get_dupes() grouped data error fix

[jzadra]

Description

If data is grouped:

1. Saves the grouping structure and then removes it before checking for duplicates.
2. One warning per session is issued notifying user that get_dupes() isn't checking within groups but rather over the whole data set, however will still return an object with the same grouping structure.
3. Groups are reapplied before returning dupes.

Added tests that check that

1. The grouping structure of the original data is the same as the result.
2. The output for grouped data is the same as the output for ungrouped data other than the grouping structure.

Related Issue

Addresses #329.

Example

x <- mtcars %>% group_by(cyl, gear)
get_dupes(x)

[codecov]

Codecov Report

Merging #345 into master will not change coverage.
The diff coverage is 100%.

@@          Coverage Diff          @@
##           master   #345   +/-   ##
=====================================
  Coverage     100%   100%           
=====================================
  Files          26     26           
  Lines         996   1002    +6     
=====================================
+ Hits          996   1002    +6

Impacted Files	Coverage Δ
R/get_dupes.R	100% <100%> (ø)	⬆️

[jzadra]

I'm not clear on exactly what codecov is if anyone can provide some feedback on what I've done wrong!

[sfirke]

codecov checks which lines of code are run when all the tests are run. Your PR has two "misses", that is there are two lines that are not run in any test: https://codecov.io/gh/sfirke/janitor/pull/345/diff?src=pr&el=tree#diff-Ui9nZXRfZHVwZXMuUg Looks like it's the grouped_df warning, that could be checked with a new test that looks at expect_message.

[sfirke]

Code itself looks great though! 👍 After reviewing it, now I wonder if it's impossible to test that chunk that codecov says is missing coverage, because the message only prints if interactive() and in the test environment interactive() returns false.

I don't see a way to fake an interactive environment to run that warning. The only idea I have is to just not have tests for those two lines and acknowledge that by ending those two lines with # nocov to tell codecov to ignore them.

[jzadra]

Looks like you were correct about the interactive issue. I'll add #nocov to those lines.

[sfirke]

Looks like it didn't run, I'm guessing because interactive() would be FALSE? I think you can just add `# nocov to those two lines in the function itself.

[sfirke]

Excellent! Thanks @jzadra