I'm looking to pivot a data frame using pivot_longer. Generally, I select a column using some pattern match (e.g. pivot_longer(cols = contains(...)) ). Now I wanted to select columns based on the data types they contain which did not work as expected. For example, to pivot columns containing logical data I tried:
df <- data.frame(nums = seq(1,6,1),
chrs = letters[1:6],
logis = rep(c(TRUE, FALSE), 3),
logis2 = rep(c(TRUE, FALSE), 3))
pivot_longer(df,
cols = sapply(df, is.logical))
This results in an error telling me that the cols argument only accepts numeric or character data types.
A simple solution would be to identify the columns containing a certain data type, get the names of these columns and then pivot the data using the vector of characters:
vars_to_pivot <- names(df)[sapply(df, is.logical)]
pivot_longer(df,
cols = all_of(vars_to_pivot)) %>%
head
# A tibble: 6 × 4
nums chrs name value
<dbl> <chr> <chr> <lgl>
1 1 a logis TRUE
2 1 a logis2 TRUE
3 2 b logis FALSE
4 2 b logis2 FALSE
5 3 c logis TRUE
6 3 c logis2 TRUE
I'm however curious whether there is a dplyr/tidyverse verb for column selection within pivot functions based on logical vectors?
You can tidyselect::where() here, which selects the variables for which a function returns TRUE:
pivot_longer(df,
cols = where(is.logical)
)
# A tibble: 12 × 4
nums chrs name value
<dbl> <chr> <chr> <lgl>
1 1 a logis TRUE
2 1 a logis2 TRUE
3 2 b logis FALSE
4 2 b logis2 FALSE
5 3 c logis TRUE
6 3 c logis2 TRUE
7 4 d logis FALSE
8 4 d logis2 FALSE
9 5 e logis TRUE
10 5 e logis2 TRUE
11 6 f logis FALSE
12 6 f logis2 FALSE
This is a helper function that is part tidyselect, a domain-specific language for selecting variables. The full list (which includes all_of() that you use in your question) is here.