[GitHub] [spark] sadikovi commented on a diff in pull request #37228: [SPARK-37980][SQL] Extend METADATA column to support row indexes for Parquet files

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

sadikovi commented on code in PR #37228:
URL: https://github.com/apache/spark/pull/37228#discussion_r944003632


##########
sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/FileMetadataStructRowIndexSuite.scala:
##########
@@ -0,0 +1,233 @@
+/*
+ * 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.spark.sql.execution.datasources
+
+import org.apache.spark.sql.{AnalysisException, DataFrame, QueryTest}
+import org.apache.spark.sql.functions.{col, lit}
+import org.apache.spark.sql.internal.SQLConf
+import org.apache.spark.sql.test.SharedSparkSession
+import org.apache.spark.sql.types.{LongType, StructField, StructType}
+
+class FileMetadataStructRowIndexSuite extends QueryTest with SharedSparkSession {
+  import testImplicits._
+
+  val EXPECTED_ROW_ID_COL = "expected_row_idx"
+  val EXPECTED_EXTRA_COL = "expected_extra_col"
+  val EXPECTED_PARTITION_COL = "experted_pb_col"
+  val NUM_ROWS = 100
+
+  def withReadDataFrame(
+         format: String,
+         partitionCol: String = null,
+         extraCol: String = "ec",
+         extraSchemaFields: Seq[StructField] = Seq.empty)
+      (f: DataFrame => Unit): Unit = {
+    withTempPath { path =>
+      val baseDf = spark.range(0, NUM_ROWS, 1, 1).toDF("id")
+        .withColumn(extraCol, $"id" + lit(1000 * 1000))
+        .withColumn(EXPECTED_EXTRA_COL, col(extraCol))
+      val writeSchema: StructType = if (partitionCol != null) {
+        val writeDf = baseDf
+          .withColumn(partitionCol, ($"id" / 10).cast("int") + lit(1000))
+          .withColumn(EXPECTED_PARTITION_COL, col(partitionCol))
+          .withColumn(EXPECTED_ROW_ID_COL, $"id" % 10)
+        writeDf.write.format(format).partitionBy(partitionCol).save(path.getAbsolutePath)
+        writeDf.schema
+      } else {
+        val writeDf = baseDf
+          .withColumn(EXPECTED_ROW_ID_COL, $"id")
+        writeDf.write.format(format).save(path.getAbsolutePath)
+        writeDf.schema
+      }
+      val readSchema: StructType = new StructType(writeSchema.fields ++ extraSchemaFields)
+      val readDf = spark.read.format(format).schema(readSchema).load(path.getAbsolutePath)
+      f(readDf)
+    }
+  }
+
+  private val allMetadataCols = Seq(
+    FileFormat.FILE_PATH,
+    FileFormat.FILE_SIZE,
+    FileFormat.FILE_MODIFICATION_TIME,
+    FileFormat.ROW_INDEX
+  )
+
+  /** Identifies the names of all the metadata columns present in the schema. */
+  private def collectMetadataCols(struct: StructType): Seq[String] = {
+    struct.fields.flatMap { field => field.dataType match {
+      case s: StructType => collectMetadataCols(s)
+      case _ if allMetadataCols.contains(field.name) => Some(field.name)
+      case _ => None
+    }}
+  }
+
+  for (useVectorizedReader <- Seq(false, true))
+  for (useOffHeapMemory <- Seq(useVectorizedReader, false).distinct)
+  for (partitioned <- Seq(false, true)) {
+    val label = Seq(
+        { if (useVectorizedReader) "vectorized" else "parquet-mr"},
+        { if (useOffHeapMemory) "off-heap" else "" },
+        { if (partitioned) "partitioned" else "" }
+      ).filter(_.nonEmpty).mkString(", ")
+    test(s"parquet ($label) - read _metadata.row_index") {
+      withSQLConf(
+          SQLConf.PARQUET_VECTORIZED_READER_ENABLED.key -> useVectorizedReader.toString,
+          SQLConf.COLUMN_VECTOR_OFFHEAP_ENABLED.key -> useOffHeapMemory.toString) {
+        withReadDataFrame("parquet", partitionCol = "pb") { df =>
