b) { return std::ranges::equal(std::forward<R1>(a), std::forward<R2>(b)); } // Converts characters in a range [|first|, |last|) to lowercase, and // stores the result in another range, beginning at |result|. It // returns an output iterator to the element past the last element // stored. template <std::input_iterator I, std::weakly_incrementable O> requires(std::indirectly_copyable<I, O>) constexpr O tolower(I first, I last, O result) { return std::ranges::transform(std::move(first), std::move(last), std::move(result), lowcase) .out; } // Converts characters in a range |r| to lowercase, and stores the // result in another range, beginning at |result|. It returns an // output iterator to the element past the last element stored. template <std::ranges::input_range R, std::weakly_incrementable O> requires(std::indirectly_copyable<std::ranges::iterator_t<R>, O> && !std::is_array_v<std::remove_cvref_t<R>>) constexpr O tolower(R &&r, O result) { return std::ranges::transform(std::forward<R>(r), std::move(result), lowcase) .out; } // Returns string representation of |n| with 2 fractional digits. std::string dtos(double n); inline constexpr auto count_digit_tbl = [] { std::array<uint64_t, std::numeric_limits<uint64_t>::digits10> tbl; uint64_t x = 1; for (size_t i = 0; i < tbl.size(); ++i) { x *= 10; tbl[i] = x - 1; } return tbl; }(); // count_digit returns the minimum number of digits to represent |x| // in base 10. // // credit: // https://lemire.me/blog/2025/01/07/counting-the-digits-of-64-bit-integers/ template <std::unsigned_integral T> constexpr size_t count_digit(T x) { auto y = static_cast<size_t>(19 * (std::numeric_limits<T>::digits - 1 - std::countl_zero(static_cast<T>(x | 1))) >> 6); y += x > count_digit_tbl[y]; return y + 1; } inline constexpr auto utos_digits = [] { std::array<char, 200> a; for (size_t i = 0; i < 100; ++i) { a[i * 2] = '0' + static_cast<char>(i / 10); a[i * 2 + 1] = '0' + static_cast<char>(i % 10); } return a; }(); struct UIntFormatter { template <std::unsigned_integral T, std::weakly_incrementable O> requires(std::indirectly_writable<O, char>) constexpr O operator()(T n, O result) { using result_type = std::iter_value_t<O>; if (n < 10) { *result++ = static_cast<result_type>('0' + static_cast<char>(n)); return result; } if (n < 100) { return std::ranges::copy_n(utos_digits.data() + n * 2, 2, result).out; } std::ranges::advance(result, as_signed(count_digit(n))); auto p = result; for (; n >= 100; n /= 100) { std::ranges::advance(p, -2); std::ranges::copy_n(utos_digits.data() + (n % 100) * 2, 2, p); } if (n < 10) { *--p = static_cast<result_type>('0' + static_cast<char>(n)); return result; } std::ranges::advance(p, -2); std::ranges::copy_n(utos_digits.data() + n * 2, 2, p); return result; } }; template <std::unsigned_integral T, std::weakly_incrementable O> requires(std::indirectly_writable<O, char>) constexpr O utos(T n, O result) { return UIntFormatter{}(std::move(n), std::move(result)); } template <std::unsigned_integral T> constexpr std::string utos(T n) { using namespace std::literals; if (n == 0) { return "0"s; } std::string res; res.resize_and_overwrite(count_digit(n), [n](auto p, auto len) { return std::ranges::distance(p, utos(n, p)); }); return res; } template <std::unsigned_integral T> std::string_view make_string_ref_uint(BlockAllocator &balloc, T n) { auto iov = make_byte_ref( balloc, count_digit(static_cast<std::make_unsigned_t<T>>(n)) + 1); auto p = std::ranges::begin(iov); p = util::utos(n, p); *p = '\0'; return as_string_view(std::ranges::begin(iov), p); } template <std::unsigned_integral T> constexpr std::string utos_unit(T n) { char u; if (n >= (1 << 30)) { u = 'G'; n /= (1 << 30); } else if (n >= (1 << 20)) { u = 'M'; n /= (1 << 20); } else if (n >= (1 << 10)) { u = 'K'; n /= (1 << 10); } else { return utos(n); } return utos(n) + u; } // Like