Using unit 'week' requires that a relative reference time be given, but none was provided

[azzamsa]

Hi, 👋🏽

Why this is the case?
Why the chrono code works but jiff doesn't.

    let dt = 1.weeks();
    dbg!(&dt);
    let dt = dt.total(Unit::Second);
    dbg!(&dt);

    let dt = chrono::Duration::weeks(1);
    dbg!(&dt);
    let dt = dt.num_seconds();
    dbg!(&dt);

[src/main.rs:21:5] &dt = P1w
[src/main.rs:23:5] &dt = Err(
    Error {
        inner: ErrorInner {
            kind: Adhoc(
                using unit 'week' requires that a relative reference time be given, but none was provided,
            ),
            cause: None,
        },
    },
)
[src/main.rs:26:5] &dt = TimeDelta {
    secs: 604800,
    nanos: 0,
}
[src/main.rs:28:5] &dt = 604800

[azzamsa]

Jiff specifically avoids returning "unbalanced" spans like 2 months 35 days. So add a date to the mix and things work fine:

Explained here

[BurntSushi]

Yeah basically you need to have a relative date to do transformations like this. Jiff is preventing you from introducing bugs in subtle corner cases. Indeed, chrono-humanize just assumes that all months are 30 days, which is obviously wrong:

use chrono_humanize::{Accuracy, HumanTime, Tense};

fn main() {
    let dt = chrono::Duration::days(31);
    let ht = HumanTime::from(dt);
    let text = ht.to_text_en(Accuracy::Precise, Tense::Present);
    println!("{text}");
}

Has this output:

1 month and 1 day

Some months are 31 days. Some are 29. Some are 28. Some are 30. The only correct way to know the duration of a month is to interpret it relative to a particular date.

[azzamsa]

I have successfully ported chrono-humanize to jiff. But I still disable some tests due to this.

azzamsa/jiffy@a16f8c6

[BurntSushi]

It looks like you've copied chrono-humanize's mistake. :-( I would strongly urge you not to do that. Jiff expends so much effort in its API design to prevent folks from making that exact mistake.

[azzamsa]

It looks like you've copied chrono-humanize's mistake. :-( I would strongly urge you not to do that. Jiff expends so much effort in its API design to prevent folks from making that exact mistake.

You're right. Just after you mentioned this:

Indeed, chrono-humanize just assumes that all months are 30 days, which is obviously wrong:

I started thinking:

• I'll ask for your input on any corner cases I should be aware of.
• I'll simply use all of jiff's functions like get_days, get_months, etc. This way, it's just a thin wrapper, and jiff does all the heavy lifting. Jiff already provides everything I need, so there's no need to compute anything on my own.

[BurntSushi]

I'll simply use all of jiff's functions like get_days, get_months, etc. This way, it's just a thin wrapper, and jiff does all the heavy lifting. Jiff already provides everything I need, so there's no need to compute anything on my own.

Right! That is exactly the benefit of Span. It is not the normal "Integer number of nanoseconds" that you are probably used to. It keeps track of individual units. It was designed by the Temporal project to be compatible with RFC 5545, and so that any code for localizing the Span wouldn't actually need to do any duration math or make choices about which units to use.