[GitHub] [arrow-rs] ritchie46 commented on a change in pull request #341: Fix filter UB and add fast path

[GitBox <gi...@apache.org>]

ritchie46 commented on a change in pull request #341:
URL: https://github.com/apache/arrow-rs/pull/341#discussion_r638488867



##########
File path: arrow/src/compute/kernels/filter.rs
##########
@@ -225,38 +245,40 @@ pub fn filter(array: &Array, filter: &BooleanArray) -> Result<ArrayRef> {
     if filter.null_count() > 0 {
         // this greatly simplifies subsequent filtering code
         // now we only have a boolean mask to deal with
-        let array_data = filter.data_ref();
-        let null_bitmap = array_data.null_buffer().unwrap();
-        let mask = filter.values();
-        let offset = filter.offset();
-
-        let new_mask = buffer_bin_and(mask, offset, null_bitmap, offset, filter.len());
-
-        let array_data = ArrayData::builder(DataType::Boolean)
-            .len(filter.len())
-            .add_buffer(new_mask)
-            .build();
-        let filter = BooleanArray::from(array_data);
+        let filter = prep_null_mask_filter(filter);
         // fully qualified syntax, because we have an argument with the same name
         return crate::compute::kernels::filter::filter(array, &filter);
     }
 
     let iter = SlicesIterator::new(filter);
-
-    let mut mutable =
-        MutableArrayData::new(vec![array.data_ref()], false, iter.filter_count);
-    iter.for_each(|(start, end)| mutable.extend(0, start, end));
-    let data = mutable.freeze();
-    Ok(make_array(data))
+    if iter.filter_count == array.len() {
+        let data = array.data().clone();
+        Ok(make_array(data))

Review comment:
       `dyn Array` is a trait object and does not implement `Sized`




-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org