Skip to content

Spark: Enable Comet Scan in Iceberg #13793

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

Closed
hsiang-c wants to merge 5 commits into from
Closed

Spark: Enable Comet Scan in Iceberg #13793

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
cfcde40
Enable Comet Scan in Iceberg Spark
hsiang-c Aug 12, 2025
35b4dde
Fallback to Spark whenever there are delete files
hsiang-c Aug 13, 2025
92c9376
Use useCometBatchReads
hsiang-c Aug 13, 2025
90b4b4f
Use Comet 0.10.0 w/ deletion support
hsiang-c Sep 16, 2025
6578841
Merge branch 'main' into comet_scan
hsiang-c Sep 24, 2025
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
2 changes: 1 addition & 1 deletion gradle/libs.versions.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,7 @@ awssdk-s3accessgrants = "2.3.0"
bson-ver = "4.11.5"
caffeine = "2.9.3"
calcite = "1.40.0"
comet = "0.8.1"
comet = "0.10.0"
datasketches = "6.2.0"
delta-standalone = "3.3.2"
delta-spark = "3.3.2"
Expand Down
4 changes: 2 additions & 2 deletions spark/v3.4/spark/src/main/java/org/apache/iceberg/spark/source/SparkBatch.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,7 +151,7 @@ private OrcBatchReadConf orcBatchReadConf() {
// - Parquet vectorization is enabled
// - only primitives or metadata columns are projected
// - all tasks are of FileScanTask type and read only Parquet files
private boolean useParquetBatchReads() {
protected boolean useParquetBatchReads() {
return readConf.parquetVectorizationEnabled()
&& expectedSchema.columns().stream().allMatch(this::supportsParquetBatchReads)
&& taskGroups.stream().allMatch(this::supportsParquetBatchReads);
Expand All @@ -175,7 +175,7 @@ private boolean supportsParquetBatchReads(Types.NestedField field) {
return field.type().isPrimitiveType() || MetadataColumns.isMetadataColumn(field.fieldId());
}

private boolean useCometBatchReads() {
protected boolean useCometBatchReads() {
return readConf.parquetVectorizationEnabled()
&& readConf.parquetReaderType() == ParquetReaderType.COMET
&& expectedSchema.columns().stream().allMatch(this::supportsCometBatchReads)
Expand Down
9 changes: 8 additions & 1 deletion spark/v3.4/spark/src/main/java/org/apache/iceberg/spark/source/SparkScan.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,7 @@
import java.util.Optional;
import java.util.function.Supplier;
import java.util.stream.Collectors;
import org.apache.comet.parquet.SupportsComet;
import org.apache.iceberg.BlobMetadata;
import org.apache.iceberg.ScanTask;
import org.apache.iceberg.ScanTaskGroup;
Expand Down Expand Up @@ -95,7 +96,7 @@
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

abstract class SparkScan implements Scan, SupportsReportStatistics {
abstract class SparkScan implements Scan, SupportsReportStatistics, SupportsComet {
private static final Logger LOG = LoggerFactory.getLogger(SparkScan.class);
private static final String NDV_KEY = "ndv";

Expand Down Expand Up @@ -351,4 +352,10 @@ protected long adjustSplitSize(List<? extends ScanTask> tasks, long splitSize) {
return splitSize;
}
}

@Override
public boolean isCometEnabled() {
SparkBatch batch = (SparkBatch) this.toBatch();
return batch.useCometBatchReads();
}
}
16 changes: 2 additions & 14 deletions spark/v3.5/spark/src/main/java/org/apache/iceberg/spark/source/SparkBatch.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,7 +151,7 @@ private OrcBatchReadConf orcBatchReadConf() {
// - Parquet vectorization is enabled
// - only primitives or metadata columns are projected
// - all tasks are of FileScanTask type and read only Parquet files
private boolean useParquetBatchReads() {
protected boolean useParquetBatchReads() {
return readConf.parquetVectorizationEnabled()
&& expectedSchema.columns().stream().allMatch(this::supportsParquetBatchReads)
&& taskGroups.stream().allMatch(this::supportsParquetBatchReads);
Expand All @@ -175,19 +175,7 @@ private boolean supportsParquetBatchReads(Types.NestedField field) {
return field.type().isPrimitiveType() || MetadataColumns.isMetadataColumn(field.fieldId());
}

private boolean useCometBatchReads() {
return readConf.parquetVectorizationEnabled()
&& readConf.parquetReaderType() == ParquetReaderType.COMET
&& expectedSchema.columns().stream().allMatch(this::supportsCometBatchReads)
&& taskGroups.stream().allMatch(this::supportsParquetBatchReads);
}

private boolean supportsCometBatchReads(Types.NestedField field) {
return field.type().isPrimitiveType()
&& !field.type().typeId().equals(Type.TypeID.UUID)
&& field.fieldId() != MetadataColumns.ROW_ID.fieldId()
&& field.fieldId() != MetadataColumns.LAST_UPDATED_SEQUENCE_NUMBER.fieldId();
}
q

