nsivabalan commented on code in PR #19205:
URL: https://github.com/apache/hudi/pull/19205#discussion_r3735254270


##########
hudi-spark-datasource/hudi-spark/src/test/java/org/apache/hudi/functional/TestMetaFieldsMode.java:
##########
@@ -0,0 +1,469 @@
+/*
+ * 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.
+ */
+
+package org.apache.hudi.functional;
+
+import org.apache.hudi.DataSourceWriteOptions;
+import org.apache.hudi.SparkAdapterSupport$;
+import org.apache.hudi.common.config.HoodieMetadataConfig;
+import org.apache.hudi.common.model.HoodieRecord;
+import org.apache.hudi.common.model.MetaFieldsMode;
+import org.apache.hudi.common.table.HoodieTableConfig;
+import org.apache.hudi.common.table.HoodieTableMetaClient;
+import org.apache.hudi.testutils.SparkClientFunctionalTestHarness;
+
+import org.apache.spark.sql.Dataset;
+import org.apache.spark.sql.Row;
+import org.apache.spark.sql.RowFactory;
+import org.apache.spark.sql.SaveMode;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.types.StructType;
+import org.junit.jupiter.api.Test;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertNotNull;
+import static org.junit.jupiter.api.Assertions.assertNull;
+import static org.junit.jupiter.api.Assertions.assertThrows;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+/**
+ * Spark-datasource end-to-end tests for the {@code hoodie.meta.fields.mode} 
property on CoW tables.
+ * Every {@link MetaFieldsMode} value is exercised via a write / re-read round 
trip; on-disk column
+ * population is verified by reading the parquet files back and inspecting the 
meta-column values.
+ */
+class TestMetaFieldsMode extends SparkClientFunctionalTestHarness {

Review Comment:
   Done — and you were right about the cost of it being missing. Added five 
incremental cases in `TestMetaFieldsModeE2E`: `COMMIT_TIME_ONLY` returning 
exactly the second commit's rows, a from-creation read, 
`COMMIT_TIME_AND_FILE_NAME`, and rejection for the two modes without a commit 
time.
   
   Writing them surfaced two live bugs, both in the read path you flagged as 
load-bearing:
   
   1. `HoodieCopyOnWriteIncrementalHadoopFsRelationFactory` builds a 
`MergeOnReadIncrementalRelation` to back its file index, so the MoR guard was 
rejecting **every** selective-mode CoW table — the whole feature. Now scoped to 
`MERGE_ON_READ`.
   2. The CoW commit-time guard lived in `IncrementalRelationV1/V2`, which back 
only the streaming source. A datasource read of a `FILE_NAME_ONLY`/`NONE` table 
therefore returned zero rows *silently* — precisely the failure the guard 
exists to prevent. Added to the factory the datasource actually uses.
   
   So the feature's headline promise did not work before this, and nothing 
would have told us.
   
   One caveat on the assertions: they read whole rows rather than `count()` or 
a projection. `count()` returns 0 on a CoW incremental query for **every** mode 
including the default `ALL`, and it reproduces on unmodified `master` at table 
version 9 — unrelated pre-existing bug, details in the PR comment, filing 
separately.
   



