Skip to content

style question: returning multiple arguments - structure or list

10 messages · Kevin Murphy, Ross Ihaka, Brian Ripley +4 more

Original
Kevin Murphy
# 
I have a question about what is considered good style in R.

I have a function  which returns several arguments, say [x,y,z]=f(),
where sometimes I only care about the first few
(so in Matlab, I can write e.g., [x,y] = f()).

In R, there seem to be 2 ways to write this, both unsatisfying (to me):

LIST
f <- function() {
  ...
  list(x=x,y=y,z=z)
}
res <- f()
x <- res$x; y <- res$y; z <- res$z


STRUCTURE
f <- function() {
  ...
  structure(x,y=y,z=z)
}
x <- f()
y <- attr(x,"y"); z <- attr(x, "z")

Which  is considered better style, or does it just depend on the
problem?


Kevin
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Ross Ihaka
#
Kevin Murphy wrote:
The second of these looks a little odd to me.  In the first x, y, and z
are treated symmetrically while the second is asymmetric.

You might also consider whether you want to unpackage the x, y and z
values from the list at all.  If you are just using the returned values
you may as well refer to them as res$x, res$y and res$z.

	Ross
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Brian Ripley
#
On Fri, 27 Jul 2001, Kevin Murphy wrote:

            

      
      
      
    

        
The latter.  There's a third option, to create a new class and return an
object of that class, and write appropriate extractor functions.

If you look in the code base you will see many more uses of lists than
attributes, and heavy use of classes for returned lists.

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Neil Osborne
    

    

      
      #
    

  


  

      
Dear All,

I have been trying to build R using Microsoft Visual C++ (Ver 6) for the
last three and a half months. This has proved to be UNBELIEVABLY
frustrating. Unfortunately, I wasn't helped much by comments along the lines
of "RTFM" when I first asked for help from this group.

I decided to do it by myself, as the impression I got was that everybody was
to busy to help. However, after having wasted the last three and a half
months, I'm forced to come back, cup in hand. I would be very grateful to
hear from anyone who has managed to build R under VC++ 6.0, and even more
grateful, if they could e-mail me a VC project tthat builds R. It would not
be neccessary to include R the source files, since I can download these
myself. What would be useful though, would be the CORRECT tools (I spent
almost a month following links in the very useful Readme files) etc.

I  happened to stumble on R shortly after my Masters degree in Applied
Statistics. I think it's an AMAZING product, and the fact that it is open
source is even more amazing. I take off my hat to all those that helped to
make it a reality. My only gripe is that of "user-friendliness" when it
comes to building R. The impression I get is that this is a closely guarded
secret by old grey haired be-spectapled men in white coats with 50 or more
years of hacking UNIX, who don't have time to explain anything to any one,
and our stuck up in their own little world  of academia - Sorry ;-).

This is just the impression I get, and many students I've spoken to have the
same impression and so don't post to this group if they have any queries. I
expect anytime, to have some Professor e-mailing to "tell me off" for being
a naughty boy ! (and to kindly inform me that it's all in the FM!).

Anyway, I've had my rant (it's been threee and a half month's coming). If
you're too busy to bother about helping this "little boy" then DO NOT feel
obliged to REPLY to this mail. However, if there are any kind souls out
there who would like to give a helping hand to a young student who is
willing to learn, I will be most grateful to hear from you. As I mentioned
earlier, if you could e-mail a MSVC project that can just get me up and
running with building R, I will be forever grateful for your time, patience
andd kindness.

Sincerely,


N. Osborne

nc_osborne at hotmail.com
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

  
  


  


      

    

    

      
      

        


  

        

      Peter Dalgaard
    

    

      
      #
    

  


  

      
"Neil Osborne" <nc_osborne at hotmail.com> writes:

            

      
      
      
    

        
Well, sixteen years to be precise in my case. However, you need to
know that the grey-haired men (some more than others, none of us wear
white coats though, and some of us can still manage without reading
glasses) *also* gave up on Visual C++ many moons ago. The project file
structure of VC++ is simply not strong enough to build the auxiliary
files needed for R, and we only achieved the cross-platform
portability by using the gnuwin tools. This is simply the traditional
clash between two kinds of user-friendliness: making simple things
easy vs. making difficult things possible (and automatable). (You can
read the book or wait for the movie, but in academia the latter option
may require a very long wait!)

