[ncl-talk] Coordinate subscripting with mixed types

[Dave Allured - NOAA Affiliate]

Mary,

The "totype" workaround is working fine, it's just that the issue keeps
coming back whenever we encounter type variations in new data and program
combinations.  For this and my original reasons, it would be helpful to
have this feature in the language.

--Dave

On Wed, Nov 19, 2014 at 3:37 PM, Mary Haley <haley at ucar.edu> wrote:

> Hi Dave,
>
> Our JIRA ticket database is down for maintenance at the moment, so I can't
> check the status.  I'm pretty sure nothing has been done on it yet.  Dave
> Brown will be the one to answer this, but he's not in this week.
>
> You had mentioned that "totype" would work as a temporary fix.  Is this
> still working for you for now?
>
> --Mary
>
>
> On Wed, Nov 19, 2014 at 11:22 AM, Dave Allured - NOAA Affiliate <
> dave.allured at noaa.gov> wrote:
>
>> NCL support,
>>
>> Is there any progress on this request?  This issue keeps coming up in my
>> work group.  We perform a lot of generalized file handling.  Thank you.
>>
>> --Dave
>>
>> On Thu, Aug 22, 2013 at 12:30 PM, David Brown <dbrown at ucar.edu> wrote:
>>
>>> Hi Dave,
>>> This is a very good suggestion. We will create a ticket for it.
>>>  -dave
>>>
>>> On Aug 22, 2013, at 12:06 PM, Dave Allured - NOAA Affiliate <
>>> dave.allured at noaa.gov> wrote:
>>>
>>> > NCL support,
>>> >
>>> > This is a request for improved support for mixed types in coordinate
>>> > subscripting.  For example, it is legal to subscript a double
>>> > coordinate with floats, but going the other way is an error.  This
>>> > shows that type handling in simple NCL relational operators is more
>>> > liberal than in coordinate subscripting:
>>> >
>>> > lats = (/ 0., 10., 20., 30., 40., 50., 60. /)
>>> > x    = (/ 0,  111, 222, 333, 444, 555, 666 /)
>>> > x!0  = "lat"
>>> > y    = x
>>> >
>>> > x&lat = tofloat  (lats)
>>> > y&lat = todouble (lats)
>>> >
>>> > f1 = 20.     ; floats
>>> > f2 = 40.
>>> > d1 = 20d     ; doubles
>>> > d2 = 40d
>>> >
>>> > ind1 = ind (y&lat .ge. f1 .and. y&lat .le. f2)
>>> > print (ind1+"")
>>> > print ("")
>>> >
>>> > ind2 = ind (x&lat .ge. d1 .and. x&lat .le. d2)
>>> > print (ind2+"")
>>> > print ("")
>>> >
>>> > sub1 = y({f1:f2})    ; float subscripts double
>>> > print (sub1&lat+" "+typeof(sub1&lat))
>>> > print ("")
>>> >
>>> > sub2 = x({d1:d2})    ; double subscripts float
>>> > print (sub2&lat+" "+typeof(sub2&lat))
>>> >
>>> > Result:
>>> >
>>> > NCAR Command Language Version 6.1.2
>>> > (0)   2
>>> > (1)   3
>>> > (2)   4
>>> > (0)
>>> > (0)   2
>>> > (1)   3
>>> > (2)   4
>>> > (0)
>>> > (0)   20 double
>>> > (1)   30 double
>>> > (2)   40 double
>>> > (0)
>>> > fatal:Coordinate subscript type mismatch. Subscript (0) can not be
>>> > coerced to type of coordinate variable
>>> > fatal:["Execute.c":8128]:Execute: Error occurred at or near line 26 in
>>> > file mixed-mode-coord-subscripts.ncl
>>> >
>>> > I think it would be generally helpful if coordinate subscripting could
>>> > be broadened to support all reasonable cases of mixed types.  At a
>>> > minimum, please support the same combinations as for the relational
>>> > operators.  I know that there are simple workarounds, but this
>>> > improvement would benefit simpler user code and generalized file
>>> > support.  Thanks for your consideration.
>>> >
>>> > --Dave A.
>>> > NOAA/PSD/CIRES Climate Analysis Branch
>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20141119/b234f939/attachment.html