Разбор CSV файла с многострочными полями и двойными кавычками

Ответ 1

В ваших данных выборки есть одна двойная кавычка сразу после "double qoute" (sic), которая завершает двойное кавычное начало поля, но вы должны продолжить чтение до следующей запятой или концом строки. Это неправильное многострочное поле, но вы видите, что делает Excel с ним. Вы можете найти описание этого в (отличной) книге Практика программирования (dead link > ; копия интернет-архива, прямая ссылка на Princeton University), которая включает в себя библиотеку синтаксического анализа CSV в C и повторная реализация на С++, и которая обсуждает этот уровень детализации. Также существует стандарт для CSV в RFC 4180 "Общий формат и тип MIME для разделенных запятыми значений", а также вы можете изучить Wikipedia по этому вопросу.

В другом ответе приведен пример кода, который все еще находится под тестированием. Он довольно устойчив в пределах своих ограничений. Здесь он модифицируется в SSCCE (Short, Self-Contained, Correct Example).

#include <stdbool.h>
#include <wchar.h>
#include <wctype.h>

extern const wchar_t *nextCsvField(const wchar_t *p, wchar_t sep, bool *newline);

// Returns a pointer to the start of the next field,
// or zero if this is the last field in the CSV
// p is the start position of the field
// sep is the separator used, i.e. comma or semicolon
// newline says whether the field ends with a newline or with a comma
const wchar_t *nextCsvField(const wchar_t *p, wchar_t sep, bool *newline)
{
    // Parse quoted sequences
    if ('"' == p[0]) {
        p++;
        while (1) {
            // Find next double-quote
            p = wcschr(p, L'"');
            // If we don't find it or it the last symbol
            // then this is the last field
            if (!p || !p[1])
                return 0;
            // Check for "", it is an escaped double-quote
            if (p[1] != '"')
                break;
            // Skip the escaped double-quote
            p += 2;
        }
    }

    // Find next newline or comma.
    wchar_t newline_or_sep[4] = L"\n\r ";
    newline_or_sep[2] = sep;
    p = wcspbrk(p, newline_or_sep);

    // If no newline or separator, this is the last field.
    if (!p)
        return 0;

    // Check if we had newline.
    *newline = (p[0] == '\r' || p[0] == '\n');

    // Handle "\r\n", otherwise just increment
    if (p[0] == '\r' && p[1] == '\n')
        p += 2;
    else
        p++;

    return p;
}

static void dissect(const wchar_t *line)
{
    const wchar_t *start = line;
    const wchar_t *next;
    bool     eol;
    wprintf(L"Input: %d [%.*ls]\n", wcslen(line), wcslen(line)-1, line);
    while ((next = nextCsvField(start, L',', &eol)) != 0)
    {
        wprintf(L"Field: [%.*ls] (eol = %d)\n", (next - start - eol), start, eol);
        start = next;
    }
}

static const wchar_t multiline[] =
   L"First field of first row,\"This field is multiline\n"
    "\n"
    "but that OK because it enclosed in double quotes, and this\n"
    "is an escaped \"\" double quote\" but this one \"\" is not\n"
    "   \"This is second field of second row, but it is not multiline\n"
    "   because it doesn't start \n"
    "   with an immediate double quote\"\n"
    ;

int main(void)
{
    wchar_t line[1024];

    while (fgetws(line, sizeof(line)/sizeof(line[0]), stdin))
        dissect(line);
    dissect(multiline);

    return 0;
}

Пример вывода

$ cat csv.data
a,bb, c ,d""e,f
1,"2","",,"""",4
$ ./wcsv < csv.data
Input: 16 [a,bb, c ,d""e,f]
Field: [a,] (eol = 0)
Field: [bb,] (eol = 0)
Field: [ c ,] (eol = 0)
Field: [d""e,] (eol = 0)
Field: [f] (eol = 1)
Input: 17 [1,"2","",,"""",4]
Field: [1,] (eol = 0)
Field: ["2",] (eol = 0)
Field: ["",] (eol = 0)
Field: [,] (eol = 0)
Field: ["""",] (eol = 0)
Field: [4] (eol = 1)
Input: 296 [First field of first row,"This field is multiline

but that OK because it enclosed in double quotes, and this
is an escaped "" double quote" but this one "" is not
   "This is second field of second row, but it is not multiline
   because it doesn't start 
   with an immediate double quote"]
Field: [First field of first row,] (eol = 0)
Field: ["This field is multiline

but that OK because it enclosed in double quotes, and this
is an escaped "" double quote" but this one "" is not] (eol = 1)
Field: [   "This is second field of second row,] (eol = 0)
Field: [ but it is not multiline] (eol = 1)
Field: [   because it doesn't start ] (eol = 1)
Field: [   with an immediate double quote"] (eol = 1)
$

