Soften rules for interpolating values

Partially reverts #1368. See bcgov/bcdata#338 for details. If we do this again in the future, I think we have to do argument checking in individual translations. That's a lot of work but allows a whole suite of potentially useful translations for more complex R objects.

NEWS.md

@@ -37,9 +37,6 @@
 
 * The databricks backend now supports creating non-temporary tables too (#1418).
 
-* Clearer error if you attempt to embed non-atomic vectors inside of a generated 
-  query (#1368).
-
 * `x$name` never attempts to evaluate `name` (#1368).
 
 * `rows_patch(in_place = FALSE)` now works when more than one column should be

R/tidyeval.R

@@ -166,10 +166,11 @@ partial_eval_sym <- function(sym, data, env) {
       error_embed("shiny inputs", paste0(name, "$x"))
     }
 
+
     if (is_sql_literal(val)) {
       unname(val)
     } else {
-      error_embed(obj_type_friendly(val), name)
+      val
     }
   } else {
     cli::cli_abort(

tests/testthat/_snaps/backend-.md

@@ -26,11 +26,8 @@
     Code
       lazy_frame(x = 1) %>% filter(x == y$id)
     Condition
-      Error in `filter()`:
-      i In argument: `x == y$id`
-      Caused by error:
-      ! Cannot translate a list to SQL.
-      i Do you want to force evaluation in R with (e.g.) `!!y` or `local(y)`?
+      Error in `<list: id = 1>$id`:
+      ! `$` can only subset database columns, not inlined values.
 
 # useful error if $ used with inlined value
 

tests/testthat/_snaps/tidyeval.md

@@ -16,16 +16,4 @@
       Error:
       ! Cannot translate a data.frame to SQL.
       i Do you want to force evaluation in R with (e.g.) `!!df$x` or `local(df$x)`?
-    Code
-      capture_dot(lf, l)
-    Condition
-      Error:
-      ! Cannot translate an empty list to SQL.
-      i Do you want to force evaluation in R with (e.g.) `!!l` or `local(l)`?
-    Code
-      capture_dot(lf, mean)
-    Condition
-      Error:
-      ! Cannot translate a function to SQL.
-      i Do you want to force evaluation in R with (e.g.) `!!mean` or `local(mean)`?
 

tests/testthat/test-tidyeval.R

@@ -26,15 +26,12 @@ test_that("other objects get informative error", {
 
   input <- structure(list(), class = "reactivevalues")
   x <- structure(function() "y", class = "reactive")
-  l <- list()
   df <- data.frame(x = 1)
 
   expect_snapshot({
     capture_dot(lf, input)
     capture_dot(lf, x())
     capture_dot(lf, df)
-    capture_dot(lf, l)
-    capture_dot(lf, mean)
   }, error = TRUE)
 })