cloud-fan commented on code in PR #55938: URL: https://github.com/apache/spark/pull/55938#discussion_r3274453292
########## sql/catalyst/src/main/java/org/apache/spark/sql/catalyst/expressions/ArithmeticUtils.java: ########## @@ -0,0 +1,88 @@ +/* + * 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.expressions; + +import org.apache.spark.sql.errors.QueryExecutionErrors; + +/** + * Static helpers used by {@code BinaryArithmetic.doGenCode} for ANSI + * overflow-checked {@code byte} and {@code short} arithmetic. These mirror + * {@code ByteExactNumeric} / {@code ShortExactNumeric} in {@code numerics.scala} + * (which the eval path uses); the Java helpers exist only because Scala + * {@code object} methods can't be called from generated code without going + * through the {@code references[]} array. Primitive {@code int} / {@code long} / + * {@code float} / {@code double} arithmetic stays inline -- routing those + * single-bytecode operations through a static method would be a runtime + * regression. + */ Review Comment: The rationale here is contradicted by existing code. `MathUtils` and `IntervalMathUtils` are both top-level Scala `object`s and are called directly from `BinaryArithmetic.doGenCode` via `getCanonicalName.stripSuffix("$")` (`arithmetic.scala:301, 324`) — no `references[]` needed. The trick works because the Scala compiler emits public static forwarders on top-level objects' companion class. `ByteExactNumeric` and `ShortExactNumeric` are also top-level objects, and `javap` on the compiled classes confirms the forwarders exist: ``` $ javap .../org/apache/spark/sql/types/ByteExactNumeric.class public final class org.apache.spark.sql.types.ByteExactNumeric { public static byte plus(byte, byte); public static byte minus(byte, byte); public static byte times(byte, byte); ... ``` (`private[sql]` is enforced at Scala compile time only; at the bytecode level the class is `public final` and the methods are `public static`.) So the codegen could call them directly, e.g.: ```scala case ByteType | ShortType if failOnError => val methodName = symbol match { case "+" => "plus" case "-" => "minus" case "*" => "times" case _ => throw SparkException.internalError( s"Unexpected symbol '$symbol' for Byte/Short BinaryArithmetic") } val numericObj = (if (dataType == ByteType) ByteExactNumeric else ShortExactNumeric) .getClass.getCanonicalName.stripSuffix("$") defineCodeGen(ctx, ev, (eval1, eval2) => s"$numericObj.$methodName($eval1, $eval2)") ``` This would eliminate the new file entirely and resolve the duplication @viirya raised, rather than documenting it. Did you try this path and hit a concrete problem? If so it'd be worth capturing in the Javadoc — as written, the justification is misleading. -- 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]
