Re: deleting variables

Hi Mateus,
Yes these are good ideas. We have had thoughts about how to make the
'list_' family of
procedures more useful. It's somewhat unfortunate that NCL's rather
rigid differentiation
between function and procedure does not allow for invoking a function
without specifying
a return variable. We cannot maintain backwards compatibility if we
change a procedure into
a procedure into a function.
So we may need to make a new set of corresponding functions to
accomplish your suggestion.
  -dave

On Mar 2, 2007, at 1:10 PM, Mateus da Silva Teixeira wrote:

> David,
>
> A new built-in procedure would be interesting. So putting it to accept
> arguments like isvar function will make life easier.
> Since we talking about manipulation of variables I would like to
> suggest a modification in the procedures list_vars, list_procfuncs,
> and other that provide usefull information about the NCL environment.
> Instead of procedures, they could be "promoted" to functions. So,
> users could obtain useful information about variables, procedures and
> functions already defined and work with this information. It would
> increase the possibility for more general user-created functions.
> With those "new" tools, we'll be able to make things like:
>
> list = list_vars() ; returning a array with variable names in the
> ; first column and their types in the
> second column
> vars_to_delete = ind( ... ) ; the indexes of the variables that
> should be deleted, provided a selection rule and
> delete( list( vars_to_delete) )
>
> Abbreviating the subject, there is some procedures that have their
> utility underestimated. So, modifications like that could improve the
> power of NCL language.
>
> Mateus
>
> David Ian Brown escreveu:
>> Mateus and Dennis,
>>
>> Last things first: the "then" in if-then-[else]-end if constructions
>> is indeed optional "syntactic sugar" (and it is an undocumented
>> feature of NCL).
>> We should probably update the documentation.
>>
>> Regarding the possibility of a procedure that could delete multiple
>> variables,
>> presumably passed in as some kind of list.
>> Consider the following:
>>
>> a = 1
>> b = a
>> delete(b)
>>
>> Since NCL, like many languages, maintains reference counts, deleting
>> 'b'
>> does not get rid of 'a', and I think everyone would agree that this
>> is the proper
>> behavior. It is somewhat (though not exactly) the same situation if
>> you pass a
>> reference to a variable in some kind of parameter list. Anyway NCL
>> doesn't
>> like it if you have a procedure such as:
>>
>> delete_var(a)
>> begin
>> delete (a)
>> end
>>
>> What you could do, given NCL's existing state, is have a procedure
>> without
>> parameters that deletes variables defined at a higher scope. If you
>> typically use
>> a certain set of temporary variable names, you could, for instance,
>> define a
>> procedure like this:
>>
>> procedure delete_vars()
>> begin
>> if (isvar("temp_var_1")) then
>> delete(temp_var_1)
>> if (isvar("temp_var_2")) then
>> delete(temp_var_2)
>> ;etc.
>> end
>>
>> do while (.not. done)
>> temp_var_1 = <some array>
>> temp_var_2 = <another array>
>> <do stuff>
>> delete_vars()
>> end do
>>
>> This is not really satisfying however. In order to implement
>> something that would
>> be general, we would need to implement something else. Options
>> include:
>>
>> Implementing a new built-in procedure
>> procedure delete_vars(varnames : string)
>>
>> Implementing a function (and I don't know how hard this would be) that
>> returns a reference to a variable given its name without incrementing
>> the
>> reference count:
>>
>> var = get_primary_ref(varname : string)
>>
>> Implementing new syntax similar to the string ref syntax for file
>> variables that
>> would work for regular variables. Probably harder.
>> Then you could write, e.g.:
>> delete($varname$) where varname is a string
>>
>> With either of these last 2 options users could write their own
>> delete_vars type
>> procedures.
>>
>> But probably the simplest thing for us to do would be a new built-in
>> proc.
>>
>> -dave
>>
>> On Mar 2, 2007, at 10:07 AM, Dennis Shea wrote:
>>
>>>> In some scripts I have to delete many variables within DO loops. I
>>>> found
>>>> only one procedure that delete variables and I was thinking in make
>>>> a
>>>> procedure that delete a group of variables specified by a list.
>>>> But, I'm
>>>> not sure if I can access variables of the main program with a
>>>> function
>>>> and delete them. Is it possible?
>>>
>>> I am sure there is an NCL "rule" that states
>>> that variables not local to a function or
>>> procedure can not be deleted within a called procedure/function.
>>>
>>> In the past, I have asked for the "delete"
>>> procedure to be expanded
>>>
>>> a =
>>> b =
>>> c =
>>>
>>> delete(a)
>>> delete(b)
>>> delete(c)
>>>
>>> to
>>> delete(a, b, c)
>>>
>>>
>>> I was told this was not possible. :-(
>>>
>>> A delete that could operate on multiple variables
>>> would lead to cleaner codes.
>>>
>>> Note: variables locally created within procedures or functions
>>> are automatically deleted by NCL. The user need not
>>> explicitly delete them. This is a nice feature.
>>>
>>>
>>>>
>>>> By the way: in the NCL reference guide (online), in the part that
>>>> talk
>>>> about IF statement the syntax provided is:
>>>>
>>>> if(/scalar_logical_expression/) then
>>>>
>>>> / statement list /
>>>>
>>>> end if
>>>> if(/ scalar_logical_expression/) then
>>>>
>>>> / statement list /
>>>>
>>>> else
>>>>
>>>> / statement list /
>>>>
>>>> end if
>>>>
>>>> Seeing carefully one of my scripts I verify that my IF statement
>>>> hadn't
>>>> the THEN word, but the scripts still works. So, is the reference
>>>> manual
>>>> wrong don't showing that THEN is optional? Or is there a
>>>> malfunction in
>>>> IF statement?
>>>> _______________________________________________
>>>
>>>
>>> A developer will have to answer this one.
>>> I think the "then" is optional but
>>> the "end if" is required to complete the "block"
>>>
>>> fortran: if (s.eq.5) pi = 4*atan(1) ! no need for an "end if"
>>>
>>> NCL: if (s.eq.5) then
>>> pi = ..
>>> end if
>>> also
>>> if (s.eq.5) pi = 4*atan(1)
>>> end if ; <=== required
>>>
>>> if (isvar("pi")) then ; just in case s.ne.5 :-)
>>> print(pi)
>>> end if
>>>
>>>
>>> _______________________________________________
>>> ncl-talk mailing list
>>> ncl-talk_at_ucar.edu
>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>
>>
>

_______________________________________________
ncl-talk mailing list
ncl-talk_at_ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Fri Mar 02 2007 - 16:01:04 MST