sql escape for 'non-translated' function in window context

[talegari]

In the context of the following issue: #1527

Something like this does not work at the moment (generated sql does ignores window options) for spark sql backend:

dbplyr::lazy_frame(rep_df) |> 
  group_by(user_id) %>%
  dbplyr::window_order(dates) %>%
  dbplyr::window_frame(-Inf, -1) %>%
  mutate(list_amount = sql("collect_list(amount)")) %>% 
  dbplyr::sql_render()

In general, would it be better to generate the sql one would expect by replacing whatever that it there in the sql call.

In this case, it is:

SELECT
`df`.*,
collect_list(`amount`) OVER (PARTITION BY `user_id` ORDER BY `dates` ROWS BETWEEN UNBOUNDED PRECEDING AND 1 PRECEDING) AS `list_amount`
FROM `df`

Why is this required: There is no way to generate this sql chunk via dbplyr and thereby breaks one's workflow. Its practically impossible to cover all "translations" that some backend offers. Would it make sense to create a meaningful "escape hatch"?

Personally, for a serious dbplyr user like me, I will be forced to switch to some other tool say pyspark (which I do not want to) for day-to-day work or do some monkey patching with sdf_sql with handwritten sql (I choose dbplyr for convenience and elegance).