用于解析CSV数据的库

Question

请查看我的C++ CSV解析类。

我有一些具体的问题：

1. get_next_record应该是一个静态函数吗？
2. CsvParser意味着值将用逗号分隔，那么字段分隔符构造函数是否在顶部？
3. record.clear()在get_next_record的开头。任何其他方法来解决删除最后记录的问题。我意识到您可以返回记录，但是接下来您会遇到如何处理EOF或流错误的问题。

CsvParser.hpp

#ifndef CSV_PARSER_HPP_
#define CSV_PARSER_HPP_

#include <iostream>
#include <string>
#include <vector>

using Field = std::string;
using Record = std::vector<Field>;
using Records = std::vector<Record>;

class CsvParser {
public:
    CsvParser(char field_separator = ',');
    bool get_next_record(std::istream& istrm, Record& record) const;

private:
    char field_separator_char;
};

#endif // CSV_PARSER_HPP_

CsvParser.cpp

#include "CsvParser.hpp"

CsvParser::CsvParser(char field_separator) : field_separator_char(field_separator) {}

bool CsvParser::get_next_record(std::istream& istrm, Record& record) const {

    // Having to clear record because otherwise the program will keep pushing back
    // fields into the vector feels dirty.  How could this be improved?
    record.clear();
    bool in_quotes = false;
    Field field;
    int ch;
    while (istrm) {
        ch = istrm.get();
        if (ch == EOF || (ch == '\n' && !in_quotes)) {
            if (ch == EOF && record.empty() && field.empty()) {
                return false;
            }
            else {
                record.push_back(field);
                return true;
            }
        }
        else if (ch == field_separator_char && !in_quotes) {
            record.push_back(field);
            field.clear();
        }
        else if (ch == '"') {
            if (!in_quotes) {
                in_quotes = true;
            }
            else {
                // Could be an embedded " if next symbol not comma
                int nextch = istrm.peek();
                if (nextch != field_separator_char && nextch != '\n' && nextch != EOF) {
                    field += static_cast<char>(ch);
                }
                else {
                    in_quotes = false;
                }
            }
        }
        else if (ch == '\r') {
            if (in_quotes) {
                field += static_cast<char>(ch);
            }
        }
        else {
            field += static_cast<char>(ch);
        }
    }
    return false;
}

使用谷歌测试进行锻炼：

#include <gtest/gtest.h>

#include "CsvParser.hpp"
#include <sstream>
#include <string>

class CsvParserTest : public ::testing::Test {
public:
    CsvParser parser;
};

TEST_F(CsvParserTest, EmptyRecord) {

    const std::string csv{ "" };
    std::stringstream strm(csv);
    Record record;

    EXPECT_FALSE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 0u);
}

TEST_F(CsvParserTest, SimpleSingleRecord) {

  const std::string csv{ "AA,BB,CC" };
  std::stringstream strm(csv);
  Record record;

  EXPECT_TRUE(parser.get_next_record(strm, record));
  EXPECT_EQ(record.size(), 3u);
  EXPECT_EQ(record[0], "AA");
  EXPECT_EQ(record[1], "BB");
  EXPECT_EQ(record[2], "CC");
}

TEST_F(CsvParserTest, SimpleTwoRecord) {

    const std::string csv{ "AA,BB,CC\r\nDD,EE,FF" };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "AA");
    EXPECT_EQ(record[1], "BB");
    EXPECT_EQ(record[2], "CC");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "DD");
    EXPECT_EQ(record[1], "EE");
    EXPECT_EQ(record[2], "FF");
}

TEST_F(CsvParserTest, SimpleQuotedField) {

    const std::string csv{ "\"A\",BB,CCC" };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "A");
    EXPECT_EQ(record[1], "BB");
    EXPECT_EQ(record[2], "CCC");
}

TEST_F(CsvParserTest, QuotesEmbeddedInQuotedField) {

    const std::string csv{ "\"\"A\"\",BB,CCC" };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "\"A\"");
    EXPECT_EQ(record[1], "BB");
    EXPECT_EQ(record[2], "CCC");
}

TEST_F(CsvParserTest, LinefeedEmbeddedInQuotedField) {

    const std::string csv{ "\"\"A\n\"\",BB,CCC" };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "\"A\n\"");
    EXPECT_EQ(record[1], "BB");
    EXPECT_EQ(record[2], "CCC");
}

TEST_F(CsvParserTest, CommaEmbeddedInQuotedField) {

    const std::string csv{ R"(""A,"",BB,CCC)" };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], R"("A,")");
    EXPECT_EQ(record[1], "BB");
    EXPECT_EQ(record[2], "CCC");
}

TEST_F(CsvParserTest, EmptyRow) {

    const std::string csv{ ",," };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0].size(), 0u);
    EXPECT_EQ(record[1].size(), 0u);
    EXPECT_EQ(record[2].size(), 0u);
}

