[ncl-talk] Issue with cd_inv_calendar return values

Dennis Shea shea at ucar.edu
Thu Jul 30 10:26:59 MDT 2015


David Pierce's (Scripps) WWW

http://meteora.ucsd.edu/~pierce/


provides links to

CalCalcs routines

He provides some interesting information.

D

On Thu, Jul 30, 2015 at 10:12 AM, Mary Haley <haley at ucar.edu> wrote:
> Hi Saji,
>
> There are definitely issues when getting down to the seconds level, because
> the internal C code doesn't consistently handle the case where you have a
> seconds that are an even multiple of 60. Sometimes it "resets" to 0 and
> increases the minute value by one, and sometimes it doesn't. I have a note
> in to the developers of this code, but haven't heard anything yet.
>
> I can't say for certain that ut_inv_calendar handles this better, but if
> it's working for you, and you don't need special calendars, then you should
> be okay.
>
> --Mary
>
> On Wed, Jul 22, 2015 at 7:54 PM, Saji Hameed <saji.nh at gmail.com> wrote:
>>
>> Hi Kyle,
>>
>> Thanks.
>>
>> > How are you calling cd_inv_calendar in your code? Are you certain that
>> > partial days (e.g., hours, minutes, and/or seconds not equal to 0) aren't >
>> > being fed into the function?
>>
>> There is no issue with the parameters that I pass on to cd_inv_calendar,
>> but one can never be absolutely sure. The hours, minutes and seconds are
>> always 0 in my case.  Since then I have switched back to ut_inv_calendar and
>> do not face the same problem again.
>>
>> > I've used this function countless times with "hours since xxx" and "days
>> > since xxxx" and never experienced the issue you are describing.
>>
>> Sure - this happens rarely. I guess it  must be some kind of memory leak
>> issue. The funny thing is that I am not able to simulate this "error"
>> by using a test script.
>>
>> saji
>>
>> On Wed, Jul 22, 2015 at 11:13 PM, Kyle Griffin <ksgriffin2 at wisc.edu>
>> wrote:
>>>
>>> Hi Saji,
>>>
>>> How are you calling cd_inv_calendar in your code? Are you certain that
>>> partial days (e.g., hours, minutes, and/or seconds not equal to 0) aren't
>>> being fed into the function? I've used this function countless times with
>>> "hours since xxx" and "days since xxxx" and never experienced the issue you
>>> are describing.
>>>
>>>
>>> Kyle
>>>
>>> ----------------------------------------
>>> Kyle S. Griffin
>>> Department of Atmospheric and Oceanic Sciences
>>> University of Wisconsin - Madison
>>> Room 1421
>>> 1225 W Dayton St, Madison, WI 53706
>>> Email: ksgriffin2 at wisc.edu
>>>
>>> On Tue, Jul 21, 2015 at 6:32 AM, Saji Hameed <saji.nh at gmail.com> wrote:
>>>>
>>>> Dear NCL developers,
>>>>
>>>> I have a library that uses cd_inv_calendar for some critical parts. Once
>>>> in a blue moon, my
>>>> script crashes due to problems with values returned by cd_inv_calendar.
>>>> Instead of returning
>>>> a whole number (for example 37785 days since 1900-01-01 return value is
>>>> double), it sometimes returns a fraction (37784.xxxx).
>>>>
>>>> I use NCL version 6.3.0 on a Linux box.
>>>>
>>>> It is quite difficult to reproduce this issue because it happens rarely
>>>> and randomly. I notice that
>>>> http://www.ncl.ucar.edu/current_release.shtml#KnownBugs6.3.0  notes a bug
>>>> with cd_calendar when the units are specified as "minutes since" or "seconds
>>>> since".  In my case the units are
>>>> "days since 1900-01-01".
>>>>
>>>> I will use ut_calendar for the time being, but thought that I would
>>>> notify the developers about this issue.
>>>>
>>>> saji
>>>>
>>>> _______________________________________________
>>>> ncl-talk mailing list
>>>> ncl-talk at ucar.edu
>>>> List instructions, subscriber options, unsubscribe:
>>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>>>
>>>
>>
>>
>> _______________________________________________
>> ncl-talk mailing list
>> ncl-talk at ucar.edu
>> List instructions, subscriber options, unsubscribe:
>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>
>
>
> _______________________________________________
> ncl-talk mailing list
> ncl-talk at ucar.edu
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>


More information about the ncl-talk mailing list