morningman commented on code in PR #46911:
URL: https://github.com/apache/doris/pull/46911#discussion_r1966788091
##########
fe/fe-core/src/main/java/org/apache/doris/datasource/paimon/source/PaimonScanNode.java:
##########
@@ -295,33 +292,27 @@ public List<Split> getSplits(int numBackends) throws
UserException {
splitStats.add(splitStat);
}
+ // if applyCountPushdown is true, calcute row count for count pushdown
+ if (applyCountPushdown) {
Review Comment:
In this block, you traverse the `paimonSplits` 3 times, which is not
efficient.
I suggest to use a simple way to do this instead of using `stream()`
```
boolean applyTableCountPushdown = true;
long totalCount = 0;
long deletionVectorCount = 0;
for (org.apache.paimon.table.source.Split s : paimonSplits) {
totalCount += s.rowCount();
Optional<List<DeletionFile>> deletionFiles = s.deletionFiles();
if (deletionFiles.isPresent()) {
for (DeletionFile dv : deletionFiles.get()) {
if (dv != null) {
if (dv.cardinality() == null) {
applyTableCountPushdown = false;
break;
}
deletionVectorCount += dv.cardinality();
}
}
}
if (!applyTableCountPushdown) {
break;
}
}
if (applyTableCountPushdown) {
assignCountToSplits(splits, totalCount - deletionVectorCount);
}
```
##########
fe/fe-core/src/main/java/org/apache/doris/datasource/paimon/source/PaimonScanNode.java:
##########
@@ -245,38 +246,34 @@ public List<Split> getSplits(int numBackends) throws
UserException {
splitStat.setType(SplitReadType.NATIVE);
splitStat.setRawFileConvertable(true);
List<RawFile> rawFiles = optRawFiles.get();
- if (optDeletionFiles.isPresent()) {
- List<DeletionFile> deletionFiles =
optDeletionFiles.get();
- for (int i = 0; i < rawFiles.size(); i++) {
- RawFile file = rawFiles.get(i);
- DeletionFile deletionFile = deletionFiles.get(i);
- LocationPath locationPath = new
LocationPath(file.path(),
- source.getCatalog().getProperties());
- try {
- List<Split> dorisSplits =
FileSplitter.splitFile(
- locationPath,
- getRealFileSplitSize(0),
- null,
- file.length(),
- -1,
- true,
- null,
-
PaimonSplit.PaimonSplitCreator.DEFAULT);
- for (Split dorisSplit : dorisSplits) {
- // the element in DeletionFiles might be
null
- if (deletionFile != null) {
- splitStat.setHasDeletionVector(true);
- ((PaimonSplit)
dorisSplit).setDeletionFile(deletionFile);
- }
- splits.add(dorisSplit);
+ for (int i = 0; i < rawFiles.size(); i++) {
+ RawFile file = rawFiles.get(i);
+ LocationPath locationPath = new
LocationPath(file.path(),
+ source.getCatalog().getProperties());
+ try {
+ List<Split> dorisSplits = FileSplitter.splitFile(
+ locationPath,
+ // if applyCountPushdown is true, we can't
to split the file
+ // becasue the raw file and deletion
vector is one-to-one mapping
+ getRealFileSplitSize(applyCountPushdown ?
Long.MAX_VALUE : 0),
+ null,
+ file.length(),
+ -1,
+ true,
+ null,
+ PaimonSplit.PaimonSplitCreator.DEFAULT);
+ for (Split dorisSplit : dorisSplits) {
+ // try to set deletion file
+ if (optDeletionFiles.isPresent() &&
optDeletionFiles.get().get(i) != null) {
+ ((PaimonSplit)
dorisSplit).setDeletionFile(optDeletionFiles.get().get(i));
+ splitStat.setHasDeletionVector(true);
}
- ++rawFileSplitNum;
- } catch (IOException e) {
- throw new UserException("Paimon error to split
file: " + e.getMessage(), e);
}
+ splits.addAll(dorisSplits);
+ ++rawFileSplitNum;
+ } catch (IOException e) {
+ throw new UserException("Paimon error to split
file: " + e.getMessage(), e);
}
- } else {
- createRawFileSplits(rawFiles, splits,
applyCountPushdown ? Long.MAX_VALUE : 0);
Review Comment:
Why do not need this anymore?