TEST_F(CsvParserTest, QuotedFollowedByTwoEmptyFields) {

    const std::string csv{ "\"A\n\n\nB\",," };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "A\n\n\nB");
    EXPECT_EQ(record[1].size(), 0u);
    EXPECT_EQ(record[2].size(), 0u);
}

TEST_F(CsvParserTest, EmptyThenQuotedThenEmptyField) {

    const std::string csv{ ",\"A\n\n\nB\"," };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0].size(), 0u);
    EXPECT_EQ(record[1], "A\n\n\nB");
    EXPECT_EQ(record[2].size(), 0u);
}

TEST_F(CsvParserTest, EmptyEmptyThenQuoted) {

    const std::string csv{ ",,\"A\n\n\nB\"" };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0].size(), 0u);
    EXPECT_EQ(record[1].size(), 0u);
    EXPECT_EQ(record[2], "A\n\n\nB");
}

TEST_F(CsvParserTest, CRLFEndOfLIne) {

    const std::string csv{ "A,B,C\r\nD,E,F" };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "A");
    EXPECT_EQ(record[1], "B");
    EXPECT_EQ(record[2], "C");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "D");
    EXPECT_EQ(record[1], "E");
    EXPECT_EQ(record[2], "F");
}

TEST_F(CsvParserTest, EmbeddedCRLF) {

    const std::string csv{ "A,\"B\r\nC\",D\r\nE,F,G" };
    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "A");
    EXPECT_EQ(record[1], "B\r\nC");
    EXPECT_EQ(record[2], "D");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "E");
    EXPECT_EQ(record[1], "F");
    EXPECT_EQ(record[2], "G");
}

TEST_F(CsvParserTest, Complex) {

    const std::string csv = "AAA,BB,CCC\nDDD,EE,FFF\n\"A A\",\"B\nB\",CC\n\"A,B,C\",\"D         E\",F\n\"Billy \"Da Man\" Hooker\",,\n,,\n,,\"Yo bitches!\"\n,,\"Holler if you luv dem \"hat\" bitches\"\n,\"These are my long\nnotes on a load\nof stuff\n fancy some commas:,,,,,,,,,,,,,,,,,,,,,,,,,,,\",";

    std::stringstream strm(csv);
    Record record;

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "AAA");
    EXPECT_EQ(record[1], "BB");
    EXPECT_EQ(record[2], "CCC");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "DDD");
    EXPECT_EQ(record[1], "EE");
    EXPECT_EQ(record[2], "FFF");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "A A");
    EXPECT_EQ(record[1], "B\nB");
    EXPECT_EQ(record[2], "CC");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "A,B,C");
    EXPECT_EQ(record[1], "D         E");
    EXPECT_EQ(record[2], "F");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "Billy \"Da Man\" Hooker");
    EXPECT_EQ(record[1], "");
    EXPECT_EQ(record[2], "");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "");
    EXPECT_EQ(record[1], "");
    EXPECT_EQ(record[2], "");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "");
    EXPECT_EQ(record[1], "");
    EXPECT_EQ(record[2], "Yo bitches!");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "");
    EXPECT_EQ(record[1], "");
    EXPECT_EQ(record[2], "Holler if you luv dem \"hat\" bitches");

    EXPECT_TRUE(parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "");
    EXPECT_EQ(record[1], "These are my long\nnotes on a load\nof stuff\n fancy some commas:,,,,,,,,,,,,,,,,,,,,,,,,,,,");
    EXPECT_EQ(record[2], "");
}

TEST_F(CsvParserTest, TabSeparated) {

    const std::string csv{ "AA\tBB\tCC\nDD\tEE\tFF" };
    std::stringstream strm(csv);
    Record record;
    CsvParser tab_parser('\t');

    EXPECT_TRUE(tab_parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "AA");
    EXPECT_EQ(record[1], "BB");
    EXPECT_EQ(record[2], "CC");

    EXPECT_TRUE(tab_parser.get_next_record(strm, record));
    EXPECT_EQ(record.size(), 3u);
    EXPECT_EQ(record[0], "DD");
    EXPECT_EQ(record[1], "EE");
    EXPECT_EQ(record[2], "FF");
}

Answer 1

代码级别太低，缺乏功能。

使用std::getline并从流中提取整条线将更加惯用(而且可能会根据具体情况而更有效)。然后，您可以通过字符串的,函数一个接一个地找到分隔符find_first_of，并将行分隔成字符串数组。

当csv文件中有大行时，可能缺少这种方法，但是获取字符串向量的整个设计是这种情况下更大的受害者。考虑使用容器缓冲区和字符串视图向量。

缺乏功能:有时您了解csv文件的格式，并希望执行内部转换，而不是获取字符串的向量。如果您能够创建几个满足常见情况的方法，您需要支持它，那么它将把您的csv库转换成一个可用的库。