Skip to content

Rule for accessing attributes?

11 messages · Tribo Laboy, Hadley Wickham, Christos Hatzis +1 more

Tribo Laboy
# 
Hi !

I am a new user and quite confused by R-indexing.

Make a list and get the attributes
lst <- list(x = 1:3, y = 4:6, z = 7:9)
attributes(lst)

This returns:

$names
[1] "x" "y" "z"

I can easily do:

nm <-names(lst)

or

nm <-attr(lst,"names")

which both return the assigned names of the named list 'lst', but why
then this doesn't work:

lst$names

?

I am confused ... Moreover, I noticed that some of the objects (e.g.
plot objects returned by ggplot) also have attributes when queried by
the 'attributes' function, but they are accessible by the $ notation.
(e.g.

xydf <- data.frame(x = 1:5, y = 11:15)
plt <- ggplot(data = xydf, aes(x = x,y = y)) + geom_point()
attributes(plt)

Now we can change the title:

plt$title <- "My Title"
plt

So is it some inconsistency or am I missing something important?
Christos Hatzis
# 
You need to use the '@' operator to directly access attributes (not
elements) of objects:
[1] "x" "y" "z"

See ?'@' for more details.

-Christos
Tribo Laboy
# 
Thanks Christos, for your reply and for sharing how to access the
attributes?, slots? in a shorthand notation with @, but after reading
the help for @, I became even more confused. Is there any place that
has collected wisdom on R indexing?
On Thu, Mar 27, 2008 at 1:35 PM, Christos Hatzis <christos at nuverabio.com> wrote:
Brian Ripley
# 
Oh please don't recommend misuse of @ to those already confused.

@ is for accessing slots in S4 objects.  This 'works' because they happen 
to be stored as attributes.  See the help page (and the warning that it 
does no checking - we may change that).

Similarly,

plt$title <- "My Title"

works because the package maintainer (of ggplot2, unmentioned?) has chosen 
to set things up that way.  R is very flexible, and there is plenty of 
scope for package authors to do confusing things.
On Thu, 27 Mar 2008, Christos Hatzis wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Tribo Laboy
    

    

      
      #
    

  


  

      
So am I to understand that the only realy _correct_ and _recommended_
way of accessing the attributes is through

attr(someobject, "attributename") ?


Regards,

TL

On Thu, Mar 27, 2008 at 4:16 PM, Prof Brian Ripley

        
<ripley at stats.ox.ac.uk> wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Tribo Laboy
    

    

      
      #
    

  


  

      
Now, how is it that I can access the contents of a named list by
dynamically computed name?

To go back to my previous example I have a list and I know the names.
Now I want do something with that named data in a loop.

lst <- list(x = 1:3, y = 4:6, z = 7:9)
nm <-names(lst)
nm
[1] "x" "y" "z"

I can access the list elements by name directly:

lst$x; lst$y; lst$z,

But I want to do

for (k in 1:3) {
     lst$nm[k]
 }

But this doesn't work, basically because
lst$nm[1] returns a NULL.

So what do I do?

Thanks for helping,

TL

        
On Thu, Mar 27, 2008 at 4:30 PM, Tribo Laboy <tribolaboy at gmail.com> wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Brian Ripley
    

    

      
      #
    

  


  

      
Use [[ ]].

It seems it is time for you to study a good introduction to R.

        
On Thu, 27 Mar 2008, Tribo Laboy wrote:

        

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Tribo Laboy
    

    

      
      #
    

  


  

      
Thanks, this really solved my problem.

Actually, I do have a good introduction to R - a  book co-authored by
some W.N. Venables and some B.D. Ripley, colloquially called 'MASS' is
on my desk.  I find it really very helpful. Still, as it is a book on
statistics, some details on R are only mentioned in passing.
I also use some other books and online resources, but obviously will
need to delve a bit deeper. One of the reasons I am moving to R is
because it has a great user community. I hope I am not abusing it with
too many questions.

Regards,

TL


On Thu, Mar 27, 2008 at 5:16 PM, Prof Brian Ripley

        
<ripley at stats.ox.ac.uk> wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Hadley Wickham
    

    

      
      #
    

  


  

          

      
      
      
    

        
This is rather poorly documented, but the preferred way of setting the
title is now:

p + opts(title = "My Title")

All options should be documented under ?opts (but aren't, unfortunately).

Hadley

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Christos Hatzis
    

    

      
      #
    

  


  

      
Yes, indeed.  The help page says that @ extracts the contents of a slot in
S4 objects.  But you mention below that this 'works' for S3 objects because
S4 slots are stored as attributes.  Doesn't this mean that @ is currently
implemented to access attributes of objects in general (attributes of S3
objects or slots of S4 objects that are implemented as attributes)?  I
realize that this might change in the future...

-Christos

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Brian Ripley
    

    

      
      #
    

  


  

      
On Thu, 27 Mar 2008, Christos Hatzis wrote:

        

            

      
      
      
    

        
Yes, but it will change, quite possibly for 2.7.0.