##########
regression-test/suites/external_table_p0/paimon/test_paimon_deletion_vector.groovy:
##########
@@ -0,0 +1,58 @@
+// 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.
+
+suite("test_paimon_deletion_vector",
"p0,external,doris,external_docker,external_docker_doris") {
+
+ logger.info("start paimon test")
+ String enabled = context.config.otherConfigs.get("enablePaimonTest")
+ if (enabled == null || !enabled.equalsIgnoreCase("true")) {
+ logger.info("disabled paimon test")
+ return
+ }
+
+ try {
+ String catalog_name = "test_paimon_deletion_vector"
+ String hdfs_port = context.config.otherConfigs.get("hive2HdfsPort")
+ String externalEnvIp = context.config.otherConfigs.get("externalEnvIp")
+ sql """drop catalog if exists ${catalog_name}"""
+ sql """create catalog if not exists ${catalog_name} properties (
+ "type" = "paimon",
+ "paimon.catalog.type"="filesystem",
+ "warehouse" =
"hdfs://${externalEnvIp}:${hdfs_port}/user/doris/paimon1"
+ );"""
+ sql """use `${catalog_name}`.`db1`"""
+
+ def test_cases = { String force ->
+ sql """ set force_jni_scanner=${force} """
+ qt_1 """select count(*) from deletion_vector_orc;"""
Review Comment:
2 questions:
1. Is your paimon table generated by paimon 1.0?
2. How to make sure that `count(*)` optimization is work? better test it by,
eg, checking the info in `explain` result?
##########
fe/fe-core/src/main/java/org/apache/doris/datasource/paimon/source/PaimonScanNode.java:
##########
@@ -295,33 +292,27 @@ public List<Split> getSplits(int numBackends) throws
UserException {
splitStats.add(splitStat);
}
+ // if applyCountPushdown is true, calcute row count for count pushdown
+ if (applyCountPushdown) {
Review Comment:
And please add UT for this logic, here is the test case generated by Cursor,
I didn't run it, but maybe can help you:
```
package org.apache.doris.datasource.paimon.source;
import org.apache.doris.analysis.TupleDescriptor;
import org.apache.doris.planner.PlanNodeId;
import org.apache.doris.qe.SessionVariable;
import org.apache.doris.spi.Split;
import org.apache.doris.thrift.TPushAggOp;
import org.apache.paimon.table.source.DeletionFile;
import org.junit.Before;
import org.junit.Test;
import org.mockito.Mock;
import org.mockito.MockitoAnnotations;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.List;
import java.util.Optional;
import static org.junit.Assert.assertEquals;
import static org.mockito.Mockito.mock;
import static org.mockito.Mockito.when;
public class PaimonScanNodeTest {
@Mock
private TupleDescriptor tupleDescriptor;
@Mock
private SessionVariable sessionVariable;
private PaimonScanNode scanNode;
@Before
public void setUp() {
MockitoAnnotations.initMocks(this);
scanNode = new PaimonScanNode(new PlanNodeId(1), tupleDescriptor,
false, sessionVariable);
}
@Test
public void testNoDeleteVector() {
// Prepare test data
List<org.apache.paimon.table.source.Split> paimonSplits = new
ArrayList<>();
// Create 3 splits without deletion files
for (int i = 0; i < 3; i++) {
org.apache.paimon.table.source.Split split =
mock(org.apache.paimon.table.source.Split.class);
when(split.rowCount()).thenReturn(100L);
when(split.deletionFiles()).thenReturn(Optional.empty());
paimonSplits.add(split);
}
List<Split> splits = createPaimonSplits(paimonSplits);
// Test count pushdown
scanNode.assignCountToSplits(splits, 300L);
// Verify results
for (Split split : splits) {
assertEquals(100L, ((PaimonSplit)
split).getRowCount().get().longValue());
}
}
@Test
public void testAllDeleteVectorsWithCardinality() {
List<org.apache.paimon.table.source.Split> paimonSplits = new
ArrayList<>();