// conditions for using ORC batch reads:
// - ORC vectorization is enabled
Expand Down
9 changes: 8 additions & 1 deletion spark/v3.5/spark/src/main/java/org/apache/iceberg/spark/source/SparkScan.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,7 @@
import java.util.Optional;
import java.util.function.Supplier;
import java.util.stream.Collectors;
import org.apache.comet.parquet.SupportsComet;
import org.apache.iceberg.BlobMetadata;
import org.apache.iceberg.ScanTask;
import org.apache.iceberg.ScanTaskGroup;
Expand Down Expand Up @@ -95,7 +96,7 @@
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

abstract class SparkScan implements Scan, SupportsReportStatistics {
abstract class SparkScan implements Scan, SupportsReportStatistics, SupportsComet {
private static final Logger LOG = LoggerFactory.getLogger(SparkScan.class);
private static final String NDV_KEY = "ndv";

Expand Down Expand Up @@ -351,4 +352,10 @@ protected long adjustSplitSize(List<? extends ScanTask> tasks, long splitSize) {
return splitSize;
}
}

@Override
public boolean isCometEnabled() {
SparkBatch batch = (SparkBatch) this.toBatch();
return batch.useCometBatchReads();
}
}
4 changes: 2 additions & 2 deletions spark/v4.0/build.gradle
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,7 +80,7 @@ project(":iceberg-spark:iceberg-spark-${sparkMajorVersion}_${scalaVersion}") {
exclude group: 'org.roaringbitmap'
}

compileOnly "org.apache.datafusion:comet-spark-spark3.5_2.13:${libs.versions.comet.get()}"
compileOnly "org.apache.datafusion:comet-spark-spark${sparkMajorVersion}_2.13:${libs.versions.comet.get()}"

implementation libs.parquet.column
implementation libs.parquet.hadoop
Expand Down Expand Up @@ -190,7 +190,7 @@ project(":iceberg-spark:iceberg-spark-extensions-${sparkMajorVersion}_${scalaVer
testImplementation libs.avro.avro
testImplementation libs.parquet.hadoop
testImplementation libs.awaitility
testImplementation "org.apache.datafusion:comet-spark-spark3.5_2.13:${libs.versions.comet.get()}"
testImplementation "org.apache.datafusion:comet-spark-spark${sparkMajorVersion}_2.13:${libs.versions.comet.get()}"
testImplementation(testFixtures(project(':iceberg-parquet')))

// Required because we remove antlr plugin dependencies from the compile configuration, see note above
Expand Down
2 changes: 1 addition & 1 deletion spark/v4.0/spark/src/main/java/org/apache/iceberg/spark/source/SparkBatch.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -175,7 +175,7 @@ private boolean supportsParquetBatchReads(Types.NestedField field) {
return field.type().isPrimitiveType() || MetadataColumns.isMetadataColumn(field.fieldId());
}

private boolean useCometBatchReads() {
protected boolean useCometBatchReads() {
return readConf.parquetVectorizationEnabled()
&& readConf.parquetReaderType() == ParquetReaderType.COMET
&& expectedSchema.columns().stream().allMatch(this::supportsCometBatchReads)
Expand Down
9 changes: 8 additions & 1 deletion spark/v4.0/spark/src/main/java/org/apache/iceberg/spark/source/SparkScan.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,7 @@
import java.util.Optional;
import java.util.function.Supplier;
import java.util.stream.Collectors;
import org.apache.comet.parquet.SupportsComet;
import org.apache.iceberg.BlobMetadata;
import org.apache.iceberg.ScanTask;
import org.apache.iceberg.ScanTaskGroup;
Expand Down Expand Up @@ -95,7 +96,7 @@
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

abstract class SparkScan implements Scan, SupportsReportStatistics {
abstract class SparkScan implements Scan, SupportsReportStatistics, SupportsComet {
private static final Logger LOG = LoggerFactory.getLogger(SparkScan.class);
private static final String NDV_KEY = "ndv";

Expand Down Expand Up @@ -351,4 +352,10 @@ protected long adjustSplitSize(List<? extends ScanTask> tasks, long splitSize) {
return splitSize;
}
}

@Override
public boolean isCometEnabled() {
SparkBatch batch = (SparkBatch) this.toBatch();
return batch.useCometBatchReads();
}
}
Loading