csv.h

// Copyright: (2012-2015) Ben Strasser <code@ben-strasser.net>
// License: BSD-3
//
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are met:
//
// 1. Redistributions of source code must retain the above copyright notice,
//    this list of conditions and the following disclaimer.
//
// 2. Redistributions in binary form must reproduce the above copyright notice,
//    this list of conditions and the following disclaimer in the documentation
//    and/or other materials provided with the distribution.
//
// 3. Neither the name of the copyright holder nor the names of its contributors
//    may be used to endorse or promote products derived from this software
//    without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
// ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
// LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
// CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
// SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
// INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
// CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
// ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
// POSSIBILITY OF SUCH DAMAGE.

#ifndef CSV_H
#define CSV_H

#include <algorithm>
#include <cstdio>
#include <cstring>
#include <exception>
#include <string>
#include <utility>
#include <vector>
#ifndef CSV_IO_NO_THREAD
#include <condition_variable>
#include <mutex>
#include <thread>
#endif
#include <cassert>
#include <cerrno>
#include <istream>
#include <limits>
#include <memory>

namespace io {
////////////////////////////////////////////////////////////////////////////
//                                 LineReader                             //
////////////////////////////////////////////////////////////////////////////

namespace error {
struct base : std::exception {
  virtual void format_error_message() const = 0;

  const char *what() const noexcept override {
    format_error_message();
    return error_message_buffer;
  }

  mutable char error_message_buffer[2048];
};

// this only affects the file name in the error message
const int max_file_name_length = 1024;

struct with_file_name {
  with_file_name() { std::memset(file_name, 0, sizeof(file_name)); }

  void set_file_name(const char *file_name) {
    if (file_name != nullptr) {
      // This call to strncpy has parenthesis around it
      // to silence the GCC -Wstringop-truncation warning
      (strncpy(this->file_name, file_name, sizeof(this->file_name)));
      this->file_name[sizeof(this->file_name) - 1] = '\0';
    } else {
      this->file_name[0] = '\0';
    }
  }

  char file_name[max_file_name_length + 1];
};

struct with_file_line {
  with_file_line() { file_line = -1; }

  void set_file_line(int file_line) { this->file_line = file_line; }

  int file_line;
};

struct with_errno {
  with_errno() { errno_value = 0; }

  void set_errno(int errno_value) { this->errno_value = errno_value; }

  int errno_value;
};

struct can_not_open_file : base, with_file_name, with_errno {
  void format_error_message() const override {
    if (errno_value != 0)
      std::snprintf(error_message_buffer, sizeof(error_message_buffer),
                    "Can not open file \"%s\" because \"%s\".", file_name,
                    std::strerror(errno_value));
    else
      std::snprintf(error_message_buffer, sizeof(error_message_buffer),
                    "Can not open file \"%s\".", file_name);
  }
};

struct line_length_limit_exceeded : base, with_file_name, with_file_line {
  void format_error_message() const override {
    std::snprintf(
        error_message_buffer, sizeof(error_message_buffer),
        "Line number %d in file \"%s\" exceeds the maximum length of 2^24-1.",
        file_line, file_name);
  }
};
} // namespace error

class ByteSourceBase {
public:
  virtual int read(char *buffer, int size) = 0;
  virtual ~ByteSourceBase() {}
};

namespace detail {

class OwningStdIOByteSourceBase : public ByteSourceBase {
public:
  explicit OwningStdIOByteSourceBase(FILE *file) : file(file) {
    // Tell the std library that we want to do the buffering ourself.
    std::setvbuf(file, 0, _IONBF, 0);
  }

  int read(char *buffer, int size) { return std::fread(buffer, 1, size, file); }

  ~OwningStdIOByteSourceBase() { std::fclose(file); }

private:
  FILE *file;
};

class NonOwningIStreamByteSource : public ByteSourceBase {
public:
  explicit NonOwningIStreamByteSource(std::istream &in) : in(in) {}

  int read(char *buffer, int size) {
    in.read(buffer, size);
    return in.gcount();
  }

