Allow setnames to include old columns not in x

[MaxGhenis]

A allow.absent.cols option for setnames would facilitate cases when the user wants to apply a column mapping without necessarily knowing whether all columns will exist. I'm currently using the following workaround, from my SO self-answer:

Setnames <- function(x, old, new, allow.absent.cols=F) {
  if (!allow.absent.cols) {
    setnames(x, old, new)
  } else {
    old.intersect <- intersect(old, names(x))
    common.indices <- old %in% old.intersect
    new.intersect <- new[common.indices]
    setnames(x, old.intersect, new.intersect)
  }
}

[eantonya]

I'm not convinced this is a good idea, and if implemented it would need to come with a plethora of warnings.

[arunsrinivasan]

This doesn't seem like a good idea to me as well. If you need to, then you can always use column numbers instead of names.

Unless there's a strong reasonable explanation for this, we should close this.

[MaxGhenis]

My use case involves looping through lots of raw datasets, and applying a single column name mapping to clean each in a consistent way. Something like

l <- list(...)  # List of raw data.tables, which don't all have the same set of columns
raw.names <- c(...)  # Names showing up in raw data
clean.names <- c(...)  # Clean names
for (dt in l) setnames(dt, raw.names, clean.names)

Column numbers wouldn't work for me. I agree a warning for each skipped column makes sense.

[arunsrinivasan]

You just have to add 1 more line...

for (dt in l) {
    ix = match(names(dt), raw.names, 0L)
    setnames(dt, raw.names[ix], clean.names[ix])
}

[MaxGhenis]

Thanks, that's cleaner than my intersect / %in% approach above. Listed as an answer on the SO question. I'm fine closing if this need is unusual.

[arunsrinivasan]

Max, glad that helped. :+1:

[MaxGhenis]

This is now implemented directly as setnames(..., skip_absent=TRUE).

[MichaelChirico]

Closed in https://github.com/Rdatatable/data.table/pull/3111, the other issue was #3030