fix: fix cse bug when window function is nested

crates/polars-plan/src/logical_plan/optimizer/cse_expr.rs

@@ -361,6 +361,9 @@ impl Visitor for ExprIdentifierVisitor<'_> {
 
     fn pre_visit(&mut self, node: &Self::Node) -> PolarsResult<VisitRecursion> {
         if skip_pre_visit(node.to_aexpr(), self.is_group_by) {
+            // Still add to the stack so that a parent becomes invalidated.
+            self.visit_stack
+                .push(VisitRecord::SubExprId(Identifier::new(), false));
             return Ok(VisitRecursion::Skip);
         }
 

py-polars/tests/unit/test_cse.py

@@ -1,5 +1,6 @@
 import re
-from datetime import date, datetime
+import typing
+from datetime import date, datetime, timedelta
 from tempfile import NamedTemporaryFile
 from typing import Any
 
@@ -594,3 +595,52 @@ def test_cse_11958() -> None:
         "diff3": [None, None, None, 30, 30],
         "diff4": [None, None, None, None, 40],
     }
+
+
+@typing.no_type_check
+def test_cse_14047() -> None:
+    ldf = pl.LazyFrame(
+        {
+            "timestamp": pl.datetime_range(
+                datetime(2024, 1, 12),
+                datetime(2024, 1, 12, 0, 0, 0, 150_000),
+                "10ms",
+                eager=True,
+                closed="left",
+            ),
+            "price": list(range(15)),
+        }
+    )
+
+    def count_diff(
+        price: pl.Expr, upper_bound: float = 0.1, lower_bound: float = 0.001
+    ):
+        span_end_to_curr = (
+            price.count()
+            .cast(int)
+            .rolling("timestamp", period=timedelta(seconds=lower_bound))
+        )
+        span_start_to_curr = (
+            price.count()
+            .cast(int)
+            .rolling("timestamp", period=timedelta(seconds=upper_bound))
+        )
+        return (span_start_to_curr - span_end_to_curr).alias(
+            f"count_diff_{upper_bound}_{lower_bound}"
+        )
+
+    def s_per_count(count_diff, span) -> pl.Expr:
+        return (span[1] * 1000 - span[0] * 1000) / count_diff
+
+    spans = [(0.001, 0.1), (1, 10)]
+    count_diff_exprs = [count_diff(pl.col("price"), span[0], span[1]) for span in spans]
+    s_per_count_exprs = [
+        s_per_count(count_diff, span).alias(f"zz_{span}")
+        for count_diff, span in zip(count_diff_exprs, spans)
+    ]
+
+    exprs = count_diff_exprs + s_per_count_exprs
+    ldf = ldf.with_columns(*exprs)
+    assert_frame_equal(
+        ldf.collect(comm_subexpr_elim=True), ldf.collect(comm_subexpr_elim=False)
+    )