  ~NonOwningIStreamByteSource() {}

private:
  std::istream &in;
};

class NonOwningStringByteSource : public ByteSourceBase {
public:
  NonOwningStringByteSource(const char *str, long long size)
      : str(str), remaining_byte_count(size) {}

  int read(char *buffer, int desired_byte_count) {
    int to_copy_byte_count = desired_byte_count;
    if (remaining_byte_count < to_copy_byte_count)
      to_copy_byte_count = remaining_byte_count;
    std::memcpy(buffer, str, to_copy_byte_count);
    remaining_byte_count -= to_copy_byte_count;
    str += to_copy_byte_count;
    return to_copy_byte_count;
  }

  ~NonOwningStringByteSource() {}

private:
  const char *str;
  long long remaining_byte_count;
};

#ifndef CSV_IO_NO_THREAD
class AsynchronousReader {
public:
  void init(std::unique_ptr<ByteSourceBase> arg_byte_source) {
    std::unique_lock<std::mutex> guard(lock);
    byte_source = std::move(arg_byte_source);
    desired_byte_count = -1;
    termination_requested = false;
    worker = std::thread([&] {
      std::unique_lock<std::mutex> guard(lock);
      try {
        for (;;) {
          read_requested_condition.wait(guard, [&] {
            return desired_byte_count != -1 || termination_requested;
          });
          if (termination_requested)
            return;

          read_byte_count = byte_source->read(buffer, desired_byte_count);
          desired_byte_count = -1;
          if (read_byte_count == 0)
            break;
          read_finished_condition.notify_one();
        }
      } catch (...) {
        read_error = std::current_exception();
      }
      read_finished_condition.notify_one();
    });
  }

  bool is_valid() const { return byte_source != nullptr; }

  void start_read(char *arg_buffer, int arg_desired_byte_count) {
    std::unique_lock<std::mutex> guard(lock);
    buffer = arg_buffer;
    desired_byte_count = arg_desired_byte_count;
    read_byte_count = -1;
    read_requested_condition.notify_one();
  }

  int finish_read() {
    std::unique_lock<std::mutex> guard(lock);
    read_finished_condition.wait(
        guard, [&] { return read_byte_count != -1 || read_error; });
    if (read_error)
      std::rethrow_exception(read_error);
    else
      return read_byte_count;
  }

  ~AsynchronousReader() {
    if (byte_source != nullptr) {
      {
        std::unique_lock<std::mutex> guard(lock);
        termination_requested = true;
      }
      read_requested_condition.notify_one();
      worker.join();
    }
  }

private:
  std::unique_ptr<ByteSourceBase> byte_source;

  std::thread worker;

  bool termination_requested;
  std::exception_ptr read_error;
  char *buffer;
  int desired_byte_count;
  int read_byte_count;

  std::mutex lock;
  std::condition_variable read_finished_condition;
  std::condition_variable read_requested_condition;
};
#endif

class SynchronousReader {
public:
  void init(std::unique_ptr<ByteSourceBase> arg_byte_source) {
    byte_source = std::move(arg_byte_source);
  }

  bool is_valid() const { return byte_source != nullptr; }

  void start_read(char *arg_buffer, int arg_desired_byte_count) {
    buffer = arg_buffer;
    desired_byte_count = arg_desired_byte_count;
  }

  int finish_read() { return byte_source->read(buffer, desired_byte_count); }

private:
  std::unique_ptr<ByteSourceBase> byte_source;
  char *buffer;
  int desired_byte_count;
};
} // namespace detail

class LineReader {
private:
  static const int block_len = 1 << 20;
  std::unique_ptr<char[]> buffer; // must be constructed before (and thus
                                  // destructed after) the reader!
#ifdef CSV_IO_NO_THREAD
  detail::SynchronousReader reader;
#else
  detail::AsynchronousReader reader;
#endif
  int data_begin;
  int data_end;

  char file_name[error::max_file_name_length + 1];
  unsigned file_line;

  static std::unique_ptr<ByteSourceBase> open_file(const char *file_name) {
    // We open the file in binary mode as it makes no difference under *nix
    // and under Windows we handle \r\n newlines ourself.
    FILE *file = std::fopen(file_name, "rb");
    if (file == 0) {
      int x = errno; // store errno as soon as possible, doing it after
                     // constructor call can fail.
      error::can_not_open_file err;
      err.set_errno(x);
      err.set_file_name(file_name);
      throw err;
    }
    return std::unique_ptr<ByteSourceBase>(
        new detail::OwningStdIOByteSourceBase(file));
  }

  void init(std::unique_ptr<ByteSourceBase> byte_source) {
    file_line = 0;

    buffer = std::unique_ptr<char[]>(new char[3 * block_len]);
    data_begin = 0;
    data_end = byte_source->read(buffer.get(), 2 * block_len);

