Data.table: A named vector version of setnames [request]

[and3k]

Hi!

I’m really missing a named vector version of setnames, e.g.:

setnames(DT, "b", "B") # current
setnames(DT, c("b" = "B")) # same as above

E.g., in https://github.com/hadley/plyr this is done by two different funtions: revalue and mapvalues, where revalue just calls mapvalues:

revalue <- function (x, replace = NULL, warn_missing = TRUE) 
{
    if (!is.null(x) && !is.factor(x) && !is.character(x)) {
        stop("x is not a factor or a character vector.")
    }
    mapvalues(x, from = names(replace), to = replace, warn_missing = warn_missing)
}

What do you think? Thanks!

Bela

[arunsrinivasan]

In what way do you miss it :-)? Are you finding it harder to type, remember, program with, understand, read etc.. Because the way I see it, I don't really miss it :-).

[gsee]

FWIW, you could do this

vec <- c(b="B")
setnames(DT, names(vec), vec)

[and3k]

I would find it easier to read when renaming several (like more than 10) names, which is a common task I have to do, and it would be easier for other people to read. I can always define my own function, but I thought it would be a nice addition for everybody. And I guess I’m not alone, since a convenience function of such a sort exits in plyr.

[franknarf1]

Yeah, I like to make named lists in my code for cleaning data. I don't rename variables very often, but if I'm importing from somewhere with annoying naming conventions and a lot of variables, like the Census, it's especially useful. I mean like...

my_renames <- c(
  badname  = goodname,
  badname2 = goodname2
)

The workaround GSee mentions is fine, but I would use the "named vector" syntax if it existed.

[arunsrinivasan]

I'm still not getting the huge advantage part. But marked as FR as it should be straightforward to implement AFAICT and wouldn't break anything.

[and3k]

Thank you! :D

[gsee]

The only thing that might break is if someone is currently passing in a named vector expecting the names not to be used -- the names could be in a different order, or they could not correspond to any of the current names of the data.table.

The risk of that seems relatively low, but personally I don't always bother using unname when passing a named vector into functions that accept vectors as input but don't use the names.

[franknarf1]

@gsee That wouldn't be a problem if the new functionality were put in a wrapper with a different name, like setnamesv, right?

[jangorecki]

@franknarf1 this would make setnames and setnamesv inconsistent to setkeyv, set2keyv or setorderv in terms of _symbols_ or _character_ input.
If changing setnames function then maybe an _option_ to keep 1.9.4 behaviour.

[franknarf1]

@jangorecki Yeah, I'm not saying that should be the name, but creating a separate function would be the most straightforward. I think the current behavior should remain the default for setnames.

[jangorecki]

It would be a breaking change if someone is already supplying names character into setnames

d=data.table(a=1,b=2)
setnames(d, c("b"="y","a"="x"))
d
#   y x
#1: 1 2

I think usage of setnames(d, names(nm), nm) is short enough to consider this feature request as redundant.