Re: Cleaning up C++ input validation of strings

[John W. Eaton]

On 02/17/2015 12:32 PM, Rik wrote:

> I'm sure I haven't gotten everywhere, but I have been trying to improve the
> input validation in Octave for the reasons above.  In cases where the
> function is expecting an option string, but gets something else, Octave
> should issue an error and stop so the problem can be corrected.

I could be mistaken, but the notion of conversions was there from the 
beginning because at the time I think Matlab often allowed numbers to be 
converted to strings.  So there is the possibility that something that 
would not return true for "is_string" (like a numeric array) would 
nevertheless have a string_value method that could produce a string. 
But that doesn't seem terribly useful.

> In the cases I mean to clean up, the is_string() function is already in
> place to do this:
>
> if (arg.is_string ())
> ...
> else
>    error ("OPTION must be a string")
>
> It's in the ellipsis code that I don't think it is worth checking
> error_state because I already know the arg is a string and string_value
> should therefore be able to extract a value without problems and without
> the need to check error_state.

Yes, I agree.

jwe