Rename all columns except id column by adding a prefix using dplyr

Question

I have a dataframe with columns id, feature_1, feature_2, feature_3 as follows.

df = data.frame(
  id = sample(letters, 5),
  feature_1 = sample(1:10, 5),
  feature_2 = runif(5),
  feature_3 = rnorm(5)
)

I want to rename all the feature columns by adding a prefix. The following line doesn't work and output error.

df %>%
  rename_with(~(ifelse(names(.x) == "id", paste0("source_", names(.x)), "id")))

Error in names[cols] <- .fn(names[cols], ...) : 
  replacement has length zero

Any hint on how to modify this? What does .x represent inside rename_with? Thanks in advance!

Oozy

Answer 1

library(dplyr)

df %>% 
  rename_with(~ paste0("source_", .), -id)

The third argument to rename_with is .cols, where you can use tidyselect syntax to select the columns. Here -id excludes this column.

---

Per the comments the . syntax is a cleaner/simpler style than an writing an anonymous function, but you could accomplish this equivalently as:

df %>% 
  rename_with(function(x) paste0("source_", x), -id)

# R >= 4.1.0
df %>% 
  rename_with(\(x) paste0("source_", x), -id)

Handfasting

Thanks LMc! What does the dot represent inside rename_with? I notice some dplyr functions use dot and some others use .x. What are the differences between them? – Oozy

The dot syntax refers to the column over which the function is being applied. Either . or .x will be recognized as this reference. – Handfasting

For rename_with, it seems the dot represents the column name? If I change the command to df %>% rename_with(~(ifelse(.x == "id", paste0("source_", .x), "id"))), it doesn't work either. – Oozy

In this specific function, but more generally that's what the dots represent. For example df %>% mutate(across(starts_with("feature"), ~ . + 1)) the dot does not represent the column name. – Handfasting

Thanks LMc! For my purposes I needed to exclude multiple columns using column id, so here's the syntax in case others need it. df %>% rename_with(~ paste0("source_", .), .cols = -c(1:2,4)) – Galenism

Answer 2

names(df) <- paste0("Source_", names(df))

This is a model for the solution. To apply the modification to a subset of names, use indexing, as in

i <- grep("feature", names(df))
names(df)[i] <-  paste0("Source_", names(df)[i])

Here, i is an array of indexes of the names to modify. It doesn't matter how it's created, but grep is a great way to do so when the criterion is based on the name itself.

Autostrada

Thanks whuber! I'm curious about the implementation in dplyr using pipe format r code and the functionality of .x passing inside rename_with function. – Oozy

That seems like (at best) a roundabout way of doing something that is simple and directly supported in base R. – Autostrada