Re: 'fstat' compatibility issue

[Michael D. Godfrey]

On 2/17/10 3:10 PM, John W. Eaton wrote:
> On 17-Feb-2010, S ren Hauberg wrote:
>
> | It was mentioned on the Octave-Forge list that the 'statistics' package
> | has a function called 'fstat' that clashes with the function of the same
> | name in Octave. This is bad.
> | 
> | Matlab has a function called 'fstat' that does the same as the function
> | in the 'statistics' package. For compatibility, I think we should rename
> | the 'fstat' function in core Octave into something else. This will,
> | however, break applications that use the 'fstat' function from core
> | Octave.
> | 
> | Any thoughts on how to deal with this issue?
>
> If we had namespaces, I'd be willing to move all the OS-related
> functions into something like os.FCN or posix.FCN or similar.
>
> jwe
>   
One approach for this specific case is to use the fact that
the Octave fstat requires one argument (a FID). The matlab fstat
and, I assume, the Octave-forge one require 2 arguments (V1,V2).

So, the Octave fstat could be modified so that if it gets one
argument it acts as now, but with a warning that it has been replaced
by ???. If it gets 2 arguments it does the Octave-forge stats function.

This would minimize the problems for "legacy" Octave code. Also, I
am not sure that namespaces, though very nice, would provide a
similar fix what would not require some change to legacy code.

Michael