Re: Constants in functions?

[Kai Torben Ohlhus]

On 3/7/20 7:13 AM, Nicklas Karlsson wrote:
>> On 3/6/20 3:36 AM, Sergei Steshenko via Help list for GNU Octave wrote:
>>>
>>> On 05/03/2020 20:12, Windhorn, Allen E [ACIM/LSA/MKT] wrote:
>>>> What's the accepted way to define a constant inside a function?  Example:
>>>> ____________________________________
>>>> function [sc, err] = symmcomp(abc)
>>>>    % Function accepts [abc] complex components and delivers symmetrical
>>>>    % components as a vector [x0, x1, x2] of complex values
>>>>    persistent a = complex(-0.5, sqrt(3)/2);    % Rotation vector
>>>>    %
>>>>    % Build conversion matrix_type (constant)
>>>>    persistent A = [1, 1, 1;
>>>>            1, a, a^2;
>>>>            1, a^2, a]./3;
>>>>    %
>>>>    % Make abc a column vector
>>>>    if (length(abc)>3)
>>>>      sc = [];
>>>>      err = "Too many components!";
>>>>    elseif (length(abc)<3)
>>>>      sc = [];
>>>>      err = "Too few components!";
>>>>    else
>>>>      abc = reshape(abc,[3,1]);
>>>>      sc = A*abc;
>>>>      err = "";
>>>>    endif
>>>> endfunction
>>>> ________________________________________
>>>>
>>>> Here a is a constant and A is a constant matrix.  I used persistent in
>>>> hopes
>>>> it would prevent them from being redefined every time the function is
>>>> called.
>>>> In C++ I would use the "const" keyword.
>>>>
>>>> Regards,
>>>> Allen
>>>
>>>
>>> An ugly (because IIRC Octave doesn't have nested functions) way:
>>>
>>> function retval = TWO()
>>>
>>>   retval = 2;
>>>
>>> endfunction
>>>
>>> .
>>>
>>> The above function will return 2. You can similarly return matrix, row o
>>> column vector, etc.
>>>
>>>
>>> --Sergei.
>>>
>>
>> Dear Allen,
>>
>> Your code seems fine to me.  Using "persistent" [1] avoids the variables
>> to be redefined each time the function is invoked.  And if you do not
>> change it's value, it is "constant".  So what it is exactly you are
>> worried about?
> 
> If it's supposed to not change Octave will make complaint in case of a 
> misstake. It is very common in other programming languages and if there is 
> compiler it is usually able to figure this already before program is run.
> 
>> ...
>> Neither Octave nor Matlab have a concept of C++ "const" variables.  Only
>> within classdef classes you can define constant properties [4].  But the
>> overhead may not be justified.
> 
> Consider this a missing feature though it would break compability with 
> Matlab. Then value is not supposed to change I usually always make it a 
> constant and consider it a good programming practice.
> 
> 
> Nicklas Karlsson
> 


Please keep the mailing list in the CC so others can benefit from our
conversation.

You can file a bug report about this.  But as it is incompatible with
Matlab and the interpreter currently does not benefit from a constant
declaration, I do not think many people are keen to work on this extension.

Workarounds to avoid mistakes are discussed in this thread.

HTH,
Kai