ARROW-13739: [R] Support dplyr::count() and tally()

[thisisnic]

No description provided.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-13739

[github-actions]

⚠️ Ticket has not been started in JIRA, please click 'Start Progress'.

[nealrichardson]

There's also a skipped test-dataset-something.R test about count, you should unskip it in this PR

[thisisnic]

The failing test here fails once per session with something like

Failure (test-dplyr-count-tally.R:43:3): count/tally with wt and grouped data
`via_batch <- rlang::eval_tidy(expr, rlang::new_data_mask(rlang::env(input = record_batch(tbl))))` threw an unexpected warning.
Message: Subsetting quosures with `[[` is deprecated as of rlang 0.4.0
Please use `quo_get_expr()` instead.
This warning is displayed once per session.
Class: deprecatedWarning/warning/condition

but on subsequent runs in the same session, it's fine and passes. I'm a bit unsure of what to do here as I looked for the line of code which triggers it but see no [[ there.

Is this something we've seen before @nealrichardson ?

[nealrichardson]

The failing test here fails once per session with something like

Failure (test-dplyr-count-tally.R:43:3): count/tally with wt and grouped data
`via_batch <- rlang::eval_tidy(expr, rlang::new_data_mask(rlang::env(input = record_batch(tbl))))` threw an unexpected warning.
Message: Subsetting quosures with `[[` is deprecated as of rlang 0.4.0
Please use `quo_get_expr()` instead.
This warning is displayed once per session.
Class: deprecatedWarning/warning/condition

but on subsequent runs in the same session, it's fine and passes. I'm a bit unsure of what to do here as I looked for the line of code which triggers it but see no [[ there.

Is this something we've seen before @nealrichardson ?

"This warning is displayed once per session" is why it doesn't happen the next time.

I've seen it before but not lately. I'll have to pull and take a look.

[nealrichardson]

Turns out we're getting quosures inside of quosures still. I'll suggest another patch, which might help for any others who are building NSE wrappers, and if that doesn't work, we can go back to the expr() building that dplyr does.

[nealrichardson]

I've fixed the tests, 🤞 we get a passing build now

[nealrichardson]

@jonkeane How do I make this lint message go away?

[jonkeane]

There are a few options: having more words in front of the code or # nolint at the end. We can also turn this one particular lintr off (or it's possible there are settings for it that makes it less sensitive) by adding commented_code_linter = NULL to https://github.com/apache/arrow/blob/master/r/.lintr#L25

I've been meaning to send a PR to lintr to make backticks like this defuse commented code — though their track record of merging these has been spotty so I haven't made it a priority.

[nealrichardson]

Just need to fix the style/lint stuff but then LGTM!