+          val res = df.select("*", s"${FileFormat.METADATA_NAME}.${FileFormat.ROW_INDEX}")
+            .where(s"$EXPECTED_ROW_ID_COL != ${FileFormat.ROW_INDEX}")
+          assert(res.count() == 0)
+        }
+      }
+    }
+  }
+
+  test("supported file format - read _metadata struct") {
+    withReadDataFrame("parquet") { df =>
+      val withMetadataStruct = df.select("*", FileFormat.METADATA_NAME)
+
+      // `_metadata.row_index` column is present when selecting `_metadata` as a whole.
+      val metadataCols = collectMetadataCols(withMetadataStruct.schema)
+      assert(metadataCols.contains(FileFormat.ROW_INDEX))
+    }
+  }
+
+  test("unsupported file format - read _metadata struct") {
+    withReadDataFrame("orc") { df =>
+      val withMetadataStruct = df.select("*", FileFormat.METADATA_NAME)
+
+      // Metadata struct can be read without an error.
+      withMetadataStruct.collect()
+
+      // Schema does not contain row index column, but contains all the remaining metadata columns.
+      val metadataCols = collectMetadataCols(withMetadataStruct.schema)
+      assert(!metadataCols.contains(FileFormat.ROW_INDEX))
+      assert(allMetadataCols.intersect(metadataCols).size == allMetadataCols.size - 1)
+    }
+  }
+
+  test("unsupported file format - read _metadata.row_index") {
+    withReadDataFrame("orc") { df =>
+      val ex = intercept[AnalysisException] {
+        df.select("*", s"${FileFormat.METADATA_NAME}.${FileFormat.ROW_INDEX}")
+      }
+      assert(ex.getMessage.contains("No such struct field row_index"))
+    }
+  }
+
+  for (useVectorizedReader <- Seq(true, false)) {
+    val label = if (useVectorizedReader) "vectorized" else "parquet-mr"
+    test(s"parquet ($label) - use mixed case for column name") {
+      withSQLConf(
+          SQLConf.PARQUET_VECTORIZED_READER_ENABLED.key -> useVectorizedReader.toString) {
+        withReadDataFrame("parquet") { df =>
+          val mixedCaseRowIndex = "RoW_InDeX"
+          assert(mixedCaseRowIndex.toLowerCase() == FileFormat.ROW_INDEX)
+
+          assert(df.select("*", s"${FileFormat.METADATA_NAME}.$mixedCaseRowIndex")
+            .where(s"$EXPECTED_ROW_ID_COL != $mixedCaseRowIndex")
+            .count == 0)
+        }
+      }
+    }
+  }
+
+  test(s"reading ${FileFormat.ROW_INDEX_TEMPORARY_COLUMN_NAME} - not present in a table") {
+    // File format supporting row index generation populates the column with row indexes.
+    withReadDataFrame("parquet", extraSchemaFields =
+        Seq(StructField(FileFormat.ROW_INDEX_TEMPORARY_COLUMN_NAME, LongType))) { df =>
+      assert(df
+          .where(col(EXPECTED_ROW_ID_COL) === col(FileFormat.ROW_INDEX_TEMPORARY_COLUMN_NAME))
+          .count == NUM_ROWS)
+    }
+
+    // File format not supporting row index generation populates missing column with nulls.
+    withReadDataFrame("json", extraSchemaFields =
+        Seq(StructField(FileFormat.ROW_INDEX_TEMPORARY_COLUMN_NAME, LongType)))  { df =>
+      assert(df
+          .where(col(FileFormat.ROW_INDEX_TEMPORARY_COLUMN_NAME).isNull)
+          .count == NUM_ROWS)
+    }
+  }
+
+  test(s"reading ${FileFormat.ROW_INDEX_TEMPORARY_COLUMN_NAME} - present in a table") {
+    withReadDataFrame("parquet", extraCol = FileFormat.ROW_INDEX_TEMPORARY_COLUMN_NAME) { df =>
+      // Column values are read from the file, rather than populated with generated row indexes.
+      // FIXME

Review Comment:
   @ala would you be able to create a follow-up ticket to address this so we don't forget? 🙂



-- 
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: reviews-unsubscribe@spark.apache.org

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


---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org