PERF-modin-project#5709: Avoid re-putting a distributed Series to the…

… engine's object store at `.map()` (modin-project#5704) Co-authored-by: Iaroslav Igoshev <[email protected]> Signed-off-by: Dmitry Chigarev <[email protected]>
mvashishtha · Feb 27, 2023 · d724802 · d724802
1 parent 3d38d37
commit d724802
Showing 1 changed file with 8 additions and 0 deletions.
diff --git a/modin/pandas/series.py b/modin/pandas/series.py
@@ -1266,6 +1266,14 @@ def map(self, arg, na_action=None):  # noqa: PR01, RT01, D200
         """
         Map values of Series according to input correspondence.
         """
+        if isinstance(arg, type(self)):
+            # HACK: if we don't cast to pandas, then the execution engine will try to
+            # propagate the distributed Series to workers and most likely would have
+            # some performance problems.
+            # TODO: A better way of doing so could be passing this `arg` as a query compiler
+            # and broadcast accordingly.
+            arg = arg._to_pandas()
+
         if not callable(arg) and hasattr(arg, "get"):
             mapper = arg