##########
hudi-spark-datasource/hudi-spark/src/test/java/org/apache/hudi/functional/TestMetaFieldsModeE2E.java:
##########
@@ -0,0 +1,469 @@
+/*
+ * 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.
+ */
+
+package org.apache.hudi.functional;
+
+import org.apache.hudi.DataSourceWriteOptions;
+import org.apache.hudi.SparkAdapterSupport$;
+import org.apache.hudi.common.config.HoodieMetadataConfig;
+import org.apache.hudi.common.model.HoodieRecord;
+import org.apache.hudi.common.model.MetaFieldsMode;
+import org.apache.hudi.common.table.HoodieTableConfig;
+import org.apache.hudi.common.table.HoodieTableMetaClient;
+import org.apache.hudi.testutils.SparkClientFunctionalTestHarness;
+
+import org.apache.spark.sql.Dataset;
+import org.apache.spark.sql.Row;
+import org.apache.spark.sql.RowFactory;
+import org.apache.spark.sql.SaveMode;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.types.StructType;
+import org.junit.jupiter.api.Test;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertNotNull;
+import static org.junit.jupiter.api.Assertions.assertNull;
+import static org.junit.jupiter.api.Assertions.assertThrows;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+/**
+ * Spark-datasource end-to-end tests for the {@code hoodie.meta.fields.mode} 
property on CoW tables.
+ * Every {@link MetaFieldsMode} value is exercised via a write / re-read round 
trip; on-disk column
+ * population is verified by reading the parquet files back and inspecting the 
meta-column values.
+ */
+class TestMetaFieldsMode extends SparkClientFunctionalTestHarness {
+
+  private static StructType simpleSchema() {
+    return DataTypes.createStructType(new StructField[]{
+        DataTypes.createStructField("column1", DataTypes.StringType, true),
+        DataTypes.createStructField("column2", DataTypes.StringType, true),
+        DataTypes.createStructField("column3", DataTypes.StringType, true)
+    }).asNullable();
+  }
+
+  private Map<String, String> baseOptions() {
+    Map<String, String> opts = new HashMap<>();
+    opts.put(DataSourceWriteOptions.RECORDKEY_FIELD().key(), "column1");
+    opts.put(DataSourceWriteOptions.PARTITIONPATH_FIELD().key(), "column2");
+    opts.put(DataSourceWriteOptions.ORDERING_FIELDS().key(), "column3");
+    opts.put(HoodieTableConfig.NAME.key(), "test_meta_fields_mode");
+    opts.put(DataSourceWriteOptions.TABLE_TYPE().key(), "COPY_ON_WRITE");
+    opts.put(HoodieMetadataConfig.ENABLE.key(), "false");

Review Comment:
   Added `selectiveModeWritesSucceedWithTheMetadataTableEnabled` — a 
`COMMIT_TIME_ONLY` write with MDT on, asserting the write succeeds, the data 
table's columns are as expected, and the MDT's own table config resolves to 
`NONE`.
   
   Your read of the mechanism was right: 
`HoodieMetadataWriteUtils#createMetadataWriteConfig` builds a *fresh* 
`HoodieWriteConfig.newBuilder()` and sets `populate.meta.fields=false` itself 
(`DEFAULT_METADATA_POPULATE_META_FIELDS`), copying only named options — never 
`withProps(dataWriteConfig.getProps())`. So the mode cannot leak today, and the 
test now pins that.
   
   This also mattered for the write-client change in this push: the new gate 
rejects any stated mismatch, so if the MDT *had* inherited the data table's 
