[FLINK-35293][hive] Hive source supports dynamic parallelism inference

docs/content.zh/docs/connectors/table/hive/hive_read_write.md

@@ -136,16 +136,21 @@ Flink 允许你灵活的配置并发推断策略。你可以在 `TableConfig` 
   </thead>
   <tbody>
     <tr>
-        <td><h5>table.exec.hive.infer-source-parallelism</h5></td>
-        <td style="word-wrap: break-word;">true</td>
-        <td>Boolean</td>
-        <td>如果是 true，会根据 split 的数量推断 source 的并发度。如果是 false，source 的并发度由配置决定。</td>
+        <td><h5>table.exec.hive.infer-source-parallelism.mode</h5></td>
+        <td style="word-wrap: break-word;">dynamic</td>
+        <td>InferMode</td>
+        <td>选择Hive Source并行度推断模式的选项，根据splits数推断并行度。
+            'static' 代表静态推断，它会在作业创建阶段推断Source并行度。
+            'dynamic' 代表动态推断，它会在作业执行阶段利用运行时信息更准确地推断Source并行度。
+            'none' 代表禁用并行度推断。
+            注意，它仍然受到已弃用选项 'table.exec.hive.infer-source-parallelism' 的影响，需要其值为 true 才能启用并行度推断。
+        </td>
     </tr>
     <tr>
         <td><h5>table.exec.hive.infer-source-parallelism.max</h5></td>
         <td style="word-wrap: break-word;">1000</td>
         <td>Integer</td>
-        <td>设置 source operator 推断的最大并发度。</td>
+        <td>设置 source operator 推断的最大并发度。请注意，默认值仅在静态并行性推断模式下有效。</td>
     </tr>
   </tbody>
 </table>

docs/content/docs/connectors/table/hive/hive_read_write.md

@@ -152,16 +152,23 @@ following parameters in `TableConfig` (note that these parameters affect all sou
   </thead>
   <tbody>
     <tr>
-        <td><h5>table.exec.hive.infer-source-parallelism</h5></td>
-        <td style="word-wrap: break-word;">true</td>
-        <td>Boolean</td>
-        <td>If is true, source parallelism is inferred according to splits number. If is false, parallelism of source are set by config.</td>
+        <td><h5>table.exec.hive.infer-source-parallelism.mode</h5></td>
+        <td style="word-wrap: break-word;">dynamic</td>
+        <td>InferMode</td>
+        <td>An option for selecting the hive source parallelism inference mode to infer parallelism according to splits number.
+            'static' represents static inference, which will infer source parallelism at job creation stage.
+            'dynamic' represents dynamic inference, which will infer parallelism at job execution stage and could more accurately infer the source parallelism.
+            'none' represents disabling parallelism inference.
+            Note that it is still affected by the deprecated option 'table.exec.hive.infer-source-parallelism', requiring its value to be true for enabling parallelism inference.
+        </td>
     </tr>
     <tr>
         <td><h5>table.exec.hive.infer-source-parallelism.max</h5></td>
         <td style="word-wrap: break-word;">1000</td>
         <td>Integer</td>
-        <td>Sets max infer parallelism for source operator.</td>
+        <td>Sets max infer parallelism for source operator.
+            Note that the default value is effective only in the static parallelism inference mode.
+        </td>
     </tr>
   </tbody>
 </table>

flink-connectors/flink-connector-files/src/main/java/org/apache/flink/connector/file/src/AbstractFileSource.java

@@ -100,7 +100,7 @@ protected AbstractFileSource(
     //  Getters
     // ------------------------------------------------------------------------
 
-    FileEnumerator.Provider getEnumeratorFactory() {
+    protected FileEnumerator.Provider getEnumeratorFactory() {
         return enumeratorFactory;
     }
 

flink-connectors/flink-connector-files/src/main/java/org/apache/flink/connector/file/table/LimitableBulkFormat.java

@@ -84,6 +84,10 @@ private boolean reachLimit() {
         return globalNumberRead != null && globalNumberRead.get() >= limit;
     }
 
+    public long getLimit() {
+        return this.limit;
+    }
+
     @Override
     public boolean isSplittable() {
         return format.isSplittable();

flink-connectors/flink-connector-hive/src/main/java/org/apache/flink/connectors/hive/HiveDynamicParallelismInferenceFactory.java

