Inconsistency between period_to_seconds and seconds_to_period

[larmarange]

period_to_seconds considers that a year is equal 365 days while seconds_to_period considers that a year is equal to 365.25 days.

Therefore:

> seconds_to_period(period_to_seconds(years(2)))
[1] "1y 0m 364d 18H 0M 0S"

[vspinu]

I don't quite understand why to use years in seconds_to_period in the first place. The code is not using months. I think it should not use years either.

[vspinu]

Ok. I am removing years from there. We will probably need to add an optional argument that users could also use an approximate conversion to years and months. I am not sure on a clean interface yet. Probably something like units = c("day", "hour", "minute", "second") which is a bit verbose, and/or units = "dhms".