include/multipart_parser.hpp

#ifndef THIRD_PARTY_GBMCWEB_INCLUDE_MULTIPART_PARSER_H_
#define THIRD_PARTY_GBMCWEB_INCLUDE_MULTIPART_PARSER_H_

#include <cstdint>
#include <string>
#include <cstddef>
#include <string_view>
#include <vector>
#include <algorithm>
#include <array>

#include "boost/beast/http/fields.hpp"  // NOLINT
#include "http_request.hpp"

enum class ParserError : uint8_t {
  PARSER_SUCCESS,
  ERROR_BOUNDARY_FORMAT,
  ERROR_BOUNDARY_CR,
  ERROR_BOUNDARY_LF,
  ERROR_BOUNDARY_DATA,
  ERROR_EMPTY_HEADER,
  ERROR_HEADER_NAME,
  ERROR_HEADER_VALUE,
  ERROR_HEADER_ENDING,
  ERROR_UNEXPECTED_END_OF_HEADER,
  ERROR_UNEXPECTED_END_OF_INPUT,
  ERROR_OUT_OF_RANGE
};

enum class State : uint8_t {
  START,
  START_BOUNDARY,
  HEADER_FIELD_START,
  HEADER_FIELD,
  HEADER_VALUE_START,
  HEADER_VALUE,
  HEADER_VALUE_ALMOST_DONE,
  HEADERS_ALMOST_DONE,
  PART_DATA_START,
  PART_DATA,
  END
};

enum class Boundary : std::uint8_t {
  NON_BOUNDARY,
  PART_BOUNDARY,
  END_BOUNDARY,
};

struct FormPart {
  boost::beast::http::fields fields;
  std::string content;
};

class MultipartParser {
 public:
  MultipartParser() = default;

  [[nodiscard]] ParserError parse(const crow::Request& req) {
    std::string_view contentType = req.getHeaderValue("content-type");

    const std::string boundaryFormat = "multipart/form-data; boundary=";
    if (!contentType.starts_with(boundaryFormat)) {
      return ParserError::ERROR_BOUNDARY_FORMAT;
    }

    std::string_view ctBoundary = contentType.substr(boundaryFormat.size());

    boundary = "\r\n--";
    boundary += ctBoundary;
    indexBoundary();
    lookbehind.resize(boundary.size() + 8);
    state = State::START;

    const char* buffer = req.body().data();
    size_t len = req.body().size();
    char cl = 0;

    for (size_t i = 0; i < len; i++) {
      // NOLINTNEXTLINE(cppcoreguidelines-pro-bounds-pointer-arithmetic)
      char c = buffer[i];
      switch (state) {
        case State::START:
          index = 0;
          state = State::START_BOUNDARY;
          [[fallthrough]];
        case State::START_BOUNDARY:
          if (index == boundary.size() - 2) {
            if (c != cr) {
              return ParserError::ERROR_BOUNDARY_CR;
            }
            index++;
            break;
          } else if (index - 1 == boundary.size() - 2) {
            if (c != lf) {
              return ParserError::ERROR_BOUNDARY_LF;
            }
            index = 0;
            mime_fields.push_back({});
            state = State::HEADER_FIELD_START;
            break;
          }
          if (c != boundary[index + 2]) {
            return ParserError::ERROR_BOUNDARY_DATA;
          }
          index++;
          break;
        case State::HEADER_FIELD_START:
          currentHeaderName.resize(0);
          state = State::HEADER_FIELD;
          headerFieldMark = i;
          index = 0;
          [[fallthrough]];
        case State::HEADER_FIELD:
          if (c == cr) {
            headerFieldMark = 0;
            state = State::HEADERS_ALMOST_DONE;
            break;
          }

          index++;
          if (c == hyphen) {
            break;
          }

          if (c == colon) {
            if (index == 1) {
              return ParserError::ERROR_EMPTY_HEADER;
            }

            // NOLINTNEXTLINE(cppcoreguidelines-pro-bounds-pointer-arithmetic)
            currentHeaderName.append(buffer + headerFieldMark,
                                     i - headerFieldMark);
            state = State::HEADER_VALUE_START;
            break;
          }
          cl = lower(c);
          if (cl < 'a' || cl > 'z') {
            return ParserError::ERROR_HEADER_NAME;
          }
          break;
        case State::HEADER_VALUE_START:
          if (c == space) {
            break;
          }
          headerValueMark = i;
          state = State::HEADER_VALUE;
          [[fallthrough]];
        case State::HEADER_VALUE:
          if (c == cr) {
            // NOLINTNEXTLINE(cppcoreguidelines-pro-bounds-pointer-arithmetic)
            std::string_view value(buffer + headerValueMark,
                                   i - headerValueMark);
            mime_fields.rbegin()->fields.set(currentHeaderName, value);
            state = State::HEADER_VALUE_ALMOST_DONE;
          }
          break;
        case State::HEADER_VALUE_ALMOST_DONE:
          if (c != lf) {
            return ParserError::ERROR_HEADER_VALUE;
          }
          state = State::HEADER_FIELD_START;
          break;
        case State::HEADERS_ALMOST_DONE:
          if (c != lf) {
            return ParserError::ERROR_HEADER_ENDING;
          }
          if (index > 0) {
            return ParserError::ERROR_UNEXPECTED_END_OF_HEADER;
          }
          state = State::PART_DATA_START;
          break;
        case State::PART_DATA_START:
          state = State::PART_DATA;
          partDataMark = i;
          [[fallthrough]];
        case State::PART_DATA: {
          if (index == 0) {
            skipNonBoundary(buffer, len, boundary.size() - 1, i);

