如何在C中读取任意长行？

Question

我想在C ++中使用等效（在C中） getline ：

std::string s;
getline(std::cin,s); // reads an arbitrarily long line and
                     // inserts its contents on s

有没有办法在C中做这样的事情？ 我正在寻找看起来像这样的东西：

char* s;
getline(stdin,s); // allocates the space necessary to fit the read
                  // line and make s point to it

编辑 ：我决定在最后使用POSIX getline函数，因为我在Linux上（如果你不知道我在说什么的话，运行man getline ），但Michael Burr提供了getline的实现，它可用于其他操作默认情况下getline不可用的系统。 即使他的实施不是最有效的可以设想的，它也能按照我的要求完成工作，所以我将其标记为我的问题的答案。

Answer 1

如果您无法访问POSIX getline()实现，这里是我所处的公共域实现。

我添加了一个getline_simple()函数， getline_simple()返回动态分配的缓冲区中的下一行。 如果您对详细的错误处理不感兴趣，可以使用该函数逐行读取文件：

#define _CRT_SECURE_NO_WARNINGS 1

#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
#include <errno.h>
#include <limits.h>
#include <sys/types.h>


#if !__GNUC__
#if _WIN64
typedef long long ssize_t;
#else
typedef long ssize_t;
#endif
#endif


#if !defined(SSIZE_MAX)
#define SSIZE_MAX ((ssize_t)(SIZE_MAX/2))
#endif

#if !defined(EOVERFLOW)
#define EOVERFLOW (ERANGE)      /* is there something better to use? */
#endif


/*
    nx_getdelim()

    a version of the POSIX getdelim() function that return error codes directly
    instead of messing with the global `errno` value.
*/
ssize_t nx_getdelim(char **lineptr, size_t *n, int delim, FILE *stream);



/*
    getdelim_calc_new_alloc()

    Helper function for getdelim() to figure out an appropriate new
    allocation size that's not too small or too big.

    These numbers seem to work pretty well for most text files.

    returns the input value if it decides that new allocation block
    would be just too big (the caller should handle this as 
    an error).
*/
static
size_t nx_getdelim_get_realloc_size( size_t current_size)
{
    enum {
        k_min_realloc_inc = 32,
        k_max_realloc_inc = 1024,
    };

    if (SSIZE_MAX < current_size) return current_size;

    if (current_size <= k_min_realloc_inc) return current_size + k_min_realloc_inc;

    if (current_size >= k_max_realloc_inc) return current_size + k_max_realloc_inc;

    return current_size * 2;
}



/*
    getdelim_append() 

    a helper function for getdelim() that adds a new character to 
    the outbuffer, reallocating as necessary to ensure the character
    and a following null terminator can fit

*/
static
int nx_getdelim_append( char** lineptr, size_t* bufsize, size_t count, char ch)
{
    char* tmp = NULL;
    size_t tmp_size = 0;

    // assert the contracts for this functions inputs
    assert( lineptr != NULL);
    assert( bufsize != NULL);

    if (count >= (((size_t) SSIZE_MAX) + 1)) {
        // writing more than SSIZE_MAX to the buffer isn't supported
        return -1;
    }

    tmp = *lineptr;
    tmp_size = tmp ? *bufsize : 0;

    // need room for the character plus the null terminator
    if ((count + 2) > tmp_size) {
        tmp_size = nx_getdelim_get_realloc_size( tmp_size);

        tmp = (char*) realloc( tmp, tmp_size);

        if (!tmp) {
            return -1;
        }
    }

    *lineptr = tmp;
    *bufsize = tmp_size;

    // remember, the reallocation size calculation might not have 
    // changed the block size, so we have to check again
    if (tmp && ((count+2) <= tmp_size)) {
        tmp[count++] = ch;
        tmp[count] = 0;
        return 1;
    }

    return -1;
}


/*
    nx_getdelim()

    A getdelim() function modeled on the Linux/POSIX/GNU 
    function of the same name.

    Read data into a dynamically resizable buffer until 
    EOF or until a delimiter character is found.  The returned
    data will be null terminated (unless there's an error allocating
    memory that prevents it).



    params:

        lineptr -   a pointer to a char* allocated by malloc() 
                    (actually any pointer that can legitimately be
                    passed to free()).  *lineptr will be updated 
                    by getdelim() if the memory block needs to be 
                    reallocated to accommodate the input data.

                    *lineptr can be NULL (though lineptr itself cannot),
                    in which case the function will allocate any necessary 
                    buffer.

        n -         a pointer to a size_t object that contains the size of 
                    the buffer pointed to by *lineptr (if non-NULL).

                    The size of whatever buff the resulting data is 
                    returned in will be passed back in *n

        delim -     the delimiter character.  The function will stop
                    reading one this character is read form the stream.

                    It will be included in the returned data, and a
                    null terminator character will follow it.

        stream -    A FILE* stream object to read data from.

    Returns:

        The number of characters placed in the returned buffer, including
        the delimiter character, but not including the terminating null.

        If no characters are read and EOF is set (or attempting to read 
        from the stream on the first attempt caused the eof indication 
        to be set), a null terminator will be written to the buffer and
        0 will be returned.

        If an error occurs while reading the stream, a 0 will be returned.
        A null terminator will not necessarily be at the end of the data 
        written.

        On the following error conditions, the negative value of the error 
        code will be returned:

            ENOMEM:     out of memory
            EOVERFLOW:  SSIZE_MAX character written to te buffer before 
                        reaching the delimiter
                        (on Windows, EOVERFLOW is mapped to ERANGE)

         The buffer will not necessarily be null terminated in these cases.


    Notes:

        The returned data might include embedded nulls (if they exist
        in the data stream) - in that case, the return value of the
        function is the only way to reliably determine how much data
        was placed in the buffer.

