Here’s a bug/misfeature is was bitten by today. Basically, I was trying to extract the label associated with each variable in a tibble, using (bascially) map(df, attr_getter("label")).
However, for some of the variables, the value of the labels (note the s) attribute, a ‘> 1’-length vector, were extracted instead. It turned out that this happens when the label attribute is missing. The reason is partial matching on the attribute name, which unfortunately is the default in attr().
It’s to late to change the base attr() function, but partial matching is dangerous ☠️, and I see no reason that this should be the default in attr_getter() too!
To fix this, one only needs to add exact = TRUE in the call to the attr function in get_attr().
Reprex:
library(purrr)
x = c(0, 1, 1, 0)
attr(x, "labels") = c(`0` = "Female", `1` = "Male")
get_label = attr_getter("label")
get_label(x) # Expected output: NULL
#> 0 1
#> "Female" "Male"
Here’s a bug/misfeature is was bitten by today. Basically, I was trying to extract the
labelassociated with each variable in a tibble, using (bascially)map(df, attr_getter("label")).However, for some of the variables, the value of the
labels(note thes) attribute, a ‘> 1’-length vector, were extracted instead. It turned out that this happens when thelabelattribute is missing. The reason is partial matching on the attribute name, which unfortunately is the default inattr().It’s to late to change the base
attr()function, but partial matching is dangerous ☠️, and I see no reason that this should be the default inattr_getter()too!To fix this, one only needs to add
exact = TRUEin the call to theattrfunction inget_attr().Reprex: