Wrong class of input dataset gives unclear error message during harmo_process()

[zchenmr]

I got the error below when trying to run harmo_process, which was resolved by converting the input dataset into a tibble. However, it was unclear from the error message and traceback where the issue came from. The class of the original input dataset was "data.frame" (instead of ""tbl_df" "tbl" "data.frame"). When I ran the is_dataset function, it returned TRUE so I initially thought that the problem was with other processing elements and not the dataset.

[GuiFabre]

@a-trottier : here is the same bug you face, I'll keep you posted if anything comes.

@zchenmr : can you give me an example of a dataset that generates the problem ?
Thanks

[zchenmr]

The dataset is on an external server so I can't share it, but I can send a summary report if that helps. I'm not sure if it's actually the class causing the problem, I've run harmo_process with other datasets of class "data.frame" and those still worked without any issues.

[GuiFabre]

hello @zchenmr. I think I found the problem, which might not be due to the package.
In a nutshell, across() function from dplyr, and widely used in the package does not work with the grouping variable.

tidyverse/dplyr#6127

# does not work and throw an error
iris %>% group_by(Species) %>% mutate(across(Species,as.character))

# does not work but silently bypass Species (everything is coerced into a character, except Species)
iris %>% group_by(Species) %>% mutate(across(everything(),as.character))

# does work
iris %>% group_by(Species) %>% mutate(Species = as.character(Species))

I'll check across the package when the dataset is grouped and ensure it is ungroup() before

@