How to use sep2 = "" on data.table-CodePudding

Is there a way to use data.table::fwrite to write the values of a column without any separation between them?

For example:

library("data.table")
geno <- data.table(
  IID = 1:10,
  SNP = lapply(1:10, function(i) sample(0:2, 10, replace = TRUE))
)
fwrite(geno, "Geno.txt", col.names = FALSE, sep = " ", sep2 = c("","",""))

But the sep2 does not allow it and gives me the following error:

Error in fwrite(geno, "Geno.txt", col.names = FALSE, row.names = FALSE,  : 
  is.character(sep2) && length(sep2) == 3L && nchar(sep2[2L]) ==  .... is not TRUE

I would like to have the following result, without having to collapse all values before writing it to a file.

1 2221210202
2 0020010221
3 1010022212
4 0120121221
5 1212211202
6 2100002010
7 1110011210
8 1212012121
9 2221121021
10 1122220101

Thank you.

CodePudding user response：

According to ?fwrite, sep2[2] must be a single character. Therefore you have to collapse the list, rather than use sep2.

You can use

fwrite(geno[, .(IID, SNP=sapply(SNP, paste0, collapse=''))], 'test.txt', sep=' ')

CodePudding user response：

Alternative: write it with a character known to not exist in the data and then remove it programmatically on the file. The second step here can be done in R, but frankly command-line tools are much faster at this. I'll use tr here, as it is likely to be the fastest.

fwrite(geno, "Geno.txt", col.names = FALSE, sep = " ", sep2 = c("","\037",""))
readLines("Geno.txt", n=2)
# [1] "1 1\0372\0371\0372\0372\0370\0371\0370\0372\0370" "2 1\0370\0372\0372\0371\0370\0372\0372\0371\0370"

system2("tr", c("-d", "\037"), stdin="Geno.txt", stdout="Geno2.txt")
readLines("Geno2.txt", n=2)
# [1] "1 1212201020" "2 1022102210"

tr should be available on all unix-like OSes including MacOS, and within Rtools-4.0 for windows under "c:\\rtools40\\usr\\bin\\tr.exe" or whichever path is closest for your install.

For this, I chose the unicode \037 which is used by many things as a "Delimiter", and seems unlikely to be found in most datasets. However, others will work just as easily, including sep2 = c("", "|", "") with "system2(tr", c("-d", "|"), ...).