mbutrovich commented on code in PR #2671:
URL: https://github.com/apache/iceberg-rust/pull/2671#discussion_r3715740272
##########
crates/integrations/datafusion/tests/integration_datafusion_test.rs:
##########
@@ -95,6 +112,531 @@ fn get_table_creation(
Ok(creation)
}
+async fn get_multi_file_table_context(
+ namespace_name: &str,
+ table_name: &str,
+ data_file_count: usize,
+) -> Result<(Arc<MemoryCatalog>, NamespaceIdent, String)> {
+ let iceberg_catalog = Arc::new(get_iceberg_catalog().await);
+ let namespace = NamespaceIdent::new(namespace_name.to_string());
+ set_test_namespace(&iceberg_catalog, &namespace).await?;
+
+ let creation = get_table_creation(temp_path(), table_name, None)?;
+ iceberg_catalog.create_table(&namespace, creation).await?;
+
+ let write_ctx = SessionContext::new_with_config(
+ SessionConfig::new().with_target_partitions(data_file_count),
+ );
+ let arrow_schema = Arc::new(ArrowSchema::new(vec![
+ Field::new("foo1", DataType::Int32, false),
+ Field::new("foo2", DataType::Utf8, false),
+ ]));
+
+ let batches: Vec<RecordBatch> = (1..=data_file_count as i32)
+ .map(|idx| {
+ RecordBatch::try_new(arrow_schema.clone(), vec![
+ Arc::new(Int32Array::from(vec![idx])) as ArrayRef,
+ Arc::new(StringArray::from(vec![format!("row-{idx}")])) as
ArrayRef,
+ ])
+ })
+ .collect::<std::result::Result<_, _>>()?;
+
+ let partitions = batches.into_iter().map(|batch| vec![batch]).collect();
+ let source_table = Arc::new(MemTable::try_new(arrow_schema,
partitions).unwrap());
+ write_ctx
+ .register_table("source_table", source_table)
+ .unwrap();
+
+ let catalog =
Arc::new(IcebergCatalogProvider::try_new(iceberg_catalog.clone()).await?);
+ write_ctx.register_catalog("catalog", catalog);
+
+ let insert_sql =
+ format!("INSERT INTO catalog.{namespace_name}.{table_name} SELECT *
FROM source_table");
+ let batches = write_ctx
+ .sql(&insert_sql)
+ .await
+ .unwrap()
+ .collect()
+ .await
+ .unwrap();
+ assert_eq!(batches.len(), 1);
+
+ let rows_inserted = batches[0]
+ .column(0)
+ .as_any()
+ .downcast_ref::<UInt64Array>()
+ .unwrap();
+ assert_eq!(rows_inserted.value(0), data_file_count as u64);
+
+ Ok((iceberg_catalog, namespace, table_name.to_string()))
+}
+
+async fn get_multi_row_group_table_context(
+ namespace_name: &str,
+ table_name: &str,
+) -> Result<(Arc<MemoryCatalog>, NamespaceIdent, String)> {
+ let iceberg_catalog = Arc::new(get_iceberg_catalog().await);
+ let namespace = NamespaceIdent::new(namespace_name.to_string());
+ set_test_namespace(&iceberg_catalog, &namespace).await?;
+
+ let creation = get_table_creation(temp_path(), table_name, None)?;
+ let table = iceberg_catalog.create_table(&namespace, creation).await?;
+ let arrow_schema: Arc<ArrowSchema> = Arc::new(
+ table
+ .metadata()
+ .current_schema()
+ .as_ref()
+ .try_into()
+ .unwrap(),
+ );
+
+ let file_rows = [
+ vec![
+ (1, "row-1"),
+ (2, "row-2"),
+ (100, "row-100"),
+ (101, "row-101"),
+ ],
+ vec![
+ (50, "row-50"),
+ (150, "row-150"),
+ (151, "row-151"),
+ (152, "row-152"),
+ ],
+ vec![
+ (99, "row-99"),
+ (1000, "row-1000"),
+ (1001, "row-1001"),
+ (1002, "row-1002"),
+ ],
+ ];
+
+ let location_generator =
DefaultLocationGenerator::new(table.metadata()).unwrap();
+ let writer_properties = WriterProperties::builder()
+ .set_max_row_group_row_count(Some(2))
+ .build();
+ let mut data_files = Vec::with_capacity(file_rows.len());
+
+ for (file_idx, rows) in file_rows.into_iter().enumerate() {
+ let parquet_writer_builder = ParquetWriterBuilder::new(
