nzchar

NoteAdded in 0.5.0

What it does

Checks for usage of x != "" or x == "", or comparisons of nchar(x) with zero, such as nchar(x) == 0, instead of nzchar(x) or !nzchar(x).

Why is this bad?

x == "" is less efficient than !nzchar(x) when x is a large vector of long strings.

One crucial difference is in the default handling of NA_character_, i.e., missing strings. nzchar(NA_character_) is TRUE, while NA_character_ == "" is NA. Generated fixes use nzchar(x, keepNA = TRUE) to preserve missing values.

This rule comes with an unsafe fix because nzchar() can still differ for factors or classed objects and does not preserve attributes such as names and dimensions.

Example

x <- sample(c("abcdefghijklmn", "", "opqrstuvwyz"), 1e7, TRUE)
x[x == ""]

Use instead:

x <- sample(c("abcdefghijklmn", "", "opqrstuvwyz"), 1e7, TRUE)
x[!nzchar(x, keepNA = TRUE)]

References

See ?nzchar