This is an automated email from the ASF dual-hosted git repository. dheres pushed a commit to branch empty_projection in repository https://gitbox.apache.org/repos/asf/arrow-datafusion.git
commit 69eba492244574316138ea1fe16ce37d6a297466 Author: Daniël Heres <[email protected]> AuthorDate: Thu Jul 27 17:08:12 2023 +0200 Change empty projection to not add an extra column --- datafusion/optimizer/src/push_down_projection.rs | 25 ++++++++---------------- 1 file changed, 8 insertions(+), 17 deletions(-) diff --git a/datafusion/optimizer/src/push_down_projection.rs b/datafusion/optimizer/src/push_down_projection.rs index eb9ae3c981..4545234864 100644 --- a/datafusion/optimizer/src/push_down_projection.rs +++ b/datafusion/optimizer/src/push_down_projection.rs @@ -495,23 +495,14 @@ fn push_down_scan( .filter_map(ArrowResult::ok) .collect(); - if projection.is_empty() { - if has_projection && !schema.fields().is_empty() { - // Ensure that we are reading at least one column from the table in case the query - // does not reference any columns directly such as "SELECT COUNT(1) FROM table", - // except when the table is empty (no column) - projection.insert(0); - } else { - // for table scan without projection, we default to return all columns - projection = scan - .source - .schema() - .fields() - .iter() - .enumerate() - .map(|(i, _)| i) - .collect::<BTreeSet<usize>>(); - } + if projection.is_empty() && !has_projection { + // for table scan without projection, we default to return all columns + projection = schema + .fields() + .iter() + .enumerate() + .map(|(i, _)| i) + .collect::<BTreeSet<usize>>(); } // Building new projection from BTreeSet