+ writer_properties.clone(),
+ table.metadata().current_schema().clone(),
+ );
+ let file_name_generator = DefaultFileNameGenerator::new(
+ format!("multi-row-group-{file_idx}"),
+ None,
+ iceberg::spec::DataFileFormat::Parquet,
+ );
+ let rolling_file_writer_builder =
RollingFileWriterBuilder::new_with_default_file_size(
+ parquet_writer_builder,
+ table.file_io().clone(),
+ location_generator.clone(),
+ file_name_generator,
+ );
+ let data_file_writer_builder =
DataFileWriterBuilder::new(rolling_file_writer_builder);
+ let mut data_file_writer = data_file_writer_builder.build(None).await?;
+ let batch = RecordBatch::try_new(arrow_schema.clone(), vec![
+ Arc::new(Int32Array::from(
+ rows.iter().map(|(foo1, _)| *foo1).collect::<Vec<_>>(),
+ )) as ArrayRef,
+ Arc::new(StringArray::from(
+ rows.iter().map(|(_, foo2)| *foo2).collect::<Vec<_>>(),
+ )) as ArrayRef,
+ ])?;
+
+ data_file_writer.write(batch).await?;
+ let file_data_files = data_file_writer.close().await?;
+ assert_eq!(file_data_files.len(), 1);
+ data_files.extend(file_data_files);
+ }
+
+ for data_file in &data_files {
+ let file_path = data_file
+ .file_path()
+ .strip_prefix("file://")
+ .unwrap_or(data_file.file_path());
+ let reader = SerializedFileReader::new(File::open(file_path)?)?;
+ assert!(
+ reader.metadata().num_row_groups() > 1,
+ "expected multiple row groups for {}",
+ data_file.file_path()
+ );
+ }
+
+ let tx = Transaction::new(&table);
+ let action = tx.fast_append().add_data_files(data_files);
+ let tx = action.apply(tx)?;
+ tx.commit(iceberg_catalog.as_ref()).await?;
+
+ Ok((iceberg_catalog, namespace, table_name.to_string()))
+}
+
+async fn get_read_context(
+ catalog: Arc<MemoryCatalog>,
+ target_partitions: usize,
+ enable_eager_scan_planning: Option<bool>,
+) -> Result<SessionContext> {
+ let ctx = SessionContext::new_with_config(read_session_config(
+ target_partitions,
+ enable_eager_scan_planning,
+ ));
+ let catalog = Arc::new(IcebergCatalogProvider::try_new(catalog).await?);
+ ctx.register_catalog("catalog", catalog);
+ Ok(ctx)
+}
+
+fn read_session_config(
+ target_partitions: usize,
+ enable_eager_scan_planning: Option<bool>,
+) -> SessionConfig {
+ let config =
SessionConfig::new().with_target_partitions(target_partitions);
+
+ match enable_eager_scan_planning {
+ Some(enabled) => {
+ let mut iceberg_config = IcebergDataFusionConfig::default();
+ iceberg_config.enable_eager_scan_planning = enabled;
+ config.with_option_extension(iceberg_config)
+ }
+ None => config,
+ }
+}
+
+async fn scan_plan(
+ ctx: &SessionContext,
+ namespace: &NamespaceIdent,
+ table_name: &str,
+) -> Arc<dyn ExecutionPlan> {
+ let provider = ctx.catalog("catalog").unwrap();
+ let namespace_name = &namespace[0];
+ let schema = provider.schema(namespace_name).unwrap();
+ let table = schema.table(table_name).await.unwrap().unwrap();
+
+ let state = ctx.state();
+ table.scan(&state, None, &[], None).await.unwrap()
+}
+
+async fn scan_partition_count(
+ ctx: &SessionContext,
+ namespace: &NamespaceIdent,
+ table_name: &str,
+) -> usize {
+ let plan = scan_plan(ctx, namespace, table_name).await;
+ plan.downcast_ref::<IcebergTableScan>()
+ .expect("Expected IcebergTableScan");
+ plan.properties().output_partitioning().partition_count()
+}
+
+fn find_iceberg_scan(plan: &dyn ExecutionPlan) -> Option<&IcebergTableScan> {
+ if let Some(scan) = plan.downcast_ref::<IcebergTableScan>() {
+ return Some(scan);
+ }
+
+ plan.children()
