Skip to content

feat(python): integrate Ballista through DataFusion FFI #2252

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
timsaucer wants to merge 13 commits into
base: main
Choose a base branch
from
Draft

feat(python): integrate Ballista through DataFusion FFI #2252

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
50b1b9a
chore(deps): track DataFusion git main and port to its APIs
andygrove Jul 2, 2026
2904c4d
Merge apache/main into chore/track-datafusion-git-main; bump datafusi…
andygrove Jul 24, 2026
25f9cbd
chore: cargo fmt
andygrove Jul 24, 2026
6868a2b
chore: sync vendored datafusion_common.proto and python Cargo.lock
andygrove Jul 24, 2026
b1b5f48
test: regenerate TPC-H plan-stability golden files for DataFusion main
andygrove Jul 24, 2026
29317fb
Merge apache/main into chore/track-datafusion-git-main
andygrove Aug 6, 2026
bd6b63c
chore(deps): bump DataFusion rev to 70c26a06 and port to its APIs
andygrove Aug 6, 2026
1aa4065
test: refresh AQE snapshot for DataFusion's null_aware join display
andygrove Aug 6, 2026
c91233f
chore(python): regenerate Cargo.lock for the bumped DataFusion rev
andygrove Aug 6, 2026
f2bcf3c
feat(python): integrate Ballista FFI query planner
timsaucer Aug 7, 2026
9e81351
feat(python): expose Ballista extension codecs
timsaucer Aug 7, 2026
44169f1
Update example to use new way to export codecs and query planner to d…
timsaucer Aug 7, 2026
0a07776
feat(python): use atomic Ballista session extensions
timsaucer Aug 7, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1,306 changes: 576 additions & 730 deletions Cargo.lock
View file
Open in desktop

Large diffs are not rendered by default.

25 changes: 12 additions & 13 deletions Cargo.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,21 +35,20 @@ edition = "2024"
rust-version = "1.88.0"

[workspace.dependencies]
arrow = { version = "58.4", features = ["ipc_compression"] }
arrow-flight = { version = "58.4", features = ["flight-sql-experimental"] }
arrow = { version = "59.2.0", features = ["ipc_compression"] }
arrow-flight = { version = "59.2.0", features = ["flight-sql-experimental"] }
clap = { version = "4.5", features = ["derive", "cargo"] }

# Keep these as minor-version requirements rather than exact patch pins. The
# python workspace shares these crates through path dependencies, and
# datafusion-python lags behind datafusion patch releases, so pinning a patch
# here can break the wheel build. Cargo.lock records the exact patch version.
datafusion = "54"
datafusion-cli = "54"
datafusion-functions-aggregate-common = "54"
datafusion-proto = "54"
datafusion-proto-common = "54"
datafusion-spark = "54"
datafusion-substrait = "54"
# DataFusion crates are pinned to a git rev of apache/datafusion main rather than
# a crates.io release. Bump this rev to advance to newer DataFusion commits.
datafusion = { git = "https://github.com/apache/datafusion.git", rev = "e08aed1e5de41dcf81d529140dae07723b942a5e" }
datafusion-cli = { git = "https://github.com/apache/datafusion.git", rev = "e08aed1e5de41dcf81d529140dae07723b942a5e" }
datafusion-ffi = { git = "https://github.com/apache/datafusion.git", rev = "e08aed1e5de41dcf81d529140dae07723b942a5e" }
datafusion-functions-aggregate-common = { git = "https://github.com/apache/datafusion.git", rev = "e08aed1e5de41dcf81d529140dae07723b942a5e" }
datafusion-proto = { git = "https://github.com/apache/datafusion.git", rev = "e08aed1e5de41dcf81d529140dae07723b942a5e" }
datafusion-proto-common = { git = "https://github.com/apache/datafusion.git", rev = "e08aed1e5de41dcf81d529140dae07723b942a5e" }
datafusion-spark = { git = "https://github.com/apache/datafusion.git", rev = "e08aed1e5de41dcf81d529140dae07723b942a5e" }
datafusion-substrait = { git = "https://github.com/apache/datafusion.git", rev = "e08aed1e5de41dcf81d529140dae07723b942a5e" }

ctor = { version = "1.0" }
insta = "1.47"
Expand Down
34 changes: 17 additions & 17 deletions ballista/client/tests/context_checks.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1180,23 +1180,23 @@ mod supported {
)?;

