将来自各种数据类型的值编码为二进制文件（python），然后在c中对其进行解码

Question

I need to store values coming from various data types in a binary file (in python) and then decode this binary file in c and reconstruct the values. For more clarification, lets assume we have three variables as below

a = [1, 2, 3], dtype = int
b = ['sky', 'chair', 'book', 'desk']
c = [3.56, 4.69, 55.0, 1.698], dtype = float32

step1: export all data values to a binary file (data)

step2: import the binary file in c and then reconstruct corresponding values.

//load binary file ?
...

// declaration
int a[3];
char *b[4];
double c[4];

// decode the binary file to have the same values in c ?
...
a = [1, 2, 3];
b = {'sky', 'chair', 'book', 'desk'};
c = [3.56, 4.69, 55.0, 1.698];

Thanks in advance for your helps,

I'm trying to have something that dynamically exports the data in a binary file and also dynamically retrieve in C; a data structure like this: [number of integers, all integers values, number of floats, all float values, number of strings, all strings]

For example: [3, 1, 2, 3, 4, 0.2, 0.65, 0.56, 0.33, 2,'sky','desk']

In summary, I also need to pass the number of elements for each data type for decoder in C; If I know that values are stored in such an order (ints, floats, strs)

• step 1: Read first number (3), now we know that there are 3 int values.
• step 2: Read the next tree numbers as ints.
• step 3: Read the number of float values (4).
• step 4: read four float numbers.
• step 5: read the number of strings (2),
• step 6: read the two strings.

I know that I can use struct package to create binary packs and then write them, but I don't know what Generic strategy should I follow with respect to the unpacking process which is being implemented in C (!!!).

Answer 1

a data structure like this: [number of integers, all integers values, number of floats, all float values, number of strings, all strings]

In preparation for writing the strings to the file, I'd convert them to bytes , eg:

bb = [s.encode() for s in b]

The store part in Python amounts to providing the numbers and values along with an appropriate format string.

• integers: The straightforward format string 'I%di'%len(a) covers the number of integers I and all integers values %di , where %d is replaced by the number of items in a .
• floats: The straightforward format string 'I%df'%len(c) covers the number of floats I and all float values %df , where %d is replaced by the number of items in c .
• strings: The format string is a little less straightforward because struct.pack doesn't allow a repeat count for strings, but requires the string length. ''.join(['%ds'%(len(s)+1) for s in bb]) constructs the format for all string values %ds , where %d is replaced by the number of bytes in each string plus one for the terminating NUL. (Failing detailed specification, I chose to store the strings in C form.)

This gives:

data = struct.pack('I%di'%len(a)+'I%df'%len(c)+'I'+''.join(['%ds'%(len(s)+1) for s in bb]),
                    len(a), *a,   len(c), *c,   len(bb),       *bb)
open('data', 'wb').write(data)

The decode part in C is not very complicated, following the steps you outlined, eg:

#include <stdio.h>
#include <stdlib.h>
int geti(FILE *stream)
{   // helper function to read an integer
    int i;
    if (!fread(&i, sizeof i, 1, stream)) exit(EXIT_FAILURE);
    return i;
}
…
    // step 1: Read first number of int values.
    int ni = geti(stdin);
    int a[ni];
    // step 2: Read the next 'ni' numbers as ints.
    for (int i = 0; i < ni; ) a[i++] = geti(stdin);
    // step 3: Read the number of float values.
    int nf = geti(stdin);
    double c[nf];
    // step 4: read 'nf' float numbers.
    for (int i = 0; i < nf; )
    { float f; fread(&f, sizeof f, 1, stdin); c[i++] = f; }
    // step 5: read the number of strings,
    int ns = geti(stdin);
    char *b[ns];
    // step 6: read the 'ns' strings.
    for (int i = 0, j, c; i < ns; ++i)
    {
        b[i] = NULL, j = 0;
        do
        {
            b[i] = realloc(b[i], j+1);
            c = getc(stdin);
            if (c == EOF) exit(EXIT_FAILURE);
        } while (b[i][j++] = c);
    }

Note that in this example

1. of course you could use a stream other than stdin ,
2. not all errors (read failure or out of memory) are checked,
3. no provisions are made for the case that the number representations in source and destination are different; for this, you could utilize struct.pack 's byte order and size indication and/or ntohl() in C.