#4331 Updated distinct_if and distinct_at to include .keep_all argument

[beansrowning]

As @DaniMori suggested in #4331, this simply tacks the .keep_all argument onto the scoped variants of distinct().

I wasn't sure if this was needed (or desired) but I have also wanted a scoped variant of distinct() which would allow you the option to keep all columns. Adding it to distinct_all() seemed superfluous.

My use case was reading in many data files where I am appending a file column to track provenance. Since my data is very raw, I want to check that there isn't a duplicate observation across two or more files.

Something similar to this:

# Read in many excel files with many columns
# In this case, they're 1 row each
to_process <- dir("<my_dir>", pattern = ".*\\.xlsx$", full.names = TRUE)

# Originally, I assumed one could just do this:
# (but it's not possible due to distinct's mutate() semantics)
raw_data <- lapply(to_process, my_parsing_function) %>%
  bind_rows() %>%
  mutate(file = to_process) %>%
  distinct(-file, .keep_all = TRUE)

# Since I can't do that, I have to do some variation of this 
# (though there are likely better ways):
raw_data <- lapply(to_process, my_parsing_function) %>%
  bind_rows()

distinct_vars <- lapply(names(raw_data), as.name)

raw_data <- raw_data %>%
  mutate(file = to_process) %>%
  distinct(!!!distinct_vars, .keep_all = TRUE)

# Ideally I would just run the scoped version `distinct_at()`
# on all columns, save for file which I append later:
raw_data <- lapply(to_process, my_parsing_function) %>%
  bind_rows() %>%
  mutate(file = to_process) %>%
  distinct_at(vars(-file), .keep_all = TRUE)

# So not ~really~ that big of a deal, but it seems logical ¯\_(ツ)_/¯

[romainfrancois]

Still think it makes sense to add it to distinct_all() too

[DaniMori]

Still think it makes sense to add it to distinct_all() too

I agree that it's superfluous as @beansrowning said, but it wouldn't be the only "superfluous" case to be in dplyr. The most obvious one is select_all(). I can't see a use case where these superfluous functions or parameters would be useful, but there must be some for sure, so they may apply to distinct_all(..., .keep_all = TRUE).

Do they? I really don't know, probably someone else can state a good case in favor/against it.

[beansrowning]

@romainfrancois @DaniMori Fair point, it's reasonable to have it consistent across.

Added.

[beansrowning]

Looks like it fails the check for distinct_all() since I made it a recursive default argument.

I'll have to think about it.

EDIT: Sike. After a little bit more sleep it is clear the issue lies between the chair and the keyboard 🤦‍♂ . Should be fixed now.

[romainfrancois]

closes #4331

[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/