[Rd] bug in rect (PR#9307)

[Duncan Murdoch]

On 10/19/2006 7:22 PM, Peter Dalgaard wrote:
> Duncan Murdoch <murdoch at stats.uwo.ca> writes:
> 
>> On 10/19/2006 5:34 PM, jgvcqa at rit.edu wrote:
>>> Full_Name: Joe Voelkel
>>> Version: 2.4.0
>>> OS: Windows XP
>>> Submission from: (NULL) (129.21.11.37)
>>>
>>>
>>> Using "border=NA" in the call to rect generates an error. But the help file says
>>> border=NA is acceptable. 
>>>
>>> (The problem is that border=NA evaluates to TRUE for "is.logical(border)" inside
>>> the function. Using rect with border=FALSE works fine)
>> I don't see this.  Could you post some code to illustrate?
> 
> He should, but it is actually plain to see in the code:
> 
>     if (!is.null(density) && !is.null(angle)) {
>         if (is.logical(border)) {
>             if (border)
>                 border <- col
>             else border <- NA
>         }
> 
> so it requires non-default density and angle to bite. The code should
> probably read 
> 
> ...
>    if(!is.na(border) && is.logical(border)) 
> ... 
> 

Right.  Looks like there might be a similar bug lurking in lpolygon and 
lrect in lattice, so I'm cc'ing Deepayan.

Are you fixing rect?

Duncan Murdoch