[GitHub] [arrow-rs] alamb commented on a change in pull request #585: use exponential search to speed up lexico partition

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

alamb commented on a change in pull request #585:
URL: https://github.com/apache/arrow-rs/pull/585#discussion_r676125985



##########
File path: arrow/src/compute/kernels/partition.rs
##########
@@ -73,6 +73,30 @@ impl<'a> LexicographicalPartitionIterator<'a> {
     }
 }
 
+/// Exponential search is to remedy for the case when array size and cardinality are both large
+/// see <https://en.wikipedia.org/wiki/Exponential_search>
+#[inline]
+fn exponential_search(
+    indices: &[usize],
+    target: &usize,
+    comparator: &LexicographicalComparator<'_>,
+) -> usize {
+    let mut bound = 1;
+    while bound < indices.len()
+        && comparator.compare(&indices[bound], target) != Ordering::Greater
+    {
+        bound *= 2;
+    }
+    // invariant after while loop:
+    // indices[bound / 2] <= target < indices[min(indices.len(), bound + 1)]
+    // where <= and < are defined by the comparator;
+    // note here we have right = min(indices.len(), bound + 1) because indices[bound] might
+    // actually be considered and must be included.
+    (bound / 2)
+        + indices[(bound / 2)..indices.len().min(bound + 1)]

Review comment:
       I wonder if some of the performance improvement also comes from potentially making a smaller sequence -- aka `(bound/2) .. len` rather than `partition_point .. len`.
   
   




-- 
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: github-unsubscribe@arrow.apache.org

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