Dear mailing list,
let's suggest I have a data.frame full of annotation data. In the first
column I have Identifiers present in the data. In all other columns I have
another annotaion for this identifier column.
However, on some rows there is no 1:1 mapping but let's say a 1:3 mapping
or something like this:
ID Symbol
1 Bla
2 Foo
3 XYZ // xyz /// xyz01
4 abc
I want to "stretch" the line which has multiple annotation tags. I know I
can split it with "strsplit", this get's me here:
ID Symbol
1 Bla
2 Foo
3 "XYZ" "xyz" "xyz01"
4 abc
But how can I get to this:
ID Symbol
1 Bla
2 Foo
3 XYZ
3 xyz
3 xyz01
4 abc
I too, have a solution. Mine relies on the idiom of "do.call(rbind, FUN)" (or cbind), which efficiently rbinds/cbinds the results from FUN. For this to work, FUN must return a list.
First we define the function that does the actual splitting, returning a (sub-)matrix for each row, `splitit`. The we use `apply` to apply `splitit` on each row in the matrix; this returns the desired list that we can put into the do.call idiom.
splitit <- function(row) {
# Split the symbols (res is a normal vector)
res <- strsplit(row[2], ' // ', fixed=TRUE)[[1]]
# Bind the ID to the symbol vector, clean up the result and return
res <- t(rbind(row[1], res)) # If the original data contains several columns, just add them here.
dimnames(res) <- list(c(), c())
return(res)
}
# Test data from Heidi Dvinge and James W. MacDonald
test <- data.frame(A=c("x", "y", "z"), B=c("X", "Y1 // Y2 // Y3", "Z"), stringsAsFactors=FALSE)
df <- data.frame(ID = 1:4, Symbol = I(c("Bla","Foo","XYZ // xyz // xyz01", "abc")))
do.call(rbind, apply(test, 1, splitit))
do.call(rbind, apply(df, 1, splitit))
Kind regards,
Stefan McKinnon H?j-Edwards Dept. of Genetics and Biotechnology
PhD student Faculty of Agricultural Sciences
stefan.hoj-edwards at agrsci.dk Aarhus University
Tel.: +45 8999 1291 Blichers All? 20, Postboks 50
Web: www.iysik.com DK-8830?Tjele
Tel.: +45 8999 1900
Web: www.agrsci.au.dk