I have a tibble (data.frame) that I need to apply a number of type updates to. I have a readr::col_spec object that describes the desired types, but since the data does not originate as a csv file, I cannot use read_csv(..., col_types=cspec) to apply the changes to the specified columns.
Since col_spec is a data structure designed exactly to specify desired data types, I would nevertheless to use it directly as an input to a function that applies the changes for me, rather than writing a long custom script to apply the different columns. See the following example:
library(tidyverse)
# Subset starwars to get sw (comparable to my input data)
sw <- starwars %>%
select(name, height, ends_with("_color")) %>%
slice(c(1,4,5,19))
sw
#> # A tibble: 4 × 5
#> name height hair_color skin_color eye_color
#> <chr> <int> <chr> <chr> <chr>
#> 1 Luke Skywalker 172 blond fair blue
#> 2 Darth Vader 202 none white yellow
#> 3 Leia Organa 150 brown light brown
#> 4 Yoda 66 white green brown
# The col_spec that I have
cspec <- cols(
hair_color = col_factor(c("brown", "blond", "white", "none")),
skin_color = col_factor(c( "green", "light", "fair", "white")),
eye_color = col_factor(c("blue", "brown", "yellow"))
)
# I would like to apply the col_spec directly to sw
# A not so great workaround is to use a tempfile
tf <- tempfile()
sw %>% write_csv(tf)
sw_fct <- read_csv(tf, col_types=cspec)
# This is more or less the result I am after:
# But note how info on other columns (height) is lost in the roundtrip
sw_fct
#> # A tibble: 4 × 5
#> name height hair_color skin_color eye_color
#> <chr> <dbl> <fct> <fct> <fct>
#> 1 Luke Skywalker 172 blond fair blue
#> 2 Darth Vader 202 none white yellow
#> 3 Leia Organa 150 brown light brown
#> 4 Yoda 66 white green brown
We may do this by extracting the elements from the object by looping overs the cols
library(readr)
library(purrr)
sw[names(cspec$cols)] <- imap(cspec$cols, ~ parse_factor(sw[[.y]],
levels = .x$levels, ordered = .x$ordered, include_na = .x$include_na))
-checking the output
> sw
# A tibble: 4 × 5
name height hair_color skin_color eye_color
<chr> <int> <fct> <fct> <fct>
1 Luke Skywalker 172 blond fair blue
2 Darth Vader 202 none white yellow
3 Leia Organa 150 brown light brown
4 Yoda 66 white green brown
> str(sw)
tibble [4 × 5] (S3: tbl_df/tbl/data.frame)
$ name : chr [1:4] "Luke Skywalker" "Darth Vader" "Leia Organa" "Yoda"
$ height : int [1:4] 172 202 150 66
$ hair_color: Factor w/ 4 levels "brown","blond",..: 2 4 1 3
$ skin_color: Factor w/ 4 levels "green","light",..: 3 4 2 1
$ eye_color : Factor w/ 3 levels "blue","brown",..: 1 3 2 2
If we also need the attributes of 'spec', do the assignment
attr(sw, "spec") <- cspec
-checking the str
> str(sw)
tibble [4 × 5] (S3: tbl_df/tbl/data.frame)
$ name : chr [1:4] "Luke Skywalker" "Darth Vader" "Leia Organa" "Yoda"
$ height : int [1:4] 172 202 150 66
$ hair_color: Factor w/ 4 levels "brown","blond",..: 2 4 1 3
$ skin_color: Factor w/ 4 levels "green","light",..: 3 4 2 1
$ eye_color : Factor w/ 3 levels "blue","brown",..: 1 3 2 2
- attr(*, "spec")=
.. cols(
.. hair_color = col_factor(levels = c("brown", "blond", "white", "none"), ordered = FALSE, include_na = FALSE),
.. skin_color = col_factor(levels = c("green", "light", "fair", "white"), ordered = FALSE, include_na = FALSE),
.. eye_color = col_factor(levels = c("blue", "brown", "yellow"), ordered = FALSE, include_na = FALSE)
.. )