parse_factor trims x automatically but not levels

[adrtod]

parse_factor(x, levels) trims x automatically but not levels.
It is inconsistent with the factor function.
I don't know if this behavior should be fixed or documented but it can be confusing.

library(readr)

factor(c("a", "a "), levels = c("a")) %>% as.integer() # 1 NA
parse_factor(c("a", "a "), levels = c("a")) %>% as.integer() # 1 1

factor(c("a", "a "), levels = c("a ")) %>% as.integer() # NA 1
parse_factor(c("a", "a "), levels = c("a ")) %>% as.integer() # NA NA

factor(c("a", "a "), levels = c("a", "a ")) %>% as.integer() # 1 2
parse_factor(c("a", "a "), levels = c("a", "a ")) %>% as.integer() # 1 1

factor(c("a", "a "), levels = c("a ", "a")) %>% as.integer() # 2 1
parse_factor(c("a", "a "), levels = c("a ", "a")) %>% as.integer() # 2 2

[jimhester]

I guess this would mean parse_factor should have a trim argument to control trimming.

[jimhester]

So the real issue here was the parse_() functions always trimmed all of their values before parsing. I added a trim_ws parameter to control this, so you can now get identical behavior to factor in these examples.

library(readr)
library(magrittr)

factor(c("a", "a "), levels = c("a")) %>% as.integer()
#> [1]  1 NA
parse_factor(c("a", "a "), levels = c("a"), trim_ws = FALSE) %>% as.integer()
#> Warning: 1 parsing failure.
#> row col           expected actual
#>   2  -- value in level set     a
#> [1]  1 NA

factor(c("a", "a "), levels = c("a ")) %>% as.integer()
#> [1] NA  1
parse_factor(c("a", "a "), levels = c("a "), trim_ws = FALSE) %>% as.integer()
#> Warning: 1 parsing failure.
#> row col           expected actual
#>   1  -- value in level set      a
#> [1] NA  1

factor(c("a", "a "), levels = c("a", "a ")) %>% as.integer()
#> [1] 1 2
parse_factor(c("a", "a "), levels = c("a", "a "), trim_ws = FALSE) %>% as.integer()
#> [1] 1 2

factor(c("a", "a "), levels = c("a ", "a")) %>% as.integer()
#> [1] 2 1
parse_factor(c("a", "a "), levels = c("a ", "a"), trim_ws = FALSE) %>% as.integer()
#> [1] 2 1

Created on 2017-12-11 by the reprex package (v0.1.1.9000).

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/