utos_unit(), but 2 digits fraction part is followed. template <std::unsigned_integral T> constexpr std::string utos_funit(T n) { char u; int b; if (n >= (1 << 30)) { u = 'G'; b = 30; } else if (n >= (1 << 20)) { u = 'M'; b = 20; } else if (n >= (1 << 10)) { u = 'K'; b = 10; } else { return utos(n); } return dtos(static_cast<double>(n) / (1 << b)) + u; } struct CompactHexFormatter { template <std::integral T, std::weakly_incrementable O> requires(std::indirectly_writable<O, char>) O operator()(T n, O result) { using result_type = std::iter_value_t<O>; if (n == 0) { *result++ = '0'; return result; } if constexpr (std::endian::native == std::endian::little) { auto end = reinterpret_cast<uint8_t *>(&n); auto p = end + sizeof(n); for (; p != end && *(p - 1) == 0; --p) ; // Workaround for bogus UBSAN error assert(p != end); if (*(p - 1) < 16) { *result++ = static_cast<result_type>(UPPER_XDIGITS[(*--p) & 0xF]); } for (; p != end; --p) { result = format_upper_hex_uint8(*(p - 1), result); } } else { auto p = reinterpret_cast<uint8_t *>(&n); auto end = p + sizeof(n); for (; p != end && *p == 0; ++p) ; if (*p < 16) { *result++ = static_cast<result_type>(UPPER_XDIGITS[(*p++) & 0xF]); } for (; p != end; ++p) { result = format_upper_hex_uint8(*p, result); } } return result; } }; template <std::integral T, std::weakly_incrementable O> requires(std::indirectly_writable<O, char>) O utox(T n, O result) { return CompactHexFormatter{}(std::move(n), std::move(result)); } void to_token68(std::string &base64str); std::string_view to_base64(BlockAllocator &balloc, std::string_view token68str); void show_candidates(const char *unkopt, const option *options); bool has_uri_field(const urlparse_url &u, urlparse_url_fields field); bool fieldeq(const char *uri1, const urlparse_url &u1, const char *uri2, const urlparse_url &u2, urlparse_url_fields field); bool fieldeq(const char *uri, const urlparse_url &u, urlparse_url_fields field, const char *t); bool fieldeq(const char *uri, const urlparse_url &u, urlparse_url_fields field, std::string_view t); std::string_view get_uri_field(const char *uri, const urlparse_url &u, urlparse_url_fields field); uint16_t get_default_port(const char *uri, const urlparse_url &u); bool porteq(const char *uri1, const urlparse_url &u1, const char *uri2, const urlparse_url &u2); bool numeric_host(const char *hostname); bool numeric_host(const char *hostname, int family); // Returns numeric address string of |addr|. If getnameinfo() is // failed, "unknown" is returned. std::string numeric_name(const struct sockaddr *sa, socklen_t salen); // Returns string representation of numeric address and port of // |addr|. If address family is AF_UNIX, this return path to UNIX // domain socket. Otherwise, the format is like <HOST>:<PORT>. For // IPv6 address, address is enclosed by square brackets ([]). std::string to_numeric_addr(const Address *addr); std::string to_numeric_addr(const struct sockaddr *sa, socklen_t salen); // Returns true if |port| is prohibited as a QUIC client port. bool quic_prohibited_port(uint16_t port); // Returns ASCII dump of |data| of length |len|. Only ASCII printable // characters are preserved. Other characters are replaced with ".". std::string ascii_dump(const uint8_t *data, size_t len); // Returns absolute path of executable path. If argc == 0 or |cwd| is // nullptr, this function returns nullptr. If argv[0] starts with // '/', this function returns argv[0]. Otherwise return cwd + "/" + // argv[0]. If non-null is returned, it is NULL-terminated string and // dynamically allocated by malloc. The caller is responsible to free // it. char *get_exec_path(size_t argc, char **const argv, const char *cwd); // Validates path so that it does not contain directory traversal // vector. Returns true if path is safe. The |path| must start with // "/" otherwise returns false. This