mode it would have failed hard. It does not, and 
`TestSparkRDDMetadataWriteClient` passes.
   



##########
hudi-spark-datasource/hudi-spark/src/test/java/org/apache/hudi/functional/TestMetaFieldsModeE2E.java:
##########
@@ -0,0 +1,469 @@
+/*
+ * 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.
+ */
+
+package org.apache.hudi.functional;
+
+import org.apache.hudi.DataSourceWriteOptions;
+import org.apache.hudi.SparkAdapterSupport$;
+import org.apache.hudi.common.config.HoodieMetadataConfig;
+import org.apache.hudi.common.model.HoodieRecord;
+import org.apache.hudi.common.model.MetaFieldsMode;
+import org.apache.hudi.common.table.HoodieTableConfig;
+import org.apache.hudi.common.table.HoodieTableMetaClient;
+import org.apache.hudi.testutils.SparkClientFunctionalTestHarness;
+
+import org.apache.spark.sql.Dataset;
+import org.apache.spark.sql.Row;
+import org.apache.spark.sql.RowFactory;
+import org.apache.spark.sql.SaveMode;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.types.StructType;
+import org.junit.jupiter.api.Test;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertNotNull;
+import static org.junit.jupiter.api.Assertions.assertNull;
+import static org.junit.jupiter.api.Assertions.assertThrows;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+/**
+ * Spark-datasource end-to-end tests for the {@code hoodie.meta.fields.mode} 
property on CoW tables.
+ * Every {@link MetaFieldsMode} value is exercised via a write / re-read round 
trip; on-disk column
+ * population is verified by reading the parquet files back and inspecting the 
meta-column values.
+ */
+class TestMetaFieldsMode extends SparkClientFunctionalTestHarness {
+
+  private static StructType simpleSchema() {
+    return DataTypes.createStructType(new StructField[]{
+        DataTypes.createStructField("column1", DataTypes.StringType, true),
+        DataTypes.createStructField("column2", DataTypes.StringType, true),
+        DataTypes.createStructField("column3", DataTypes.StringType, true)
+    }).asNullable();
+  }
+
+  private Map<String, String> baseOptions() {
+    Map<String, String> opts = new HashMap<>();
+    opts.put(DataSourceWriteOptions.RECORDKEY_FIELD().key(), "column1");
+    opts.put(DataSourceWriteOptions.PARTITIONPATH_FIELD().key(), "column2");
+    opts.put(DataSourceWriteOptions.ORDERING_FIELDS().key(), "column3");
+    opts.put(HoodieTableConfig.NAME.key(), "test_meta_fields_mode");
+    opts.put(DataSourceWriteOptions.TABLE_TYPE().key(), "COPY_ON_WRITE");
+    opts.put(HoodieMetadataConfig.ENABLE.key(), "false");
+    return opts;
+  }
+
+  private void writeRows(List<Row> records, StructType schema, Map<String, 
String> options, String path, SaveMode mode) {
+    spark().createDataset(records,
+            
SparkAdapterSupport$.MODULE$.sparkAdapter().getCatalystExpressionUtils().getEncoder(schema))
+        .write()
+        .format("hudi")
+        .options(options)
+        .mode(mode)
+        .save(path);
+  }
+
+  private HoodieTableConfig writeSampleAndGetTableConfig(Map<String, String> 
options, String path) {
+    writeRows(Arrays.asList(
+            RowFactory.create("k1", "p1", "v1"),
+            RowFactory.create("k2", "p1", "v2")),
+        simpleSchema(), options, path, SaveMode.Overwrite);

