python struct.pack equivalent in c++

I want a fixed length string from a number just like struct.pack present in python but in c++. I thought of itoa (i,buffer,2) but problem can be that its length will depend on platform. Is there any way to make 开发者_C百科it independent of platform ?

---

If you're looking for a complete solution similar to Python's struct package, you might check out Google's Protocol Buffers Library. Using that will take care of a lot of issues (e.g. endian-ness, language-portability, cross-version compatibility) for you.

---

Here's a start:

typedef std::vector<uint8_t> byte_buffer;

template <std::size_t N>
void append_fixed_width(byte_buffer& buf, uintmax_t val) {
    int shift = ((N - 1) * 8);
    while (shift >= 0) {
        uintmax_t mask = (0xff << shift);
        buf.push_back(uint8_t((val & mask) >> shift));
        shift -= 8;
    }
}

template <typename IntType>
void append_bytes(byte_buffer& buf, IntType val) {
    append_fixed_width<sizeof(IntType)>(buf, uintmax_t(val));
}

int main() { // usage example
     byte_buffer bytes;
     append_bytes(bytes, 1);   // appends sizeof(int) bytes
     append_bytes(bytes, 1ul); // appends sizeof(unsigned long) bytes
     append_bytes(bytes, 'a'); // appends sizeof(int) bytes :p
     append_bytes(bytes, char('a')); // appends 1 byte
     return 0;
}

Append_bytes will append any integer type into a byte buffer represented using a std::vector<uint8_t>. Values are packed in big endian byte order. If you need to change this, then tweak append_fixed_width to traverse the value in a different order.

These functions build a raw byte buffer so whomever is decoding it is responsible for knowing what is in there. IIRC, this is what struct.pack does as well; in other words, the caller of struct.unpack needs to provide the same format string. You can write a variant of append_fixed_width to pack a TLV instead:

template <typename TagType, typename ValueType>
void append_tlv(byte_buffer& buf, TagType t, ValueType val) {
    append_fixed_width<sizeof(TagType)>(buf, uintmax_t(t));
    append_fixed_width<sizeof(std::size_t)>(buf, uintmax_t(sizeof(ValueType)));
    append_fixed_width<sizeof(ValueType)>(buf, uintmax_t(val));
}

I would take a serious look at Jeremy's suggestion though. I wish that it had existed when I wrote all of the binary packing code that I have now.

---

You need to define an exact-width integer type through a typedef; you do that in a platform-specific manner. If you use C99, int16_t is predefined in <stdint.h>. You can then cast to that type, and type the memory representation of a variable:

int16_t val = (int16_t) orig_val;
void *buf = &val;

Notice that you still need to deal with endianness.

If you don't have C99, you can either use compile-time or run-time size tests. For compile-time tests, consider using autoconf, which already computes the sizes of the various primitive types, so that you can select a good type at compile time. At run-time, just have a series of sizeof tests. Notice that this is somewhat inappropriate for run-time, as the test will always come out with the same result. As an alternative to autoconf, you can also use compiler/system identification macros for a compile-time test.

---

The C++ way would be to use stringstream:

stringstream ss;
int number=/*your number here*/;
ss<<number;

and to get the buffer you'd use ss.str().c_str().