Re: improved compatibility

[Andy Adler]

On Wed, 9 Jul 2003, John W. Eaton wrote:
> I would like to eliminate the problems associated with the built-in
> variables do_fortran_indexing, implicit_str_to_num_ok,
> implicit_num_to_str_ok, and ok_to_lose_imaginary_part by replacing
> them as follows:
>
>   do_fortran_indexing        -->  warn_fortran_indexing
>   implicit_num_to_str_ok     -->  warn_num_to_str
>   implicit_str_to_num_ok     -->  warn_str_to_num
>   do_to_lose_imaginary_part  -->  warn_imag_to_real

    ok_to_lose ...
>
> The code would be changed to work as if the old variables were all set
> to TRUE, except that users could request warnings by setting the new
> variables to TRUE (the default would be FALSE).
>
> This change would improve compatibility with Matlab and eliminate some
> of the problems we currently have with trying to write code that can
> work no matter what the settings for these variables happen to be.

Yes. This would be great!

Also, another thing that gets requested is a way to turn off warnings.

  For example, I often don't want
    octave:2> inv(zeros(2))
--> warning: inverse: matrix singular to machine precision, rcond = 0

So my questions is how these warnings may relate to a possible
variable to control issued warnings?


Andy