+ .into_iter()
+ .find_map(|child| find_iceberg_scan(child.as_ref()))
+}
+
+#[tokio::test]
+async fn test_multi_file_scan_produces_multiple_partitions() -> Result<()> {
+ let data_file_count = 3;
+ // Ask for more partitions than files to verify scan planning does not
expose empty partitions.
+ let target_partitions = data_file_count + 1;
+ let (iceberg_catalog, namespace, table_name) =
get_multi_file_table_context(
+ "test_multi_file_scan_partitions",
+ "my_table",
+ data_file_count,
+ )
+ .await?;
+ let ctx = get_read_context(iceberg_catalog, target_partitions,
Some(true)).await?;
+ let plan = scan_plan(&ctx, &namespace, &table_name).await;
+ plan.downcast_ref::<IcebergTableScan>()
+ .expect("Expected IcebergTableScan");
+
+ let actual_partition_count =
plan.properties().output_partitioning().partition_count();
+ assert_eq!(actual_partition_count, data_file_count);
+
+ // Pins the eager plan's task-group accounting. The reader settings it
derives from
+ // the retained TableScan are not introspectable, so their equivalence
with the lazy
+ // path is covered behaviorally by
test_multi_partition_scan_matches_single_partition_results.
+ let display = datafusion::physical_plan::displayable(plan.as_ref())
+ .one_line()
+ .to_string();
+ assert!(
+ display.contains("task_groups:[3] tasks:[3]"),
+ "unexpected eager scan display: {display}"
+ );
+
+ Ok(())
+}
+
+#[tokio::test]
+async fn test_multi_file_scan_defaults_to_single_lazy_partition() ->
Result<()> {
+ let data_file_count = 3;
+ let target_partitions = data_file_count + 1;
+ let (iceberg_catalog, namespace, table_name) =
get_multi_file_table_context(
+ "test_multi_file_scan_default_lazy",
+ "my_table",
+ data_file_count,
+ )
+ .await?;
+ let ctx = get_read_context(iceberg_catalog, target_partitions,
None).await?;
+
+ let actual_partition_count = scan_partition_count(&ctx, &namespace,
&table_name).await;
+
+ assert_eq!(actual_partition_count, 1);
+
+ Ok(())
+}
+
+#[tokio::test]
+async fn test_set_enable_eager_scan_planning() -> Result<()> {
+ let data_file_count = 3;
+ let target_partitions = data_file_count + 1;
+ let (iceberg_catalog, namespace, table_name) =
+ get_multi_file_table_context("test_set_eager_scan_planning",
"my_table", data_file_count)
+ .await?;
+ let ctx = get_read_context(iceberg_catalog, target_partitions,
Some(false)).await?;
+
+ ctx.sql("SET iceberg.enable_eager_scan_planning = true")
+ .await
+ .unwrap()
+ .collect()
+ .await
+ .unwrap();
+
+ let actual_partition_count = scan_partition_count(&ctx, &namespace,
&table_name).await;
+
+ assert_eq!(actual_partition_count, data_file_count);
+
+ Ok(())
+}
+
+#[tokio::test]
+async fn test_multi_partition_scan_enforces_global_limit() -> Result<()> {
+ let data_file_count = 3;
+ let limit = 2;
+ let target_partitions = data_file_count + 1;
+ let (iceberg_catalog, namespace, table_name) =
get_multi_file_table_context(
+ "test_multi_partition_scan_limit",
+ "my_table",
+ data_file_count,
+ )
+ .await?;
+ let ctx = get_read_context(iceberg_catalog, target_partitions,
Some(true)).await?;
+ let namespace_name = &namespace[0];
+ let query =
+ format!("SELECT foo1, foo2 FROM catalog.{namespace_name}.{table_name}
LIMIT {limit}");
+ let dataframe = ctx.sql(&query).await.unwrap();
+ let plan = dataframe.create_physical_plan().await.unwrap();
+
+ // The physical optimizer absorbs the initial GlobalLimitExec into
+ // CoalescePartitionsExec; its fetch enforces the global bound in the
final plan.