// Create 3 splits with deletion files that have cardinality
for (int i = 0; i < 3; i++) {
org.apache.paimon.table.source.Split split =
mock(org.apache.paimon.table.source.Split.class);
when(split.rowCount()).thenReturn(100L);
DeletionFile deletionFile = mock(DeletionFile.class);
when(deletionFile.cardinality()).thenReturn(20L);
when(split.deletionFiles()).thenReturn(Optional.of(Arrays.asList(deletionFile)));
paimonSplits.add(split);
}
List<Split> splits = createPaimonSplits(paimonSplits);
// Total count should be 300 - (3 * 20) = 240
scanNode.assignCountToSplits(splits, 240L);
// Verify results (240 / 3 = 80 per split)
for (Split split : splits) {
assertEquals(80L, ((PaimonSplit)
split).getRowCount().get().longValue());
}
}
@Test
public void testSomeDeleteVectorsWithoutCardinality() {
List<org.apache.paimon.table.source.Split> paimonSplits = new
ArrayList<>();
// Create 2 splits with cardinality and 1 without
for (int i = 0; i < 3; i++) {
org.apache.paimon.table.source.Split split =
mock(org.apache.paimon.table.source.Split.class);
when(split.rowCount()).thenReturn(100L);
DeletionFile deletionFile = mock(DeletionFile.class);
// Third split has no cardinality
when(deletionFile.cardinality()).thenReturn(i == 2 ? null : 20L);
when(split.deletionFiles()).thenReturn(Optional.of(Arrays.asList(deletionFile)));
paimonSplits.add(split);
}
List<Split> splits = createPaimonSplits(paimonSplits);
// Should not apply count pushdown when some deletion vectors don't
have cardinality
for (Split split : splits) {
assertEquals(Optional.empty(), ((PaimonSplit)
split).getRowCount());
}
}
@Test
public void testMixedDeleteVectors() {
List<org.apache.paimon.table.source.Split> paimonSplits = new
ArrayList<>();
// First split: no deletion file
org.apache.paimon.table.source.Split split1 =
mock(org.apache.paimon.table.source.Split.class);
when(split1.rowCount()).thenReturn(100L);
when(split1.deletionFiles()).thenReturn(Optional.empty());
paimonSplits.add(split1);
// Second split: with deletion file and cardinality
org.apache.paimon.table.source.Split split2 =
mock(org.apache.paimon.table.source.Split.class);
when(split2.rowCount()).thenReturn(100L);
DeletionFile deletionFile2 = mock(DeletionFile.class);
when(deletionFile2.cardinality()).thenReturn(20L);
when(split2.deletionFiles()).thenReturn(Optional.of(Arrays.asList(deletionFile2)));
paimonSplits.add(split2);
// Third split: with null deletion file
org.apache.paimon.table.source.Split split3 =
mock(org.apache.paimon.table.source.Split.class);
when(split3.rowCount()).thenReturn(100L);
when(split3.deletionFiles()).thenReturn(Optional.of(Arrays.asList((DeletionFile)
null)));
paimonSplits.add(split3);
List<Split> splits = createPaimonSplits(paimonSplits);
// Total count should be 300 - 20 = 280
scanNode.assignCountToSplits(splits, 280L);
// Verify results (280 / 3 ≈ 93.33 per split)
assertEquals(93L, ((PaimonSplit)
splits.get(0)).getRowCount().get().longValue());
assertEquals(93L, ((PaimonSplit)
splits.get(1)).getRowCount().get().longValue());
assertEquals(94L, ((PaimonSplit)
splits.get(2)).getRowCount().get().longValue());
}
private List<Split>
createPaimonSplits(List<org.apache.paimon.table.source.Split> paimonSplits) {
List<Split> splits = new ArrayList<>();
for (org.apache.paimon.table.source.Split paimonSplit :
paimonSplits) {
splits.add(new PaimonSplit(paimonSplit));
}
return splits;
}
}
```
--
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: commits-unsubscr...@doris.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For additional commands, e-mail: commits-h...@doris.apache.org