@@ -0,0 +1,67 @@
+/*
+ * 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.flink.connectors.hive;
+
+import org.apache.flink.api.common.ExecutionConfig;
+import org.apache.flink.table.catalog.ObjectPath;
+
+import org.apache.hadoop.mapred.JobConf;
+
+/**
+ * The factory class for {@link HiveParallelismInference} to support Hive source dynamic parallelism
+ * inference.
+ */
+class HiveDynamicParallelismInferenceFactory implements HiveParallelismInference.Provider {
+
+    private final ObjectPath tablePath;
+    private final JobConf jobConf;
+    private final int globalMaxParallelism;
+
+    HiveDynamicParallelismInferenceFactory(
+            ObjectPath tablePath, JobConf jobConf, int globalMaxParallelism) {
+        this.tablePath = tablePath;
+        this.jobConf = jobConf;
+        this.globalMaxParallelism = globalMaxParallelism;
+    }
+
+    @Override
+    public HiveParallelismInference create() {
+        boolean inferEnabled =
+                jobConf.getBoolean(
+                        HiveOptions.TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM.key(),
+                        HiveOptions.TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM.defaultValue());
+        HiveOptions.InferMode inferMode =
+                jobConf.getEnum(
+                        HiveOptions.TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM_MODE.key(),
+                        HiveOptions.TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM_MODE.defaultValue());
+        // This logic should be fixed if config option `table.exec.hive.infer-source-parallelism`
+        // is removed.
+        boolean infer = inferEnabled && inferMode == HiveOptions.InferMode.DYNAMIC;
+        int inferMaxParallelism =
+                Math.min(
+                        (int)
+                                jobConf.getLong(
+                                        HiveOptions.TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM_MAX
+                                                .key(),
+                                        globalMaxParallelism),
+                        globalMaxParallelism);
+        int parallelism = ExecutionConfig.PARALLELISM_DEFAULT;
+        return new HiveParallelismInference(tablePath, infer, inferMaxParallelism, parallelism);
+    }
+}

flink-connectors/flink-connector-hive/src/main/java/org/apache/flink/connectors/hive/HiveOptions.java

@@ -18,6 +18,7 @@
 
 package org.apache.flink.connectors.hive;
 
+import org.apache.flink.annotation.PublicEvolving;
 import org.apache.flink.configuration.ConfigOption;
 import org.apache.flink.configuration.DescribedEnum;
 import org.apache.flink.configuration.MemorySize;
@@ -50,6 +51,8 @@ public class HiveOptions {
                                     "If it is true, flink will read the files of partitioned hive table from subdirectories under the partition directory to be read.\n"
                                             + "If it is false, an exception that 'not a file: xxx' will be thrown when the partition directory contains any sub-directory.");
 
+    /** @deprecated Use {@link #TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM_MODE} instead. */
+    @Deprecated
     public static final ConfigOption<Boolean> TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM =
             key("table.exec.hive.infer-source-parallelism")
                     .booleanType()
@@ -58,11 +61,33 @@ public class HiveOptions {
                             "If is false, parallelism of source are set by config.\n"
                                     + "If is true, source parallelism is inferred according to splits number.\n");
 
+    @PublicEvolving
+    public static final ConfigOption<InferMode> TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM_MODE =
+            key("table.exec.hive.infer-source-parallelism.mode")
+                    .enumType(InferMode.class)
+                    .defaultValue(InferMode.DYNAMIC)
+                    .withDescription(
+                            Description.builder()
+                                    .text(
+                                            "An option for selecting the hive source parallelism inference mode to infer parallelism according to splits number.")
+                                    .list(
+                                            text(
+                                                    "'static' represents static inference, which will infer source parallelism at job creation stage."),
+                                            text(
+                                                    "'dynamic' represents dynamic inference, which will infer parallelism at job execution stage and could more accurately infer the source parallelism."),
+                                            text(
+                                                    "'none' represents disabling parallelism inference."),
+                                            text(
+                                                    "Note that it is still affected by the deprecated option 'table.exec.hive.infer-source-parallelism', requiring its value to be true for enabling parallelism inference."))
+                                    .build());
+
     public static final ConfigOption<Integer> TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM_MAX =
             key("table.exec.hive.infer-source-parallelism.max")
                     .intType()
                     .defaultValue(1000)
-                    .withDescription("Sets max infer parallelism for source operator.");
+                    .withDescription(
+                            "Sets max infer parallelism for source operator. "
+                                    + "Note that the default value is effective only in the static parallelism inference mode.");
 
