From owner-bug-octave at bevo dot che dot wisc dot edu  Wed Jan 22 11:24:28 1997
Subject: Re: erfinv.m
From: Andreas Weingessel <Andreas dot Weingessel at ci dot tuwien dot ac dot at>
To: "John W. Eaton" <jwe at bevo dot che dot wisc dot edu>
Cc: bug-octave at bevo dot che dot wisc dot edu
Date: Wed, 22 Jan 1997 18:26:00 +0100

>>>>> On Wed, 22 Jan 1997 11:04:11 -0600,
>>>>> John W Eaton wrote:

> On 22-Jan-1997, Andreas Weingessel <weingessel at ci dot tuwien dot ac dot at> wrote:
> : The function "erfinv" yields an error of the form
> : error: A([]) = X: X must also be an empty matrix
> : error: evaluating assignment expression near line 40, column 22
> : error: called from `erfinv' in file `/usr/local/lib/octave/2.0.1/m/specfun/erfinv.m'
> : 
> : This is caused by the fact that an assignment of a 0x1-matrix to a
> : 0x0-matrix worked in octave-1.1.1, but does not work now.
> : 
> : Repeat-By:
> : ---------
> : 
> : octave:1> erfinv(0.3)
> : error: A([]) = X: X must also be an empty matrix
> : error: evaluating assignment expression near line 40, column 22
> : error: called from `erfinv' in file `/usr/local/lib/octave/2.0.1/m/specfun/erfinv.m'
> : 
> : 
> : Fix:
> : ---
> : 
> : Replace the lines
> : 
> :    y (find (x == -1)) = (-Inf) * ones (sum (x == -1), 1);
> :    y (find (x == 1)) = Inf * ones (sum (x == 1), 1);
> : 
> : by
> : 
> :   y (find (x == -1)) = (-Inf) * ones (sum (x == -1));
> :   y (find (x == 1)) = Inf * ones (sum (x == 1));

> I don't think this fix is right though.  If

>   sum (x == 1)

> is nonzero, then

>   ones (sum (x == 1))

> will produce a matrix instead of a vector, and you will get another
> error.

> Perhaps it should be:

>   t = (x == -1);
>   n = sum (t);
>   if (n > 0)
>     y (find (t)) = -Inf * ones (n, 1);
>   endif

> instead?

> I realize that this is not quite as concise...

> Another possibility would be to write

>   t = find (x == -1);
>   y (t) = -Inf * ones (size (t));

> What do you think?

I think the 2nd possibility looks fine. If one prefers the original
form one could write

    y (find (x == -1)) = (-Inf) * ones (sum (x == -1), (sum(x == -1)>0));
    y (find (x == 1)) = Inf * ones (sum (x == 1), (sum(x == 1)>0));

-aw