[GitHub] [spark] srielau commented on a change in pull request #35855: [SPARK-38336][SQL] Support DEFAULT column values in CREATE/REPLACE TABLE and INSERT INTO statements

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

srielau commented on a change in pull request #35855:
URL: https://github.com/apache/spark/pull/35855#discussion_r828188039



##########
File path: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/DefaultColumns.scala
##########
@@ -0,0 +1,287 @@
+/*
+ * 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.catalyst.analysis
+
+import org.apache.spark.sql.AnalysisException
+import org.apache.spark.sql.catalyst.TableIdentifier
+import org.apache.spark.sql.catalyst.catalog._
+import org.apache.spark.sql.catalyst.expressions.{Expression, _}
+import org.apache.spark.sql.catalyst.optimizer.ConstantFolding
+import org.apache.spark.sql.catalyst.parser.{CatalystSqlParser, ParseException}
+import org.apache.spark.sql.catalyst.plans.logical._
+import org.apache.spark.sql.types._
+
+/**
+ * This class contains logic for processing DEFAULT columns in statements such as CREATE TABLE.
+ */
+object DefaultColumns {
+  val default = "default"
+  val analysisPrefix =
+    " has a DEFAULT value which fails to resolve to a valid constant expression: "
+  val columnDefaultNotFound = "Column 'default' does not exist"
+  lazy val parser = new CatalystSqlParser()
+  lazy val analyzer =
+    new Analyzer(new SessionCatalog(new InMemoryCatalog, FunctionRegistry.builtin))
+
+  /**
+   * Finds DEFAULT expressions in CREATE/REPLACE TABLE commands and constant-folds then.
+   *
+   * Example:
+   * CREATE TABLE T(a INT, b INT DEFAULT 5 + 5) becomes
+   * CREATE TABLE T(a INT, b INT DEFAULT 10)
+   *
+   * @param tableSchema represents the names and types of the columns of the statement to process.
+   * @param statementType name of the statement being processed, such as INSERT; useful for errors.
+   * @return a copy of `tableSchema` with field metadata updated with the constant-folded values.
+   */
+  def ConstantFoldDefaultExpressions(tableSchema: StructType, statementType: String): StructType = {
+    // Get the list of column indexes in the CREATE TABLE command with DEFAULT values.
+    val (fields: Array[StructField], indexes: Array[Int]) =
+      tableSchema.fields.zipWithIndex.filter { case (f, i) => f.metadata.contains(default) }.unzip
+    // Extract the list of DEFAULT column values from the CreateTable command.
+    val colNames: Seq[String] = fields.map { _.name }
+    val defaults: Seq[String] = fields.map { _.metadata.getString(default) }
+    // Extract the list of DEFAULT column values from the CreateTable command.
+    val exprs: Seq[Expression] = colNames.zip(defaults).map {
+      case (name, text) => Parse(name, text, statementType)
+    }
+    // Analyze and constant-fold each parse result.
+    val analyzed: Seq[Expression] = (exprs, defaults, colNames).zipped.map {
+      case (expr, default, name) => Analyze(expr, default, name, statementType)
+    }
+    // Create a map from the column index of each DEFAULT column to its type.
+    val indexMap: Map[Int, StructField] = (indexes, fields, analyzed).zipped.map {
+      case (index, field, expr) =>
+        val newMetadata: Metadata = new MetadataBuilder().withMetadata(field.metadata)
+          .putString(default, expr.sql).build()
+        (index, field.copy(metadata = newMetadata))
+    }.toMap
+    // Finally, replace the original struct fields with the new ones.
+    val newFields: Seq[StructField] =
+      tableSchema.fields.zipWithIndex.map { case (f, i) => indexMap.getOrElse(i, f) }
+    StructType(newFields)
+  }
+
+  /**
+   * Adds a projection over the plan in `insert.query` generating missing default column values.
+   *
+   * @param insert the INSERT INTO statement to add missing DEFAULT column references to.
+   * @param catalog the catalog to use for looking up the schema of the INSERT INTO table object.
+   * @return the updated statement with missing DEFAULT column values appended to the list.
+   */
+  def AddProjectionForMissingDefaultColumnValues(
+      insert: InsertIntoStatement, catalog: SessionCatalog): InsertIntoStatement = {
+    // Compute the number of attributes returned by the INSERT INTO statement.
+    val numQueryOutputs: Int = insert.query match {
+      case table: UnresolvedInlineTable
+        if table.rows.nonEmpty && table.rows.forall(_.size == table.rows(0).size) =>
+        table.rows(0).size
+      case project: Project => project.projectList.size
+      case _ => return insert
+    }
+    // The table value provides the DEFAULT column values as text; analyze them into expressions.
+    val schema: StructType = getInsertTableSchema(insert, catalog).getOrElse(return insert)
+    val schemaWithoutPartitionCols = StructType(schema.fields.dropRight(insert.partitionSpec.size))
+    val coerced: Seq[Expression] = for {
+      field <- schemaWithoutPartitionCols.fields.drop(numQueryOutputs)
+      name: String = field.name
+      text: String =
+        if (field.metadata.contains(default)) field.metadata.getString(default) else "NULL"
+      // Parse the DEFAULT column expression. If the parsing fails, throw an error to the user.
+      expr: Expression = Parse(name, text, "INSERT")
+      // Analyze and constant-fold each result.
+      analyzed: Expression = Analyze(expr, text, name, "INSERT")
+      // Perform implicit coercion from the provided expression type to the required column type.
+      errorPrefix = "Failed to execute INSERT command because the destination table column "
+      coerced: Expression =
+        if (field.dataType == analyzed.dataType) {
+          analyzed
+        } else if (Cast.canUpCast(analyzed.dataType, field.dataType)) {
+          Cast(analyzed, field.dataType)
+        } else {
+          throw new AnalysisException(errorPrefix +
+            s"$name has a DEFAULT value with type ${field.dataType}, but the " +
+            s"query provided a value of incompatible type ${analyzed.dataType}")
+        }
+    } yield coerced
+    // Finally, return a projection of the original `insert.query` output attributes plus new
+    // aliases over the DEFAULT column values.
+    // If the insertQuery is an existing Project, flatten them together.
+    val newQuery = insert.query match {
+      case Project(projectList, child) =>
+        val newAliases: Seq[NamedExpression] = coerced.zip(schemaWithoutPartitionCols.fields).map {
+          case (expr, field) => Alias(expr, field.name)() }
+        Project(projectList ++ newAliases, child)
+      case table: UnresolvedInlineTable =>
+        val newNames: Seq[String] =
+          schemaWithoutPartitionCols.fields.drop(numQueryOutputs).map { _.name }
+        table.copy(names = table.names ++ newNames, rows = table.rows.map { row => row ++ coerced })
+      case _ => insert.query
+    }
+    insert.copy(query = newQuery)
+  }
+
+  /**
+   * Replaces unresolved "DEFAULT" column references with matching default column values.
+   *
+   * Background: CREATE TABLE and ALTER TABLE invocations support setting column default values for
+   * later operations. Following INSERT, and INSERT MERGE commands may then reference the value
+   * using the DEFAULT keyword as needed.
+   *
+   * Example:
+   * CREATE TABLE T(a INT DEFAULT 4, b INT NOT NULL DEFAULT 5);

Review comment:
       I just want to clarify that we MUST NOT  limit ourselves to constant expressions.
   E.g. there are common use cases around e.g. CURRENT_DATE() a well as using NON DETERMINISTIC functions
   E.g. to generate UUIDs  for primary keys.
   We should design accordingly.
   




-- 
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