are_na() should not work on lists

[njtierney]

Hello!

At the moment in is.nan and is.infinite don't work on dataframes, so I wonder if it would be useful to create is_inf, are_inf, and is_nan and are_nan equivalents?

Happy to submit a PR for this if you think it would be in scope

is.infinite(airquality)
#> Error in is.infinite(airquality): default method not implemented for type 'list'
is.nan(airquality)
#> Error in is.nan(airquality): default method not implemented for type 'list'

purrr::map_df(airquality, is.infinite)
#> # A tibble: 153 x 6
#>    Ozone Solar.R Wind  Temp  Month Day  
#>    <lgl> <lgl>   <lgl> <lgl> <lgl> <lgl>
#>  1 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  2 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  3 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  4 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  5 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  6 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  7 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  8 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  9 FALSE FALSE   FALSE FALSE FALSE FALSE
#> 10 FALSE FALSE   FALSE FALSE FALSE FALSE
#> # ... with 143 more rows
purrr::map_df(airquality, is.nan)
#> # A tibble: 153 x 6
#>    Ozone Solar.R Wind  Temp  Month Day  
#>    <lgl> <lgl>   <lgl> <lgl> <lgl> <lgl>
#>  1 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  2 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  3 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  4 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  5 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  6 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  7 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  8 FALSE FALSE   FALSE FALSE FALSE FALSE
#>  9 FALSE FALSE   FALSE FALSE FALSE FALSE
#> 10 FALSE FALSE   FALSE FALSE FALSE FALSE
#> # ... with 143 more rows

Created on 2018-07-08 by the reprex package (v0.2.0).

[lionel-]

In general we consider it is not good UI to auto-map atomic vector functions over lists.

[njtierney]

Sorry if I wasn't clear, what I was hoping would be possible would be are_inf and is_inf in the same way that rlang currently has is_na and are_na functions - the map part was an illustration of the output they might produce.

[lionel-]

I am not sure I understand because are_na() works on atomic vectors, not on lists/dfs?

[njtierney]

But are_na does work on data.frames and vectors - which is great!

> rlang::are_na(head(airquality))
  Ozone Solar.R  Wind  Temp Month   Day
1 FALSE   FALSE FALSE FALSE FALSE FALSE
2 FALSE   FALSE FALSE FALSE FALSE FALSE
3 FALSE   FALSE FALSE FALSE FALSE FALSE
4 FALSE   FALSE FALSE FALSE FALSE FALSE
5  TRUE    TRUE FALSE FALSE FALSE FALSE
6 FALSE    TRUE FALSE FALSE FALSE FALSE

> rlang::are_na(head(airquality$Ozone))
[1] FALSE FALSE FALSE FALSE  TRUE FALSE

There is not consistent behaviour in Base R for is.infinite and is.nan - I would expect this to return the same output for a dataframe as does is.na. It does not:

> is.infinite(airquality)
Error in is.infinite(airquality) : 
  default method not implemented for type 'list'

It only works on a column, or a list

is.infinite(airquality$Ozone)

  [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [10] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [19] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [28] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [37] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [46] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [55] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [64] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [73] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [82] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [91] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[100] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[109] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[118] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[127] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[136] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[145] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE

Same as for is.nan

> is.nan(airquality)
Error in is.nan(airquality) : 
  default method not implemented for type 'list'
> is.nan(airquality$Ozone)
  [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [10] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [19] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [28] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [37] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [46] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [55] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [64] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [73] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [82] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
 [91] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[100] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[109] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[118] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[127] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[136] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[145] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE

Do you think rlang would support more consistency across these functions

• is.infinite and is.nan ?

[lionel-]

But are_na does work on data.frames and vectors - which is great!

Well I think that's a bug ;)

[lionel-]

I think our current viewpoint is that it's sloppy typing for a function to work on both T and Coll<T>. But the connection between vector functions and df functions is something that we plan to explore as part of the vctrs and tidyr package, IIRC.

cc @hadley @jennybc

[hadley]

NULL is the equivalent of NA for a list, so I think are_na(list(1, NULL)) would return c(FALSE, TRUE). But we need to think this through. It certainly should always return a logical vector, not sometimes a matrix.

[njtierney]

I think it would be helpful to have a verb/function that does always return a data.frame / matrix:

• is_na returns TRUE/FALSE (scalar)
• are_na returns vector of TRUE/FALSE (vector)
• are_na_<class> returns class <class>. e.g., are_na_df returns a data.frame of NA?

What do you think?

[hadley]

Yes, but I think that would be out of scope for rlang, and sounds more like a vctrs function to me.

[njtierney]

OK fair enough - should I add this to the existing issue on sum(is.na()) helpers?

[hadley]

Yes - I think the principle that we could now follow is that are_na(x) has type logical, but shape that matches x. I think that is a succinct description of the behaviour that you desire.

[hadley]

(And actually that's a nice description of a vectorised function - the shape of the output matches the shape of the input(s))