(datafusion) branch main updated: perf: defer nulls creation until needed in `ArrowBytesViewMap` (#25014)

[gi...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

github-merge-queue[bot] pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/datafusion.git


The following commit(s) were added to refs/heads/main by this push:
     new 7b89517295 perf: defer nulls creation until needed in `ArrowBytesViewMap` (#25014)
7b89517295 is described below

commit 7b89517295a9c65af5c17437b4abe9efd5972fd5
Author: Raz Luvaton <16...@users.noreply.github.com>
AuthorDate: Thu Sep 10 12:29:41 2026 +0000

    perf: defer nulls creation until needed in `ArrowBytesViewMap` (#25014)
    
    ## Which issue does this PR close?
    
    N/A
    
    ## Rationale for this change
    No need to hold `NullBufferBuilder` and keep appending to it when we can
    create it once much faster
    
    ## What changes are included in this PR?
    
    remove `NullBufferBuilder` from `ArrowBytesViewMap` and create when
    converting to state
    
    ## What is the testing strategy for this PR?
    existing tests
    
    ## Are there any user-facing changes?
    no
---
 datafusion/physical-expr-common/src/binary_map.rs      |  2 +-
 datafusion/physical-expr-common/src/binary_view_map.rs | 18 +++++-------------
 2 files changed, 6 insertions(+), 14 deletions(-)

diff --git a/datafusion/physical-expr-common/src/binary_map.rs b/datafusion/physical-expr-common/src/binary_map.rs
index 16e00b6549..c82e473f99 100644
--- a/datafusion/physical-expr-common/src/binary_map.rs
+++ b/datafusion/physical-expr-common/src/binary_map.rs
@@ -685,7 +685,7 @@ where
 }
 
 /// Returns a `NullBuffer` with a single null value at the given index
-fn single_null_buffer(num_values: usize, null_index: usize) -> NullBuffer {
+pub(crate) fn single_null_buffer(num_values: usize, null_index: usize) -> NullBuffer {
     let mut null_builder = NullBufferBuilder::new(num_values);
     null_builder.append_n_non_nulls(null_index);
     null_builder.append_null();
diff --git a/datafusion/physical-expr-common/src/binary_view_map.rs b/datafusion/physical-expr-common/src/binary_view_map.rs
index 29f4014c5f..0483ed5044 100644
--- a/datafusion/physical-expr-common/src/binary_view_map.rs
+++ b/datafusion/physical-expr-common/src/binary_view_map.rs
@@ -17,8 +17,7 @@
 
 //! [`ArrowBytesViewMap`] and [`ArrowBytesViewSet`] for storing maps/sets of values from
 //! `StringViewArray`/`BinaryViewArray`.
-use crate::binary_map::OutputType;
-use arrow::array::NullBufferBuilder;
+use crate::binary_map::{OutputType, single_null_buffer};
 use arrow::array::cast::AsArray;
 use arrow::array::{
     Array, ArrayRef, BinaryViewArray, BinaryViewBuilder, ByteView, StringViewBuilder,
@@ -148,8 +147,6 @@ where
     in_progress: Vec<u8>,
     /// Completed buffers containing string data
     completed: Vec<Buffer>,
-    /// Tracks null values (true = null)
-    nulls: NullBufferBuilder,
 
     /// random state used to generate hashes
     random_state: RandomState,
@@ -194,7 +191,6 @@ where
             views: Vec::new(),
             in_progress: Vec::new(),
             completed: Vec::new(),
-            nulls: NullBufferBuilder::new(0),
             random_state: RandomState::default(),
             hashes_buffer: vec![],
             null: None,
@@ -330,7 +326,6 @@ where
                     let payload = make_payload_fn(None);
                     let null_index = self.views.len();
                     self.views.push(0);
-                    self.nulls.append_null();
                     self.null = Some((payload, null_index));
                     payload
                 };
@@ -455,7 +450,10 @@ where
         }
 
         // Build null buffer if we have any nulls
-        let null_buffer = self.nulls.finish();
+        let null_buffer = match self.null {
+            Some((_payload, index)) => Some(single_null_buffer(self.views.len(), index)),
+            None => None,
+        };
 
         let views = ScalarBuffer::from(self.views);
         let array =
@@ -529,7 +527,6 @@ where
     /// allocation in those buffers.
     unsafe fn append_inline_view(&mut self, view: u128) -> u128 {
         self.views.push(view);
-        self.nulls.append_non_null();
         view
     }
 
@@ -556,7 +553,6 @@ where
         };
 
         self.views.push(view);
-        self.nulls.append_non_null();
         view
     }
 
@@ -590,13 +586,11 @@ where
         let in_progress_size = self.in_progress.allocated_size();
         let completed_size = self.completed.allocated_size()
             + self.completed.iter().map(Buffer::capacity).sum::<usize>();
-        let nulls_size = self.nulls.allocated_size();
 
         map_size
             + views_size
             + in_progress_size
             + completed_size
-            + nulls_size
             + self.hashes_buffer.allocated_size()
     }
 }
@@ -967,7 +961,6 @@ mod tests {
             + map.in_progress.allocated_size()
             + map.completed.allocated_size()
             + map.completed.iter().map(Buffer::capacity).sum::<usize>()
-            + map.nulls.allocated_size()
             + map.hashes_buffer.allocated_size();
         assert_eq!(map.size(), expected_size);
 
@@ -976,7 +969,6 @@ mod tests {
             + map.views.len() * size_of::<u128>()
             + map.in_progress.capacity()
             + map.completed.iter().map(Buffer::len).sum::<usize>()
-            + map.nulls.allocated_size()
             + map.hashes_buffer.allocated_size();
         let retained_capacity_delta = (map.views.capacity() - map.views.len())
             * size_of::<u128>()


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@datafusion.apache.org
For additional commands, e-mail: commits-help@datafusion.apache.org