Review Comment:
   Added `appendWithoutRestatingTheModeKeepsTheTableSelective`: commit 1 
`Overwrite` with the mode, commit 2 `Append` restating nothing, then asserting 
the table is still `COMMIT_TIME_ONLY` and that rows from both commits carry a 
commit time.
   
   Your point about `SaveMode.Overwrite` masking the inheritance path was the 
important one, and it turned out to matter more than a test gap. The 
inheritance rule is now centralized in 
`BaseHoodieWriteClient#validateAgainstTableProperties`: a writer that states 
neither meta-field property inherits the table's mode, one that states either 
is compared and rejected on mismatch. That covers the datasource, the streamer, 
and direct write-client users in one place, rather than depending on 
`mergeParamsAndGetHoodieConfig`'s blanket copy — which, as you noted, does not 
run for `Overwrite` and is bypassed by `validateTableConfig` there too.
   



##########
hudi-spark-datasource/hudi-spark/src/test/java/org/apache/hudi/functional/TestMetaFieldsMode.java:
##########
@@ -0,0 +1,469 @@
+/*
+ * 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.
+ */
+
+package org.apache.hudi.functional;
+
+import org.apache.hudi.DataSourceWriteOptions;
+import org.apache.hudi.SparkAdapterSupport$;
+import org.apache.hudi.common.config.HoodieMetadataConfig;
+import org.apache.hudi.common.model.HoodieRecord;
+import org.apache.hudi.common.model.MetaFieldsMode;
+import org.apache.hudi.common.table.HoodieTableConfig;
+import org.apache.hudi.common.table.HoodieTableMetaClient;
+import org.apache.hudi.testutils.SparkClientFunctionalTestHarness;
+
+import org.apache.spark.sql.Dataset;
+import org.apache.spark.sql.Row;
+import org.apache.spark.sql.RowFactory;
+import org.apache.spark.sql.SaveMode;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.types.StructType;
+import org.junit.jupiter.api.Test;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertNotNull;
+import static org.junit.jupiter.api.Assertions.assertNull;
+import static org.junit.jupiter.api.Assertions.assertThrows;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+/**
+ * Spark-datasource end-to-end tests for the {@code hoodie.meta.fields.mode} 
property on CoW tables.
+ * Every {@link MetaFieldsMode} value is exercised via a write / re-read round 
trip; on-disk column
+ * population is verified by reading the parquet files back and inspecting the 
meta-column values.
+ */
+class TestMetaFieldsMode extends SparkClientFunctionalTestHarness {
+
+  private static StructType simpleSchema() {
+    return DataTypes.createStructType(new StructField[]{
+        DataTypes.createStructField("column1", DataTypes.StringType, true),
+        DataTypes.createStructField("column2", DataTypes.StringType, true),
+        DataTypes.createStructField("column3", DataTypes.StringType, true)
+    }).asNullable();
+  }
+
+  private Map<String, String> baseOptions() {
+    Map<String, String> opts = new HashMap<>();
+    opts.put(DataSourceWriteOptions.RECORDKEY_FIELD().key(), "column1");
+    opts.put(DataSourceWriteOptions.PARTITIONPATH_FIELD().key(), "column2");
+    opts.put(DataSourceWriteOptions.ORDERING_FIELDS().key(), "column3");
+    opts.put(HoodieTableConfig.NAME.key(), "test_meta_fields_mode");
+    opts.put(DataSourceWriteOptions.TABLE_TYPE().key(), "COPY_ON_WRITE");
+    opts.put(HoodieMetadataConfig.ENABLE.key(), "false");
+    return opts;
+  }
+
+  private void writeRows(List<Row> records, StructType schema, Map<String, 
String> options, String path, SaveMode mode) {
+    spark().createDataset(records,
+            
SparkAdapterSupport$.MODULE$.sparkAdapter().getCatalystExpressionUtils().getEncoder(schema))
+        .write()
+        .format("hudi")
+        .options(options)
+        .mode(mode)
+        .save(path);
+  }
+
+  private HoodieTableConfig writeSampleAndGetTableConfig(Map<String, String> 
options, String path) {
+    writeRows(Arrays.asList(
+            RowFactory.create("k1", "p1", "v1"),
+            RowFactory.create("k2", "p1", "v2")),
+        simpleSchema(), options, path, SaveMode.Overwrite);
+    HoodieTableMetaClient metaClient =
+        
HoodieTableMetaClient.builder().setBasePath(path).setConf(storageConf()).build();
+    return metaClient.getTableConfig();
+  }
+
+  /**
+   * End-to-end assertion of the on-disk meta columns after a write. Reads the 
parquet files back
+   * (bypassing Hudi's own read path so we see the raw column values) and 
asserts which meta
+   * columns are non-null.
+   */
+  private void assertMetaColumnPopulation(String path, MetaFieldsMode 
expectedMode) {
+    Dataset<Row> raw = spark().read().parquet(path + "/*/*.parquet");
+    Row first = raw.select(
+        HoodieRecord.COMMIT_TIME_METADATA_FIELD,
+        HoodieRecord.COMMIT_SEQNO_METADATA_FIELD,
+        HoodieRecord.RECORD_KEY_METADATA_FIELD,
+        HoodieRecord.PARTITION_PATH_METADATA_FIELD,
+        HoodieRecord.FILENAME_METADATA_FIELD).first();
+
+    if (expectedMode.isCommitTimePopulated()) {
+      assertNotNull(first.get(0), "expected _hoodie_commit_time to be 
populated for mode " + expectedMode);
+    } else {
+      assertNull(first.get(0), "expected _hoodie_commit_time to be null for 