    // Ignore UTF-8 BOM
    if (data_end >= 3 && buffer[0] == '\xEF' && buffer[1] == '\xBB' &&
        buffer[2] == '\xBF')
      data_begin = 3;

    if (data_end == 2 * block_len) {
      reader.init(std::move(byte_source));
      reader.start_read(buffer.get() + 2 * block_len, block_len);
    }
  }

public:
  LineReader() = delete;
  LineReader(const LineReader &) = delete;
  LineReader &operator=(const LineReader &) = delete;

  explicit LineReader(const char *file_name) {
    set_file_name(file_name);
    init(open_file(file_name));
  }

  explicit LineReader(const std::string &file_name) {
    set_file_name(file_name.c_str());
    init(open_file(file_name.c_str()));
  }

  LineReader(const char *file_name,
             std::unique_ptr<ByteSourceBase> byte_source) {
    set_file_name(file_name);
    init(std::move(byte_source));
  }

  LineReader(const std::string &file_name,
             std::unique_ptr<ByteSourceBase> byte_source) {
    set_file_name(file_name.c_str());
    init(std::move(byte_source));
  }

  LineReader(const char *file_name, const char *data_begin,
             const char *data_end) {
    set_file_name(file_name);
    init(std::unique_ptr<ByteSourceBase>(new detail::NonOwningStringByteSource(
        data_begin, data_end - data_begin)));
  }

  LineReader(const std::string &file_name, const char *data_begin,
             const char *data_end) {
    set_file_name(file_name.c_str());
    init(std::unique_ptr<ByteSourceBase>(new detail::NonOwningStringByteSource(
        data_begin, data_end - data_begin)));
  }

  LineReader(const char *file_name, FILE *file) {
    set_file_name(file_name);
    init(std::unique_ptr<ByteSourceBase>(
        new detail::OwningStdIOByteSourceBase(file)));
  }

  LineReader(const std::string &file_name, FILE *file) {
    set_file_name(file_name.c_str());
    init(std::unique_ptr<ByteSourceBase>(
        new detail::OwningStdIOByteSourceBase(file)));
  }

  LineReader(const char *file_name, std::istream &in) {
    set_file_name(file_name);
    init(std::unique_ptr<ByteSourceBase>(
        new detail::NonOwningIStreamByteSource(in)));
  }

  LineReader(const std::string &file_name, std::istream &in) {
    set_file_name(file_name.c_str());
    init(std::unique_ptr<ByteSourceBase>(
        new detail::NonOwningIStreamByteSource(in)));
  }

  void set_file_name(const std::string &file_name) {
    set_file_name(file_name.c_str());
  }

  void set_file_name(const char *file_name) {
    if (file_name != nullptr) {
      strncpy(this->file_name, file_name, sizeof(this->file_name) - 1);
      this->file_name[sizeof(this->file_name) - 1] = '\0';
    } else {
      this->file_name[0] = '\0';
    }
  }

  const char *get_truncated_file_name() const { return file_name; }

  void set_file_line(unsigned file_line) { this->file_line = file_line; }

  unsigned get_file_line() const { return file_line; }

  char *next_line() {
    if (data_begin == data_end)
      return nullptr;

    ++file_line;

    assert(data_begin < data_end);
    assert(data_end <= block_len * 2);

    if (data_begin >= block_len) {
      std::memcpy(buffer.get(), buffer.get() + block_len, block_len);
      data_begin -= block_len;
      data_end -= block_len;
      if (reader.is_valid()) {
        data_end += reader.finish_read();
        std::memcpy(buffer.get() + block_len, buffer.get() + 2 * block_len,
                    block_len);
        reader.start_read(buffer.get() + 2 * block_len, block_len);
      }
    }

    int line_end = data_begin;
    while (line_end != data_end && buffer[line_end] != '\n') {
      ++line_end;
    }

    if (line_end - data_begin + 1 > block_len) {
      error::line_length_limit_exceeded err;
      err.set_file_name(file_name);
      err.set_file_line(file_line);
      throw err;
    }

    if (line_end != data_end && buffer[line_end] == '\n') {
      buffer[line_end] = '\0';
    } else {
      // some files are missing the newline at the end of the
      // last line
      ++data_end;
      buffer[line_end] = '\0';
    }

