Repository navigation

Issues

[DISCUSSION] 2026 Q4 Roadmap
#26684 · alamb opened on Oct 23, 2026
11

Wrong results: a filter like k / 10 = 5 makes ORDER BY and GROUP BY treat k as constant
Bug
apache/datafusion#26779 · opened on Oct 26, 2026
Clustered aggregates don't release memory after emitting completed groups under partial group clustering
Feature
apache/datafusion#26777 · opened on Oct 26, 2026
WindowAggExec and grouped TopK can exceed batch_size
apache/datafusion#26772 · opened on Oct 26, 2026
Window ordered by (total DESC, k) fails SanityCheckPlan when k is a function of a filtered constant column
apache/datafusion#26762 · opened on Oct 26, 2026
7
Three slow tests set the length of the extended and hash-collision CI jobs
apache/datafusion#26759 · opened on Oct 26, 2026
2
Avoid unnecessary clones in logical planning and expression simplification
apache/datafusion#26756 · opened on Oct 26, 2026
2
datafusion-cli drops rows from CSV, TSV and JSON output when a result has more than --maxrows rows
apache/datafusion#26754 · opened on Oct 26, 2026
1
Support writing NULL partition values to the Hive default partition directory
apache/datafusion#26749 · opened on Oct 25, 2026
Skip all-null inputs when merging min/max in Statistics::try_merge_iter
apache/datafusion#26748 · opened on Oct 25, 2026
Wrong results: order-preserving RepartitionExec keeps orderings of its input that do not hold after the merge
apache/datafusion#26745 · opened on Oct 25, 2026
1
DISTINCT ON with nested payloads can fail memory reservation after spilling
Bug
apache/datafusion#26744 · opened on Oct 25, 2026
TopK queries can fail plan validation after exact sort pushdown
Bug
apache/datafusion#26739 · opened on Oct 25, 2026
proto: recursion protection for physical plan serde
Feature
apache/datafusion#26738 · opened on Oct 25, 2026
proto: recursion protection for deserializing logical plans
Feature
apache/datafusion#26737 · opened on Oct 25, 2026
1
Fully sorted partial aggregate still fails with ResourcesExhausted after emitting completed groups, because the table keeps its capacity
apache/datafusion#26725 · opened on Oct 24, 2026
[EPIC] Simplify memory-limited (spilling) execution
apache/datafusion#26711 · opened on Oct 24, 2026
[dbolympics] Correlated scalar subqueries with ORDER BY ... LIMIT 1 fail with "Physical plan does not support logical expression ScalarSubquery"
apache/datafusion#26710 · opened on Oct 24, 2026
6
WindowAggExec and BoundedWindowAggExec reserve no memory, so a memory limit does not bound them
apache/datafusion#26707 · opened on Oct 24, 2026
1
Support reusing the physical plan in prepared statements
Feature
apache/datafusion#26706 · opened on Oct 24, 2026
Partial aggregate state reaches the final aggregate without a memory reservation (median, array_agg)
apache/datafusion#26705 · opened on Oct 24, 2026
1
[dbolympics] Subqueries in an aggregate (GROUP BY key, aggregate argument or FILTER) fail with "Physical plan does not support logical expression"
apache/datafusion#26704 · opened on Oct 24, 2026
1
Final hash aggregate fails with ResourcesExhausted when its materialized output is larger than its table
apache/datafusion#26701 · opened on Oct 24, 2026
1
Run the sqllogictest corpus across a matrix of batch sizes
apache/datafusion#26700 · opened on Oct 24, 2026
7
[EPIC] Custom ExecutionPlan nodes: correct results, and the same optimizations as built-in nodes
apache/datafusion#26697 · opened on Oct 24, 2026
Hash aggregate output slices keep the whole dictionary, so downstream work grows with the group count
apache/datafusion#26696 · opened on Oct 24, 2026
4