function should be called after // percent-decode was performed. bool check_path(const std::string &path); // Returns the |tv| value as 64 bit integer using a microsecond as an // unit. int64_t to_time64(const timeval &tv); // Returns true if ALPN ID |proto| is supported HTTP/2 protocol // identifier. bool check_h2_is_selected(std::string_view proto); // Selects h2 protocol ALPN ID if one of supported h2 versions are // present in |in| of length inlen. Returns true if h2 version is // selected. bool select_h2(const unsigned char **out, unsigned char *outlen, const unsigned char *in, unsigned int inlen); // Selects protocol ALPN ID if one of identifiers contained in |protolist| is // present in |in| of length inlen. Returns true if identifier is // selected. bool select_protocol(const unsigned char **out, unsigned char *outlen, const unsigned char *in, unsigned int inlen, std::vector<std::string> proto_list); // Parses delimited strings in |s| and returns the array of substring, // delimited by |delim|. The any white spaces around substring are // treated as a part of substring. std::vector<std::string> parse_config_str_list(std::string_view s, char delim = ','); // Parses delimited strings in |s| and returns Substrings in |s| // delimited by |delim|. The any white spaces around substring are // treated as a part of substring if |delim| is not a white space. If // |s| is an empty string, this function returns an empty vector. std::vector<std::string_view> split_str(std::string_view s, char delim); // Behaves like split_str, but this variant allocates new string using // |balloc|. The each element of the returned container is a view // over the allocated string. std::vector<std::string_view> split_str(BlockAllocator &balloc, std::string_view s, char delim); // Behaves like split_str, but this variant splits at most |n| - 1 // times and returns at most |n| sub-strings. |n| must be greater // than zero. If |s| is an empty string, this function returns an // empty vector. std::vector<std::string_view> split_str(std::string_view s, char delim, size_t n); // Writes given time |tp| in Common Log format (e.g., // 03/Jul/2014:00:19:38 +0900) in buffer pointed by |out|. The buffer // must be at least 27 bytes, including terminal NULL byte. This // function returns std::string_view wrapping the buffer pointed by |out|, // and this string is terminated by NULL. std::string_view format_common_log(char *out, std::chrono::system_clock::time_point tp); #ifdef HAVE_STD_CHRONO_TIME_ZONE // Works like above but with a given time zone. std::string_view format_common_log(char *out, std::chrono::system_clock::time_point tp, const std::chrono::time_zone *tz); #endif // defined(HAVE_STD_CHRONO_TIME_ZONE) // Returns given time |tp| in ISO 8601 format (e.g., // 2014-11-15T12:58:24.741Z or 2014-11-15T12:58:24.741+09:00). std::string format_iso8601(std::chrono::system_clock::time_point tp); // Writes given time |tp| in ISO 8601 format (e.g., // 2014-11-15T12:58:24.741Z or 2014-11-15T12:58:24.741+09:00) in // buffer pointed by |out|. The buffer must be at least 30 bytes, // including terminal NULL byte. This function returns std::string_view // wrapping the buffer pointed by |out|, and this string is terminated // by NULL. std::string_view format_iso8601(char *out, std::chrono::system_clock::time_point tp); #ifdef HAVE_STD_CHRONO_TIME_ZONE // Works like above but with a given time zone. std::string_view format_iso8601(char *out, std::chrono::system_clock::time_point tp, const std::chrono::time_zone *tz); #endif // defined(HAVE_STD_CHRONO_TIME_ZONE) // Writes given time |tp| in ISO 8601 basic format (e.g., // 20141115T125824.741Z or 20141115T125824.741+0900) in buffer pointed // by |out|. The buffer must be at least 25 bytes, including terminal // NULL byte. This function returns std::string_view wrapping