Skip to content

problem (and solution) to rle on vector with NA values

3 messages · Cormac Long, Peter Ehlers, Nick Sabbe

#
Hello there R-help,

I'm not sure if this should be posted here - so apologies if this is the case.
I've found a problem while using rle and am proposing a solution to the issue.

Description:
I ran into a niggle with rle today when working with vectors with NA values
(using R 2.31.0 on Windows 7 x64). It transpires that a run of NA values
is not encoded in the same way as a run of other values. See the following
example as an illustration:

Example:
The example
??????? rv<-c(1,1,NA,NA,3,3,3);rle(rv)
Returns
??????? Run Length Encoding
??????? ? lengths: int [1:4] 2 1 1 3
??????? ? values : num [1:4] 1 NA NA 3
not
??????? Run Length Encoding
??????? ? lengths: int [1:3] 2 2 3
??????? ? values : num [1:3] 1 NA 3
as I expected. This caused my code to fail later (unsurprising).

Analysis:
The problem stems from the test
? ? ? ?? y <- x[-1L] != x[-n]
in line 7 of the rle function body. In this test, NA values return logical NA
values, not TRUE/FALSE (again, unsurprising).

Resolution:
I modified the rle function code as included below. As far as I tested, this
modification appears safe. The convoluted construction of naMaskVal
should guarantee that the NA masking value is always different from
any value in the vector and should be safe regardless of the input vector
form (a raw vector is not handled since the NA values do not apply here).

rle<-function (x)
{
??? if (!is.vector(x) && !is.list(x))
??????? stop("'x' must be an atomic vector")
??? n <- length(x)
??? if (n == 0L)
??????? return(structure(list(lengths = integer(), values = x),
??????????? class = "rle"))

??? #### BEGIN NEW SECTION PART 1 ####
??? naRepFlag<-F
??? if(any(is.na(x))){
??????? naRepFlag<-T
??????? IS_LOGIC<-ifelse(typeof(x)=="logical",T,F)

??????? if(typeof(x)=="logical"){
??????????? x<-as.integer(x)
??????????? naMaskVal<-2
??????? }else if(typeof(x)=="character"){
??????????? naMaskVal<-paste(sample(c(letters,LETTERS,0:9),32,replace=T),collapse="")
??????? }else{
??????????? naMaskVal<-max(0,abs(x[!is.infinite(x)]),na.rm=T)+1
??????? }

??????? x[which(is.na(x))]<-naMaskVal
??? }
??? #### END NEW SECTION PART 1 ####

??? y <- x[-1L] != x[-n]
??? i <- c(which(y), n)

??? #### BEGIN NEW SECTION PART 2 ####
??? if(naRepFlag)
??????? x[which(x==naMaskVal)]<-NA

??? if(IS_LOGIC)
??????? x<-as.logical(x)
??? #### END NEW SECTION PART 2 ####

??? structure(list(lengths = diff(c(0L, i)), values = x[i]),
??????? class = "rle")
}

Conclusion:
I think that the proposed code modification is an improvement on the existing
implementation of rle. Is it impertinent to suggest this R-modification to the
gurus at R?

Best wishes (in flame-war trepidation),
Dr. Cormac Long.
#
On 2011-06-23 06:44, Cormac Long wrote:
Well, it's not worth a flame, but ...
from the help page (see 'Details'):

  "Missing values are regarded as unequal to the previous value,
   even if that is also missing."

Peter Ehlers
#
Hello Cormac.

Not having thoroughly checked whether your code actually works, the behavior
of rle you describe is the one documented (check the details of ?rle) and
makes sense as the missingness could have different reasons.
As such, changing this type of behavior would probably break a lot of
existing code that is built on top of rle.

There are other peculiarities and disputabilities about some base R
functions (the order of the arguments for sample trips me every time), but
unless the argument is really strong or a downright bug, I doubt people will
be willing to change this. Perhaps making the new behavior optional (through
a new parameter na.action or similar, with the default the original
behavior) is an option?

Feel free to run your own version of rle in any case. I suggest you rename
it, though, as it may cause problems for some packages.


Nick Sabbe
--
ping: nick.sabbe at ugent.be
link: http://biomath.ugent.be
wink: A1.056, Coupure Links 653, 9000 Gent
ring: 09/264.59.36

-- Do Not Disapprove