I had to spend about 1 hour yesterday because R did something completely unpredictable (for my taste). It dropped an attribute without a warning.
df <- data.frame(x = rep(c(1, 2), 20))
attr(df$x, "label") <- "This is clearly a label"
df$x
## [1] 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1
## [36] 2 1 2 1 2
## attr(,"label")
## [1] "This is clearly a label"
The label is clearly there. To my surprise, if I subset this data frame, R drops the attribute.
new_df <- df[df$x == 2, , drop = FALSE]
new_df$x
## [1] 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
It doesn’t matter if it’s using bracket subsetting ([
) or subset
.
new_df <- subset(df, x == 2)
new_df$x
## [1] 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
That’s not good. R’s dropping attributes silently. For my specific purpose I ended up using dplyr::filter
which safely enough preserves attributes.
library(dplyr)
df %>%
filter(df, x == 2) %>%
pull(x)
## [1] 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
## attr(,"label")
## [1] "This is clearly a label"