    // handle windows \r\n-line breaks
    if (line_end != data_begin && buffer[line_end - 1] == '\r')
      buffer[line_end - 1] = '\0';

    char *ret = buffer.get() + data_begin;
    data_begin = line_end + 1;
    return ret;
  }
};

////////////////////////////////////////////////////////////////////////////
//                                 CSV                                    //
////////////////////////////////////////////////////////////////////////////

namespace error {
const int max_column_name_length = 63;
struct with_column_name {
  with_column_name() {
    std::memset(column_name, 0, max_column_name_length + 1);
  }

  void set_column_name(const char *column_name) {
    if (column_name != nullptr) {
      std::strncpy(this->column_name, column_name, max_column_name_length);
      this->column_name[max_column_name_length] = '\0';
    } else {
      this->column_name[0] = '\0';
    }
  }

  char column_name[max_column_name_length + 1];
};

const int max_column_content_length = 63;

struct with_column_content {
  with_column_content() {
    std::memset(column_content, 0, max_column_content_length + 1);
  }

  void set_column_content(const char *column_content) {
    if (column_content != nullptr) {
      std::strncpy(this->column_content, column_content,
                   max_column_content_length);
      this->column_content[max_column_content_length] = '\0';
    } else {
      this->column_content[0] = '\0';
    }
  }

  char column_content[max_column_content_length + 1];
};

struct extra_column_in_header : base, with_file_name, with_column_name {
  void format_error_message() const override {
    std::snprintf(error_message_buffer, sizeof(error_message_buffer),
                  R"(Extra column "%s" in header of file "%s".)", column_name,
                  file_name);
  }
};

struct missing_column_in_header : base, with_file_name, with_column_name {
  void format_error_message() const override {
    std::snprintf(error_message_buffer, sizeof(error_message_buffer),
                  R"(Missing column "%s" in header of file "%s".)", column_name,
                  file_name);
  }
};

struct duplicated_column_in_header : base, with_file_name, with_column_name {
  void format_error_message() const override {
    std::snprintf(error_message_buffer, sizeof(error_message_buffer),
                  R"(Duplicated column "%s" in header of file "%s".)",
                  column_name, file_name);
  }
};

struct header_missing : base, with_file_name {
  void format_error_message() const override {
    std::snprintf(error_message_buffer, sizeof(error_message_buffer),
                  "Header missing in file \"%s\".", file_name);
  }
};

struct too_few_columns : base, with_file_name, with_file_line {
  void format_error_message() const override {
    std::snprintf(error_message_buffer, sizeof(error_message_buffer),
                  "Too few columns in line %d in file \"%s\".", file_line,
                  file_name);
  }
};

struct too_many_columns : base, with_file_name, with_file_line {
  void format_error_message() const override {
    std::snprintf(error_message_buffer, sizeof(error_message_buffer),
                  "Too many columns in line %d in file \"%s\".", file_line,
                  file_name);
  }
};

struct escaped_string_not_closed : base, with_file_name, with_file_line {
  void format_error_message() const override {
    std::snprintf(error_message_buffer, sizeof(error_message_buffer),
                  "Escaped string was not closed in line %d in file \"%s\".",
                  file_line, file_name);
  }
};

struct integer_must_be_positive : base,
                                  with_file_name,
                                  with_file_line,
                                  with_column_name,
                                  with_column_content {
  void format_error_message() const override {
    std::snprintf(
        error_message_buffer, sizeof(error_message_buffer),
        R"(The integer "%s" must be positive or 0 in column "%s" in file "%s" in line "%d".)",
        column_content, column_name, file_name, file_line);
  }
};

struct no_digit : base,
                  with_file_name,
                  with_file_line,
                  with_column_name,
                  with_column_content {
  void format_error_message() const override {
    std::snprintf(
        error_message_buffer, sizeof(error_message_buffer),
        R"(The integer "%s" contains an invalid digit in column "%s" in file "%s" in line "%d".)",
        column_content, column_name, file_name, file_line);
  }
};

