Tidyverse function: drop_na

[gregleleu]

Could you please add {tidyr}'s drop_na function to the tidyverse function for sf?
I know it's just a wrapper for filter(is.na(.)) but it's a convenient one :) and also for people who use it all the time it's always a source of error because drop_na currently drops the geometry silently.

Thanks!

[KiranmayiV]

Hi @gregleleu ,

Could you give reprex? I did this, the geometry column doesn't seem to drop.

df <- sf::st_as_sf(
  tibble::tibble(x = c(1, 2, NA), 
             y = c("a", NA, "b"),
             lng = c(20.1, 20.2, 20.1),
             lat = c(-1.1, -1.2, -1.2) ),
coords = c("lng", "lat"))

tidyr::drop_na(df)
#> # A tibble: 1 x 3
#>       x y        geometry
#>   <dbl> <chr>     <POINT>
#> 1     1 a     (20.1 -1.1)

^{Created on 2021-01-15 by the reprex package (v0.3.0)}

[edzer]

It looks like the class is dropped.

[KiranmayiV]

It looks like the class is dropped.

Aaah, ok thanks. I will look into it.

[KiranmayiV]

If this looks good for drop_na.sf, I can put in a PR to include this in tidyverse.R . Please let me know if you prefer any changes.

library(roxygen2)
library(sf)
#> Linking to GEOS 3.7.2, GDAL 2.4.2, PROJ 5.2.0
library(tibble)
library(tidyr)

#' @name tidyverse
#' @param … see original function docs
#' @examples
#' library(tidyr)
#' df <- st_as_sf(
#' tibble::tibble(x = c(1, 2, NA,3),
#' y = c(“a”, NA, “b”, “a”),
#' lng = c(20.1, 20.2, 20.1, 20.4),
#' lat = c(-1.1, -1.2, -1.2, -1.4)),
#' coords = c(“lng”, “lat”))
#' drop_na(df, x)
#' drop_na(df)

t_n <- function(data, ...){
  if (!requireNamespace("tidyr", quietly = TRUE))
    stop("tidyr required: install first?")
  
  class(df) <- setdiff(class(df), "sf")
  sf::st_as_sf(tidyr::drop_na(df, ...))
  
}

df <- sf::st_as_sf(
  tibble::tibble(x = c(1, 2, NA,3), 
                 y = c("a", NA, "b", "a"),
                 lng = c(20.1, 20.2, 20.1, 20.4),
                 lat = c(-1.1, -1.2, -1.2, -1.4)),
  coords = c("lng", "lat"))

t_n(df)
#> Simple feature collection with 2 features and 2 fields
#> geometry type:  POINT
#> dimension:      XY
#> bbox:           xmin: 20.1 ymin: -1.4 xmax: 20.4 ymax: -1.1
#> CRS:            NA
#> # A tibble: 2 x 3
#>       x y        geometry
#>   <dbl> <chr>     <POINT>
#> 1     1 a     (20.1 -1.1)
#> 2     3 a     (20.4 -1.4)

^{Created on 2021-01-18 by the reprex package (v0.3.0)}

Thanks,

[edzer]

No doubt this will mostly work, but I'd appreciate if you'd follow the pattern e.g. given by arrange.sf, as that would also catch some corner cases (e.g. multiple geometry columns) that your solution doesn't - or else point out why you divert.

[KiranmayiV]

Sorry about that, I am new to open source contribution and learning. Yes, happy to include for corner cases - following arrange.sf pattern.

[ilikegitlab]

Any progress? Just noticed that drop_na mysteriously caused some things to fail. Took me a while to see it was changing the class of my custom tibble....

[gregleleu]

I just tried making a PR (it's a pretty simple change) but I can't get devtools::check to work (it's not finding sqlite3 apparently, and I could find how to set --with_sqlite3_lib) but it installs ok

[gregleleu]

Close via #1975

[ilikegitlab]

thanks! Actually noticed i meant to complain about tidyr::drop_na, not r-spatial. But still nice something got fixed...