[GitHub] [iceberg] flyrain commented on a diff in pull request #6012: Spark 3.3: Add a procedure to generate table changes

[GitBox <gi...@apache.org>]

flyrain commented on code in PR #6012:
URL: https://github.com/apache/iceberg/pull/6012#discussion_r1067563712


##########
spark/v3.3/spark/src/main/java/org/apache/iceberg/spark/procedures/GenerateChangesProcedure.java:
##########
@@ -0,0 +1,267 @@
+/*
+ * 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.iceberg.spark.procedures;
+
+import java.sql.Timestamp;
+import java.util.Arrays;
+import java.util.List;
+import java.util.Map;
+import java.util.stream.Collectors;
+import org.apache.iceberg.MetadataColumns;
+import org.apache.iceberg.relocated.com.google.common.collect.Maps;
+import org.apache.iceberg.spark.ChangelogIterator;
+import org.apache.iceberg.spark.SparkReadOptions;
+import org.apache.iceberg.spark.source.SparkChangelogTable;
+import org.apache.spark.api.java.function.MapPartitionsFunction;
+import org.apache.spark.sql.Column;
+import org.apache.spark.sql.DataFrameReader;
+import org.apache.spark.sql.Dataset;
+import org.apache.spark.sql.Row;
+import org.apache.spark.sql.catalyst.InternalRow;
+import org.apache.spark.sql.catalyst.encoders.RowEncoder;
+import org.apache.spark.sql.connector.catalog.TableCatalog;
+import org.apache.spark.sql.connector.iceberg.catalog.ProcedureParameter;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.Metadata;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.types.StructType;
+import org.apache.spark.unsafe.types.UTF8String;
+import org.jetbrains.annotations.NotNull;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+import scala.runtime.BoxedUnit;
+
+/**
+ * A procedure that creates a view for changed rows.
+ *
+ * <p>The procedure computes update-rows and removes the carry-over rows by default. You can disable
+ * them through parameters to get better performance.
+ *
+ * <p>Carry-over rows are the result of a removal and insertion of the same row within an operation
+ * because of the copy-on-write mechanism. For example, given a file which contains row1 (id=1,
+ * data='a') and row2 (id=2, data='b'). A copy-on-write delete of row2 would require erasing this
+ * file and preserving row1 in a new file. The change-log table would report this as (id=1,
+ * data='a', op='DELETE') and (id=1, data='a', op='INSERT'), despite it not being an actual change
+ * to the table. The iterator finds the carry-over rows and removes them from the result.
+ *
+ * <p>An update-row is converted from a pair of delete row and insert row. Identifier columns are
+ * needed for identifying whether they refer to the same row. You can either set Identifier Field
+ * IDs as the table properties or input them as the procedure parameters. Here is an example of
+ * update-row with an identifier column(id). A pair of delete row and insert row with the same id:
+ *
+ * <ul>
+ *   <li>(id=1, data='a', op='DELETE')
+ *   <li>(id=1, data='b', op='INSERT')
+ * </ul>
+ *
+ * <p>will be marked as update-rows:
+ *
+ * <ul>
+ *   <li>(id=1, data='a', op='UPDATE_BEFORE')
+ *   <li>(id=1, data='b', op='UPDATE_AFTER')
+ * </ul>
+ */
+public class GenerateChangesProcedure extends BaseProcedure {
+  private static final Logger LOG = LoggerFactory.getLogger(GenerateChangesProcedure.class);
+
+  private static final ProcedureParameter[] PARAMETERS =
+      new ProcedureParameter[] {
+        ProcedureParameter.required("table", DataTypes.StringType),
+        ProcedureParameter.optional("table_change_view", DataTypes.StringType),
+        ProcedureParameter.optional("options", STRING_MAP),
+        ProcedureParameter.optional("compute_updates", DataTypes.BooleanType),
+        ProcedureParameter.optional("remove_carryovers", DataTypes.BooleanType),
+        ProcedureParameter.optional("identifier_columns", DataTypes.StringType),
+      };
+
+  private static final StructType OUTPUT_TYPE =
+      new StructType(
+          new StructField[] {
+            new StructField("view_name", DataTypes.StringType, false, Metadata.empty())
+          });
+
+  public static SparkProcedures.ProcedureBuilder builder() {
+    return new BaseProcedure.Builder<GenerateChangesProcedure>() {
+      @Override
+      protected GenerateChangesProcedure doBuild() {
+        return new GenerateChangesProcedure(tableCatalog());
+      }
+    };
+  }
+
+  private GenerateChangesProcedure(TableCatalog tableCatalog) {
+    super(tableCatalog);
+  }
+
+  @Override
+  public ProcedureParameter[] parameters() {
+    return PARAMETERS;
+  }
+
+  @Override
+  public StructType outputType() {
+    return OUTPUT_TYPE;
+  }
+
+  @Override
+  public InternalRow[] call(InternalRow args) {
+    String tableName = args.getString(0);
+
+    // Read data from the table.changes
+    Dataset<Row> df = changelogRecords(tableName, args);
+
+    // compute updated rows and remove carry-over rows by default
+    boolean computeUpdatedRow = args.isNullAt(3) ? true : args.getBoolean(3);
+    boolean removeCarryoverRow = args.isNullAt(4) ? true : args.getBoolean(4);
+
+    if (computeUpdatedRow) {
+      if (hasIdentifierColumns(args)) {
+        String[] identifierColumns = args.getString(5).split(",");
+
+        if (identifierColumns.length > 0) {
+          Column[] repartitionColumns = getRepartitionExpr(df, identifierColumns);
+          df = transform(df, repartitionColumns);
+        }
+      } else {
+        LOG.warn("Cannot compute the update-rows because identifier columns are not set");
+        if (removeCarryoverRow) {
+          df = removeCarryoverRows(df);
+        }
+      }
+    } else if (removeCarryoverRow) {
+      df = removeCarryoverRows(df);
+    }
+
+    String viewName = viewName(args, tableName);
+
+    // Create a view for users to query
+    df.createOrReplaceTempView(viewName);
+
+    return toOutputRows(viewName);
+  }
+
+  private Dataset<Row> removeCarryoverRows(Dataset<Row> df) {
+    Column[] repartitionColumns =
+        Arrays.stream(df.columns())
+            .filter(c -> !c.equals(MetadataColumns.CHANGE_TYPE.name()))
+            .map(df::col)
+            .toArray(Column[]::new);
+    return transform(df, repartitionColumns);

Review Comment:
   Reused the same changelog iterator for removing carry-over rows only. I think we can optimize it here, for example, using the windows function. WDYT?



-- 
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: issues-unsubscribe@iceberg.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@iceberg.apache.org
For additional commands, e-mail: issues-help@iceberg.apache.org