struct integer_overflow : base,
                          with_file_name,
                          with_file_line,
                          with_column_name,
                          with_column_content {
  void format_error_message() const override {
    std::snprintf(
        error_message_buffer, sizeof(error_message_buffer),
        R"(The integer "%s" overflows in column "%s" in file "%s" in line "%d".)",
        column_content, column_name, file_name, file_line);
  }
};

struct integer_underflow : base,
                           with_file_name,
                           with_file_line,
                           with_column_name,
                           with_column_content {
  void format_error_message() const override {
    std::snprintf(
        error_message_buffer, sizeof(error_message_buffer),
        R"(The integer "%s" underflows in column "%s" in file "%s" in line "%d".)",
        column_content, column_name, file_name, file_line);
  }
};

struct invalid_single_character : base,
                                  with_file_name,
                                  with_file_line,
                                  with_column_name,
                                  with_column_content {
  void format_error_message() const override {
    std::snprintf(
        error_message_buffer, sizeof(error_message_buffer),
        R"(The content "%s" of column "%s" in file "%s" in line "%d" is not a single character.)",
        column_content, column_name, file_name, file_line);
  }
};
} // namespace error

using ignore_column = unsigned int;
static const ignore_column ignore_no_column = 0;
static const ignore_column ignore_extra_column = 1;
static const ignore_column ignore_missing_column = 2;

template <char... trim_char_list> struct trim_chars {
private:
  constexpr static bool is_trim_char(char) { return false; }

  template <class... OtherTrimChars>
  constexpr static bool is_trim_char(char c, char trim_char,
                                     OtherTrimChars... other_trim_chars) {
    return c == trim_char || is_trim_char(c, other_trim_chars...);
  }

public:
  static void trim(char *&str_begin, char *&str_end) {
    while (str_begin != str_end && is_trim_char(*str_begin, trim_char_list...))
      ++str_begin;
    while (str_begin != str_end &&
           is_trim_char(*(str_end - 1), trim_char_list...))
      --str_end;
    *str_end = '\0';
  }
};

struct no_comment {
  static bool is_comment(const char *) { return false; }
};

template <char... comment_start_char_list> struct single_line_comment {
private:
  constexpr static bool is_comment_start_char(char) { return false; }

  template <class... OtherCommentStartChars>
  constexpr static bool
  is_comment_start_char(char c, char comment_start_char,
                        OtherCommentStartChars... other_comment_start_chars) {
    return c == comment_start_char ||
           is_comment_start_char(c, other_comment_start_chars...);
  }

public:
  static bool is_comment(const char *line) {
    return is_comment_start_char(*line, comment_start_char_list...);
  }
};

struct empty_line_comment {
  static bool is_comment(const char *line) {
    if (*line == '\0')
      return true;
    while (*line == ' ' || *line == '\t') {
      ++line;
      if (*line == 0)
        return true;
    }
    return false;
  }
};

template <char... comment_start_char_list>
struct single_and_empty_line_comment {
  static bool is_comment(const char *line) {
    return single_line_comment<comment_start_char_list...>::is_comment(line) ||
           empty_line_comment::is_comment(line);
  }
};

template <char sep> struct no_quote_escape {
  static const char *find_next_column_end(const char *col_begin) {
    while (*col_begin != sep && *col_begin != '\0')
      ++col_begin;
    return col_begin;
  }

  static void unescape(char *&, char *&) {}
};

template <char sep, char quote> struct double_quote_escape {
  static const char *find_next_column_end(const char *col_begin) {
    while (*col_begin != sep && *col_begin != '\0')
      if (*col_begin != quote)
        ++col_begin;
      else {
        do {
          ++col_begin;
          while (*col_begin != quote) {
            if (*col_begin == '\0')
              throw error::escaped_string_not_closed();
            ++col_begin;
          }
          ++col_begin;
        } while (*col_begin == quote);
      }
    return col_begin;
  }