mode " + expectedMode);
+    }
+    if (expectedMode.isFileNamePopulated()) {
+      assertNotNull(first.get(4), "expected _hoodie_file_name to be populated 
for mode " + expectedMode);
+    } else {
+      assertNull(first.get(4), "expected _hoodie_file_name to be null for mode 
" + expectedMode);
+    }
+    // Record key, partition path, and commit seq no are ALL-only.
+    if (expectedMode == MetaFieldsMode.ALL) {
+      assertNotNull(first.get(2), "record key must be populated in ALL mode");
+      assertNotNull(first.get(3), "partition path must be populated in ALL 
mode");
+      assertNotNull(first.get(1), "commit seq no must be populated in ALL 
mode");
+    } else {
+      assertNull(first.get(2), "record key must be null outside ALL mode, got: 
" + first.get(2));
+      assertNull(first.get(3), "partition path must be null outside ALL mode, 
got: " + first.get(3));
+      assertNull(first.get(1), "commit seq no must be null outside ALL mode, 
got: " + first.get(1));
+    }
+  }
+
+  @Test
+  void allModePersistsAndPopulatesAllColumns() {
+    Map<String, String> options = baseOptions();
+    // ALL is the default; no need to set the mode explicitly.
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.BULK_INSERT_OPERATION_OPT_VAL());
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+
+    assertTrue(tc.populateMetaFields());
+    assertEquals(MetaFieldsMode.ALL, tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), MetaFieldsMode.ALL);
+  }
+
+  @Test
+  void noneModePersistsAndLeavesAllColumnsNull() {
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "false");
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.BULK_INSERT_OPERATION_OPT_VAL());
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+
+    assertFalse(tc.populateMetaFields());
+    assertEquals(MetaFieldsMode.NONE, tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), MetaFieldsMode.NONE);
+  }
+
+  @Test
+  void commitTimeOnlyModePopulatesOnlyCommitTime() {
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "false");
+    options.put(HoodieTableConfig.META_FIELDS_MODE.key(), 
MetaFieldsMode.COMMIT_TIME_ONLY.name());
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.BULK_INSERT_OPERATION_OPT_VAL());
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+
+    assertEquals(MetaFieldsMode.COMMIT_TIME_ONLY.name(),
+        tc.getProps().getProperty(HoodieTableConfig.META_FIELDS_MODE.key()));
+    assertEquals(MetaFieldsMode.COMMIT_TIME_ONLY, tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), MetaFieldsMode.COMMIT_TIME_ONLY);
+  }
+
+  @Test
+  void fileNameOnlyModePopulatesOnlyFileName() {
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "false");
+    options.put(HoodieTableConfig.META_FIELDS_MODE.key(), 
MetaFieldsMode.FILE_NAME_ONLY.name());
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.BULK_INSERT_OPERATION_OPT_VAL());
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+
+    assertEquals(MetaFieldsMode.FILE_NAME_ONLY, tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), MetaFieldsMode.FILE_NAME_ONLY);
+  }
+
+  @Test
+  void commitTimeAndFileNameModePopulatesBoth() {
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "false");
+    options.put(HoodieTableConfig.META_FIELDS_MODE.key(), 
MetaFieldsMode.COMMIT_TIME_AND_FILE_NAME.name());
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.BULK_INSERT_OPERATION_OPT_VAL());
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+
+    assertEquals(MetaFieldsMode.COMMIT_TIME_AND_FILE_NAME, 
tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), 
MetaFieldsMode.COMMIT_TIME_AND_FILE_NAME);
+  }
+
+  @Test
+  void selectiveModeWinsOverLegacyPopulateTrue() {
+    // hoodie.meta.fields.mode is the source of truth: an explicit mode is 
honored regardless of
+    // the deprecated boolean, so this combination is no longer ambiguous and 
is not rejected.
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "true");
+    options.put(HoodieTableConfig.META_FIELDS_MODE.key(), 
MetaFieldsMode.COMMIT_TIME_ONLY.name());
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.BULK_INSERT_OPERATION_OPT_VAL());
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+
+    assertEquals(MetaFieldsMode.COMMIT_TIME_ONLY, tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), MetaFieldsMode.COMMIT_TIME_ONLY);
+    // ...and hoodie.properties must not contradict the mode. A pre-1.3.0 
reader ignores the mode
+    // property entirely, so leaving populate.meta.fields=true here would make 
it treat a
+    // selectively-written table as ALL.
+    assertFalse(tc.populateMetaFields(),
+        "legacy populate.meta.fields must be derived from the mode, not 
carried through verbatim");
+  }
+
+  @Test
+  void noneModePersistsLegacyBooleanAsFalse() {
+    // The unsafe case: an old incremental reader that sees 
populate.meta.fields=true on a NONE
+    // table would run against all-null commit times and silently return zero 
rows.
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "true");
+    options.put(HoodieTableConfig.META_FIELDS_MODE.key(), 
MetaFieldsMode.NONE.name());
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.BULK_INSERT_OPERATION_OPT_VAL());
