C++ floats and doubles on different platforms/architectures

Question

For context, I'm developing a library for Arduino to transfer data packets over LoRa receivers.

Question I'm having is, the library is placing the raw bits from floats and doubles into uint8_t arrays, sending them and then tranferring them back into floats and doubles.

This of course works when the compiler, platform and architecture is the same, but will I run into problems when unpacking them and simply copying the data straight to a float on other platforms like C++ on Windows or Linux?

I could of course use integers and have some factor to retain decimal places but this adds complexity I'd rather somehow work around.

Answer 1

If your implementation uses ieee754 floating point numbers, and also has unsigned 32 and 64 bit integers, then most likely there is a one-to-one correspondence between float and 32 bit integers, and between double and 64 bit integers. And that applies to exactly 100% of implementations that I’ve used in the last ten years. (Long double is more variable).

So you can move a float into a 32 bit unsigned int, store the four bytes in some fixed order by storing x>>24, x>>16, x>>8 and x as bytes in any fixed order, and read them back in the same order. Same for double with 64 bit.

Alternatively, you store the numbers as text as JSON does.