'unknown or uninitialised column' warning is annoying

[kevinushey]

It's very common to check if a list element exists with code of the form:

is.null(l$x)

But doing so with a tibble emits warnings:

library(tibble)
options(warn = 1)
tbl <- tibble(a = 1)
is.null(tbl$c)
#> Warning: Unknown or uninitialised column: 'c'.
#> [1] TRUE

This sort of thing can happen when writing generic code that accepts all list-like R objects (of which tibbles are one).

See also: https://community.rstudio.com/t/bug-report-unknown-column-warning-when-using-tibbles/12246

[jennybc]

I thought has_name() was the semi-official thing one is supposed to do here:

library(tidyverse)
options(warn = 1)
tbl <- tibble(a = 1)
is.null(tbl$c)
#> Warning: Unknown or uninitialised column: 'c'.
#> [1] TRUE

tbl %>% has_name("c")
#> [1] FALSE
has_name(tbl, "c")
#> [1] FALSE

Created on 2018-08-07 by the reprex package (v0.2.0.9000).

has_name() seems to reflect what we're actually doing better than checking for NULL-ness, which requires one to know that's what happens when you ask for a non-existent component element of a list by name.

[hadley]

Yeah, this is a trade-off that we deliberately made (it was an error in very early versions which utterly failed).

I think it's better to be a little noisy here because, due to the forgiving nature of many summary functions, you otherwise get silently incorrect results (e.g. sum(mtcars$x)). IMO, that's worse than occasionally getting a warning that you don't need (with known work around).

[github-actions]

This old thread has been automatically locked. If you think you have found something related to this, please open a new issue and link to this old issue if necessary.