        If the function returns 0 use feof() and/or ferror() to determine
        which case caused the return.

        If EOF is returned after having written one or more characters
        to the buffer, a normal count will be returned (but there will 
        be no delimiter character in the buffer).  

        If 0 is returned and ferror() returns a non-zero value,
        the data buffer may not be null terminated.

        In other cases where a negative value is returned, the data
        buffer is not necessarily null terminated and there 
        is no reliable means to determining what data in the buffer is
        valid.

        The pointer returned in *lineptr and the buffer size
        returned in *n will be valid on error returns unless
        NULL pointers are passed in for one or more of these
        parameters (in which case the return value will be -EINVAL).

*/
ssize_t nx_getdelim(char **lineptr, size_t *n, int delim, FILE *stream)
{
    int retval = 0;

    ssize_t result = 0;    
    char* line = NULL;
    size_t size = 0;
    size_t count = 0;
    int err = 0;
    int ch = 0;

    if (!lineptr || !n) {
        return -EINVAL;
    }

    line = *lineptr;
    size = *n;

    for (;;) {
        ch = fgetc( stream);

        if (ch == EOF) {
            break;
        }

        result = nx_getdelim_append( &line, &size, count, ch);

        // check for error adding to the buffer (ie., out of memory)
        if (result < 0) {
            err = -ENOMEM;
            break;
        }

        ++count;

        // check if we're done because we've found the delimiter
        if ((unsigned char)ch == (unsigned char)delim) {
            break;
        }

        // check if we're passing the maximum supported buffer size
        if (count > SSIZE_MAX) {
            err = -EOVERFLOW;
            break;
        }
    }

    // update the caller's data
    *lineptr = line;
    *n = size;

    // check for various error returns
    if (err != 0) {
        return err;
    }

    if (ferror(stream)) {
        return 0;
    }

    if (feof(stream) && (count == 0)) {
        if (nx_getdelim_append( &line, &size, count, 0) < 0) {
            return -ENOMEM;
        }
    }

    return count;
}




ssize_t nx_getline(char **lineptr, size_t *n, FILE *stream)
{
    return nx_getdelim( lineptr, n, '\n', stream);
}



/*
    versions of getline() and getdelim() that attempt to follow
    POSIX semantics (ie. they set errno on error returns and
    return -1 when the stream error indicator or end-of-file
    indicator is set (ie., ferror() or feof() would return
    non-zero).
*/
ssize_t getdelim(char **lineptr, size_t *n, char delim, FILE *stream)
{
    ssize_t retval = nx_getdelim( lineptr, n, delim, stream);

    if (retval < 0) {
        errno = -retval;
        retval = -1;
    }

    if (retval == 0) {
        retval = -1;
    }

    return retval;
}

ssize_t getline(char **lineptr, size_t *n, FILE *stream)
{
    return getdelim( lineptr, n, '\n', stream);
}


/* 
    A simple function to return the next line of text in a dynamically 
    allocated buffer

    On error a NULL pointer is returned.  When the caller no longer needs the 
    returned data, the pointer returned should be passed to `free()`.


 */
char* getline_simple( FILE* stream)
{
    char* p = NULL;
    size_t size = 0;

    ssize_t result = getline( &p, &size, stream);

    if (result < 0) {
        free(p);
        p = NULL;
    }

    return p;
}

免责声明 - 此代码适用于我的目的，但我不保证会出现这种情况。 如果打算使用此代码，请使用尽职调查。

Answer 2

你需要在循环中逐步执行它，使用fgets()读取数据块，并且realloc()如果你没有捕到行的末尾则扩大内存缓冲区。

但是，可能有一些库函数可以帮助您。

#define BLOCKSIZE 1024

char *readAllocLine(FILE *fp) {
    char    *line = NULL;
    size_t  maxlength = 0;
    assert(fp != NULL);
    for(;;) { // Read the line in BLOCKSIZE -blocks.
        char *crlf, *block;

        maxlength += BLOCKSIZE;
        // This exploits realloc behaviour upon hitting NULL
        if (NULL == (line = realloc(line, maxlength+1))) {
            break; // realloc error returns NULL.
        }
        block = line + maxlength - BLOCKSIZE;
        // BLOCKSIZE+1 to accommodate final zero
        if (NULL == fgets(block, BLOCKSIZE+1, fp)) {
            // TODO: rewind fp in case of error.
            if (block == line) {
                    // Error.
                    free(line); line = NULL;
            }
            break;
        }
        // This was the last block iff we find a CRLF inside.
        if (NULL != (crlf = strchr(block, '\n'))) {
            *crlf = 0x0;
            if (crlf != block) {
                if ('\r' == *(--crlf))
                    *crlf = 0x0;
            }
            break;
        } /* if */
    } /* for */
    return line;
}

用这个主要测试

int main(int argc, char **argv) {
    FILE *fp;
    if (argc !=2 ) {
            fprintf(stderr, "Syntax: %s testfile.txt\n", argv[0]);
            return -1;
    }
    fp = fopen(argv[1], "r");
    while(!feof(fp)) {
        char *s = readAllocLine(fp);
        if (NULL != s) {
            printf("\"%s\"\n", s);
            free(s);
        } else {
            printf("--- end of file ---\n");
            break;
        }
    }
    fclose(fp);
    return 0;
}

和这个脚本：

for i in $( seq 1020 1028 ); do
    # Didn't want to think over 
    yes x | tr -d "\n" | dd of=test bs=1 count=$i 2>/dev/null
    ./readline test | wc -c | tr "\n" ","
    echo "" >> test
    ./readline test | wc -c | tr "\n" ","
    echo "" >> test
    ./readline test | wc -c
done