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+
+    assertEquals(MetaFieldsMode.NONE, tc.getMetaFieldsMode());
+    assertFalse(tc.populateMetaFields(),
+        "NONE must persist populate.meta.fields=false so pre-1.3.0 readers do 
not treat it as ALL");
+  }
+
+  @Test
+  void allModePersistsLegacyBooleanAsTrue() {
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "false");
+    options.put(HoodieTableConfig.META_FIELDS_MODE.key(), 
MetaFieldsMode.ALL.name());
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.BULK_INSERT_OPERATION_OPT_VAL());
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+
+    assertEquals(MetaFieldsMode.ALL, tc.getMetaFieldsMode());
+    assertTrue(tc.populateMetaFields(),
+        "ALL must persist populate.meta.fields=true for pre-1.3.0 readers");
+  }
+
+  @Test
+  void unknownModeValueIsRejected() {
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "false");
+    options.put(HoodieTableConfig.META_FIELDS_MODE.key(), "SOMETHING_BOGUS");
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.BULK_INSERT_OPERATION_OPT_VAL());
+
+    Throwable thrown = assertThrows(Throwable.class, () ->
+        writeRows(Collections.singletonList(RowFactory.create("k1", "p1", 
"v1")),
+            simpleSchema(), options, basePath(), SaveMode.Overwrite));
+
+    String rootMessage = rootMessageOf(thrown);
+    assertTrue(rootMessage.contains("SOMETHING_BOGUS"),
+        "Expected error to name the rejected value, got: " + rootMessage);
+  }
+
+  // -------------------------------------------------------------------------
+  // Non-row-writer path coverage. Bulk insert with row.writer.enable=false 
forces the
+  // HoodieAvroParquetWriter path (via HoodieCreateHandle) instead of the 
internal-row writer path.
+  // Both paths must respect the mode identically.
+  // -------------------------------------------------------------------------
+
+  @Test
+  void nonRowWriterPathAllMode() {
+    Map<String, String> options = baseOptions();
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.INSERT_OPERATION_OPT_VAL());
+    options.put("hoodie.datasource.write.row.writer.enable", "false");
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+    assertEquals(MetaFieldsMode.ALL, tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), MetaFieldsMode.ALL);
+  }
+
+  @Test
+  void nonRowWriterPathNoneMode() {
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "false");
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.INSERT_OPERATION_OPT_VAL());
+    options.put("hoodie.datasource.write.row.writer.enable", "false");
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+    assertEquals(MetaFieldsMode.NONE, tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), MetaFieldsMode.NONE);
+  }
+
+  @Test
+  void nonRowWriterPathCommitTimeOnly() {
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "false");
+    options.put(HoodieTableConfig.META_FIELDS_MODE.key(), 
MetaFieldsMode.COMMIT_TIME_ONLY.name());
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.INSERT_OPERATION_OPT_VAL());
+    options.put("hoodie.datasource.write.row.writer.enable", "false");
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+    assertEquals(MetaFieldsMode.COMMIT_TIME_ONLY, tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), MetaFieldsMode.COMMIT_TIME_ONLY);
+  }
+
+  @Test
+  void nonRowWriterPathFileNameOnly() {
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "false");
+    options.put(HoodieTableConfig.META_FIELDS_MODE.key(), 
MetaFieldsMode.FILE_NAME_ONLY.name());
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.INSERT_OPERATION_OPT_VAL());
+    options.put("hoodie.datasource.write.row.writer.enable", "false");
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+    assertEquals(MetaFieldsMode.FILE_NAME_ONLY, tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), MetaFieldsMode.FILE_NAME_ONLY);
+  }
+
+  @Test
+  void nonRowWriterPathCommitTimeAndFileName() {
+    Map<String, String> options = baseOptions();
+    options.put(HoodieTableConfig.POPULATE_META_FIELDS.key(), "false");
+    options.put(HoodieTableConfig.META_FIELDS_MODE.key(), 
MetaFieldsMode.COMMIT_TIME_AND_FILE_NAME.name());
+    options.put(DataSourceWriteOptions.OPERATION().key(), 
DataSourceWriteOptions.INSERT_OPERATION_OPT_VAL());
+    options.put("hoodie.datasource.write.row.writer.enable", "false");
+
+    HoodieTableConfig tc = writeSampleAndGetTableConfig(options, basePath());
+    assertEquals(MetaFieldsMode.COMMIT_TIME_AND_FILE_NAME, 
tc.getMetaFieldsMode());
+    assertMetaColumnPopulation(basePath(), 
MetaFieldsMode.COMMIT_TIME_AND_FILE_NAME);
+  }
+
+  // -------------------------------------------------------------------------
+  // Clustering coverage. Inline clustering rewrites files through the 
create/merge handles which
+  // delegate to the same underlying HoodieAvroParquetWriter / 
HoodieRowCreateHandle we exercise
+  // in the write tests. Verifies clustered files preserve the mode's column 
population semantics.
+  // -------------------------------------------------------------------------
+
+  @Test
+  void clusteringPreservesCommitTimeOnlyMode() {

