Add ignore_class argument and ignore grouped_df if check_groups = FALSE

[rossellhayes]

• Adds argument ignore_class to class checks (and passed by table, vector, and column checks), specifying class differences to ignore.
  • Unnamed elements are dropped from the list of classes for the purpose of comparison: e.g. ignore_class = "glue" will ignore cases where .result is glue/character and .solution is character, or vice versa.
  • Named elements are treated as equivalent to their names: e.g. ignore_class = c("integer" = "numeric") will ignore cases where .result is integer and .solution is numeric, or vice versa.
• Table checking now ignores the grouped_df class if check_groups = FALSE.
• Classes are now checked with setequal() instead of identical(), so objects with the same classes in different orders will not trigger a class problem.

library(dplyr)
library(tblcheck)

.result   <- data.frame(a = 1, b = 2)
.solution <- tibble(a = 1, b = 2)

tbl_check_class(ignore_class = c("tbl_df", "tbl"))
# No output

.result   <- 1L
.solution <- 1

vec_check_class(ignore_class = c("integer" = "numeric"))
# No output

.result   <- tibble(a = 1, b = 2)
.solution <- tibble(a = 1, b = 2) %>% group_by(b)

tbl_check(check_groups = FALSE)
# No output

^{Created on 2022-02-08 by the reprex package (v2.0.1)}

Closes #108.

[gadenbuie]

I bumped into this unusual behavior, I don't think it's intended.

library(dplyr)
library(tblcheck)

.result <- tibble(a = 1L, b = "2", c = "3")
.solution <- tibble(a = 1, b = 2, c = 3L)

tbl_check(ignore_class = c("numeric" = "integer", "character" = "numeric"))
#> <tblcheck problem>
#> Your `a` column should be a number (class `numeric`), but it is an integer (class `integer`).
#> $ type           : chr "class"
#> $ expected       : chr "numeric"
#> $ actual         : chr "integer"
#> $ expected_length: int 1
#> $ actual_length  : int 1
#> $ location       : chr "column"
#> $ column         : chr "a"

tbl_check(ignore_class = c("numeric" = "integer", "numeric" = "character"))
#> <tblcheck problem>
#> Your `b` column should be a number (class `numeric`), but it is a text string (class `character`).
#> $ type           : chr "class"
#> $ expected       : chr "numeric"
#> $ actual         : chr "character"
#> $ expected_length: int 1
#> $ actual_length  : int 1
#> $ location       : chr "column"
#> $ column         : chr "b"

Also after thinking about this more, I wonder if we should skip class checking altogether if, after removing the ignored classes from the set of classes on .solution, we have no classes left. That would let us use ignore_class = "numeric" to ignore instances where the .solution is numeric without having to specify each class/pair that might go with a numeric. (i.e. skip a and b)

.result <- tibble(a = 1L, b = "2", c = "3")
.solution <- tibble(a = 1, b = 2, c = 3L)

tbl_check(ignore_class = "numeric")

[rossellhayes]

Do this in a loop to avoid problem of ignore_class("integer" = "numeric", "numeric" = "character")

[rossellhayes]

@gadenbuie

• 4b5d6aa should resolve the bug you found
• 85af52e skips class checking if all classes of expected are included in ignore_class

[gadenbuie]

Nice work! 🥳