gengliangwang commented on code in PR #58298:
URL: https://github.com/apache/spark/pull/58298#discussion_r4008007054


##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/v2/CapturedSchemaProjection.scala:
##########
@@ -0,0 +1,310 @@
+/*
+ * 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.v2
+
+import org.apache.spark.SparkException
+import org.apache.spark.sql.catalyst.SQLConfHelper
+import org.apache.spark.sql.catalyst.analysis.Resolver
+import org.apache.spark.sql.catalyst.expressions.{Alias, ArrayTransform, 
AttributeReference, CreateNamedStruct, Expression, GetStructField, If, IsNull, 
KnownNotNull, LambdaFunction, Literal, MetadataAttributeWithLogicalName, 
NamedLambdaVariable, TaggingExpression, TransformKeys, TransformValues, 
UnresolvedNamedLambdaVariable}
+import org.apache.spark.sql.catalyst.plans.logical.{LogicalPlan, Project}
+import org.apache.spark.sql.catalyst.util.MetadataColumnHelper
+import org.apache.spark.sql.types.{ArrayType, DataType, MapType, Metadata, 
StructType}
+
+/**
+ * Rebinds a relation that reads a current table schema to output attributes 
captured from an
+ * earlier compatible schema. The current schema is exposed by the relation so 
its output remains
+ * aligned with the physical scan, while a projection recreates the captured 
output for the
+ * already-analyzed parent plan.
+ */
+private[sql] object CapturedSchemaProjection extends SQLConfHelper {
+
+  /**
+   * Prevents [[CreateNamedStruct]] from inheriting metadata from a field 
value while leaving the
+   * value's type, nullability, evaluation, and code generation unchanged.
+   */
+  private case class MetadataPropagationBarrier(child: Expression) extends 
TaggingExpression {
+    override protected def withNewChildInternal(
+        newChild: Expression): MetadataPropagationBarrier = copy(child = 
newChild)
+  }
+
+  def rebindToCapturedSchema(relation: DataSourceV2Relation): LogicalPlan = {
+    // The relation still carries the output captured at analysis time; only 
its table has been
+    // swapped for the current one.
+    val capturedOutput = relation.output
+    val resolver = conf.resolver
+    val current = DataSourceV2Relation.create(
+      relation.table,
+      relation.catalog,
+      relation.identifier,
+      relation.options,
+      relation.timeTravelSpec)
+    val currentMetadataOutput = current.metadataOutput
+    val currentMetadata = capturedOutput.filter(_.isMetadataCol).map { 
captured =>
+      val logicalName = metadataLogicalName(captured)
+      matchName(currentMetadataOutput, logicalName, 
resolver)(metadataLogicalName)
+        .map(pos => currentMetadataOutput(pos))
+        .getOrElse {
+          // The connector still reports this metadata column, so it can only 
be absent here
+          // because a data column has taken its name and the connector 
suppresses rather than
+          // renames the conflict (`canRenameConflictingMetadataColumns`). 
Validation owns
+          // rejecting that.
+          unexpectedSchemaChange(
+            s"captured metadata column $logicalName is missing from the 
current relation")
+        }
+    }
+
+    val currentOutput = current.output ++ currentMetadata
+
+    // Refresh may visit an already rebound relation. Preserve its attributes 
so the projection
+    // above it continues to reference valid expression IDs.
+    //
+    // A further schema change on such a relation adds a second projection 
instead of replacing
+    // the first. Only the cache stores a refreshed plan, so the effect is 
limited to that entry:
+    // it stops matching the single projection a query rebuilds from its own 
captured output, and
+    // is no longer reused. Results stay correct.
+    if (sameOutputShape(capturedOutput, currentOutput)) {
+      return relation
+    }
+
+    val capturedIndex = new AttributeIndex(capturedOutput, resolver)
+    val reboundOutput = currentOutput.map { currentAttr =>
+      capturedIndex.get(currentAttr).filter(canReuse(_, 
currentAttr)).getOrElse(currentAttr)
+    }
+    val reboundRelation = relation.copy(output = reboundOutput)
+
+    val reboundIndex = new AttributeIndex(reboundOutput, resolver)
+    val projectList = capturedOutput.map { capturedAttr =>
+      val currentAttr = reboundIndex.get(capturedAttr).getOrElse {
+        unexpectedSchemaChange(
+          s"captured column ${capturedAttr.name} is missing from current table 
${relation.name}")
+      }
+      if (currentAttr.exprId == capturedAttr.exprId &&
+        sameAttributeShape(currentAttr, capturedAttr)) {
+        currentAttr
+      } else {
+        if (currentAttr.nullable != capturedAttr.nullable) {
+          unexpectedSchemaChange(
+            s"nullability changed for captured column ${capturedAttr.name} in 
${relation.name}")
+        }
+        val projected = projectToType(
+          currentAttr, currentAttr.dataType, capturedAttr.dataType, resolver)
+        if (projected.dataType != capturedAttr.dataType ||
+          projected.nullable != capturedAttr.nullable) {
+          unexpectedSchemaChange(
+            s"failed to recreate captured column ${capturedAttr.name} in 
${relation.name}")
+        }
+        Alias(projected, capturedAttr.name)(
+          exprId = capturedAttr.exprId,
+          qualifier = capturedAttr.qualifier,
+          explicitMetadata = Some(capturedAttr.metadata))
+      }
+    }
+
+    Project(projectList, reboundRelation)
+  }
+
+  private[v2] def projectToType(

Review Comment:
   Can we avoid introducing a second, DSv2-only schema-alignment implementation 
here? `V2TableRefreshUtil` has already validated the captured and current 
schemas through `V2TableUtil.validateCapturedColumns` / 
`validateCapturedMetadataColumns`, which delegate to 
`SchemaUtils.validateSchemaCompatibility(..., ALLOW_NEW_FIELDS, checkIds = 
true)`. `Project.matchSchema` already contains the recursive struct/array/map 
narrowing needed to build a projection from the fresh current relation. It is 
not a direct drop-in today: it merges metadata, uses resolver-only matching, 
cannot distinguish logically renamed metadata columns, and does not preserve 
captured expression IDs or qualifiers. However, those gaps argue for extracting 
or extending a shared strict Catalyst schema-projection helper, then applying 
final aliases with the captured name, exprId, qualifier, and exact metadata. A 
plain attribute-only Project works only for top-level additions; nested 
additions still need the shared recursive
  narrowing. Please refactor this rather than maintaining another roughly 200 
lines of recursive schema-projection logic here.



-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to