Units in date2num / num2date

[dpl]

In the documentation for num2date and date2num, it says that "units" in the arguments list should be: "a string of the form 'time units since reference time' describing the time units. time units can be days, hours, minutes, seconds, milliseconds or microseconds." Now I on several occasions encountered data where the time unit is set as, e.g., "months since yyyy-mm-dd hh:mm:ss". Is there any way or planned implementation to deal with such longer-timestep units as well?

[mdecker]

The problem with time units such as "months" is that they are not well defined. In contrast to days, hours, etc. the length of a month depends on the calendar used and even varies between different months.
The alternative is to define a month exactly as 1/12th of a year as udunits does. This "solution" is not very intuitive to the user will will likely lead to misunderstandings.

A few words about this issue can be found in the CF conventions: http://cfconventions.org/Data/cf-conventions/cf-conventions-1.6/build/cf-conventions.html#time-coordinate

We recommend that the unit year be used with caution. The Udunits package defines a
year to be exactly 365.242198781 days (the interval between 2 successive passages of
the sun through vernal equinox). It is not a calendar year. Udunits includes the following
definitions for years: a common_year is 365 days, a leap_year is 366 days, a Julian_year
is 365.25 days, and a Gregorian_year is 365.2425 days.

For similar reasons the unit month , which is defined in udunits.dat to be exactly year/12,
should also be used with caution. 

[jswhit]

We could convert "months since" to "seconds since" using 1 month = year/12 = 2629743.831225 seconds. Same for "weeks since" and "years since". This is how netcdf-java does it (http://www.unidata.ucar.edu/software/thredds/current/netcdf-java/CDM/CalendarDateTime.html). It may lead to confusion though, and cause more problems than it solves.

[shoyer]

My two cents is that it's better to give an error and force users to convert such ill-defined units explicitly.

[davidhassell]

I would agree. The Udunits definition of "month" is practically never what is meant by encoders of "months since ...", so giving an error would be safe and clear.

[jswhit]

We already do throw an error, so it seems like no change in the current behavior is needed. Closing now...

[dpl]

Thanks for all replies - by the way, is there something like a users mailing list?

[jswhit]

There's no mailing list right now. Wish github had a better mechanism for discussions.

[bennoblumenthal]

While the udunits definition of month is not what is ever meant, that is a udunits flaw, not a flaw in the standard.

units of months since with a calendar of 360_day is perfectly well defined and there is no reason why this cannot be properly coded.

[jswhit]

A pull request allowing months since when calendar is 360_day would be welcome.

[rabernat]

👍 to @bennoblumenthal's comment.