lance-format / lance-format/lance
Migrate FilteredReadExec to native DataFusion partitions
@wjones127 is already working on this.
Since Jun 4, 2026.
- Dominant language
- Rust
- Stars
- 7.1k
- Forks
- 852
- Avg merge
- 3d 18h
- Merged PRs (30d)
- 272
Description
FilteredReadExec has an existing MultiplePartitions(n) threading mode that correctly advertises output_partitioning() = RoundRobinBatch(n) and uses a shared work queue across partitions for load balancing. However, this mode is opt-in — the default is OnePartitionMultipleThreads(get_num_compute_intensive_cpus()), which outputs a single partition and manages parallelism internally.
The goal is to make MultiplePartitions(target_partitions) the default, reading target_partitions from the session config at construction time. This makes FilteredReadExec a proper multi-partition DataFusion source, so that data read by partition i stays on thread i through the rest of the pipeline without crossing thread boundaries.
OnePartitionMultipleThreads mode can likely be retired once this lands.
Depends on #7081 (EnforceDistribution) to handle exchange nodes downstream automatically.
Part of the migration tracked in #6967.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.