Austin Schuh | 36244a1 | 2019-09-21 17:52:38 -0700 | [diff] [blame^] | 1 | // Copyright 2018 The Abseil Authors. |
| 2 | // |
| 3 | // Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | // you may not use this file except in compliance with the License. |
| 5 | // You may obtain a copy of the License at |
| 6 | // |
| 7 | // https://www.apache.org/licenses/LICENSE-2.0 |
| 8 | // |
| 9 | // Unless required by applicable law or agreed to in writing, software |
| 10 | // distributed under the License is distributed on an "AS IS" BASIS, |
| 11 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 12 | // See the License for the specific language governing permissions and |
| 13 | // limitations under the License. |
| 14 | |
| 15 | #ifndef ABSL_STRINGS_INTERNAL_CHARCONV_PARSE_H_ |
| 16 | #define ABSL_STRINGS_INTERNAL_CHARCONV_PARSE_H_ |
| 17 | |
| 18 | #include <cstdint> |
| 19 | |
| 20 | #include "absl/strings/charconv.h" |
| 21 | |
| 22 | namespace absl { |
| 23 | namespace strings_internal { |
| 24 | |
| 25 | // Enum indicating whether a parsed float is a number or special value. |
| 26 | enum class FloatType { kNumber, kInfinity, kNan }; |
| 27 | |
| 28 | // The decomposed parts of a parsed `float` or `double`. |
| 29 | struct ParsedFloat { |
| 30 | // Representation of the parsed mantissa, with the decimal point adjusted to |
| 31 | // make it an integer. |
| 32 | // |
| 33 | // During decimal scanning, this contains 19 significant digits worth of |
| 34 | // mantissa value. If digits beyond this point are found, they |
| 35 | // are truncated, and if any of these dropped digits are nonzero, then |
| 36 | // `mantissa` is inexact, and the full mantissa is stored in [subrange_begin, |
| 37 | // subrange_end). |
| 38 | // |
| 39 | // During hexadecimal scanning, this contains 15 significant hex digits worth |
| 40 | // of mantissa value. Digits beyond this point are sticky -- they are |
| 41 | // truncated, but if any dropped digits are nonzero, the low bit of mantissa |
| 42 | // will be set. (This allows for precise rounding, and avoids the need |
| 43 | // to store the full mantissa in [subrange_begin, subrange_end).) |
| 44 | uint64_t mantissa = 0; |
| 45 | |
| 46 | // Floating point expontent. This reflects any decimal point adjustments and |
| 47 | // any truncated digits from the mantissa. The absolute value of the parsed |
| 48 | // number is represented by mantissa * (base ** exponent), where base==10 for |
| 49 | // decimal floats, and base==2 for hexadecimal floats. |
| 50 | int exponent = 0; |
| 51 | |
| 52 | // The literal exponent value scanned from the input, or 0 if none was |
| 53 | // present. This does not reflect any adjustments applied to mantissa. |
| 54 | int literal_exponent = 0; |
| 55 | |
| 56 | // The type of number scanned. |
| 57 | FloatType type = FloatType::kNumber; |
| 58 | |
| 59 | // When non-null, [subrange_begin, subrange_end) marks a range of characters |
| 60 | // that require further processing. The meaning is dependent on float type. |
| 61 | // If type == kNumber and this is set, this is a "wide input": the input |
| 62 | // mantissa contained more than 19 digits. The range contains the full |
| 63 | // mantissa. It plus `literal_exponent` need to be examined to find the best |
| 64 | // floating point match. |
| 65 | // If type == kNan and this is set, the range marks the contents of a |
| 66 | // matched parenthesized character region after the NaN. |
| 67 | const char* subrange_begin = nullptr; |
| 68 | const char* subrange_end = nullptr; |
| 69 | |
| 70 | // One-past-the-end of the successfully parsed region, or nullptr if no |
| 71 | // matching pattern was found. |
| 72 | const char* end = nullptr; |
| 73 | }; |
| 74 | |
| 75 | // Read the floating point number in the provided range, and populate |
| 76 | // ParsedFloat accordingly. |
| 77 | // |
| 78 | // format_flags is a bitmask value specifying what patterns this API will match. |
| 79 | // `scientific` and `fixed` are honored per std::from_chars rules |
| 80 | // ([utility.from.chars], C++17): if exactly one of these bits is set, then an |
| 81 | // exponent is required, or dislallowed, respectively. |
| 82 | // |
| 83 | // Template parameter `base` must be either 10 or 16. For base 16, a "0x" is |
| 84 | // *not* consumed. The `hex` bit from format_flags is ignored by ParseFloat. |
| 85 | template <int base> |
| 86 | ParsedFloat ParseFloat(const char* begin, const char* end, |
| 87 | absl::chars_format format_flags); |
| 88 | |
| 89 | extern template ParsedFloat ParseFloat<10>(const char* begin, const char* end, |
| 90 | absl::chars_format format_flags); |
| 91 | extern template ParsedFloat ParseFloat<16>(const char* begin, const char* end, |
| 92 | absl::chars_format format_flags); |
| 93 | |
| 94 | } // namespace strings_internal |
| 95 | } // namespace absl |
| 96 | #endif // ABSL_STRINGS_INTERNAL_CHARCONV_PARSE_H_ |