path: root/src/java_uescape.cc

#include "java_uescape.hh"

#include "u16.hh"
#include "u8.hh"

#include <algorithm>
#include <cassert>
#include <cstddef>
#include <cstdlib>
#include <cstring>
#include <expected>
#include <memory>
#include <optional>
#include <utility>

namespace {

const size_t kBufferSize = 1024;

template <typename T, typename Reader, typename Writer>
class UnicodeEscapeReader {
 public:
  UnicodeEscapeReader(std::unique_ptr<Reader> reader, T backslash, T u, T zero,
                      T nine, T lc_a, T lc_f, T uc_a, T uc_f)
      : reader_(std::move(reader)),
        backslash_(backslash),
        u_(u),
        zero_(zero),
        nine_(nine),
        lc_a_(lc_a),
        lc_f_(lc_f),
        uc_a_(uc_a),
        uc_f_(uc_f),
        buffer_(std::make_unique_for_overwrite<T[]>(kBufferSize)) {}

  [[nodiscard]] std::expected<size_t, io::ReadError> read(void* dst,
                                                          size_t max) {
    State state;
    bool eof;

    state.wstart = reinterpret_cast<T*>(dst);
    state.wptr = state.wstart;
    // NOLINTNEXTLINE(bugprone-sizeof-expression)
    state.wend = state.wstart + max / sizeof(T);

    if (fill_ == 0) {
      // Optimize for the case when there are no unicode escapes.
      auto ret = reader_->read(dst, max);
      if (!ret.has_value() || ret.value() == 0) {
        return ret;
      }

      eof = false;
      state.rstart = reinterpret_cast<T*>(dst);
      // NOLINTNEXTLINE(bugprone-sizeof-expression)
      state.rend = state.rstart + (ret.value() / sizeof(T));
      state.shared = true;
    } else {
      if (error_.has_value()) {
        return std::unexpected(error_.value());
      }

      auto ret = reader_->read(buffer_.get() + fill_,
                               (kBufferSize - fill_) * sizeof(T));
      if (!ret.has_value()) {
        return ret;
      }
      eof = ret.value() == 0;
      fill_ += ret.value() / sizeof(T);

      state.rstart = buffer_.get();
      state.rend = buffer_.get() + fill_;
      state.shared = false;
    }

    while (true) {
      auto* ptr = std::find(state.rstart, state.rend, backslash_);
      if (transfer(state, ptr) || ptr == state.rend)
        return finish(state);

      auto* const first_backslash = ptr;
      do {
        ptr++;
      } while (ptr < state.rend && *ptr == backslash_);

      if (ptr == state.rend && !eof)
        return finish(state);

      if (ptr == state.rend || *ptr != u_ || (ptr - first_backslash) % 2 == 0) {
        // Even number of backslashes, or no u
        if (transfer(state, ptr))
          return finish(state);
        continue;
      }

      // auto* const first_u = ptr;
      do {
        ptr++;
      } while (ptr < state.rend && *ptr == u_);

      if (state.rend - ptr < 4) {
        if (!eof)
          return finish(state);

        // If an eligible \ is followed by u, or more than one u, and the last u is not followed by four hexadecimal digits, then a compile-time error occurs.
        return std::unexpected(error(io::ReadError::InvalidData));
      }

      auto maybe_code = unhex4(ptr);
      if (!maybe_code.has_value()) {
        // If an eligible \ is followed by u, or more than one u, and the last u is not followed by four hexadecimal digits, then a compile-time error occurs.
        return std::unexpected(error(io::ReadError::InvalidData));
      }
      ptr += 4;
      uint32_t code = maybe_code.value();

      if (code >= 0xdc00 && code <= 0xdfff) {
        // Low surrogate before (or without?), replace
        code = 0xfffd;
      }

      if (code < 0xd800 || code > 0xdfff) {
        // Not a pair
        if (write_code(state, code, ptr))
          return finish(state);
        continue;
      }

      auto* const second_first_backslash = ptr;
      do {
        ptr++;
      } while (ptr < state.rend && *ptr == backslash_);

      if (ptr == state.rend && !eof)
        return finish(state);

      if (ptr == state.rend || *ptr != u_ ||
          (ptr - second_first_backslash) % 2 == 0) {
        // High surrogate not followed by an escape, write out replacement
        // and restart.
        if (write_code(state, 0xfffd, second_first_backslash))
          return finish(state);
        continue;
      }