the buffer // pointed by |out|, and this string is terminated by NULL. std::string_view format_iso8601_basic(char *out, std::chrono::system_clock::time_point tp); #ifdef HAVE_STD_CHRONO_TIME_ZONE // Works like above but with a given time zone. std::string_view format_iso8601_basic(char *out, std::chrono::system_clock::time_point tp, const std::chrono::time_zone *tz); #endif // defined(HAVE_STD_CHRONO_TIME_ZONE) // Returns given time |tp| in HTTP Date format (e.g., Mon, 10 Oct 2016 // 10:25:58 GMT) std::string format_http_date(std::chrono::system_clock::time_point tp); // Writes given time |tp| in HTTP Date format (e.g., Mon, 10 Oct 2016 // 10:25:58 GMT) in buffer pointed by |out|. The buffer must be at // least 30 bytes, including terminal NULL byte. This function // returns std::string_view wrapping the buffer pointed by |out|, and this // string is terminated by NULL. std::string_view format_http_date(char *out, std::chrono::system_clock::time_point tp); // Return the system precision of the template parameter |Clock| as // a nanosecond value of type |Rep| template <typename Clock, typename Rep> Rep clock_precision() { std::chrono::duration<Rep, std::nano> duration = typename Clock::duration(1); return duration.count(); } #ifdef HAVE_LIBEV template <typename Duration = std::chrono::steady_clock::duration> Duration duration_from(ev_tstamp d) { return std::chrono::duration_cast<Duration>(std::chrono::duration<double>(d)); } template <typename Duration> ev_tstamp ev_tstamp_from(const Duration &d) { return std::chrono::duration<double>(d).count(); } #endif // defined(HAVE_LIBEV) int make_socket_closeonexec(int fd); int make_socket_nonblocking(int fd); int make_socket_nodelay(int fd); std::expected<int, Error> create_nonblock_socket(int family); std::expected<int, Error> create_nonblock_udp_socket(int family); std::expected<void, Error> bind_any_addr_udp(int fd, int family); bool check_socket_connected(int fd); // Returns the error code (errno) by inspecting SO_ERROR of given // |fd|. This function returns the error code if it succeeds. // Returning 0 means no error. If getsockopt fails, this function // returns errno of its call. int get_socket_error(int fd); // Returns true if |host| is IPv6 numeric address (e.g., ::1) bool ipv6_numeric_addr(const char *host); // Parses |s| as unsigned integer and returns the parsed integer. // Additionally, if |s| ends with 'k', 'm', 'g' and its upper case // characters, multiply the integer by 1024, 1024 * 1024 and 1024 * // 1024 respectively. The maximum value that this function can // recognize is std::numeric_limits<int64_t>::max(). std::expected<uint64_t, Error> parse_uint_with_unit(std::string_view s); // Parses |s| as unsigned integer and returns the parsed integer. The // maximum value that this function can recognize is // std::numeric_limits<int64_t>::max(). std::expected<uint64_t, Error> parse_uint(std::string_view s); // Parses |s| as unsigned integer and returns the parsed integer // casted to double. If |s| ends with "s", the parsed value's unit is // a second. If |s| ends with "ms", the unit is millisecond. // Similarly, it also supports 'm' and 'h' for minutes and hours // respectively. If none of them are given, the unit is second. std::expected<double, Error> parse_duration_with_unit(std::string_view s); // Returns string representation of time duration |t|. If t has // fractional part (at least more than or equal to 1e-3), |t| is // multiplied by 1000 and the unit "ms" is appended. Otherwise, |t| // is left as is and "s" is appended. std::string duration_str(double t); // Returns string representation of time duration |t|. It appends // unit after the formatting. The available units are s, ms and us. // The unit which is equal to or less than |t| is used and 2 // fractional digits follow. std::string format_duration(std::chrono::microseconds