[GitHub] [arrow] dragosmg commented on a diff in pull request #13960: ARROW-14071: [R] Try to arrow_eval user-defined functions - implicit approach

[GitBox <gi...@apache.org>]

dragosmg commented on code in PR #13960:
URL: https://github.com/apache/arrow/pull/13960#discussion_r955319688


##########
r/R/dplyr-eval.R:
##########
@@ -121,3 +128,120 @@ format_expr <- function(x) {
   }
   head(out, 1)
 }
+
+# vector of function names that do not have corresponding bindings, but we
+# shouldn't try to translate
+translation_exceptions <- c(
+  "c",
+  "$",
+  "factor",
+  # "~",
+  # "(",
+  "across",
+  ":",
+  "[",
+  "regex",
+  "fixed",
+  "list",
+  "%>%",
+  # all the types functions
+  "int8",
+  "int16",
+  "int32",
+  "int64",
+  "uint8",
+  "uint16",
+  "uint32",
+  "uint64",
+  "float16",
+  "halffloat",
+  "float32",
+  "float",
+  "float64",
+  "boolean",
+  "bool",
+  "utf8",
+  "large_utf8",
+  "binary",
+  "large_binary",
+  "fixed_size_binary",
+  "string",
+  "date32",
+  "date64",
+  "time32",
+  "time64",
+  "duration",
+  "null",
+  "timestamp",
+  "decimal",
+  "decimal128",
+  "decimal256"
+)
+
+register_user_bindings <- function(quo, .env) {
+  unknown_functions_chr <- setdiff(
+    all_funs(quo),
+    union(
+      names(.env),
+      translation_exceptions
+    )
+  )
+
+  if (length(unknown_functions_chr != 0)) {
+    # get the actual functions from the quosure's original environment or, if
+    # the call contains `::`, get the function from the namespace
+    unknown_functions <- purrr::map_if(
+      .x = unknown_functions_chr,
+      .p = ~ !grepl("::", .x),
+      .f = ~ tryCatch(as_function(.x, env = rlang::quo_get_env(quo)), error = function(e) NULL),
+      .else = ~ asNamespace(sub(":{+}.*?$", "", .x))[[sub("^.*?:{+}", "", .x)]]
+    )
+
+    # set the original quosure environment as the parent environment for the
+    # functions
+    parent.env(.env) <- rlang::quo_get_env(quo)
+    for (i in seq_along(unknown_functions)) {
+      unknown_fn_name <- unknown_functions_chr[[i]]
+      unknown_fn <- unknown_functions[[i]]
+      if (!is.null(unknown_fn) && registrable(unknown_fn, .env)) {
+        environment(unknown_fn) <- .env
+        register_binding(
+          unknown_fn_name,
+          unknown_fn,
+          registry = .env,
+          update_cache = TRUE
+        )
+      } else {
+        # if there are call we don't have bindings for, try to register them first
+        unknown_function_body <-
+          tryCatch(
+            rlang::fn_body(unknown_fn),
+            error = function(e) NULL
+          )
+        if (!is.null(unknown_function_body)) {
+          new_quo <- rlang::new_quosure(unknown_function_body[[2]], env = quo_get_env(quo))

Review Comment:
   Most of my unit tests define functions in-line, and this hasn't been a problem. For example, the definition of `nchar2`, [here](https://github.com/dragosmg/arrow/blob/917065133153273418b859d6875b29918df306ee/r/tests/testthat/test-dplyr-eval.R#L21). I think `rlang::as_function(...)` makes that not be a problem.



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org