Github user funes commented on a diff in the pull request: https://github.com/apache/spark/pull/661#discussion_r12311565 --- Diff: mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala --- @@ -136,7 +136,7 @@ object Vectors { new DenseVector(v.toArray) // Can't use underlying array directly, so make a new one } case v: BSV[Double] => - new SparseVector(v.length, v.index, v.data) + new SparseVector(v.length, v.index.slice(0, v.used), v.data.slice(0, v.used)) --- End diff -- That makes sense. And I just found that it failed the test "Should not copy data". SparseVector from VectorBuilder has this problem of size inconsistency. It uses a dynamic growing approach to resize the index/data array and fills the unused slot zeros. So I think there is no way to completely avoid new allocation.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---