work around some dbplyr::translate_sql limitations in hof_*

[yitao-li]

would be interesting to see if there is any workaround for the issue in #2513 with accessing struct fields from lambda body (aside from the mutate(dest_col = sql(...)) workaround)

[mattpollock]

Since periods are allowed in R variable names, why not make it a little more R-specific using the $ operator? In sparklyr.nested I was a bit hand-wavy and allowed x.value or x$value to be used interchangeably in sdf_select.

Something like:

new_frame <- test_frame %>%
   group_by(patients) %>%
   summarise(analyteDates = collect_list(analyteDate),
             valuesList = collect_list(struct(value))) %>%
   hof_transform(dest_col = value, expr = valuesList, x %->% (x$value))

[yitao-li]

Update: actually no work around needed -- dbplyr::translate_sql already translates a$b in R to a.b in SQL