     public static final ConfigOption<Boolean> TABLE_EXEC_HIVE_FALLBACK_MAPRED_WRITER =
             key("table.exec.hive.fallback-mapred-writer")
@@ -281,4 +306,30 @@ public InlineElement getDescription() {
             return description;
         }
     }
+
+    /** Infer mode used for {@link #TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM_MODE}. */
+    public enum InferMode implements DescribedEnum {
+        STATIC("static", text("Static parallelism inference mode.")),
+        DYNAMIC("dynamic", text("Dynamic parallelism inference mode.")),
+        NONE("none", text("Disable parallelism inference."));
+
+        private final String value;
+
+        private final InlineElement description;
+
+        InferMode(String value, InlineElement description) {
+            this.value = value;
+            this.description = description;
+        }
+
+        @Override
+        public String toString() {
+            return value;
+        }
+
+        @Override
+        public InlineElement getDescription() {
+            return description;
+        }
+    }
 }

flink-connectors/flink-connector-hive/src/main/java/org/apache/flink/connectors/hive/HiveParallelismInference.java

@@ -18,10 +18,7 @@
 
 package org.apache.flink.connectors.hive;
 
-import org.apache.flink.configuration.ReadableConfig;
-import org.apache.flink.table.api.config.ExecutionConfigOptions;
 import org.apache.flink.table.catalog.ObjectPath;
-import org.apache.flink.util.Preconditions;
 import org.apache.flink.util.function.SupplierWithException;
 
 import org.slf4j.Logger;
@@ -40,18 +37,12 @@ class HiveParallelismInference {
 
     private int parallelism;
 
-    HiveParallelismInference(ObjectPath tablePath, ReadableConfig flinkConf) {
+    HiveParallelismInference(
+            ObjectPath tablePath, boolean infer, int inferMaxParallelism, int parallelism) {
         this.tablePath = tablePath;
-        this.infer = flinkConf.get(HiveOptions.TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM);
-        this.inferMaxParallelism =
-                flinkConf.get(HiveOptions.TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM_MAX);
-        Preconditions.checkArgument(
-                inferMaxParallelism >= 1,
-                HiveOptions.TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM_MAX.key()
-                        + " cannot be less than 1");
-
-        this.parallelism =
-                flinkConf.get(ExecutionConfigOptions.TABLE_EXEC_RESOURCE_DEFAULT_PARALLELISM);
+        this.infer = infer;
+        this.inferMaxParallelism = inferMaxParallelism;
+        this.parallelism = parallelism;
     }
 
     /**
@@ -73,7 +64,8 @@ int limit(Long limit) {
 
     /**
      * Infer parallelism by number of files and number of splits. If {@link
-     * HiveOptions#TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM} is not set this method does nothing.
+     * HiveOptions#TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM} is false or {@link
+     * HiveOptions#TABLE_EXEC_HIVE_INFER_SOURCE_PARALLELISM_MODE} is none, this method does nothing.
      */
     HiveParallelismInference infer(
             SupplierWithException<Integer, IOException> numFiles,
@@ -113,4 +105,15 @@ private int logRunningTime(
                 result);
         return result;
     }
+
+    /** Factory for the {@code HiveParallelismInference}. */
+    interface Provider {
+
+        /**
+         * Creates a new {@code HiveParallelismInference}.
+         *
+         * @return a new {@code HiveParallelismInference} with designated factors.
+         */
+        HiveParallelismInference create();
+    }
 }

flink-connectors/flink-connector-hive/src/main/java/org/apache/flink/connectors/hive/HiveSource.java

