Re: error function

[Shai Ayal]

On 5/15/07, John W. Eaton <address@hidden> wrote:
> On 15-May-2007, Shai Ayal wrote:
>
> | On 5/15/07, John W. Eaton <address@hidden> wrote:
> | > The string_value method calls error if it can't perform the
> | > conversion.
> |
> | but no error messge is printed right? does it call error ("") ?
>
> Yes, it calls error and prints a message.  The only function that is
> supposed to set error_state is error.
>
> | > Maybe you need
> | >
> | >  bool rgb_property::is_none (void) const;
> | >  bool rgb_property::is_auto (void) const;
> | >  const double *rgb_property::rgb_value (void) const;
> | >
> | > and then code that uses this class will just have to check if it cares
> | > about "none" and "auto".  I don't know what else to suggest unless the
> | > rgb function can somehow return reasonable values when the state is
> | > "none" or "auto".
> |
> | I thought of is_rgb and is_radio. However I recal thatusing is_*
> | functions was the way to determine octave_value type, but then I read
> | somewhere that the recomended way was to try and convert and than
> | check error_state, as above. Can you elaborate on why this is
> | recommended for octave_value?
>
> There are different cases where different things are appropriate.  For
> example, a function that always expects a double value as its first
> argument and for which everything else is an error should use
> something like
>
>   double x = arg.double_value ();
>   if (! error_state)
>     ...
>
> because that is better than checking for all possible types that might
> have methods which allow them to be converted to double values.
>
> But if a function can accept a character string or a double value as
> its first argument, then you are forced to use something like
>
>   if (arg.is_string ())
>     {
>       std::string s = arg.string_value ();
>     }
>   else
>     {
>       double = arg.double_value ();
>
>       if (! error_state)
>         ...
>     }
OK, I think I got it now. Thanks for the explanation


> Yes, it might be better if Octave used exceptions but Octave's
> interpreter was written before C++ had exceptions (or even a standard
> library).  Changing it now would be a lot of work.

no complaints here. I just wasn't sure.

Shai