Extending Suffixes in Merge to All Non-by Columns

Question

suffixes in merge works only on common column names. Is there anyway to extend this to the rest of the columns as well without manually updating co

Answer 1

This is an interesting question, and I doubt that extending merge would be a straightforward solution unless Matt Dowle and Co. think it's something worth implementing in merge.data.table.

Here's one approach that came to mind:

DTs <- c("df1", "df2")
suffixes <- seq_along(DTs)

for (i in seq_along(DTs)) {
  Name <- setdiff(colnames(get(DTs[i])), "a")
  setnames(get(DTs[i]), Name, paste(Name, suffixes[i], sep = "."))
}

merge(df1, df2, by = "a") # Will obviously work as you expect now

Answer 2

Try the following:

colnames(
  mergeWithSuffix(df1,df2, by = 'a', suffixes = c("1","2"))
)
[1] "a"   "b.1" "d.1" "d.2"

Notice that the original data.frames are unharmed.

colnames(df1)
[1] "a" "b" "d"

colnames(df2)
[1] "a" "d"

The functions are as follows

require(data.table)

mergeWithSuffix <- function(x, y, by, suffixes=NULL, ...) {

  # Add Suffixes
  mkSuffix(x, suffixes[[1]], merge.col=by)
  mkSuffix(y, suffixes[[2]], merge.col=by)

  # Merge
  ret <- merge(x, y, by = by, suffixes = NULL, ...)

  # Remove Suffixes
  undoSuffix(x, suffixes[[1]], merge.col=by)
  undoSuffix(y, suffixes[[2]], merge.col=by)
  return(ret)
}

mkSuffix <- function(x, sfx, sep=".", merge.col=NULL)  {
  nms <- setdiff(names(x), merge.col)
  setnames(x, nms, paste(nms, sfx, sep=".") ) 
}

undoSuffix <- function(x, sfx, sep=".", merge.col=NULL) {
  nms <- setdiff(names(x), merge.col)
  setnames(x, nms, sub(paste0(get("sep"), sfx, "$"), "", nms))
}

---

Notice that setnames works by reference, so the overhead is almost negligible. Also, as discussed elsewhere, this works equally well on data.frames and data.table

Answer 3

A simple solution:

mrg<-(merge(df1,df2, by = 'a', suffixes = c("1","2")))
setnames(mrg,paste0(names(mrg),ifelse(names(mrg) %in% setdiff(names(df1),names(df2)),"1","")))
setnames(mrg,paste0(names(mrg),ifelse(names(mrg) %in% setdiff(names(df2),names(df1)),"2","")))

> names(mrg)
[1] "a"  "b1" "d1" "d2"

Edit: thanks to comments by Ricardo Saporta for cleaning this up considerably and teaching me a few new tips!