@@ -20,6 +20,8 @@
 
 import org.apache.flink.annotation.PublicEvolving;
 import org.apache.flink.api.connector.source.Boundedness;
+import org.apache.flink.api.connector.source.DynamicFilteringInfo;
+import org.apache.flink.api.connector.source.DynamicParallelismInference;
 import org.apache.flink.api.connector.source.SplitEnumerator;
 import org.apache.flink.api.connector.source.SplitEnumeratorContext;
 import org.apache.flink.connector.file.src.AbstractFileSource;
@@ -29,10 +31,12 @@
 import org.apache.flink.connector.file.src.enumerate.FileEnumerator;
 import org.apache.flink.connector.file.src.reader.BulkFormat;
 import org.apache.flink.connector.file.table.ContinuousPartitionFetcher;
+import org.apache.flink.connector.file.table.LimitableBulkFormat;
 import org.apache.flink.connectors.hive.read.HiveSourceSplit;
 import org.apache.flink.core.fs.Path;
 import org.apache.flink.core.io.SimpleVersionedSerializer;
 import org.apache.flink.table.catalog.ObjectPath;
+import org.apache.flink.table.connector.source.DynamicFilteringEvent;
 import org.apache.flink.table.data.RowData;
 import org.apache.flink.util.Preconditions;
 
@@ -55,7 +59,8 @@
  * @param <T> the type of record returned by this source
  */
 @PublicEvolving
-public class HiveSource<T> extends AbstractFileSource<T, HiveSourceSplit> {
+public class HiveSource<T> extends AbstractFileSource<T, HiveSourceSplit>
+        implements DynamicParallelismInference {
 
     private static final long serialVersionUID = 1L;
 
@@ -68,6 +73,7 @@ public class HiveSource<T> extends AbstractFileSource<T, HiveSourceSplit> {
     private final ContinuousPartitionFetcher<Partition, ?> fetcher;
     private final HiveTableSource.HiveContinuousPartitionFetcherContext<?> fetcherContext;
     private final ObjectPath tablePath;
+    private Long limit = null;
 
     HiveSource(
             Path[] inputPaths,
@@ -97,6 +103,9 @@ public class HiveSource<T> extends AbstractFileSource<T, HiveSourceSplit> {
         this.partitionBytes = partitionBytes;
         this.fetcher = fetcher;
         this.fetcherContext = fetcherContext;
+        if (readerFormat instanceof LimitableBulkFormat) {
+            limit = ((LimitableBulkFormat<?, ?>) readerFormat).getLimit();
+        }
     }
 
     @Override
@@ -186,4 +195,48 @@ private boolean continuousPartitionedEnumerator() {
                         jobConfWrapper),
                 getAssignerFactory());
     }
+
+    @Override
+    public int inferParallelism(Context dynamicParallelismContext) {
+        FileEnumerator fileEnumerator;
+        List<HiveTablePartition> partitions;
+        if (dynamicFilterPartitionKeys != null) {
+            fileEnumerator =
+                    new HiveSourceDynamicFileEnumerator.Provider(
+                                    tablePath.getFullName(),
+                                    dynamicFilterPartitionKeys,
+                                    partitionBytes,
+                                    hiveVersion,
+                                    jobConfWrapper)
+                            .create();
+            if (dynamicParallelismContext.getDynamicFilteringInfo().isPresent()) {
+                DynamicFilteringInfo dynamicFilteringInfo =
+                        dynamicParallelismContext.getDynamicFilteringInfo().get();
+                if (dynamicFilteringInfo instanceof DynamicFilteringEvent) {
+                    ((HiveSourceDynamicFileEnumerator) fileEnumerator)
+                            .setDynamicFilteringData(
+                                    ((DynamicFilteringEvent) dynamicFilteringInfo).getData());
+                }
+            }
+            partitions = ((HiveSourceDynamicFileEnumerator) fileEnumerator).getFinalPartitions();
+        } else {
+            fileEnumerator = getEnumeratorFactory().create();
+            partitions = ((HiveSourceFileEnumerator) fileEnumerator).getPartitions();
+        }
+
+        return new HiveDynamicParallelismInferenceFactory(
+                        tablePath,
+                        jobConfWrapper.conf(),
+                        dynamicParallelismContext.getParallelismInferenceUpperBound())
+                .create()
+                .infer(
+                        () ->
+                                HiveSourceFileEnumerator.getNumFiles(
+                                        partitions, jobConfWrapper.conf()),
+                        () ->
+                                HiveSourceFileEnumerator.createInputSplits(
+                                                0, partitions, jobConfWrapper.conf(), true)
+                                        .size())
+                .limit(limit);
+    }
 }