setnames could break out early if identical(old,new)

[MichaelChirico]

Am writing a function that works like

if (length(idx <- grep('^_', names(x)))) setnames(x, idx, gsub('^_', '', names(x)[idx])

It would be cleaner to just write

setnames(x, gsub('^_', '', names(x)))

And (in the !length case) let setnames realize old and new are the same & return early for efficiency.

[MichaelChirico]

Actually it already does this in the missing(new) case

data.table/R/data.table.R

Lines 2447 to 2451 in a8e926a

	w = which((nx != old) | (is.na(nx) & !is.na(old)))
	} else {
	w = which(nx != old)
	}
	if (!length(w)) return(invisible(x)) # no changes

Will file a fix for !missing(new)

[HughParsonage]

It would be cleaner to just write

setnames(x, gsub('^_', '', names(x)))

I believe this would work already, no? If the pattern in gsub is not matched the replacement is not attempted.

[MichaelChirico]

@HughParsonage it works but setnames still overwrites the names redundantly

[mattdowle]

This case is a nice one. Have been looking at it and @HughParsonage seems right to me: that case (where the pattern is not matched to any column name) does not overwrite the names redundantly. Even where only a few column names are matched, it only updates those. So if those don't touch key or index names, then the key or index name updates will be saved too. The setnames(DT, new) form seems fine then and no improvement possible.
The PR nicely improves the other form (setnames(DT, old, new)) which wasn't doing any of the redundancy checking or saving.