rtpsw commented on code in PR #14041:
URL: https://github.com/apache/arrow/pull/14041#discussion_r966128176
##########
cpp/src/arrow/compute/exec/source_node.cc:
##########
@@ -291,13 +291,194 @@ struct TableSourceNode : public SourceNode {
}
};
+template <typename This, typename Options>
+struct SchemaSourceNode : public SourceNode {
+ SchemaSourceNode(ExecPlan* plan, std::shared_ptr<Schema> schema,
+ arrow::AsyncGenerator<util::optional<ExecBatch>> generator)
+ : SourceNode(plan, schema, generator) {}
+
+ static Result<ExecNode*> Make(ExecPlan* plan, std::vector<ExecNode*> inputs,
+ const ExecNodeOptions& options) {
+ RETURN_NOT_OK(ValidateExecNodeInputs(plan, inputs, 0, This::kKindName));
+ const auto& cast_options = checked_cast<const Options&>(options);
+ auto& it_maker = cast_options.it_maker;
+ auto& schema = cast_options.schema;
+
+ auto io_executor = plan->exec_context()->executor();
+ auto it = it_maker();
+
+ RETURN_NOT_OK(ValidateSchemaSourceNodeInput(io_executor, schema,
This::kKindName));
+ ARROW_ASSIGN_OR_RAISE(auto generator, This::MakeGenerator(it, io_executor,
schema));
+ return plan->EmplaceNode<This>(plan, schema, generator);
+ }
+
+ static arrow::Status ValidateSchemaSourceNodeInput(
+ arrow::internal::Executor* io_executor, const std::shared_ptr<Schema>&
schema,
+ const char* kKindName) {
+ if (schema == NULLPTR) {
+ return Status::Invalid(kKindName, " requires schema which is not null");
+ }
+ if (io_executor == NULLPTR) {
+ return Status::Invalid(kKindName, " requires IO-executor which is not
null");
+ }
Review Comment:
IIRC, I ran into a runtime problem when creating a generator using a null
executor. I'm not sure whether the alternative of using the default executor
would work well for generators intended for IO, rather than CPU, work. I'm also
not aware of a default IO executor, but if there is one then I'd try using it
here.
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