Я сказал "в пределах своих ограничений"; каковы его ограничения?

В первую очередь, он изолирует необработанное поле, а не преобразованное поле. Таким образом, поле, которое оно изолирует, должно быть изменено для получения "реального" значения, при этом закрытые двойные кавычки отключены, а внутренние двойные двойные кавычки заменены одинарными кавычками. Преобразование необработанного поля в реальное значение имитирует довольно много кода в функции nextCsvField(). Входы - это начало поля и конца поля (символ разделителя). Здесь второй SSCCE с дополнительной функцией csvFieldData(), а функция dissect(), показанная выше, пересмотрена для ее вызова. Формат рассеченного выхода немного отличается, поэтому он выглядит лучше:

#include <stdbool.h>
#include <wchar.h>
#include <wctype.h>

extern const wchar_t *nextCsvField(const wchar_t *p, wchar_t sep, bool *newline);

// Returns a pointer to the start of the next field,
// or zero if this is the last field in the CSV
// p is the start position of the field
// sep is the separator used, i.e. comma or semicolon
// newline says whether the field ends with a newline or with a comma
const wchar_t *nextCsvField(const wchar_t *p, wchar_t sep, bool *newline)
{
    // Parse quoted sequences
    if ('"' == p[0]) {
        p++;
        while (1) {
            // Find next double-quote
            p = wcschr(p, L'"');
            // If we don't find it or it the last symbol
            // then this is the last field
            if (!p || !p[1])
                return 0;
            // Check for "", it is an escaped double-quote
            if (p[1] != '"')
                break;
            // Skip the escaped double-quote
            p += 2;
        }
    }

    // Find next newline or comma.
    wchar_t newline_or_sep[4] = L"\n\r ";
    newline_or_sep[2] = sep;
    p = wcspbrk(p, newline_or_sep);

    // If no newline or separator, this is the last field.
    if (!p)
        return 0;

    // Check if we had newline.
    *newline = (p[0] == '\r' || p[0] == '\n');

    // Handle "\r\n", otherwise just increment
    if (p[0] == '\r' && p[1] == '\n')
        p += 2;
    else
        p++;

    return p;
}

static wchar_t *csvFieldData(const wchar_t *fld_s, const wchar_t *fld_e, wchar_t *buffer, size_t buflen)
{
    wchar_t *dst = buffer;
    wchar_t *end = buffer + buflen - 1;
    const wchar_t *src = fld_s;

    if (*src == L'"')
    {
        const wchar_t *p = src + 1;
        while (p < fld_e && dst < end)
        {
            if (p[0] == L'"' && p+1 < fld_s && p[1] == L'"')
            {
                *dst++ = p[0];
                p += 2;
            }
            else if (p[0] == L'"')
            {
                p++;
                break;
            }
            else
                *dst++ = *p++;
        }
        src = p;
    }
    while (src < fld_e && dst < end)
        *dst++ = *src++;
    if (dst >= end)
        return 0;
    *dst = L'\0';
    return(buffer);
}

static void dissect(const wchar_t *line)
{
    const wchar_t *start = line;
    const wchar_t *next;
    bool     eol;
    wprintf(L"Input %3zd: [%.*ls]\n", wcslen(line), wcslen(line)-1, line);
    while ((next = nextCsvField(start, L',', &eol)) != 0)
    {
        wchar_t buffer[1024];
        wprintf(L"Raw Field: [%.*ls] (eol = %d)\n", (next - start - eol), start, eol);
        if (csvFieldData(start, next-1, buffer, sizeof(buffer)/sizeof(buffer[0])) != 0)
            wprintf(L"Field %3zd: [%ls]\n", wcslen(buffer), buffer);
        start = next;
    }
}

