[GitHub] [arrow] nealrichardson commented on a change in pull request #12564: ARROW-15818: [R] Implement initial Substrait consumer in the R bindings

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

nealrichardson commented on a change in pull request #12564:
URL: https://github.com/apache/arrow/pull/12564#discussion_r819626624



##########
File path: r/R/query-engine.R
##########
@@ -296,3 +296,39 @@ ExecNode <- R6Class("ExecNode",
     schema = function() ExecNode_output_schema(self)
   )
 )
+
+do_exec_plan_substrait <- function(.data, substrait_plan) {
+  if (is.string(substrait_plan)) {
+    substrait_plan <- engine__internal__SubstraitFromJSON(substrait_plan)
+  } else if (is.raw(substrait_plan)) {
+    substrait_plan <- buffer(substrait_plan)
+  } else {
+    abort("`substrait_plan` must be a JSON string or raw() vector")
+  }
+
+  plan <- ExecPlan$create()
+
+  if (inherits(.data, "RecordBatchReader")) {
+    source_node <- ExecNode_ReadFromRecordBatchReader(self, dataset$.data)
+  } else if (inehrits(.data, "ArrowTabular")) {
+    dataset <- InMemoryDataset$create(dataset)
+    source_node <- ExecNode_Scan(
+      plan,
+      dataset,
+      Expression$scalar(TRUE),
+      colnames %||% character(0)
+    )
+  } else if (inherits(.data, "Dataset")) {
+    source_node <- ExecNode_Scan(
+      plan,
+      .data,
+      Expression$scalar(TRUE),
+      colnames %||% character(0)
+    )
+  } else {
+    obj_desc <- paste0(class(.data), collapse = " / ")
+    abort(glue("Can't construct source node from object of type {obj_desc}"))
+  }

Review comment:
       Why isn't this just `plan$Scan(.data)`?
   
   This is probably a bigger question, but apparently (with ExecPlans) you have to express the filter and projection into the Scan node in order to get predicate pushdown to work. That might not work when consuming substrait, the substrait consumer is going to need to handle the pushdown.
   
   Also you have a typo `inehrits`




-- 
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