improved CSE (common subexpression elimination) for c.group_by mode

docs/CHANGELOG.md

@@ -1,3 +1,7 @@
+## 1.14.2 (2024-09-04)
+
+- improved CSE (common subexpression elimination) for `c.group_by` mode
+
 ## 1.14.1 (2024-09-01)
 
 - python 3.13t (free threading) support

src/convtools/__init__.py

@@ -4,4 +4,4 @@
 from ._dt import DateGrid, DateTimeGrid
 
 
-__version__ = "1.14.1"
+__version__ = "1.14.2"

src/convtools/_aggregations.py

@@ -160,17 +160,17 @@ def fuzzy_merge_aggregate_cmp(x, y):
 
 
 def no_side_effects_test(x):
-    if isinstance(x, AstCall):
-        x = x.func
-    if isinstance(x, AstAttribute):
-        x = x.value
-    elif (
+    if (
         isinstance(x, AstCompare)
         and isinstance(x.ops[0], AstIs)
         and isinstance(x.comparators[0], AstName)
         and x.comparators[0].id == "_none"
     ):
         x = x.left
+    elif isinstance(x, AstCall):
+        x = x.func
+    if isinstance(x, AstAttribute):
+        x = x.value
     return isinstance(x, AstName) and x.id.startswith("agg_data_")
 
 

tests/test_group_by.py

@@ -1015,6 +1015,23 @@ def test_group_by_reducers_reuse():
         },
     ]
 
+    converter = (
+        c.group_by(c.item("x"))
+        .aggregate(
+            {
+                "x": c.item("x"),
+                "y": c.ReduceFuncs.Sum(c.item("obj", "y")),
+                "z": c.ReduceFuncs.Sum(c.item("obj", "z")),
+            }
+        )
+        .gen_converter(debug=False)
+    )
+    code_str = format_code(get_code_str(converter))
+    assert (
+        code_str.count("row_['obj']") == 1
+        or code_str.count('row_["obj"]') == 1
+    )
+
 
 def test_aggregate_reducers_reuse():
     converter = c.aggregate(