static const wchar_t multiline[] =
   L"First field of first row,\"This field is multiline\n"
    "\n"
    "but that OK because it enclosed in double quotes, and this\n"
    "is an escaped \"\" double quote\" but this one \"\" is not\n"
    "   \"This is second field of second row, but it is not multiline\n"
    "   because it doesn't start \n"
    "   with an immediate double quote\"\n"
    ;

int main(void)
{
    wchar_t line[1024];

    while (fgetws(line, sizeof(line)/sizeof(line[0]), stdin))
        dissect(line);
    dissect(multiline);

    return 0;
}

Пример вывода

$ ./wcsv < csv.data
Input  16: [a,bb, c ,d""e,f]
Raw Field: [a,] (eol = 0)
Field   1: [a]
Raw Field: [bb,] (eol = 0)
Field   2: [bb]
Raw Field: [ c ,] (eol = 0)
Field   3: [ c ]
Raw Field: [d""e,] (eol = 0)
Field   4: [d""e]
Raw Field: [f] (eol = 1)
Field   1: [f]
Input  17: [1,"2","",,"""",4]
Raw Field: [1,] (eol = 0)
Field   1: [1]
Raw Field: ["2",] (eol = 0)
Field   1: [2]
Raw Field: ["",] (eol = 0)
Field   0: []
Raw Field: [,] (eol = 0)
Field   0: []
Raw Field: ["""",] (eol = 0)
Field   2: [""]
Raw Field: [4] (eol = 1)
Field   1: [4]
Input 296: [First field of first row,"This field is multiline

but that OK because it enclosed in double quotes, and this
is an escaped "" double quote" but this one "" is not
   "This is second field of second row, but it is not multiline
   because it doesn't start 
   with an immediate double quote"]
Raw Field: [First field of first row,] (eol = 0)
Field  24: [First field of first row]
Raw Field: ["This field is multiline

but that OK because it enclosed in double quotes, and this
is an escaped "" double quote" but this one "" is not] (eol = 1)
Field 140: [This field is multiline

but that OK because it enclosed in double quotes, and this
is an escaped " double quote" but this one "" is not]
Raw Field: [   "This is second field of second row,] (eol = 0)
Field  38: [   "This is second field of second row]
Raw Field: [ but it is not multiline] (eol = 1)
Field  24: [ but it is not multiline]
Raw Field: [   because it doesn't start ] (eol = 1)
Field  28: [   because it doesn't start ]
Raw Field: [   with an immediate double quote"] (eol = 1)
Field  34: [   with an immediate double quote"]
$ 

Я не тестировал с окончанием строки \r\n (или простой \r).

Ответ 2

Я написал функцию (которую я все еще тестирую), которая пытается разобрать это, а также может.

// Returns a pointer to the start of the next field, or zero if this is the
// last field in the CSV p is the start position of the field sep is the
// separator used, i.e. comma or semicolon newline says whether the field ends
// with a newline or with a comma
static const wchar_t* nextCsvField(const wchar_t *p, wchar_t sep, bool *newline, const wchar_t **escapedEnd)
{
    *escapedEnd = 0;
    *newline = false;

    // Parse quoted sequences
    if ('"' == p[0]) {
        p++;
        while (1) {
            // Find next double-quote.
            p = wcschr(p, L'"');
            // If we didn't find it then this is the last field.
            if (!p)
                return 0;
            // Check for "", it is an escaped double-quote.
            if (p[1] != '"') {
                *escapedEnd = p;
                break;
            }
            // If it the last symbol then this is the last field
            if (!p[1])
                return 0;
            // Skip the escaped double-quote
            p += 2;
        }
    }

    // Find next newline or comma.
    wchar_t newline_or_sep[4] = L"\n\r ";
    newline_or_sep[2] = sep;
    p = wcspbrk(p, newline_or_sep);

    // If no newline or separator, this is the last field.
    if (!p)
        return 0;

    // Check if we had newline.
    *newline = (p[0] == '\r' || p[0] == '\n');

    // Handle "\r\n", otherwise just increment
    if (p[0] == '\r' && p[1] == '\n')
        p += 2;
    else
        p++;

    return p;
}