+ let global_limit_coalescer = plan
+ .downcast_ref::<CoalescePartitionsExec>()
Review Comment:
This is a good end-to-end check that the global limit is actually enforced.
One thought: it pins today's DataFusion optimizer output shape
(`GlobalLimitExec` absorbed into `CoalescePartitionsExec::fetch`). If a future
DataFusion bump changes how that composes, this assertion could fail with no
regression in this crate's code. Might be worth a short comment here noting
that a failure after a `datafusion` version bump likely means the plan shape
changed, not that eager scanning broke, so whoever hits it later doesn't have
to rediscover that.
##########
crates/integrations/datafusion/src/physical_plan/scan_planning.rs:
##########
@@ -0,0 +1,200 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+use std::sync::Arc;
+
+use datafusion::arrow::datatypes::SchemaRef as ArrowSchemaRef;
+use datafusion::error::Result as DFResult;
+use datafusion::prelude::Expr;
+use futures::TryStreamExt;
+use iceberg::arrow::ArrowReaderBuilder;
+use iceberg::expr::Predicate;
+use iceberg::scan::{FileScanTask, TableScan};
+use iceberg::table::Table;
+
+use super::expr_to_predicate::convert_filters_to_predicate;
+use crate::to_datafusion_error;
+
+#[derive(Debug, Clone)]
+pub(crate) struct IcebergScanConfig {
+ /// Snapshot of the table to scan.
+ snapshot_id: Option<i64>,
+ /// Output schema after projection.
+ output_schema: ArrowSchemaRef,
+ /// Projection column names, None means all columns.
+ column_names: Option<Vec<String>>,
+ /// Filters to apply to the table scan.
+ predicates: Option<Predicate>,
+}
+
+impl IcebergScanConfig {
+ pub(crate) fn new(
+ schema: ArrowSchemaRef,
+ snapshot_id: Option<i64>,
+ projection: Option<&Vec<usize>>,
+ filters: &[Expr],
+ ) -> Self {
+ let output_schema = match projection {
+ None => schema.clone(),
+ Some(projection) => Arc::new(schema.project(projection).unwrap()),
+ };
+
+ Self {
+ snapshot_id,
+ output_schema,
+ column_names: get_column_names(schema, projection),
+ predicates: convert_filters_to_predicate(filters),
+ }
+ }
+
+ pub(crate) fn snapshot_id(&self) -> Option<i64> {
+ self.snapshot_id
+ }
+
+ pub(crate) fn output_schema(&self) -> ArrowSchemaRef {
+ self.output_schema.clone()
+ }
+
+ pub(crate) fn column_names(&self) -> Option<&[String]> {
+ self.column_names.as_deref()
+ }
+
+ pub(crate) fn predicates(&self) -> Option<&Predicate> {
+ self.predicates.as_ref()
+ }
+}
+
+/// Result of eager scan planning: the [`TableScan`] that planned the file scan
+/// tasks, alongside those tasks grouped per output partition.
