On 3/15/2011 3:46 AM, Jon Blower wrote:
> Could you not just use "days since" and use a variable number of days on the axis?
>
> 31 days since 1930-01-01 = 1930-02-01
> 59 days since 1930-01-01 = 1930-03-01
>
> This would correctly reflect the variable length of months, but puts a little more burden on the data provider to do the date arithmetic (not too hard given modern libraries).
>
> Or is there a use case that would demand the use of "months/years since"?
I guess this is Dave's case 1.
> (Also, are these time units really "fuzzy" as your subject line implies? I think they are just "irregularly spaced".)
fuzzy in the sense that they are not really constant, though udunits
assumes they are.
> Jon
>
> -----Original Message-----
> From: cf-metadata-bounces at cgd.ucar.edu [mailto:cf-metadata-bounces at cgd.ucar.edu] On Behalf Of John Caron
> Sent: 15 March 2011 00:31
> To: cf-metadata at cgd.ucar.edu
> Subject: [CF-metadata] udunits handling of fuzzy time units
>
> because udunits converts units like months and years to a fixed number
> of seconds, one cant really use things like months and years as units,
> since you get things like this:
>
> 0 months since 1930-01-01 == 1930-01-01T00:00:00Z
> 1 months since 1930-01-01 == 1930-01-31T10:29:03Z
> 2 months since 1930-01-01 == 1930-03-02T20:58:07Z
> 3 months since 1930-01-01 == 1930-04-02T07:27:11Z
> 4 months since 1930-01-01 == 1930-05-02T17:56:15Z
> 5 months since 1930-01-01 == 1930-06-02T04:25:19Z
> 6 months since 1930-01-01 == 1930-07-02T14:54:22Z
> 7 months since 1930-01-01 == 1930-08-02T01:23:26Z
> 8 months since 1930-01-01 == 1930-09-01T11:52:30Z
> 9 months since 1930-01-01 == 1930-10-01T22:21:34Z
> 10 months since 1930-01-01 == 1930-11-01T08:50:38Z
> 11 months since 1930-01-01 == 1930-12-01T19:19:42Z
>
> 0 years since 1850-01-01 == 1850-01-01T00:00:00Z
> 10 years since 1850-01-01 == 1860-01-01T10:07:39Z
> 20 years since 1850-01-01 == 1869-12-31T20:15:19Z
> 30 years since 1850-01-01 == 1880-01-01T06:22:59Z
> 40 years since 1850-01-01 == 1889-12-31T16:30:38Z
> 50 years since 1850-01-01 == 1900-01-01T02:38:18Z
> 60 years since 1850-01-01 == 1910-01-01T12:45:58Z
> 70 years since 1850-01-01 == 1920-01-01T22:53:38Z
> 80 years since 1850-01-01 == 1930-01-01T09:01:17Z
> 90 years since 1850-01-01 == 1940-01-01T19:08:57Z
>
> whereas you might really want:
>
> 0 months since 1930-01-01 = 1930-01-01T00:00Z
> 1 months since 1930-01-01 = 1930-02-01T00:00Z
> 2 months since 1930-01-01 = 1930-03-01T00:00Z
> 3 months since 1930-01-01 = 1930-04-01T00:00Z
> 4 months since 1930-01-01 = 1930-05-01T00:00Z
> 5 months since 1930-01-01 = 1930-06-01T00:00Z
> 6 months since 1930-01-01 = 1930-07-01T00:00Z
> 7 months since 1930-01-01 = 1930-08-01T00:00Z
> 8 months since 1930-01-01 = 1930-09-01T00:00Z
> 9 months since 1930-01-01 = 1930-10-01T00:00Z
> 10 months since 1930-01-01 = 1930-11-01T00:00Z
> 11 months since 1930-01-01 = 1930-12-01T00:00Z
>
> 0 years since 1850-01-01 = 1850-01-01T00:00Z
> 10 years since 1850-01-01 = 1860-01-01T00:00Z
> 20 years since 1850-01-01 = 1870-01-01T00:00Z
> 30 years since 1850-01-01 = 1880-01-01T00:00Z
> 40 years since 1850-01-01 = 1890-01-01T00:00Z
> 50 years since 1850-01-01 = 1900-01-01T00:00Z
> 60 years since 1850-01-01 = 1910-01-01T00:00Z
> 70 years since 1850-01-01 = 1920-01-01T00:00Z
> 80 years since 1850-01-01 = 1930-01-01T00:00Z
> 90 years since 1850-01-01 = 1940-01-01T00:00Z
>
> the solutions that occur to me are:
>
> 1) use ISO date strings
> 2) change the way udunits works
> 3) create another syntax that can be distinguished from udunits and make
> it work like the above
> 4) something else i havent thought of
>
> any ideas?
>
>
>
> _______________________________________________
> CF-metadata mailing list
> CF-metadata at cgd.ucar.edu
> http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
Received on Tue Mar 15 2011 - 06:54:09 GMT