            // NOLINTNEXTLINE(cppcoreguidelines-pro-bounds-pointer-arithmetic)
            c = buffer[i];
          }
          const ParserError ec = processPartData(buffer, i, c);
          if (ec != ParserError::PARSER_SUCCESS) {
            return ec;
          }
          break;
        }
        case State::END:
          break;
      }
    }

    if (state != State::END) {
      return ParserError::ERROR_UNEXPECTED_END_OF_INPUT;
    }

    return ParserError::PARSER_SUCCESS;
  }
  std::vector<FormPart> mime_fields;
  std::string boundary;

 private:
  void indexBoundary() {
    std::fill(boundaryIndex.begin(), boundaryIndex.end(), 0);
    for (const char current : boundary) {
      boundaryIndex[static_cast<unsigned char>(current)] = true;
    }
  }

  static char lower(char c) { return static_cast<char>(c | 0x20); }

  inline bool isBoundaryChar(char c) const {
    return boundaryIndex[static_cast<unsigned char>(c)];
  }

  void skipNonBoundary(const char* buffer, size_t len, size_t boundaryEnd,
                       size_t& i) {
    // boyer-moore derived algorithm to safely skip non-boundary data
    while (i + boundary.size() <= len) {
      // NOLINTNEXTLINE(cppcoreguidelines-pro-bounds-pointer-arithmetic)
      if (isBoundaryChar(buffer[i + boundaryEnd])) {
        break;
      }
      i += boundary.size();
    }
  }

  ParserError processPartData(const char* buffer, size_t& i, char c) {
    size_t prevIndex = index;

    if (index < boundary.size()) {
      if (boundary[index] == c) {
        if (index == 0) {
          // NOLINTNEXTLINE(cppcoreguidelines-pro-bounds-pointer-arithmetic)
          const char* start = buffer + partDataMark;
          size_t size = i - partDataMark;
          mime_fields.rbegin()->content += std::string_view(start, size);
        }
        index++;
      } else {
        index = 0;
      }
    } else if (index == boundary.size()) {
      index++;
      if (c == cr) {
        // cr = part boundary
        flags = Boundary::PART_BOUNDARY;
      } else if (c == hyphen) {
        // hyphen = end boundary
        flags = Boundary::END_BOUNDARY;
      } else {
        index = 0;
      }
    } else {
      if (flags == Boundary::PART_BOUNDARY) {
        index = 0;
        if (c == lf) {
          // unset the PART_BOUNDARY flag
          flags = Boundary::NON_BOUNDARY;
          mime_fields.push_back({});
          state = State::HEADER_FIELD_START;
          return ParserError::PARSER_SUCCESS;
        }
      }
      if (flags == Boundary::END_BOUNDARY) {
        if (c == hyphen) {
          state = State::END;
        } else {
          flags = Boundary::NON_BOUNDARY;
          index = 0;
        }
      }
    }

    if (index > 0) {
      if ((index - 1) >= lookbehind.size()) {
        // Should never happen, but when it does it won't cause crash
        return ParserError::ERROR_OUT_OF_RANGE;
      }
      lookbehind[index - 1] = c;
    } else if (prevIndex > 0) {
      // if our boundary turned out to be rubbish, the captured
      // lookbehind belongs to partData

      mime_fields.rbegin()->content += lookbehind.substr(0, prevIndex);
      partDataMark = i;

      // reconsider the current character even so it interrupted
      // the sequence it could be the beginning of a new sequence
      i--;
    }
    return ParserError::PARSER_SUCCESS;
  }

  std::string currentHeaderName;
  std::string currentHeaderValue;

  static constexpr char cr = '\r';
  static constexpr char lf = '\n';
  static constexpr char space = ' ';
  static constexpr char hyphen = '-';
  static constexpr char colon = ':';

  std::array<bool, 256> boundaryIndex{};
  std::string lookbehind;
  State state{State::START};
  Boundary flags{Boundary::NON_BOUNDARY};
  size_t index = 0;
  size_t partDataMark = 0;
  size_t headerFieldMark = 0;
  size_t headerValueMark = 0;
};

#endif  // THIRD_PARTY_GBMCWEB_INCLUDE_MULTIPART_PARSER_H_