Unexpected results from the duration format

[cbandy]

The OpenAPI definition of the duration format follows RFC 3339. The definition here is more relaxed, which is nice, but IsDuration allows text that produces unexpected values from ParseDuration and unmarshaling.

https://go.dev/play/p/0MhAO3Ife4I

I expected 1.5 hours to be 90 minutes, but this valid text is interpreted as 5 hours:

fmt.Println(time.ParseDuration("1.5h"))        // 1h30m0s <nil>
fmt.Println(strfmt.ParseDuration("1.5 hours")) // 5h0m0s <nil>

I expected 1.5 days to be 36 hours, but this valid text is interpreted as 120 hours:

fmt.Println(strfmt.ParseDuration("1.5 days")) // 120h0m0s <nil>
fmt.Println(1.5 * 24 * time.Hour)             // 36h0m0s

[fredbi]

I believe that at the time we implemented this, no such specification did exist.

I’ve also noticed the clear reference in oai 3 to the iso duration format, which honestly is awful.

Nevertheless a standard is a standard and i’d be happy with an alternate duration time (e.g “iso dutation”) that would validate strictly against the standard.

Regarding the other problem you report, about marshaling, it t comes from the fact that we don’t keep the original string. Once parsed, it has just become a time.Duration, so we have to somehow make an arbitrary choice when it comes to rendering it. Perhaps an extra method like Format(layout) would help?

[cbandy]

I understand why time.Duration presents days as hours. That's not a problem for me.

I'm reporting that ParseDuration accepts fractional numbers and then misinterprets them. For example, "1.5 hours" is parsed to 5.0*time.Hour.

[fredbi]

OK yes this use case had been entirely omited. Here is a PR that should address this limitation