Re: A standard for time and labelling problems

Unidata User Support (support@unidata.ucar.edu)
Wed, 24 Mar 93 14:14:37 -0700

>From: blincoln@SSESCO.com (Brian Lincoln)
Date: Wed, 24 Mar 93 15:14:05 -0600
Subject: Re: A standard for time and labelling problems

> >From: rakesh@lamont.ldgo.columbia.edu
> >Date: Wed, 24 Mar 93 09:49:12 -0500
> >Subject: Re: A standard for time and labelling problems
>
> We have used the Unix standard of time, i.e. seconds since Jan. 1, 1970.
> This way we can store time as one long integer (4 bytes) per time value,
> and are able to store time values in sec from year 1902 through year 2038.
> Unix utilities can then be used to convert time into any form you want, e.g.
> Date format: yyyy/mm/dd/hh:mm:ss
> Julian date format: yy/dd/hh:mm:ss
> String format: Ex: Fri May 15 16:12:19 1992

have you seen the UDUNITS stuff from ucar? its available at unidata.ucar.edu
and it has a format similar to, but not the same as you have above..
the order is a little more regular in the string and is
" units since int year int month int day int hour int min double sec int ucthr
int utcmin"
so " Seconds since 1992-5-15 16:12:19.0 -5:00"

for your example above...
and then a variable "time" is defined as a double offset.. im sure it
wouldnt be too much of a stretch to use another vardef .. NC_LONG or something
(if you think that 8 bytes is too many for a timestamp)
but the UDUNITS standard is double..

.. hope this isnt just noise...

bcl
blincoln@ssesco.com