Using `dplyr::dense_rank` without attaching `dplyr` causes the function replacement to fail

[multimeric]

Here is a strange edge case. Under the following conditions, dplyr::dense_rank is not converted to a DENSE_RANK() SQL function:

• dplyr::dense_rank() is inside a dplyr::across()
• dplyr is not attached via library
• The function is called via the namespace: dplyr::dense_rank

Here are some examples. Firstly, the failure case. Note how vec_rank (which is a dplyr, and not an SQL function) is used:

> dbplyr::lazy_frame(a=5:1, b=1:5) |> dplyr::mutate(dplyr::across(dplyr::everything(), dplyr::dense_rank)) |> dplyr::show_query()
<SQL>
SELECT
  vec_rank(`a`, 'dense' AS `ties`, 'na' AS `incomplete`) AS `a`,
  vec_rank(`b`, 'dense' AS `ties`, 'na' AS `incomplete`) AS `b`
FROM `df`
Warning messages:
1: Named arguments ignored for SQL vec_rank
2: Named arguments ignored for SQL vec_rank

However, if we simply remove the dplyr:: namespace, it works fine:

> dbplyr::lazy_frame(a=5:1, b=1:5) |> dplyr::mutate(dplyr::across(dplyr::everything(), dense_rank)) |> dplyr::show_query()
<SQL>
SELECT
  CASE
WHEN (NOT((`a` IS NULL))) THEN DENSE_RANK() OVER (PARTITION BY (CASE WHEN ((`a` IS NULL)) THEN 1 ELSE 0 END) ORDER BY `a`)
END AS `a`,
  CASE
WHEN (NOT((`b` IS NULL))) THEN DENSE_RANK() OVER (PARTITION BY (CASE WHEN ((`b` IS NULL)) THEN 1 ELSE 0 END) ORDER BY `b`)
END AS `b`
FROM `df`

Also, if we don't use across():

> dbplyr::lazy_frame(a=5:1, b=1:5) |> dplyr::mutate(a = dplyr::dense_rank(a)) |> dplyr::show_query()
<SQL>
SELECT
  CASE
WHEN (NOT((`a` IS NULL))) THEN DENSE_RANK() OVER (PARTITION BY (CASE WHEN ((`a` IS NULL)) THEN 1 ELSE 0 END) ORDER BY `a`)
END AS `a`,
  `b`
FROM `df`