u); // Just like above, but this takes |t| as seconds. std::string format_duration(double t); // The maximum buffer size including terminal NULL to store the result // of make_hostport. inline constexpr size_t max_hostport = NI_MAXHOST + /* [] for IPv6 */ 2 + /* : */ 1 + /* port */ 5 + /* terminal NULL */ 1; // Just like make_http_hostport(), but doesn't treat 80 and 443 // specially. std::string_view make_hostport(BlockAllocator &balloc, std::string_view host, uint16_t port); template <std::weakly_incrementable O> requires(std::indirectly_writable<O, char>) std::string_view make_hostport(std::string_view host, uint16_t port, O result) { auto ipv6 = ipv6_numeric_addr(host.data()); auto p = result; if (ipv6) { *p++ = '['; } p = std::ranges::copy(host, p).out; if (ipv6) { *p++ = ']'; } *p++ = ':'; p = utos(port, p); *p = '\0'; return as_string_view(result, p); } // Creates "host:port" string using given |host| and |port|. If // |host| is numeric IPv6 address (e.g., ::1), it is enclosed by "[" // and "]". If |port| is 80 or 443, port part is omitted. std::string_view make_http_hostport(BlockAllocator &balloc, std::string_view host, uint16_t port); template <std::weakly_incrementable O> requires(std::indirectly_writable<O, char>) std::string_view make_http_hostport(std::string_view host, uint16_t port, O result) { if (port != 80 && port != 443) { return make_hostport(host, port, std::move(result)); } auto ipv6 = ipv6_numeric_addr(host.data()); auto p = result; if (ipv6) { *p++ = '['; } p = std::ranges::copy(host, p).out; if (ipv6) { *p++ = ']'; } *p = '\0'; return as_string_view(result, p); } // hexdump dumps |data| of length |datalen| in the format similar to // hexdump(1) with -C option. std::expected<void, Error> hexdump(FILE *out, const void *data, size_t datalen); // Copies 2 byte unsigned integer |n| in host byte order to |buf| in // network byte order. void put_uint16be(uint8_t *buf, uint16_t n); // Copies 4 byte unsigned integer |n| in host byte order to |buf| in // network byte order. void put_uint32be(uint8_t *buf, uint32_t n); // Retrieves 2 byte unsigned integer stored in |data| in network byte // order and returns it in host byte order. uint16_t get_uint16(const uint8_t *data); // Retrieves 4 byte unsigned integer stored in |data| in network byte // order and returns it in host byte order. uint32_t get_uint32(const uint8_t *data); // Retrieves 8 byte unsigned integer stored in |data| in network byte // order and returns it in host byte order. uint64_t get_uint64(const uint8_t *data); // Reads mime types file (see /etc/mime.types), and stores extension // -> MIME type map in |res|. std::expected<std::unordered_map<std::string, std::string>, Error> read_mime_types(const char *filename); // Fills random alpha and digit byte to the range [|first|, |last|). // Returns the one beyond the |last|. template <typename OutputIt, typename Generator> OutputIt random_alpha_digit(OutputIt first, OutputIt last, Generator &gen) { // If we use uint8_t instead char, gcc 6.2.0 complains by shouting // char-array initialized from wide string. static constexpr char s[] = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789"; std::uniform_int_distribution<> dis(0, 26 * 2 + 10 - 1); for (; first != last; ++first) { *first = static_cast<std::iter_value_t<OutputIt>>(s[dis(gen)]); } return first; } // Fills random bytes to the range [|first|, |last|). template <std::input_or_output_iterator O, typename Generator> void random_bytes(O first, O last, Generator &&gen) { std::uniform_int_distribution<uint8_t> dis; std::ranges::generate(std::move(first), std::move(last), [&dis, &gen] { return dis(gen); }); } // Fills the buffer pointed by |data| of length |destlen| with the // secure random bytes. void secure_random(uint8_t *dest, size_t destlen); // Shuffles the range [|first|, |last|] by