let expected = [
"+-------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+",
"| plan_type | plan |",
"+-------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+",
"| Plan with Metrics | =========SuccessfulStage[stage_id=1, partitions=1]========= |",
"| | SortShuffleWriterExec: partitioning=Hash([id@0], 16), metrics=[output_rows=..., input_rows=..., spill_bytes=..., spill_count=..., repart_time=..., spill_time=..., write_time=...] |",
"| | AggregateExec: mode=Partial, gby=[id@0 as id], aggr=[count(Int64(1))], metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., spill_count=..., spilled_bytes=..., spilled_rows=..., skipped_aggregation_rows=..., peak_mem_used=..., aggregate_arguments_time=..., aggregation_time=..., emitting_time=..., time_calculating_group_ids=..., reduction_factor=...] |",
"| | ProjectionExec: expr=[__unnest_placeholder(make_array(Int64(1),Int64(2),Int64(3),Int64(4),Int64(5)),depth=1)@0 as id], metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., expr_0_eval_time=...] |",
"| | UnnestExec, metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., input_batches=..., input_rows=...] |",
"| | ProjectionExec: expr=[[1, 2, 3, 4, 5] as __unnest_placeholder(make_array(Int64(1),Int64(2),Int64(3),Int64(4),Int64(5)))], metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., expr_0_eval_time=...] |",
"| | PlaceholderRowExec, metrics=[...] |",
"| | |",
"| | =========SuccessfulStage[stage_id=2, partitions=16]========= |",
"| | ShuffleWriterExec: partitioning: None, metrics=[output_rows=..., input_rows=..., write_time=...] |",
"| | ProjectionExec: expr=[count(Int64(1))@1 as count(*), id@0 as id], metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., expr_0_eval_time=..., expr_1_eval_time=...] |",
"| | AggregateExec: mode=FinalPartitioned, gby=[id@0 as id], aggr=[count(Int64(1))], metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., spill_count=..., spilled_bytes=..., spilled_rows=..., peak_mem_used=..., aggregate_arguments_time=..., aggregation_time=..., emitting_time=..., time_calculating_group_ids=...] |",
"| | ShuffleReaderExec: upstream_stage: 1, partitioning: Hash([id@0], 16), metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., decoded_bytes=..., fetch_requests=..., fetch_retries=..., local_partitions=..., remote_partitions=..., fetch_time=..., local_read_time=..., permit_wait_time=...] |",
"+-------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+",
"+-------------------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+",
"| plan_type | plan |",
"+-------------------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+",
"| Plan with Metrics | =========SuccessfulStage[stage_id=1, partitions=1]========= |",
"| | SortShuffleWriterExec: partitioning=Hash([id@0], 16), metrics=[output_rows=..., input_rows=..., spill_bytes=..., spill_count=..., repart_time=..., spill_time=..., write_time=...] |",
"| | AggregateExec: mode=Partial, gby=[id@0 as id], aggr=[count(Int64(1))], metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., spill_count=..., spilled_bytes=..., spilled_rows=..., skipped_aggregation_rows=..., aggregate_arguments_time=..., aggregation_time=..., emitting_time=..., time_calculating_group_ids=..., reduction_factor=...] |",
"| | ProjectionExec: expr=[__unnest_placeholder(make_array(Int64(1),Int64(2),Int64(3),Int64(4),Int64(5)),depth=1)@0 as id], metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., expr_0_eval_time=...] |",
"| | UnnestExec, metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., input_batches=..., input_rows=...] |",
"| | ProjectionExec: expr=[[1, 2, 3, 4, 5] as __unnest_placeholder(make_array(Int64(1),Int64(2),Int64(3),Int64(4),Int64(5)))], metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., expr_0_eval_time=...] |",
"| | PlaceholderRowExec, metrics=[...] |",
"| | |",
"| | =========SuccessfulStage[stage_id=2, partitions=16]========= |",
"| | ShuffleWriterExec: partitioning: None, metrics=[output_rows=..., input_rows=..., write_time=...] |",
"| | ProjectionExec: expr=[count(Int64(1))@1 as count(*), id@0 as id], metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., expr_0_eval_time=..., expr_1_eval_time=...] |",
"| | AggregateExec: mode=FinalPartitioned, gby=[id@0 as id], aggr=[count(Int64(1))], metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., spill_count=..., spilled_bytes=..., spilled_rows=..., aggregate_arguments_time=..., aggregation_time=..., emitting_time=..., time_calculating_group_ids=...] |",
"| | ShuffleReaderExec: upstream_stage: 1, partitioning: Hash([id@0], 16), metrics=[output_rows=..., elapsed_compute=..., output_bytes=..., output_batches=..., decoded_bytes=..., fetch_requests=..., fetch_retries=..., local_partitions=..., remote_partitions=..., fetch_time=..., local_read_time=..., permit_wait_time=...] |",
"+-------------------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+",
];

