Encoding and decoding uint8_t using TinyCbor C Library - c++

Question

I am implementing C++ 11 based application and I am using TinyCbor C library for Encoding and Decoding application specific data as below:
#include "cbor.h"
#include <iostream>
using namespace std;
int main() {
struct MyTest {
uint8_t varA;
float vabB;
};
MyTest obj;
obj.varA = 100; // If I set it t0 below 20 then it works
obj.varB = 10.10;
uint8_t buff[100];
//Encode
CborEncode encoder;
CborEncode array;
cbor_encoder_init(&encoder, buff, sizeof(buff), 0);
cbor_encoder_create_array(&encode, &array, CborIndefiniteLength);
cbor_encode_simple_value(&array, obj.varA);
cbor_encode_float(&array, obj.varB);
cbor_encoder_close_container(&encoder, &array);
// Decode
CborParser parse;
CborValue value;
cbor_parser_init(buff, sizeof(buff), 0, &parser, &value);
CborValue array;
cbor_value_enter_container(&value, &array);
uint8_t val;
cbor_value_get_simple_type(&array, &val);
// This prints blank
cout << "uint8_t value: " << static_cast<int>(val) << endl;
float fval;
cbor_value_get_simple_type(&array, &fval);
cout << "float value: " << fval << endl;
return 0;
}
Above code works when I set value of uint8_t varA to below 20, I see 20 getting printed on console but if I set more than 20 then sometimes it gives error CborErrorIllegalSimpleType. Or if value is set to 21 then it returns me type as CborBooleanType or CborNullType.
What is wrong with the code
How to encode and decode uint8_t using TinyCbor.

Answer 1

You have a few things going on here.
cbor_encoder_create_array(&encode, &array, CborIndefiniteLength);
Don’t use indefinite length unless you plan on streaming the encoding. If you have all the data in front of you when encoring, use defined length. See why here: Serialize fixed size Map to CBOR also, and I’m not sure what version you are using but indefinite length objects were at least fairly recently a TO DO list item for TinyCBOR.
cbor_value_get_simple_type(&array, &val);
You don’t need simple here. Simple types are primatives that are mostly undefined. CBOR’s default type is int64_t, signed long long. Simple does allow for a natural stop at 8bit, but 20 in simple is Boolean false. 21 true , 22 null. You discovered this already. It can’t store a negative, or float, and while you could use it to represent an 8bit like “100”, you really shouldn’t. The good news about CBOR’s types is that while the default is 64bit, it uses as little memory to encode as needed. So storing 100 in CBOR is one byte, not 8, not counting the overhead for a second. EDIT: Clarification, 2 bytes to store integer 100. CBOR is one byte for 0-23, but 24 and one has a byte of overhead.
By overhead I mean, when you encode a cbor integer, the first three bits mean UNSIGNED INTEGER (binary 000), one bit is reserved, the other 4 bits are value, so if you had a small value that fit in those four bits (value 0-23) you can get it all in one byte. That reserved bit is weird, see the chart here to understand it: https://en.m.wikipedia.org/wiki/CBOR So 24-255 requires 2 bytes encoded, etc etc. A far cry from it always using 8bytes just because it can.
The short version here is that CBOR won’t use more space than needed - BUT - it’s not a strongly typed serializer! If you mean for 100 inside 8bits to be stored in that array and someone stores a 10000 to be with 16bits, it’ll look/work fine until you go to parse and store a large number in an 8bit spot. You need to cast or validate your data. I recommend parse THEN validate.
cbor_value_get_simple_type(&array, &fval);
cout << "float value: " << fval << endl;
I’d have to look in TinyCBOR’s code, but I think this is a happy accident and not technically supported to work. Because simple types use the same three major bits you are able to get the value’s 64bit precision with get_simple. You should instead examine the type and make the correct call for get full or half precision float.
TinyCBOR is pretty nice, but there are definitely a couple gotchas hidden. It really helps to undersnd the CBOR encoding even though you are trusting the serializer to do the work for you.

Answer 2

Use CborError cbor_encode_uint(CborEncoder *encoder, uint64_t value) instead. This function will encode an unsigned integer value with the smallest representation.