            

      
      
      
    

        
Well, it is... *provided* you use the suggested tools. Once the tools
are installed, you start a shell, go to the relevant directory, type
"make" and sit down and watch the magic. This might have taken you
three and a half days to get to work and three and a half weeks to
begin to understand how it works... 

If you want to use a different set of tools, in particular a
proprietary one, you really can't expect anyone else to jump up and
help you out. It might be doable - versions of R for Windows around
the 0.50 revision (c.1997) were actually made with VC++ (by Robert),
but it usually took him over a month from the Unix releases were made.
However, the functionality of R has been expanded considerably since
then and I have no clue as to where VC++ stands when it comes to
dynamic linking and suchlike. I also think some non-binary items were
actually built on Unix and copied across (but my memory is fading).

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Thomas J Vogels
    

    

      
      #
    

  


  

      
Hi,

you also have the option of "throwing out" the third result by setting it to
NULL (taking your LIST option)?

            

      
      
      
    

        
[1] "x" "y" "z"

            

      
      
      
    

        
[1] "x" "y"

and then work with res (res$x and res$y)?

Regards,
  -tom

--
Email: vogels at cmu.edu
Phone: 412.855.2096

        
On Fri, 27 Jul 2001, Kevin Murphy wrote:

        

            

      
      
      
    

        
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

  
  


  


      

    

    

      
      

        


  

        

      Peter Dalgaard
    

    

      
      #
    

  


  

      
"Thomas J Vogels" <tov at ece.cmu.edu> writes:

            

      
      
      
    

        
A little summer exercise: Can one write a list assignment function,
i.e. "list<-" so that

list(a,b,c) <- f()

would be equivalent to

r <- f()
a <- r[[1]]
b <- r[[2]]
c <- r[[3]]

Even better, do something useful with named list elements. (And what
are the odds of finding that this is really an exercise hidden
somewhere in a book by Venables and Ripley?)

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Thomas Lumley
    

    

      
      #
    

  


  

      
On 28 Jul 2001, Peter Dalgaard BSA wrote:

        

            

      
      
      
    

        
I think you'll find that the first argument of an assignment function has
to exist already -- the function is passed the evaluated first argument.

This means it isn't quite true that
  a<-"foo<-"(a,value)
is the same as
  foo(a)<-value

eg

            

      
      
      
    

        
Error: Object "a" not found

            

      
      
      
    

        
[[1]]
a

[[2]]
[1] 1

            

      
      
      
    

        
[[1]]
*tmp*

[[2]]
[1] 4


If this is correct, then you can't really write "list<-"

	-thomas

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

  
  


  


      

    

    

      
      

        


  

        

      Peter Dalgaard
    

    

      
      #
    

  


  

      
Thomas Lumley <tlumley at u.washington.edu> writes:

            

      
      
      
    

        
Dang! You're right...

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Kevin Murphy
    

    

      
      #
    

  


  

      
Re: teaching computer science - I'm surprised nobody has recommended
using the classic "Structure and Interpretation of Computer Programs",
by Abelson, Sussman, and Sussman (1984/1996)
(http://mitpress.mit.edu/sicp/sicp.html), especially since R is so
closely modelled on Scheme, the language used in this book. (SICP/Scheme
is used in the introductory CS class in Berkeley and many other
institutions.)


Re: list/structures -  I discovered a new reason why using lists is
better than structures. Suppose you create res=structure(x,y=y,z=z).
How do you just print out x, without the corresponding attributes (which
might 
be large)? If res=list(x=x,y=y,z=z), it's easy. 

There is a more subtle advantage of the Matlab-style [x,y,z]=f() syntax
- by checking how many return arguments are on the LHS, you can decide
whether to even bother computing (yet alone returning) the "optional"
results y,z.
Obviously there are other ways of doing this, but I think this is quite
slick (and I have used not infrequently).


Re: coming to R from Matlab -  I am not trying to write Matlab-like R
code (although sometimes I can't help it :), but rather to learn the
stylistic conventions of the R community (which, I am pleased to see,
seems very concerned with writing elegant code).

Kevin
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._