[Scilab-users] Strange list error

[Stefan Du Rietz]

On 2014-01-28 17:13, Adrien Vogt-Schilb wrote:
--------------------
> On 28/01/2014 16:43, Stefan Du Rietz wrote:
>> On 2014-01-28 09:54, Adrien Vogt-Schilb wrote:
>> --------------------
>>> On 28/01/2014 09:41, Stefan Du Rietz wrote:
>>>> On 2014-01-27 22:43, Serge Steer wrote:
>>>> --------------------
>>>>> Le 27/01/2014 19:05, Stefan Du Rietz a écrit :
>>>>>>
>>>>> Under pause, you cannot modify a variable which is defined in the
>>>>> calling context
>>>>> so in such a context args is not a list as you expect but an empty
>>>>> array which is implicitely created
>>>>>
>>>>> To make it work
>>>>> you must do
>>>>> -1-> args; //makes a local copy of args
>>>>> -1->args(1) = null()
>>>>>
>>>>> Serge
>>>>
>>>> What do you mean by "defined in the calling context"?
>>>> I can modify this:
>>>>
>>>> -1->isfast
>>>>  isfast  =
>>>>     0.    0.
>>>> -1->isfast(1) = 1
>>>>  isfast  =
>>>>     1.    0.
>>>>
>>>> Stefan
>>>>
>>>>
>>>
>>> Stefan, you may want to have a look at
>>> http://wiki.scilab.org/howto/global%20and%20local%20variables
>>> This page explains what Serge means i believe.
>>>
>>> Adrien
>>>
>> Thank you Adrien, but I have had many looks at it ;-)
>>
>> I still don't understand the difference between my variables args
>> and isfast. They are both defined in the function.
>>
>> args: input values of the function are assigned to the corresponding
>> local variables of the function (local variables and scoping 5.)
>>
>> isfast: local variables defined at level_N remain local at that
>> level. (local variables and scoping 2.).
>>
>> What am I missing? Am I just stupid?
>>
>> Stefan
>
> I believe the local and global variable stuff explains why you are
> able to reproduce your bug only under pause, as documented here
> http://wiki.scilab.org/howto/global%20and%20local%20variables (gosh,
> the page does not explain that pause and functions behave the same
> way... this is probably our problem here)
> This means that there is not "strange list error", only a confusing
> behaviour of scilab under pause. If i am correct, none of your
> functions suddenly stopped working.

The function suddenly stopped working because of an error by me, which 
I reported 2014-01-25 21:07. It was when I debugged my function that 
this *pause* problem showed up.

> Coming back to the original post:
>
> -1->args(1) = null();
> args(1) = null();
>                   !--error 44
> Wrong argument #2.
>
> and when I checked:
> -1->typeof(args)
>   ans  =
>   list
>
> Can anybody explain that?
>
> The explanation is that you just created a new pause environment,
> where args do no exist yet.

But why only args (from varargin)? isfast does exist. And if I add a 
variable *a* before varargin, I can change that after *pause*.

Changed in testlisterror.sci:
function isfast = testlisterror(a, varargin)

>   The line -1->args(1) = null(); (witten just after entering in pause)
> then behaves exactly the same way as if you open a new scilab consoloe
> and write args(1) = null()
>
> In other words, you cannot debug your function using pause like that.

I have long ago understood pause as a means to be able to be in the 
environment of the function and manipulate things from the command 
line, and I have mostly been successful in doing that. Here, you 
cannot do that and there is nowhere a comment on it.

> In the precise case of your function, one thing that would work is to
> pause the function just after entering (before the line args =
> varargin;), or to use the solution Serge gave you, that is replace the
> lines 8 and 9
> pause
>      args(1) = null();
> with
> pause
> args; //makes a local copy of args in the new pause environment
>      args(1) = null();
>

And how do I know which variables need that? See above about *a*.

> Please make sure you understand that nothing discussed here as nothing
> to do with your function, (which i did not read and i bet serge did
> not read).
> The bug comes from how you use pause to try to debug your function
> (and yes, this is very annoying as it prevents to use pause to debug
> scilab seamlessly.).
>
> I'm confused whether this answer your original question, and whether
> you have more questions
>
> Adrien
>
If it is so annoying, should we not report a bug?

Stefan