      // auto* const first_u = ptr;
      do {
        ptr++;
      } while (ptr < state.rend && *ptr == u_);

      if (state.rend - ptr < 4) {
        if (!eof)
          return finish(state);

        // If an eligible \ is followed by u, or more than one u, and the last u is not followed by four hexadecimal digits, then a compile-time error occurs.
        return std::unexpected(error(io::ReadError::InvalidData));
      }

      maybe_code = unhex4(ptr);
      if (!maybe_code.has_value()) {
        // If an eligible \ is followed by u, or more than one u, and the last u is not followed by four hexadecimal digits, then a compile-time error occurs.
        return std::unexpected(error(io::ReadError::InvalidData));
      }
      ptr += 4;
      uint32_t low_code = maybe_code.value();

      if (low_code >= 0xdc00 && low_code <= 0xdfff) {
        // Pair
        code = 0x10000 + (((code - 0xd800) << 10) | (low_code - 0xdc00));
        if (write_code(state, code, ptr))
          return finish(state);
      } else {
        // High surrogate not followed by an low surrogate, write out
        // replacement and restart.
        if (write_code(state, 0xfffd, second_first_backslash))
          return finish(state);
      }
    }
  }

  [[nodiscard]] std::expected<size_t, io::ReadError> skip(size_t max) {
    auto tmp = std::make_unique_for_overwrite<T[]>(max);
    return read(tmp.get(), max);
  }

 private:
  struct State {
    T* rstart;
    T* rend;

    bool shared;

    T* wstart;
    T* wptr;
    T* wend;
  };

  bool transfer(State& state, T*& ptr) {
    if (state.shared) {
      state.rstart = ptr;
      state.wptr = ptr;
      return false;
    }
    size_t ravail = ptr - state.rstart;
    size_t wavail = state.wend - state.wptr;
    size_t avail = std::min(ravail, wavail);

    memcpy(state.wptr, state.rstart, avail);
    state.wptr += avail;
    fill_ -= avail;
    memmove(buffer_.get(), ptr, fill_ * sizeof(T));

    if (avail != ravail)
      return true;

    ptr -= ravail;
    state.rstart = buffer_.get();
    state.rend = buffer_.get() + fill_;
    return false;
  }

  std::expected<size_t, io::ReadError> finish(State const& state) {
    if (state.shared) {
      size_t avail = state.rend - state.rstart;
      if (fill_ + avail > kBufferSize)
        abort();  // NOLINT(misc-include-cleaner)
      memcpy(buffer_.get() + fill_, state.rstart, avail * sizeof(T));
      fill_ += avail;
    }

    // We can't return zero, that is treated as EOF, we need to read more.
    if (state.wptr == state.wstart) {
      if (fill_ > 0) {
        return read(state.wstart, (state.wend - state.wstart) * sizeof(T));
      }
    }

    return (state.wptr - state.wstart) * sizeof(T);
  }

  bool write_code(State& state, uint32_t code, T* ptr) {
    auto* const wptr = state.wptr;
    if (!writer_(state.wptr, state.wend, code))
      return true;

    if (state.shared) {
      // Remove the extra bytes (if any)
      auto* rstart = state.rstart + (state.wptr - wptr);
      assert(ptr >= rstart);
      memmove(rstart, ptr, (state.rend - ptr) * sizeof(T));
      state.rend -= (ptr - rstart);
      state.rstart = rstart;
    } else {
      // Just drop escape from buffer
      fill_ = state.rend - ptr;
      memmove(buffer_.get(), ptr, fill_ * sizeof(T));
      state.rstart = buffer_.get();
      state.rend = state.rstart + fill_;
    }
    return false;
  }

  io::ReadError error(io::ReadError err) {
    // If read() returns an error it should continue to do so.
    error_ = err;
    fill_ = 1;
    return err;
  }

