[Bioc-devel] names, mcols and metadata are not changed in "[<-" operator of class Vector

Hi Sebastian,

This is the expected behavior and it's motivated by how [<- behaves on
ordinary vectors:

   x <- setNames(1:3, LETTERS[1:3])

Then:

   > x

   A B C
   1 2 3

Replacing the first two elements:

   x[2:1] <- x[1:2]

Then:

   > x

   A B C
   2 1 3

The names are preserved.

So we need to think of [<- as an operator that touches the values of
a vector-like object without touching its structure or its attributes.
People sometimes rely on this behavior to stuff the entire object with
a given value:

   x[] <- some_value

They wouldn't expect this to destroy the names or metadata cols that
are on 'x'.

Hope that makes sense,

Cheers,
H.


On 05/07/2014 11:59 AM, Sebastian Gibb wrote:

Dear all,

today I observed an unexpected behaviour (at least for me) using the "[<-"
operator on classes that inherited "Vector".
I want to change the order of some elements inside the Vector (in my use case it
is an AAStringSet object). If I use an index in "[<-" the names, the
elementMetadata and the metadata are not replaced by their counterparts, e.g.:

###
library("IRanges")

i <- IRanges(1:3, 4:6)
names(i) <- LETTERS[1:3]

i
# IRanges of length 3
#     start end width names
# [1]     1   4     4     A
# [2]     2   5     4     B
# [3]     3   6     4     C

i[2:1] <- i[1:2]
i
# IRanges of length 3
#     start end width names
# [1]     2   5     4     A
# [2]     1   4     4     B
# [3]     3   6     4     C


## names should be B, A, C; the elementMetadata and metadata are in the wrong
## order, too.
## expected output:
# IRanges of length 3
#     start end width names
# [1]     2   5     4     B
# [2]     1   4     4     A
# [3]     3   6     4     C

###

I tried to figure out the reason for this and ended up with the "replaceROWS"
method (in {IRanges,S4Vectors}/R/Vector-class.R).
In this method the mcols, metadata and names are just restored. To be honest I
do not really understand this function.
Why is the new value first append to the original vector (`ans <- c(x, value)`)
and subsequently extracted? Would a simple replace, e.g. `x[i] <- value` not be
enough?

In my opinion the last three lines should restore the original metadata
*and* replace the corresponding metadata by their new counterparts.

### replaceROWS from IRanges/R/Vector-class.R
setMethod("replaceROWS", "Vector",
     function(x, i, value)
     {
         idx <- seq_along(x)
         i <- extractROWS(setNames(idx, names(x)), i)
         ## Assuming that objects of class 'class(x)' can be combined with c().
         ans <- c(x, value)
         idx[i] <- length(x) + seq_len(length(value))
         ## Assuming that [ works on objects of class 'class(x)'.
         ans <- ans[idx]
         ## Restore the original decoration.
         metadata(ans) <- metadata(x)
         names(ans) <- names(x)
         mcols(ans) <- mcols(x)
         ans
     }
)
###

Kind regards,

Sebastian

--
Herv? Pag?s

Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M1-B514
P.O. Box 19024
Seattle, WA 98109-1024

E-mail: hpages at fhcrc.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319