assert_batches_eq!(expected, &[sanitized]);
Expand Down
20 changes: 11 additions & 9 deletions ballista/client/tests/context_setup.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -109,16 +109,16 @@ mod standalone {
use ballista_core::serde::BallistaPhysicalExtensionCodec;
use datafusion::{
assert_batches_eq,
catalog::Session,
common::exec_err,
execution::{
SessionState, SessionStateBuilder, TaskContext, context::QueryPlanner,
},
execution::{SessionStateBuilder, TaskContext, context::QueryPlanner},
logical_expr::LogicalPlan,
physical_plan::ExecutionPlan,
prelude::{SessionConfig, SessionContext},
};
use datafusion_proto::{
logical_plan::LogicalExtensionCodec, physical_plan::PhysicalExtensionCodec,
logical_plan::LogicalExtensionCodec,
physical_plan::{PhysicalExtensionCodec, PhysicalProtoConverterExtension},
};

#[tokio::test]
Expand Down Expand Up @@ -299,7 +299,7 @@ mod standalone {
fn try_decode_table_provider(
&self,
_buf: &[u8],
_table_ref: &datafusion::sql::TableReference,
_table_ref: &datafusion::common::TableReference,
_schema: datafusion::arrow::datatypes::SchemaRef,
_ctx: &TaskContext,
) -> datafusion::error::Result<
Expand All @@ -312,7 +312,7 @@ mod standalone {

fn try_encode_table_provider(
&self,
_table_ref: &datafusion::sql::TableReference,
_table_ref: &datafusion::common::TableReference,
_node: std::sync::Arc<dyn datafusion::catalog::TableProvider>,
_buf: &mut Vec<u8>,
) -> datafusion::error::Result<()> {
Expand Down Expand Up @@ -366,21 +366,23 @@ mod standalone {
buf: &[u8],
inputs: &[Arc<dyn datafusion::physical_plan::ExecutionPlan>],
ctx: &TaskContext,
proto_converter: &dyn PhysicalProtoConverterExtension,
) -> datafusion::error::Result<Arc<dyn datafusion::physical_plan::ExecutionPlan>>
{
self.invoked
.store(true, std::sync::atomic::Ordering::Relaxed);
self.codec.try_decode(buf, inputs, ctx)
self.codec.try_decode(buf, inputs, ctx, proto_converter)
}

fn try_encode(
&self,
node: Arc<dyn datafusion::physical_plan::ExecutionPlan>,
buf: &mut Vec<u8>,
proto_converter: &dyn PhysicalProtoConverterExtension,
) -> datafusion::error::Result<()> {
self.invoked
.store(true, std::sync::atomic::Ordering::Relaxed);
self.codec.try_encode(node, buf)
self.codec.try_encode(node, buf, proto_converter)
}
}

Expand All @@ -392,7 +394,7 @@ mod standalone {
async fn create_physical_plan(
&self,
_logical_plan: &LogicalPlan,
_session_state: &SessionState,
_session_state: &dyn Session,
) -> datafusion::error::Result<Arc<dyn ExecutionPlan>> {
exec_err!("does not work")
}
Expand Down
8 changes: 8 additions & 0 deletions ballista/core/proto/ballista.proto
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,9 +56,17 @@ message BallistaPhysicalPlanNode {
UnorderedRangeRepartitionExecNode unordered_range_repartition = 8;
OrderedRangeRepartitionExecNode ordered_range_repartition = 9;
PerPartitionFilterExecNode per_partition_filter = 10;
DistributedQueryExecNode distributed_query = 11;
}
}

message DistributedQueryExecNode {
string scheduler_url = 1;
repeated KeyValuePair settings = 2;
bytes logical_plan = 3;
string session_id = 4;
}

// Value-range router over N locally-sorted overlapping input partitions.
// Redistributes them into K range-disjoint output partitions where each
// output is fully sorted on the routing expression. Uses N × K scatter
Expand Down
Loading
Loading