Can the library parse durations in ISO8601 format?

[anthonylouisbsb]

The date library can parse durations in ISO8601 format in a similar way that already exists for dates?

Example: The user sents an input string in a P[n]Y[n]M[n]DT[n]H[n]M[n]S or P[n]W format and the library extracts the quantity of each value in period(the number of passed days, the number of passed years..), in a similar way of the java Joda Time library already does.

I checked in the wiki that is possible to create some structures related to periods, but I have to parse the input string with the period in ISO8601 format using regex.

[anthonylouisbsb]

I found an answer in StackOverflow that tells that this library does not parse periods,

[HowardHinnant]

That answer is correct. You'll have to use some other tool (such as regex) for a parse in this format. But once you get the information into arithmetic types, then you can use this library to build the duration from the arithmetic types.

[ecorm]

The C++ standard library does not have all the data structures necessary to store the information parsed from an ISO 8601 period. std::chrono::year_month_day and std::chrono::hh_mm_ss would not be suitable to store arbitrary years, months, days, hours, minutes, and seconds. It would need something like a Luxon Duration that can differentiate between what they call "calendar math" and "time math". std::tm would not be able to store non-integral values.

Another example is how Java has both Duration and Period that parse ISO 8601 periods differently.

I think it would be possible for this library (or a future std::chrono library) to parse a subset of ISO 8601 periods in the same way as Java's Period, limited to days, hours, minutes, and seconds into a std::chrono::duration without ambiguity. I don't know why Javs's Period can't parse from weeks; I don't see the ambiguity. EDIT: It's because of ISO 8601 itself.

[HowardHinnant]

Once parsed, this library can do both calendrical math and chronological math (time math as they call it). But yes, the parsing is pretty much limited to strptime parsing (with a few extensions).

[ecorm]

@HowardHinnant Do you have any idea why Java's Duration.parse does not accept ISO 8601 weeks? Sorry, there's no direct link to the method's documentation, but you can CTRL+F "parse".

[HowardHinnant]

No idea really. My best guess is that they didn't want to venture too close to months (calendrical arithmetic). :-)

[ecorm]

It turns out it's the ISO 8601 standard itself that constrains it to either PnW or PnYnMnDTnHnMnS. See 4.4.3.2 in this draft of the 2016 version of the standard.

I cannot seem to find a rationale for the exclusion of the W in the PnYnMnDTnHnMnS expression. Perhaps it's due to human factors (i.e. avoid confusion).