Question

I have a C program that receives a 64 byte array of char (which is passed via USB). Depending on the first byte (which indicates the command type) I want to ‘impose’ a structure over the char array to make the code clearer.

For example, if the command code is 10 I would expect something like:

struct
{
    uint8_t commandNumber;
    uint16_t xPos;
    uint16_t yPos;
    int32_t identificationNumber;
 } commandTen;

So I would like to cast my char packet[64] ‘onto’ commandTen and then access the fields using something like:

localVar = commandTenPacket->xPos;

How can this be achieved in C?

Thanks in advance!

Answer 1

First, as others said you’d have to ensure that your struct has no padding. Your compiler probably has an extension for that, #pragma pack or so.

Define a struct for each of your use cases, not a variable as in your example.

Then define a union

typedef union overlay overlay;

union overlay {
 uint8_t raw[64];
 struct type10 command10;
 struct type42 command42;
};

Now create a buffer of that type

overlay buffer;

Feed the “raw” part to your function that receives the data: getit(buffer.raw). And then

switch (buffer.raw[0]) {
 case 10: {
   // use buffer.command10
   break;
 }
 case 42: {
   // use buffer.command42
 }
}

This is guaranteed by the C standard to work well, since you are reading everything as uint8_t AKA unsigned char. In fact, the principle use case for unions is just this sort of “type puning”. The whole network layer with it different types of IPv4, IPv6 etc addresses works in a similar way.