divyankshah commented on code in PR #5235:
URL: https://github.com/apache/datafusion-comet/pull/5235#discussion_r3761775980


##########
native/spark-expr/src/array_funcs/nested_float_normalize.rs:
##########
@@ -0,0 +1,208 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements.  See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership.  The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License.  You may obtain a copy of the License at
+//
+//   http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied.  See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+use crate::math_funcs::internal::normalize_float;
+use arrow::array::{
+    Array, ArrayRef, AsArray, FixedSizeListArray, Float32Array, Float64Array, 
LargeListArray,
+    ListArray, StructArray,
+};
+use arrow::datatypes::{DataType, Float32Type, Float64Type};
+use std::sync::Arc;
+
+pub(super) fn has_float_leaf(dt: &DataType) -> bool {
+    match dt {
+        DataType::Float32 | DataType::Float64 => true,
+        DataType::List(field) | DataType::LargeList(field) | 
DataType::FixedSizeList(field, _) => {
+            has_float_leaf(field.data_type())
+        }
+        DataType::Struct(fields) => fields.iter().any(|f| 
has_float_leaf(f.data_type())),
+        _ => false,
+    }
+}
+
+/// Recursively rebuilds nested arrays with `-0.0` normalized to `0.0` and NaN 
canonicalized
+/// in any Float32/Float64 leaves.
+pub(super) fn normalize_nested_floats(array: &ArrayRef) -> ArrayRef {

Review Comment:
   Hi @peterxcli, thanks for digging into this and for finding the #5166 
thread, that's genuinely useful context.
   
   I think the two pieces you're describing are different sizes though. Moving 
the helper is a small refactor, but making NormalizeNaNAndZero actually recurse 
into List/Struct doesn't do much on its own, since nothing on the JVM side 
constructs it with a nested data_type today. The part that actually delivers 
the perf win needs contraintExpressions.scala to detect the ArrayTransform(arr, 
x -> NormalizeNaNAndZero(x)) shape and collapse it, and that's real serde work 
with its own tests, on a node Spark inserts broadly for joins, grouping and 
sort, not just collect_set or arrays_overlap.
   
   Does that split sound reasonable? I'd rather keep it out of this PR and open 
it as its own issue instead, referencing this thread and the one andygrove 
raised on #5166. I'll also add all the deatils to it.



-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to