Negative dates parsing

[Yomguithereal]

Hello @BurntSushi, I often work with data from the humanities and BCE dates are not uncommon here. I stumbled upon this timestamp in Wikidata: -0280-01-01T00:00:00Z and was surprised that jiff was not able to parse it. I then stumbled upon this other discussion here: #302 and I seem to understand that according to ISO 8601, this date should have been written -000280-01-01T00:00:00Z that jiff parsed correctly. So I was wondering what your opinion on this was and whether you think those kind of formats should be supported. In which case, would you accept a contribution to deal with the issue? In any case, I suppose this can very well be handled from userland.

[BurntSushi]

This is indeed intentional. Jiff follows the Temporal ISO 8601 grammar, which requires that datetimes prior to 1 BCE are represented with a leading - and 6 digits for the year.

Negative timestamps are somewhat odd and a bit of a grey area in ISO 8601 proper. According to ISO 8601, -0280-01-01T00:00:00Z is not inherently invalid, but requires "mutual agreement" among the parties to agree upon it, including a specification on the number of digits to allow for the year.

Note though that Jiff's strptime support is a bit more flexible. For example, this works:

use jiff::{civil::DateTime, tz::TimeZone};

fn main() -> anyhow::Result<()> {
    let s = "-0280-01-01T00:00:00Z";
    let dt = DateTime::strptime("%Y-%m-%dT%H:%M:%SZ", s)?;
    let ts = dt.to_zoned(TimeZone::UTC)?.timestamp();
    dbg!(ts);

    Ok(())
}

Or, if you want to specifically allow up to 6 digits for the year (it defaults to 4), then you can use %06Y:

use jiff::{civil::DateTime, tz::TimeZone};

fn main() -> anyhow::Result<()> {
    let s = "-000280-01-01T00:00:00Z";
    let dt = DateTime::strptime("%06Y-%m-%dT%H:%M:%SZ", s)?;
    let ts = dt.to_zoned(TimeZone::UTC)?.timestamp();
    dbg!(ts);

    Ok(())
}

I then stumbled upon this other discussion here: #302 and I seem to understand that according to ISO 8601, this date should have been written -000280-01-01T00:00:00Z that jiff parsed correctly.

Did the error message from parsing -0280-01-01T00:00:00Z help at all? It does say that it expects a six digit year. How could it be improved?

[Yomguithereal]

Thanks for the detailed answer @BurntSushi, I will try to go the strptime way.

The error message for such a case is:

"-0280-01-01T00:00:00Z".parse::().unwrap_err().to_string() = "failed to parse year in date "-0280-01-01T00:00:00Z": failed to parse "0280-0" as year (a six digit integer): invalid digit, expected 0-9 but got -"

It certainly hints at the error being related to the year and that there is an issue related to the - in front of it.