calling swap function // |swap| for each pair. |swap| takes 2 iterators. If |swap| is // noop, no modification is made. template <std::random_access_iterator I, typename Generator, typename Swap> void shuffle(I first, I last, Generator &&gen, Swap swap) { auto len = std::ranges::distance(first, last); if (len < 2) { return; } using dist_type = std::uniform_int_distribution<decltype(len)>; using param_type = dist_type::param_type; dist_type d; for (decltype(len) i = 0; i < len - 1; ++i) { swap(first + i, first + d(gen, param_type(i, len - 1))); } } template <std::ranges::random_access_range R, typename Generator, typename Swap> requires(!std::is_array_v<std::remove_cvref_t<R>>) void shuffle(R &&r, Generator &&gen, Swap swap) { return shuffle(std::ranges::begin(r), std::ranges::end(r), std::forward<Generator>(gen), std::move(swap)); } // Returns x**y double int_pow(double x, size_t y); uint32_t hash32(std::string_view s); // Computes SHA-256 of |s|, and stores it in |buf|. std::expected<void, Error> sha256(std::span<uint8_t, 32> dest, std::span<const uint8_t> s); template <std::ranges::contiguous_range R> requires(std::ranges::sized_range<R> && !std::is_convertible_v<R, std::span<const uint8_t>>) std::expected<void, Error> sha256(std::span<uint8_t, 32> dest, R &&r) { return sha256(dest, as_uint8_span(std::span{r})); } // Computes SHA-1 of |s|, and stores it in |buf|. std::expected<void, Error> sha1(std::span<uint8_t, 20> dest, std::span<const uint8_t> s); template <std::ranges::contiguous_range R> requires(std::ranges::sized_range<R> && !std::is_convertible_v<R, std::span<const uint8_t>>) std::expected<void, Error> sha1(std::span<uint8_t, 20> dest, R &&r) { return sha1(dest, as_uint8_span(std::span{r})); } // Returns host from |hostport|. The returned string might not be // NULL-terminated. std::expected<std::string_view, Error> extract_host(std::string_view hostport); // split_hostport splits host and port in |hostport|. Unlike // extract_host, square brackets enclosing host name is stripped. If // port is not available, it returns empty string in the second // string. The returned string might not be NULL-terminated. std::expected<std::pair<std::string_view, std::string_view>, Error> split_hostport(std::string_view hostport); // Returns new std::mt19937 object. std::mt19937 make_mt19937(); // daemonize calls daemon(3). If __APPLE__ is defined, it implements // daemon() using fork(). std::expected<void, Error> daemonize(int nochdir, int noclose); // Returns |s| from which trailing white spaces (SPC or HTAB) are // removed. If any white spaces are removed, new string is allocated // by |balloc| and returned. Otherwise, the copy of |s| is returned // without allocation. std::string_view rstrip(BlockAllocator &balloc, std::string_view s); // contains returns true if |r| contains |value|. template <std::ranges::input_range R, typename T> requires(!std::is_array_v<std::remove_cvref_t<R>>) bool contains(R &&r, const T &value) { return std::ranges::find(r, value) != std::ranges::end(r); } // contains returns true if |value| is contained in a range [|first|, // |last|). template <std::input_iterator I, typename T> constexpr bool contains(I first, I last, const T &value) { return std::ranges::find(std::move(first), last, value) != last; } // stream_error returns true if |f| is in error state. EOF is not // considered as an error. bool stream_error(const std::ifstream &f); #ifdef ENABLE_HTTP3 std::expected<Address, Error> msghdr_get_local_addr(msghdr *msg, int family); uint8_t msghdr_get_ecn(msghdr *msg, int family); // msghdr_get_udp_gro returns UDP_GRO value from |msg|. If UDP_GRO is // not found, or UDP_GRO is not supported, this function returns 0. size_t msghdr_get_udp_gro(msghdr *msg); #endif // defined(ENABLE_HTTP3) } // namespace util } // namespace nghttp2 #endif // !defined(UTIL_H)