Hi there,

I think a recent update in the devel version might have done something unexpected to cbind for GRanges values, if the object you're trying to cbind is a data.frame rather than a DataFrame. Perhaps I should always be converting to DataFrame before cbinding, but I have a lot of older code where I cbind a regular old data.frame.  

Anyway, thought I'd check in - I'm guessing this new behaviour is a bug?  Although possibly I should always be converting to DataFrame before cbinding (in that case, perhaps make it a bit clearer that cbinding a data.frame is not supported?  maybe emit an error?).   

I think the code below will explain the issue.

Thanks, as always,

Janet

----------------------

library(GenomicRanges)

#### make a GRanges object (from ?GRanges)
seqinfo <- Seqinfo(paste0("chr", 1:3), c(1000, 2000, 1500), NA, "mock1")
gr1 <-
  GRanges(seqnames =
          Rle(c("chr1", "chr2", "chr1", "chr3"), c(1, 3, 2, 4)),
          ranges = IRanges(
            1:10, width = 10:1, names = head(letters,10)),
          strand = Rle(
            strand(c("-", "+", "*", "+", "-")),
            c(1, 2, 2, 3, 2)),
          score = 1:10,
          GC = seq(1, 0, length=10),
          seqinfo=seqinfo)
gr1

#### make a data.frame that we'll add
dat <- data.frame(col1 = rep("A",10), col2=rep("B",10) )

#### cbinding after converting to a DataFrame works as expected:
gr2 <- gr1
values(gr2) <- cbind( values(gr2), DataFrame(dat))
gr2
head(gr2,2)
#GRanges object with 2 ranges and 4 metadata columns:
#    seqnames    ranges strand |     score        GC     col1     col2
#       <Rle> <IRanges>  <Rle> | <integer> <numeric> <factor> <factor>
#  a     chr1   [1, 10]      - |         1 1.0000000        A        B
#  b     chr2   [2, 10]      + |         2 0.8888889        A        B

#### but try cbinding the raw data.frame, and the result looks weird:
gr3 <- gr1
values(gr3) <- cbind( values(gr3), dat)
head(gr3,2)
#GRanges object with 2 ranges and 2 metadata columns:
#    seqnames    ranges strand |       V1      dat
#       <Rle> <IRanges>  <Rle> |   <list>   <list>
#  a     chr1   [1, 10]      - | ######## ########
#  b     chr2   [2, 10]      + | ######## ########
#### result looks fine with the release version (GenomicRanges_1.18.4)
#### with my real data, the analagous cbind command failed with this error: 
# Error in normalizeMetadataColumnsReplacementValue(value, x) : 
#   15 rows in value to replace 71038 rows
#### although dim(dat) was [1] 71038    15
#### it worked with the real data after I did DataFrame(dat), though

sessionInfo()
R Under development (unstable) (2014-10-31 r66921)
Platform: x86_64-unknown-linux-gnu (64-bit)

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats4    parallel  stats     graphics  grDevices utils     datasets 
[8] methods   base     

other attached packages:
[1] GenomicRanges_1.19.46 GenomeInfoDb_1.3.13   IRanges_2.1.43       
[4] S4Vectors_0.5.22      BiocGenerics_0.13.6  

loaded via a namespace (and not attached):
[1] XVector_0.7.4

Thanks, Herve.   You're right:  a more recent version of R-devel (2015-03-12 r67984) doesn't give that error, although the error does reproduce with 2014-10-31 r66921.    I'm having a little trouble updating my own R-devel installation at the moment (don't think I want to troubleshoot that for now), but luckily it is installed centrally now on our servers, so I can use that. 

Janet