Upgrade testthat to use 3rd edition

[nikolas-burkoff]

From https://github.com/insightsengineering/coredev-tasks/issues/154

Part of insightsengineering/NEST-roadmap#65

[Melkiades]

I think we already do this right @nikolas-burkoff ?

[nikolas-burkoff]

We have if the DESC file has Config/testthat/edition: 3 in https://testthat.r-lib.org/articles/third-edition.html#activating

[Melkiades]

I think it is worth it. Error messages with waldo are much more informative and warnings are better handled. Is testthat at version >3 on our platforms @arkadiuszbeer ? If so I suggest to complete this

[arkadiuszbeer]

@Melkiades yes, we are using [email protected]

[Melkiades]

I was wondering what should we do in case of warnings. ATM it does return the warning and not the output. For example:

>   # Check with LOW abnormality.
>   result <- testthat::expect_warning(
+     s_count_abnormal_by_baseline(
+       df = df,
+       .var = "myrange",
+       abnormal = "LOW",
+       variables = list(id = "myid", baseline = "mybase")
+     )
+   ) %>% suppressWarnings()
> result
<simpleWarning in as_factor_keep_attributes(df[[.var]], na_level = na_level): automatically converting character variable df[[.var]] to factor, better manually convert to factor to avoid failures>

Should I make a wrapper of expect_warning like custom_expect_warning that preserves the old behavior with a double call or checking the piece twice in the test itself? I do not see other solutions

[Melkiades]

btw there are hundreds of errors :D

[Melkiades]

talking with @shajoezhu and due to the amount of work needed to fix this, we put it back to the stack to be dealt with during next increment