  [[nodiscard]] std::optional<uint16_t> unhex4(T* ptr) const {
    auto a = unhex1(ptr[0]);
    auto b = unhex1(ptr[1]);
    auto c = unhex1(ptr[2]);
    auto d = unhex1(ptr[3]);
    if (a.has_value() && b.has_value() && c.has_value() && d.has_value())
      return (*a << 12) | (*b << 8) | (*c << 4) | *d;
    return std::nullopt;
  }

  [[nodiscard]] std::optional<uint16_t> unhex1(T c) const {
    if (c >= zero_ && c <= nine_)
      return c - zero_;
    if (c >= lc_a_ && c <= lc_f_)
      return 10 + (c - lc_a_);
    if (c >= uc_a_ && c <= uc_f_)
      return 10 + (c - uc_a_);
    return std::nullopt;
  }

  std::unique_ptr<Reader> reader_;
  Writer writer_;
  T const backslash_;
  T const u_;
  T const zero_;
  T const nine_;
  T const lc_a_;
  T const lc_f_;
  T const uc_a_;
  T const uc_f_;
  std::unique_ptr<T[]> buffer_;
  size_t fill_{0};
  std::optional<io::ReadError> error_;
};

}  // namespace

namespace u8::java {

namespace {

struct Writer {
  bool operator()(uint8_t*& start, uint8_t* const& end, uint32_t code) const {
    return u8::write(start, end, code);
  }
};

class ReaderImpl : public u8::Reader,
                   UnicodeEscapeReader<uint8_t, u8::Reader, Writer> {
 public:
  explicit ReaderImpl(std::unique_ptr<u8::Reader> reader)
      : UnicodeEscapeReader<uint8_t, u8::Reader, Writer>(
            std::move(reader), '\\', 'u', '0', '9', 'a', 'f', 'A', 'F') {}

  [[nodiscard]] std::expected<size_t, io::ReadError> read(void* dst,
                                                          size_t max) override {
    return UnicodeEscapeReader<uint8_t, u8::Reader, Writer>::read(dst, max);
  }

  [[nodiscard]] std::expected<size_t, io::ReadError> skip(size_t max) override {
    return UnicodeEscapeReader<uint8_t, u8::Reader, Writer>::skip(max);
  }
};

}  // namespace

[[nodiscard]]
std::unique_ptr<Reader> open(std::unique_ptr<io::Reader> reader,
                             u::ReaderConfig config) {
  return std::make_unique<ReaderImpl>(u8::open(std::move(reader), config));
}

}  // namespace u8::java

namespace u16::java {

namespace {

struct Writer {
  bool operator()(uint16_t*& start, uint16_t* const& end, uint32_t code) const {
    return u16::write(start, end, code);
  }
};

class ReaderImpl : public u16::Reader,
                   UnicodeEscapeReader<uint16_t, u16::Reader, Writer> {
 public:
  explicit ReaderImpl(std::unique_ptr<u16::Reader> reader)
      : UnicodeEscapeReader<uint16_t, u16::Reader, Writer>(
            std::move(reader), u'\\', u'u', u'0', u'9', u'a', u'f', u'A',
            u'F') {}

  [[nodiscard]] std::expected<size_t, io::ReadError> read(void* dst,
                                                          size_t max) override {
    return UnicodeEscapeReader<uint16_t, u16::Reader, Writer>::read(dst, max);
  }

  [[nodiscard]] std::expected<size_t, io::ReadError> skip(size_t max) override {
    return UnicodeEscapeReader<uint16_t, u16::Reader, Writer>::skip(max);
  }
};

}  // namespace

[[nodiscard]]
std::unique_ptr<Reader> open(std::unique_ptr<io::Reader> reader,
                             u::ReaderConfig config) {
  return std::make_unique<ReaderImpl>(u16::open(std::move(reader), config));
}

}  // namespace u16::java