  static void unescape(char *&col_begin, char *&col_end) {
    if (col_end - col_begin >= 2) {
      if (*col_begin == quote && *(col_end - 1) == quote) {
        ++col_begin;
        --col_end;
        char *out = col_begin;
        for (char *in = col_begin; in != col_end; ++in) {
          if (*in == quote && (in + 1) != col_end && *(in + 1) == quote) {
            ++in;
          }
          *out = *in;
          ++out;
        }
        col_end = out;
        *col_end = '\0';
      }
    }
  }
};

struct throw_on_overflow {
  template <class T> static void on_overflow(T &) {
    throw error::integer_overflow();
  }

  template <class T> static void on_underflow(T &) {
    throw error::integer_underflow();
  }
};

struct ignore_overflow {
  template <class T> static void on_overflow(T &) {}

  template <class T> static void on_underflow(T &) {}
};

struct set_to_max_on_overflow {
  template <class T> static void on_overflow(T &x) {
    // using (std::numeric_limits<T>::max) instead of
    // std::numeric_limits<T>::max to make code including windows.h with its max
    // macro happy
    x = (std::numeric_limits<T>::max)();
  }

  template <class T> static void on_underflow(T &x) {
    x = (std::numeric_limits<T>::min)();
  }
};

namespace detail {
template <class quote_policy>
void chop_next_column(char *&line, char *&col_begin, char *&col_end) {
  assert(line != nullptr);

  col_begin = line;
  // the col_begin + (... - col_begin) removes the constness
  col_end =
      col_begin + (quote_policy::find_next_column_end(col_begin) - col_begin);

  if (*col_end == '\0') {
    line = nullptr;
  } else {
    *col_end = '\0';
    line = col_end + 1;
  }
}

template <class trim_policy, class quote_policy>
void parse_line(char *line, char **sorted_col,
                const std::vector<int> &col_order) {
  for (int i : col_order) {
    if (line == nullptr)
      throw ::io::error::too_few_columns();
    char *col_begin, *col_end;
    chop_next_column<quote_policy>(line, col_begin, col_end);

    if (i != -1) {
      trim_policy::trim(col_begin, col_end);
      quote_policy::unescape(col_begin, col_end);

      sorted_col[i] = col_begin;
    }
  }
  if (line != nullptr)
    throw ::io::error::too_many_columns();
}

template <unsigned column_count, class trim_policy, class quote_policy>
void parse_header_line(char *line, std::vector<int> &col_order,
                       const std::string *col_name,
                       ignore_column ignore_policy) {
  col_order.clear();

  bool found[column_count];
  std::fill(found, found + column_count, false);
  while (line) {
    char *col_begin, *col_end;
    chop_next_column<quote_policy>(line, col_begin, col_end);

    trim_policy::trim(col_begin, col_end);
    quote_policy::unescape(col_begin, col_end);

    for (unsigned i = 0; i < column_count; ++i)
      if (col_begin == col_name[i]) {
        if (found[i]) {
          error::duplicated_column_in_header err;
          err.set_column_name(col_begin);
          throw err;
        }
        found[i] = true;
        col_order.push_back(i);
        col_begin = 0;
        break;
      }
    if (col_begin) {
      if (ignore_policy & ::io::ignore_extra_column)
        col_order.push_back(-1);
      else {
        error::extra_column_in_header err;
        err.set_column_name(col_begin);
        throw err;
      }
    }
  }
  if (!(ignore_policy & ::io::ignore_missing_column)) {
    for (unsigned i = 0; i < column_count; ++i) {
      if (!found[i]) {
        error::missing_column_in_header err;
        err.set_column_name(col_name[i].c_str());
        throw err;
      }
    }
  }
}

template <class overflow_policy> void parse(char *col, char &x) {
  if (!*col)
    throw error::invalid_single_character();
  x = *col;
  ++col;
  if (*col)
    throw error::invalid_single_character();
}

template <class overflow_policy> void parse(char *col, std::string &x) {
  x = col;
}

template <class overflow_policy> void parse(char *col, const char *&x) {
  x = col;
}

template <class overflow_policy> void parse(char *col, char *&x) { x = col; }

template <class overflow_policy, class T>
void parse_unsigned_integer(const char *col, T &x) {
  x = 0;
  while (*col != '\0') {
    if ('0' <= *col && *col <= '9') {
      T y = *col - '0';
      if (x > ((std::numeric_limits<T>::max)() - y) / 10) {
        overflow_policy::on_overflow(x);
        return;
      }
      x = 10 * x + y;
    } else
      throw error::no_digit();
    ++col;
  }
}

