Timestamp is not round-trip equivalent with Display and parse()

[chachi]

The + 1 at the end of the impl From<Duration> breaks stringifying and then parsing the same number back out.

uhlc-rs/src/ntp64.rs

Line 231 in 68bf159

NTP64((secs << 32) + ((nanos * FRAC_PER_SEC) / NANO_PER_SEC) + 1)

The following test fails:

    #[test]
    fn test_exact_parse() {
        use std::str::FromStr;
        let id1: ID = ID::try_from([0x01]).unwrap();
        let ts1 = Timestamp::new(NTP64(100), id1);
        assert_eq!(ts1, Timestamp::from_str(&ts1.to_string()).unwrap());
    }

[chachi]

Hmm on further investigation, this may be due to the nanosecond rounding more than the +1

[JEnoch]

Right, the root cause is the rounding.

The Display trait and thus the ToString trait convert the NTP64 to a RFC3339 representation to be more "human readable".
As the last 32 bits of a NTP64 is the fraction of 1 second, converting this fraction to any subdivision of the second requires to make a division by 2^32. This means that 1u64 as NTP64 represents approximately 0,232830644 nanoseconds.

Thus converting a NTP64 to string and then back to NTP64 will give back the same number in only for ~23,28% of the numbers.
e.g. this test succeeds:

    #[test]
    fn test_exact_parse() {
        use std::str::FromStr;
        let id1: ID = ID::try_from([0x01]).unwrap();
        let ts1 = Timestamp::new(NTP64(104), id1);
        assert_eq!(ts1, Timestamp::from_str(&ts1.to_string()).unwrap());
    }