uros-db commented on code in PR #56266:
URL: https://github.com/apache/spark/pull/56266#discussion_r3341018039


##########
sql/api/src/main/scala/org/apache/spark/sql/types/ops/TimestampNanosTypeApiOps.scala:
##########
@@ -0,0 +1,91 @@
+/*
+ * 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.types.ops
+
+import org.apache.spark.SparkException
+import org.apache.spark.sql.AnalysisException
+import org.apache.spark.sql.catalyst.encoders.AgnosticEncoder
+import org.apache.spark.sql.errors.DataTypeErrorsBase
+import org.apache.spark.sql.types.{TimestampLTZNanosType, 
TimestampNTZNanosType}
+
+/**
+ * Client-side (spark-api) operations shared by the nanosecond timestamp types
+ * (TimestampNTZNanosType and TimestampLTZNanosType).
+ *
+ * Internal values are [[org.apache.spark.unsafe.types.TimestampNanosVal]] 
(epoch micros + nanos
+ * within the micro). The two concrete subclasses differ only in their 
DataType and SQL-literal
+ * prefix; storage and formatting are identical.
+ *
+ * SCOPE (SPARK-57207): this issue wires physical representation, literals, 
row accessors, and
+ * codegen class selection. String formatting is not yet implemented: format() 
throws an internal
+ * error so callers get a clear message rather than a debug string; dedicated 
fractional-second
+ * formatters land in a follow-up issue.
+ * Dataset encoders are out of scope (SPARK-57033 and related), so getEncoder 
reports the type as
+ * unsupported, matching the legacy RowEncoder behavior.
+ *
+ * @since 4.3.0
+ */
+abstract class TimestampNanosTypeApiOps extends TypeApiOps with 
DataTypeErrorsBase {
+
+  /** SQL literal prefix for this type, e.g. "TIMESTAMP_NTZ" or 
"TIMESTAMP_LTZ". */
+  protected def sqlTypeName: String
+
+  // ==================== String Formatting ====================
+
+  // String formatting of nanosecond timestamps is not yet implemented 
(follow-up after
+  // SPARK-57207). Throw an internal error so that callers see a clear message 
rather than a
+  // debug toString from TimestampNanosVal.
+  override def format(v: Any): String =
+    throw SparkException.internalError(
+      s"Formatting of ${dataType.typeName} is not yet implemented.")
+
+  override def toSQLValue(v: Any): String = s"$sqlTypeName '${format(v)}'"
+
+  // ==================== Row Encoding ====================
+
+  // Encoders are handled in a follow-up issue (SPARK-57033). Until then, 
report the type as
+  // unsupported with the same error as the legacy RowEncoder fallback to 
preserve parity.
+  override def getEncoder: AgnosticEncoder[_] =
+    throw new AnalysisException(
+      errorClass = "UNSUPPORTED_DATA_TYPE_FOR_ENCODER",

Review Comment:
   +1 with @stevomitric
   
   getEncoder throws UNSUPPORTED_DATA_TYPE_FOR_ENCODER, claiming this matches 
the legacy RowEncoder fallback. It does not. The legacy path already supports 
these types (SPARK-57033 landed), and encoderForDataType dispatches through the 
framework first, falling back to the legacy match only when TypeApiOps returns 
None.
   
   So with the flag off, a nanos column gets 
LocalDateTimeNanosEncoder/InstantNanosEncoder and Dataset create/collect 
roundtrips work. With the flag on, TypeApiOps(dt) is now Some(...), so 
_.getEncoder is called and throws — regressing functionality that currently 
works. This is the opposite of parity, and it's @stevomitric's first comment.
   
   Fix: have getEncoder return the existing encoders (and call 
DataTypeErrors.checkTimestampNanosTypesEnabled() to mirror the legacy path 
exactly), e.g. override in the NTZ/LTZ subclasses returning 
LocalDateTimeNanosEncoder(t.precision) / InstantNanosEncoder(t.precision).
   
   Note the test suite actively blesses the bug: "getEncoder is unsupported..." 
asserts the throw, and "framework disabled falls back to identical legacy 
behavior" re-checks physical type / literal / codegen / row roundtrip but not 
getEncoder, so the divergence is invisible. An off-vs-on parity assertion on 
getEncoder would have caught it.



-- 
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