Odd parameter names of setnames

[Kodiologist]

If you call setnames with just a vector of new names, then it looks like setnames(df, c("date", "place", "value")), which is fine. But if you name the argument supplying the new names, then the call looks like setnames(old = c("date", "place", "value"), df), which is confusing because you're supplying new names, not old ones. Consider allowing old to be missing, so new can be set by name instead.

[MichaelChirico]

The goal of the current design is to allow both usages:

iris = data.table(iris)
setnames(iris, letters[1:6])
setnames(iris, letters[1:3], LETTERS[1:3])

Unfortunately unless I'm missing something (very possible), that by necessity means the second argument will be overloaded -- in the first case, it's the new names; in the second case, it's the old names. We chose old as the argument name for clarity in the latter use case.

If you have an alternative in mind, that would be great. Or, if you explain what's wrong for your use case with using implicit argument matching, that would also be helpful.

[Kodiologist]

You should be able keep the existing overloading with an approach like the following. Keep the order and names of parameters as they are, but put this near the top of the definition of setnames:

if (is.missing(old) && !is.missing(new)) {
    old = new
    new = NULL
}

This way, all existing code should work unaltered, and it will be possible to say setnames(iris, new = LETTERS[1:6]).

[MichaelChirico]

but that's not the same. it requires naming 'new' explicitly, which will break a mountain of code.

…

On Wed, Nov 13, 2019, 2:33 AM Kodi Arfer ***@***.***> wrote: You should be able keep the existing overloading with an approach like the following. Keep the order and names of parameters as they are, but put this near the top of the definition of setnames: if (is.missing(old) && !is.missing(new)) { old = new new = NULL } This way, all existing code should work unaltered, and it will be possible to say setnames(iris, new = LETTERS[1:6]). — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4041?email_source=notifications&email_token=AB2BA5NINZHA3XDY5MYHFPTQTLZHTA5CNFSM4JMF26KKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOED3LQ6I#issuecomment-553039993>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB2BA5PQQRHCIZOS3A4LFDLQTLZHTANCNFSM4JMF26KA> .

[Kodiologist]

I don't follow. Since the parameter order stays the same, setnames(iris, letters[1:6]) and setnames(iris, letters[1:3], LETTERS[1:3]) will still do what they did before. Since the parameter names stay the same, any calls that used names will still work, too. The only change should be to calls that weren't legal originally, such as setnames(iris, new = LETTERS[1:6]).

[MichaelChirico]

maybe it's best to file a PR since it seems like you have something in mind already for how it would work.