template <class overflow_policy> void parse(char *col, unsigned char &x) {
  parse_unsigned_integer<overflow_policy>(col, x);
}
template <class overflow_policy> void parse(char *col, unsigned short &x) {
  parse_unsigned_integer<overflow_policy>(col, x);
}
template <class overflow_policy> void parse(char *col, unsigned int &x) {
  parse_unsigned_integer<overflow_policy>(col, x);
}
template <class overflow_policy> void parse(char *col, unsigned long &x) {
  parse_unsigned_integer<overflow_policy>(col, x);
}
template <class overflow_policy> void parse(char *col, unsigned long long &x) {
  parse_unsigned_integer<overflow_policy>(col, x);
}

template <class overflow_policy, class T>
void parse_signed_integer(const char *col, T &x) {
  if (*col == '-') {
    ++col;

    x = 0;
    while (*col != '\0') {
      if ('0' <= *col && *col <= '9') {
        T y = *col - '0';
        if (x < ((std::numeric_limits<T>::min)() + y) / 10) {
          overflow_policy::on_underflow(x);
          return;
        }
        x = 10 * x - y;
      } else
        throw error::no_digit();
      ++col;
    }
    return;
  } else if (*col == '+')
    ++col;
  parse_unsigned_integer<overflow_policy>(col, x);
}

template <class overflow_policy> void parse(char *col, signed char &x) {
  parse_signed_integer<overflow_policy>(col, x);
}
template <class overflow_policy> void parse(char *col, signed short &x) {
  parse_signed_integer<overflow_policy>(col, x);
}
template <class overflow_policy> void parse(char *col, signed int &x) {
  parse_signed_integer<overflow_policy>(col, x);
}
template <class overflow_policy> void parse(char *col, signed long &x) {
  parse_signed_integer<overflow_policy>(col, x);
}
template <class overflow_policy> void parse(char *col, signed long long &x) {
  parse_signed_integer<overflow_policy>(col, x);
}

template <class T> void parse_float(const char *col, T &x) {
  bool is_neg = false;
  if (*col == '-') {
    is_neg = true;
    ++col;
  } else if (*col == '+')
    ++col;

  x = 0;
  while ('0' <= *col && *col <= '9') {
    int y = *col - '0';
    x *= 10;
    x += y;
    ++col;
  }

  if (*col == '.' || *col == ',') {
    ++col;
    T pos = 1;
    while ('0' <= *col && *col <= '9') {
      pos /= 10;
      int y = *col - '0';
      ++col;
      x += y * pos;
    }
  }

  if (*col == 'e' || *col == 'E') {
    ++col;
    int e;

    parse_signed_integer<set_to_max_on_overflow>(col, e);

    if (e != 0) {
      T base;
      if (e < 0) {
        base = T(0.1);
        e = -e;
      } else {
        base = T(10);
      }

      while (e != 1) {
        if ((e & 1) == 0) {
          base = base * base;
          e >>= 1;
        } else {
          x *= base;
          --e;
        }
      }
      x *= base;
    }
  } else {
    if (*col != '\0')
      throw error::no_digit();
  }

  if (is_neg)
    x = -x;
}

template <class overflow_policy> void parse(char *col, float &x) {
  parse_float(col, x);
}
template <class overflow_policy> void parse(char *col, double &x) {
  parse_float(col, x);
}
template <class overflow_policy> void parse(char *col, long double &x) {
  parse_float(col, x);
}

template <class overflow_policy, class T> void parse(char *col, T &x) {
  // Mute unused variable compiler warning
  (void)col;
  (void)x;
  // GCC evaluates "false" when reading the template and
  // "sizeof(T)!=sizeof(T)" only when instantiating it. This is why
  // this strange construct is used.
  static_assert(sizeof(T) != sizeof(T),
                "Can not parse this type. Only builtin integrals, floats, "
                "char, char*, const char* and std::string are supported");
}