Review Comment:
   Confirmed and fixed. You were right on all three points: `assertNotNull` 
passes on the stale pre-clustering value, the glob picked up the replaced file, 
and nothing asserted clustering ran.
   
   Rewritten to assert `_hoodie_file_name` equals the containing file for every 
row:
   
   ```java
   Dataset<Row> rows = spark().read().format("hudi").load(path)
       .withColumn("__containing_file", functions.input_file_name());
   // per row: containingFile.endsWith("/" + fileName)
   ```
   
   plus `assertEquals(1, ...getCompletedReplaceTimeline().countInstants())` so 
the test fails if clustering did not run. Reading through Hudi rather than 
globbing parquet also fixes the replaced-file problem you noted — no cleaning 
has run at that point.
   
   `clusteringPreservesAllMode` and `clusteringPreservesNoneMode` are dropped: 
both route through `writeRow` / `writeRowNoMetaFields` and never enter the 
branch `358fbfdd717a` touched, exactly as you said.
   
   Also added `clusteringLeavesFileNameNullUnderCommitTimeOnly`, which covers 
the direction the original tests could not express at all.
   



##########
hudi-client/hudi-client-common/src/test/java/org/apache/hudi/table/upgrade/TestTenToNineDowngradeHandler.java:
##########
@@ -24,18 +24,25 @@
 import org.junit.jupiter.api.Test;
 
 import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
 import static org.junit.jupiter.api.Assertions.assertTrue;
 
 class TestTenToNineDowngradeHandler {
 
   @Test
-  void testDowngradeRemovesStorageLayoutOnly() {
+  void testDowngradeRemovesStorageLayoutAndMetaFieldsMode() {
     UpgradeDowngrade.TableConfigChangeSet changeSet =
         new TenToNineDowngradeHandler().downgrade(null, null, null, null);

Review Comment:
   Fixed in `69a0906419d8` — you were right that passing `null` for the helper 
short-circuited to `ALL` and left the whole selective branch dead. It now uses 
a `helperFor(...)` mock factory covering all five modes plus the no-helper case.
   
   Worth noting the branch it exercises has since changed shape: per the 
discussion on the sibling thread, a selective mode no longer degrades to `NONE` 
with a warning — the downgrade **throws**. So 
`downgradeRejectsSelectiveModesRatherThanDegradingThem` now asserts that for 
all three selective modes, and 
`downgradeWritesTheLegacyBooleanDerivedFromTheMode` is scoped to `ALL`/`NONE`.
   
   I checked the tests are not vacuous: removing the throw fails exactly 3 
tests, and reverting the boolean write-back fails 7 of 10.
   



##########
hudi-spark-datasource/hudi-spark/src/test/scala/org/apache/hudi/TestHoodieSparkSqlWriter.scala:
##########
@@ -107,7 +107,7 @@ class TestHoodieSparkSqlWriter extends 
HoodieSparkWriterTestBase {
     // fetch all records from parquet files generated from write to hudi
     val actualDf = sqlContext.read.parquet(fullPartitionPaths(0), 
fullPartitionPaths(1), fullPartitionPaths(2))
     if (!populateMetaFields) {
-      List(0, 1, 2, 3, 4).foreach(i => assertEquals(0, 
actualDf.select(HoodieRecord.HOODIE_META_COLUMNS.get(i)).filter(entry => 
!(entry.mkString(",").equals(""))).count()))
+      List(0, 1, 2, 3, 4).foreach(i => assertEquals(0, 
actualDf.select(HoodieRecord.HOODIE_META_COLUMNS.get(i)).filter(entry => 
!entry.isNullAt(0) && entry.getString(0).nonEmpty).count()))

Review Comment:
   Agreed, and fixed — you were right that the relaxation went further than the 
behavior change required, and that it left the two tests disagreeing on 
strictness.
   
   Extracted `assertNoMetaFieldsPopulated(df)` into `HoodieSparkWriterTestBase` 
next to `dropMetaFields`, as you suggested, rather than duplicating the edit:
   
   ```scala
   def assertNoMetaFieldsPopulated(df: Dataset[Row]): Unit = {
     (0 until HoodieRecord.HOODIE_META_COLUMNS.size()).foreach { i =>
       val column = HoodieRecord.HOODIE_META_COLUMNS.get(i)
       assertEquals(0, df.select(column).filter(entry => 
!entry.isNullAt(0)).count(), ...)
     }
   }
   ```
   
   Strictly `NULL`, so it now agrees with the `assertNull` assertions in the 
functional test for the same scenario. Applied in both this file and 
`TestHoodieSparkSqlWriterWithTestFormat`.
   



-- 
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]

Reply via email to