Added missing connection parameter to the dbplyr_fill0 implementation

[rsund]

The connection parameter was not passed to the first translate_sql() call in the dbplyr_fill0 implementation for databases without support for IGNORE NULLS. Because of that the quotes other than ` ` were not correct for that particular translate_sql() call.

Following the example in the fill documentation the produced SQL was as follows with a database that uses " " as quotes:

<SQL>
SELECT "group", "name", "role", MAX("n_squirrels") OVER (PARTITION BY "..dbplyr_partion_1") AS "n_squirrels", 
MAX("n_squirrels2") OVER (PARTITION BY "..dbplyr_partion_2") AS "n_squirrels2", "id" FROM
(SELECT "group", "name", "role", "n_squirrels", "n_squirrels2", "id", SUM(CASE WHEN (((`n_squirrels`) IS NULL)) THEN 0 
ELSE 1 END) OVER (ORDER BY "id" ROWS UNBOUNDED PRECEDING) AS "..dbplyr_partion_1", 
SUM(CASE WHEN (((`n_squirrels2`) IS NULL))  THEN 0 ELSE 1 END) OVER (ORDER BY "id" ROWS UNBOUNDED PRECEDING) AS
"..dbplyr_partion_2" FROM "squirrels") "q01"

So following the SUM(CASE WHEN... there are wrong quotes in two places (around the variable names).

The fix was simple and only requires that the missing connection parameter is added to the translate_sql() call as shown in the commit.

[rsund]

Interestingly the tests failed. I should have been testing more carefully before sending a PR, but I couldn't anticipate that such a simple change had such consequences, especially as things were working just fine with the database with " " quotes after the commit.

Any idea what should be changed?

Maybe instead of adding a connection parameter, the current connection could be stored with the function set_current_con() in the beginning of the dbplyr_fill0() function so that the engine with " " quotes could fetch the connection (in the translate_sql() there is a line for fetching the con if parameter is missing: con <- con %||% sql_current_con() %||% simulate_dbi()).

But I still don't quite get that why adding a clearly missing (and important) parameter changes the behavior in CASE WHEN translation?!

[mgirlich]

Good catch, thanks for the PR. The tests fail due to an incorrect check when translating case_when() that only happens for SQLite. I created a PR to fix this. Once this is merged, the tests pass.

It would be great if you could also add a NEWS entry.

[rsund]

OK, good to hear. A commit for the NEWS entry added as requested.