core.time.Duration does not support ISO8601 representation

[Herringway]

Much like SysTime in Phobos, Duration should support fromISOString and toISOString methods to convert from/to ISO8601 representations. These are useful for serialization and other forms of interoperability.

One concern is that years and months are too ambiguous in this context, but it should be acceptable to reject or ignore them. There is precedence for this in other languages's standard libraries like Java's.

[ichordev]

Durations represent an amount of time, not a date. Converting them to/from dates doesn't make sense. A Duration can also represent a negative amount of time.

[Herringway]

Durations represent an amount of time, not a date. Converting them to/from dates doesn't make sense. A Duration can also represent a negative amount of time.

Dates are not the only thing covered by ISO8601. It specifies representations for durations as well, such as PT1M for 1 minute.

[ichordev]

ISO8601 […] specifies representations for durations as well, such as PT1M for 1 minute.

Sort of. It specifies time intervals which happen to contain durations, but those durations are only supposed to be used in time intervals, which include a date. As such, the durations are required to account for changing time zones, differing year/month lengths, etc. Duration in D has no concept of these things and therefore can’t really implement the standard properly except as a sort of lingua franca. I’m not sure this is massively useful in the first place though. Are there lots of durations out there saved using this format? If so, how do they tackle these issues?

[ichordev]

Also I don’t see anything about provisions for negative durations?

[Herringway]

ISO8601 […] specifies representations for durations as well, such as PT1M for 1 minute.

Sort of. It specifies time intervals which happen to contain durations, but those durations are only supposed to be used in time intervals, which include a date. As such, the durations are required to account for changing time zones, differing year/month lengths, etc. Duration in D has no concept of these things and therefore can’t really implement the standard properly except as a sort of lingua franca. I’m not sure this is massively useful in the first place though. Are there lots of durations out there saved using this format? If so, how do they tackle these issues?

Where are you getting this information from? ISO8601 very explicitly describes standalone durations. Intervals make use of these durations, but it's not exclusively part of them.

I provided a link to documentation for java's standard library support for this in the first post, which simply doesn't support years and months and additionally supports negative durations.

[ichordev]

ISO8601 very explicitly describes standalone durations.

ISO8601-1:2019 does not:

ISO8601-1:2019 explicitly defines durations as non-negative, and only as they relate to a time interval:

Durations are elaborated in 5.5.2, a subsection of 5.5 (the section on time intervals). There’s really no room to say that this document intends durations to be used independently of time intervals unless you ignore what it says and how it is structured.

I provided a link to documentation for java's standard library

This is an example of one library that adapts ISO8601’s incidental duration definition, I’m looking for examples of this being used for data exchange, like in a file format or network transmission. We can’t do stupid things just because of Java.