} // namespace detail

template <unsigned column_count, class trim_policy = trim_chars<' ', '\t'>,
          class quote_policy = no_quote_escape<','>,
          class overflow_policy = throw_on_overflow,
          class comment_policy = no_comment>
class CSVReader {
private:
  LineReader in;

  char *row[column_count];
  std::string column_names[column_count];

  std::vector<int> col_order;

  template <class... ColNames>
  void set_column_names(std::string s, ColNames... cols) {
    column_names[column_count - sizeof...(ColNames) - 1] = std::move(s);
    set_column_names(std::forward<ColNames>(cols)...);
  }

  void set_column_names() {}

public:
  CSVReader() = delete;
  CSVReader(const CSVReader &) = delete;
  CSVReader &operator=(const CSVReader &);

  template <class... Args>
  explicit CSVReader(Args &&... args) : in(std::forward<Args>(args)...) {
    std::fill(row, row + column_count, nullptr);
    col_order.resize(column_count);
    for (unsigned i = 0; i < column_count; ++i)
      col_order[i] = i;
    for (unsigned i = 1; i <= column_count; ++i)
      column_names[i - 1] = "col" + std::to_string(i);
  }

  char *next_line() { return in.next_line(); }

  template <class... ColNames>
  void read_header(ignore_column ignore_policy, ColNames... cols) {
    static_assert(sizeof...(ColNames) >= column_count,
                  "not enough column names specified");
    static_assert(sizeof...(ColNames) <= column_count,
                  "too many column names specified");
    try {
      set_column_names(std::forward<ColNames>(cols)...);

      char *line;
      do {
        line = in.next_line();
        if (!line)
          throw error::header_missing();
      } while (comment_policy::is_comment(line));

      detail::parse_header_line<column_count, trim_policy, quote_policy>(
          line, col_order, column_names, ignore_policy);
    } catch (error::with_file_name &err) {
      err.set_file_name(in.get_truncated_file_name());
      throw;
    }
  }

  template <class... ColNames> void set_header(ColNames... cols) {
    static_assert(sizeof...(ColNames) >= column_count,
                  "not enough column names specified");
    static_assert(sizeof...(ColNames) <= column_count,
                  "too many column names specified");
    set_column_names(std::forward<ColNames>(cols)...);
    std::fill(row, row + column_count, nullptr);
    col_order.resize(column_count);
    for (unsigned i = 0; i < column_count; ++i)
      col_order[i] = i;
  }

  bool has_column(const std::string &name) const {
    return col_order.end() !=
           std::find(col_order.begin(), col_order.end(),
                     std::find(std::begin(column_names), std::end(column_names),
                               name) -
                         std::begin(column_names));
  }

  void set_file_name(const std::string &file_name) {
    in.set_file_name(file_name);
  }

  void set_file_name(const char *file_name) { in.set_file_name(file_name); }

  const char *get_truncated_file_name() const {
    return in.get_truncated_file_name();
  }

  void set_file_line(unsigned file_line) { in.set_file_line(file_line); }

  unsigned get_file_line() const { return in.get_file_line(); }

private:
  void parse_helper(std::size_t) {}

  template <class T, class... ColType>
  void parse_helper(std::size_t r, T &t, ColType &... cols) {
    if (row[r]) {
      try {
        try {
          ::io::detail::parse<overflow_policy>(row[r], t);
        } catch (error::with_column_content &err) {
          err.set_column_content(row[r]);
          throw;
        }
      } catch (error::with_column_name &err) {
        err.set_column_name(column_names[r].c_str());
        throw;
      }
    }
    parse_helper(r + 1, cols...);
  }

public:
  template <class... ColType> bool read_row(ColType &... cols) {
    static_assert(sizeof...(ColType) >= column_count,
                  "not enough columns specified");
    static_assert(sizeof...(ColType) <= column_count,
                  "too many columns specified");
    try {
      try {

        char *line;
        do {
          line = in.next_line();
          if (!line)
            return false;
        } while (comment_policy::is_comment(line));

        detail::parse_line<trim_policy, quote_policy>(line, row, col_order);

        parse_helper(0, cols...);
      } catch (error::with_file_name &err) {
        err.set_file_name(in.get_truncated_file_name());
        throw;
      }
    } catch (error::with_file_line &err) {
      err.set_file_line(in.get_file_line());
      throw;
    }

    return true;
  }
};
} // namespace io
#endif