rename_with.sf doesn't pass arguments as per dplyr::rename_with help file

[andrewbaxter439]

As discovered in a SO post, when renaming an sf object with rename_with.sf, it doesn't pass the arguments to the inner function via ... as expected:

library(sf)
#> Linking to GEOS 3.11.2, GDAL 3.6.2, PROJ 9.2.0; sf_use_s2() is TRUE
library(stringr)

nc <- sf::st_read(system.file("shape/nc.shp", package = "sf"), quiet = TRUE)

# This works:
nc |> 
    dplyr:::rename_with.data.frame(str_remove, pattern = "\\d")
#> Simple feature collection with 100 features and 14 fields
#> Geometry type: MULTIPOLYGON
#> Dimension:     XY
#> Bounding box:  xmin: -84.32385 ymin: 33.88199 xmax: -75.45698 ymax: 36.58965
#> Geodetic CRS:  NAD27
#> First 10 features:
#> ...

# `replace_with.sf` method doesn't pass argument in `...`:
nc |> 
    rename_with(str_remove, pattern = "\\d")
#> Error in rename_with(nc, str_remove, pattern = "\\d"): could not find function "rename_with"

Suggested fix (pr to follow) would be to add ... to .fn() call here:

sf/R/tidyverse.R

Line 272 in 99d6565

names(agr) = .fn(names(agr))

[edzer]

I am seeing

> nc |> 
    rename_with(str_remove, pattern = "\\d")
Error in .fn(names(agr)) : argument "pattern" is missing, with no default
> traceback()
6: vctrs::vec_size_common(string = string, pattern = pattern, replacement = replacement, 
       .call = error_call)
5: check_lengths(string, pattern, replacement)
4: str_replace(string, pattern, "")
3: .fn(names(agr))
2: rename_with.sf(nc, str_remove, pattern = "\\d")
1: rename_with(nc, str_remove, pattern = "\\d")

is that the problem you are referring to?

[andrewbaxter439]

Yes sorry, I'm not sure what happened with my example code above - had assumed it had run the same as that before I posted it!

Edit - I see I forgot to run the library(dplyr) line above my code when making a reprex. That's exactly the error I got.

[edzer]

Thanks!

[andrewbaxter439]

As a brief question - in #2312 I noted that the names(agr) = .fn(names(agr), ...) doesn't seem to add much functionality and perhaps messes up the names. Is this step truly needed and is the dropping of non-rename_with-transformed names intended/acceptable?

When I added testing to the updated rename_with.sf function it explicitly failed to return the expected object as the "agr" attribute was changed. I do realise I don't know the full depths of what this line accomplishes so do ignore if not relevant.

[edzer]

The docs of rename_with say ... additional arguments passed onto ‘.fn’ - what could go wrong?

[andrewbaxter439]

It works for the column names well, but the second step (the names(agr) bit) seems to rename the selected columns in the "agr" attribute but NA the rest:

fn = function(x) paste0(x, "_renamed")

# Before renaming:
nc %>% attr("agr")
#>      AREA PERIMETER     CNTY_   CNTY_ID      NAME      FIPS    FIPSNO  CRESS_ID 
#>      <NA>      <NA>      <NA>      <NA>      <NA>      <NA>      <NA>      <NA> 
#>     BIR74     SID74   NWBIR74     BIR79     SID79   NWBIR79 
#>      <NA>      <NA>      <NA>      <NA>      <NA>      <NA> 
#> Levels: constant aggregate identity

# After renaming:
nc %>% rename_with(fn, c(FIPS, FIPSNO)) %>% attr("agr")
#>           <NA>           <NA>           <NA>           <NA>           <NA> 
#>           <NA>           <NA>           <NA>           <NA>           <NA> 
#>   FIPS_renamed FIPSNO_renamed           <NA>           <NA>           <NA> 
#>           <NA>           <NA>           <NA>           <NA>           <NA> 
#>           <NA>           <NA>           <NA>           <NA> 
#>           <NA>           <NA>           <NA>           <NA> 
#> Levels: constant aggregate identity

Had only discovered when trying to write a test which fails:

fn = function(x) paste0(x, "_renamed")

expect_identical(nc %>% rename_with(fn, c(FIPS, FIPSNO)), 
				 nc %>% rename(FIPS_renamed = FIPS, FIPSNO_renamed = FIPSNO))