LuciferYang commented on code in PR #39712: URL: https://github.com/apache/spark/pull/39712#discussion_r1088153009
########## connector/connect/client/jvm/src/test/scala/org/apache/spark/sql/connect/client/CompatibilitySuite.scala: ########## @@ -0,0 +1,153 @@ +/* + * 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.connect.client + +import java.io.File +import java.net.URLClassLoader +import java.util.regex.Pattern + +import com.typesafe.tools.mima.core._ +import com.typesafe.tools.mima.lib.MiMaLib +import org.scalatest.funsuite.AnyFunSuite // scalastyle:ignore funsuite +import org.apache.spark.sql.connect.client.util.IntegrationTestUtils._ + +/** + * This test checks the binary compatibility of the connect client API against the spark SQL API + * using MiMa. We did not write this check using a SBT build rule as the rule cannot provide the + * same level of freedom as a test. With a test we can: + * 1. Specify any two jars to run the compatibility check. + * 1. Easily make the test automatically pick up all new methods added while the client is being + * built. + * + * The test requires the following artifacts built before running: + * {{{ + * spark-sql + * spark-connect-client-jvm + * }}} + * To build the above artifact, use e.g. `sbt package` or `mvn clean install -DskipTests`. + * + * When debugging this test, if any changes to the client API, the client jar need to be built + * before running the test. An example workflow with SBT for this test: + * 1. Compatibility test has reported an unexpected client API change. + * 1. Fix the wrong client API. + * 1. Build the client jar: `sbt package` + * 1. Run the test again: `sbt "testOnly + * org.apache.spark.sql.connect.client.CompatibilitySuite"` + */ +class CompatibilitySuite extends AnyFunSuite { // scalastyle:ignore funsuite + + private lazy val clientJar: File = + findJar( + "connector/connect/client/jvm", + "spark-connect-client-jvm-assembly", + "spark-connect-client-jvm") + + private lazy val sqlJar: File = findJar("sql/core", "spark-sql", "spark-sql") + + /** + * MiMa takes an old jar (sql jar) and a new jar (client jar) as inputs and then reports all + * incompatibilities found in the new jar. The incompatibility result is then filtered using + * include and exclude rules. Include rules are first applied to find all client classes that + * need to be checked. Then exclude rules are applied to filter out all unsupported methods in + * the client classes. + */ + test("compatibility MiMa tests") { + val mima = new MiMaLib(Seq(clientJar, sqlJar)) + val allProblems = mima.collectProblems(sqlJar, clientJar, List.empty) + val includedRules = Seq( + IncludeByName("org.apache.spark.sql.Column"), + IncludeByName("org.apache.spark.sql.Column$"), + IncludeByName("org.apache.spark.sql.Dataset"), + // TODO Add the Dataset object definition + // IncludeByName("org.apache.spark.sql.Dataset$"), + IncludeByName("org.apache.spark.sql.DataFrame"), + IncludeByName("org.apache.spark.sql.SparkSession"), + IncludeByName("org.apache.spark.sql.SparkSession$")) ++ includeImplementedMethods(clientJar) + val excludeRules = Seq( + // Filter unsupported rules: + // Two sql overloading methods are marked experimental in the API and skipped in the client. + ProblemFilters.exclude[Problem]("org.apache.spark.sql.SparkSession.sql"), + // Skip all shaded dependencies in the client. + ProblemFilters.exclude[Problem]("org.sparkproject.*"), + ProblemFilters.exclude[Problem]("org.apache.spark.connect.proto.*")) + val problems = allProblems + .filter { p => + includedRules.exists(rule => rule(p)) + } + .filter { p => + excludeRules.forall(rule => rule(p)) + } + + if (problems.nonEmpty) { + fail( + s"\nComparing client jar: $clientJar\nand sql jar: $sqlJar\n" + + problems.map(p => p.description("client")).mkString("\n")) + } + } + + test("compatibility API tests: Dataset") { + val clientClassLoader: URLClassLoader = new URLClassLoader(Seq(clientJar.toURI.toURL).toArray) + val sqlClassLoader: URLClassLoader = new URLClassLoader(Seq(sqlJar.toURI.toURL).toArray) + + val clientClass = clientClassLoader.loadClass("org.apache.spark.sql.Dataset") + val sqlClass = sqlClassLoader.loadClass("org.apache.spark.sql.Dataset") + + val newMethods = clientClass.getMethods + val oldMethods = sqlClass.getMethods + + // For now we simply check the new methods is a subset of the old methods. + newMethods + .map(m => m.toString) + .foreach(method => { + assert(oldMethods.map(m => m.toString).contains(method)) + }) + } + + /** + * Find all methods that are implemented in the client jar. Once all major methods are + * implemented we can switch to include all methods under the class using ".*" e.g. + * "org.apache.spark.sql.Dataset.*" + */ + private def includeImplementedMethods(clientJar: File): Seq[IncludeByName] = { + val clsNames = Seq( + "org.apache.spark.sql.Column", + // TODO Add all overloading methods. Temporarily mute compatibility check for the Dataset Review Comment: Like above TODO, need create a jira and add the corresponding jira id to TODO -- 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]
