R devel - Jul 2008 - confusion about evaluation.

Hi All, I am confused about the following code. I thought that the  
problem stemmed from lazy evaluation and the fact that 'i' is never  
evaluated within the first lapply. However, I am then confused as to  
why it gets bound to the final element of the lapply. The environments  
of the returned functions are indeed different in both cases and each  
environment has a local binding for 'i' it just happens to be 3 --  
which I will say is wrong, but I await reeducation.

I looked for documentation concerning this, but I wasn't able to find  
anything -- I imagine that this must be documented somewhere as it  
appears like a reasonable thing to do.  Thank you as always for any  
insight. -- jim

## doesn't do what I think it should
x <- lapply(1:3, function(i) {
     function() { i^2 }
})
sapply(1:3, function(i) x[[i]]())

[1] 9 9 9

## does what I expect
x <- lapply(1:3, function(i) {
     print(i)
     function() { i^2 }
})
[1] 1
[1] 2
[1] 3

sapply(1:3, function(i) x[[i]]())

[1] 1 4 9


 > sessionInfo()
R version 2.7.1 Patched (2008-07-20 r46088)
i386-apple-darwin9.4.0

locale:
en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

On Sun, 20 Jul 2008, James Bullard wrote:
> Hi All, I am confused about the following code. I thought that the problem 
> stemmed from lazy evaluation and the fact that 'i' is never
evaluated within
> the first lapply. However, I am then confused as to why it gets bound to
the
> final element of the lapply.
Correct.  'i' in the environments is a promise, and it gets the value at
the time the promise is evaluated from the environment in which the 
promise is created.  Since you don't evaluate it until after the lapply, 
the environment in which the promise is evaluated has the final value at 
that time. BTW, the following variant

x <- vector("list", 4)
for(i in 1:3) x[[i]] <- function() i^2
sapply(1:3, function(i) x[[i]]())

works the same way.

Lazy evaluation works somewhat differently in R and in S -- I've never 
seen a comprehensive discussion as to why (nor exactly how it works in 
either system, although there is a lot about promises in the 'R
Internals'
manual).
> The environments of the returned functions are 
> indeed different in both cases and each environment has a local binding for
> 'i' it just happens to be 3 -- which I will say is wrong, but I
await
> reeducation.
>
> I looked for documentation concerning this, but I wasn't able to find 
> anything -- I imagine that this must be documented somewhere as it appears 
> like a reasonable thing to do.  Thank you as always for any insight. -- jim
> ## doesn't do what I think it should
> x <- lapply(1:3, function(i) {
>   function() { i^2 }
> })
> sapply(1:3, function(i) x[[i]]())
>
> [1] 9 9 9
>
> ## does what I expect
> x <- lapply(1:3, function(i) {
>   print(i)
>   function() { i^2 }
> })
> [1] 1
> [1] 2
> [1] 3
>
> sapply(1:3, function(i) x[[i]]())
>
> [1] 1 4 9
>
>
>> sessionInfo()
> R version 2.7.1 Patched (2008-07-20 r46088)
> i386-apple-darwin9.4.0
>
> locale:
> en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8
>
> attached base packages:
> [1] stats     graphics  grDevices utils     datasets  methods   base
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595