A subtle scoping flaw in pull()

[graphdr]

If x is not a column name in data frame d, then pull(d, x) attempts to look up the value of x in the environment instead of returning NULL or an error as I would expect.

library(dplyr)

# Create two env-variables
d <- data.frame(x = "ex", y = "why")
x <- "y"

# As expected, pulls the data-variable `x`
pull(d, x)

# If name `x` does not exist in `d`,
d <- select(d, -x)

# Not expected: finds the env-variable `x`, pulls the data-variable `y`
pull(d, x)

# Exactly as if we had written pull(d, y)
identical(pull(d, x), pull(d, y))