+#[derive(Debug)]
+pub(crate) struct EagerScanPlan {
+ /// The [`TableScan`] used to plan `task_groups`. Retained so that every
output
+ /// partition builds its reader from this same scan, instead of rebuilding
a
+ /// throwaway `TableScan` on each `execute()` call.
+ table_scan: Arc<TableScan>,
+ /// Planned file scan tasks, one group per output partition.
+ task_groups: Vec<Arc<[FileScanTask]>>,
+}
+
+impl EagerScanPlan {
+ /// Number of output partitions, i.e. the number of task groups.
+ pub(crate) fn partition_count(&self) -> usize {
+ self.task_groups.len()
+ }
+
+ /// Total number of planned file scan tasks across all partitions.
+ pub(crate) fn task_count(&self) -> usize {
+ self.task_groups.iter().map(|group| group.len()).sum()
+ }
+
+ /// Returns the task group assigned to `partition`, or `None` if out of
range.
+ pub(crate) fn task_group(&self, partition: usize) ->
Option<Arc<[FileScanTask]>> {
+ self.task_groups.get(partition).cloned()
+ }
+
+ /// Returns an [`ArrowReaderBuilder`] configured for this scan.
+ ///
+ /// This deliberately routes through [`TableScan::arrow_reader_builder`]
rather
+ /// than constructing an [`ArrowReaderBuilder`] directly: it keeps the
reader
+ /// settings (batch size, row group filtering, row selection) sourced from
the
+ /// same place as the lazy path's `TableScan::to_arrow`, so the two scan
paths
+ /// cannot silently drift apart.
+ pub(crate) fn arrow_reader_builder(&self) -> ArrowReaderBuilder {
+ self.table_scan.arrow_reader_builder()
+ }
+}
+
+pub(crate) async fn plan_eager_scan(
+ table: &Table,
+ scan_config: &IcebergScanConfig,
+ target_partitions: usize,
+) -> DFResult<EagerScanPlan> {
+ // Do not cache planned FileScanTasks in the provider in v1. They are
query-specific
+ // because projection, predicate binding, snapshot schema, and delete
planning can differ
+ // between scans. Catalog-backed providers also need fresh metadata on
each scan.
+ // TODO: Revisit provider-level caching for static tables with a precise
cache key.
+ let table_scan = Arc::new(build_table_scan(table, scan_config)?);
+
+ let tasks: Vec<FileScanTask> = table_scan
+ .plan_files()
+ .await
+ .map_err(to_datafusion_error)?
+ .try_collect::<Vec<_>>()
+ .await
+ .map_err(to_datafusion_error)?;
+
+ let task_groups = group_file_scan_tasks_round_robin(tasks,
target_partitions)
+ .into_iter()
+ .map(Arc::<[FileScanTask]>::from)
+ .collect();
+
+ Ok(EagerScanPlan {
+ table_scan,
+ task_groups,
+ })
+}
+
+fn get_column_names(
+ schema: ArrowSchemaRef,
+ projection: Option<&Vec<usize>>,
+) -> Option<Vec<String>> {
+ projection.map(|v| {
+ v.iter()
+ .map(|p| schema.field(*p).name().clone())
+ .collect::<Vec<String>>()
+ })
+}
+
+/// Groups file scan tasks into `target_partitions` groups using a naive
+/// round-robin assignment. Non-empty groups are bounded by `tasks.len()`.
+// TODO: Replace this naive round-robin grouping with size-based grouping once
the
+// first parallel scan path is stable. Keep this v1 simple and deterministic.
+fn group_file_scan_tasks_round_robin(
Review Comment:
Every test that exercises this function uses a task count equal to (or
clamped down to) the partition count, so each group ends up with exactly one
task. Would it be worth adding a direct `#[test]` here for an uneven split,
e.g. 5 tasks over 3 partitions landing as `[2, 2, 1]`? It's a small pure
function with no I/O, so a unit test seems cheap, and right now nothing would
catch a change that preserves total task/row counts but breaks the round-robin
balance itself.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
