Java8 Collections.sort（有时）不对JPA返回的列表进行排序

孔华池

2023-03-14

问题内容：

Java8在我的JPA EclipseLink 2.5.2环境中一直在做奇怪的事情。我
昨天不得不删除问题因为在这种情况下，排序受到一种奇怪的JPA行为的影响-我通过强制执行进行最终排序之前先进行排序。

仍然在带有JPA Eclipselink 2.5.2的Java
8中，以下代码有时在我的环境中无法排序（Linux，MacOSX，都使用内部版本1.8.0_25-b17）。它在JDK 1.7环境中按预期工作。

public List<Document> getDocumentsByModificationDate() {
    List<Document> docs=this.getDocuments();
    LOGGER.log(Level.INFO,"sorting "+docs.size()+" by modification date");
    Comparator<Document> comparator=new ByModificationComparator();
    Collections.sort(docs,comparator);
    return docs;
}

当从JUnit测试调用时，上述功能可以正常工作。在生产环境中进行调试时，我得到一个日志条目：

INFORMATION: sorting 34 by modification date

但在TimSort中，nRemaining <2的return语句被命中-因此不进行排序。JPA提供的IndirectList被认为是空的。

static <T> void sort(T[] a, int lo, int hi, Comparator<? super T> c,
                     T[] work, int workBase, int workLen) {
    assert c != null && a != null && lo >= 0 && lo <= hi && hi <= a.length;

    int nRemaining  = hi - lo;
    if (nRemaining < 2)
        return;  // Arrays of size 0 and 1 are always sorted

此解决方法正确排序：

   if (docs instanceof IndirectList) {
        IndirectList iList = (IndirectList)docs;
        Object sortTargetObject = iList.getDelegateObject();
        if (sortTargetObject instanceof List<?>) {
            List<Document> sortTarget=(List<Document>) sortTargetObject;
            Collections.sort(sortTarget,comparator);
        }
    } else {
        Collections.sort(docs,comparator);
    }

题：

这是JPA Eclipselink错误还是我通常可以在自己的代码中对其进行处理？

请注意-我还不能将软件更改为Java8源合规性。当前环境是Java8运行时。

我对这种行为感到惊讶-在生产环境中出现问题时，测试用例正确运行特别令人讨厌。

在https://github.com/WolfgangFahl/JPAJava8Sorting有一个示例项目，
该项目的结构与原始问题相当。

它包含一个带有JUnit测试的http://sscce.org/示例，该示例通过调用em.clear（）使问题可重现，从而分离所有对象并强制使用IndirectList。请参阅下面的JUnit案例以供参考。

渴望获取：

// https://stackoverflow.com/questions/8301820/onetomany-relationship-is-not-working
@OneToMany(cascade = CascadeType.ALL, mappedBy = "parentFolder", fetch=FetchType.EAGER)

单位案例有效。如果在JDK 8中使用了FetchType.LAZY或省略了提取类型，则其行为可能与JDK 7中的行为不同（我现在必须对此进行检查）。
为什么呢？ 目前，我假设需要指定Eager提取或在要排序的列表上进行一次迭代，基本上是在排序之前手动提取。 还有什么可以做的？

JUnit测试

可以从https://github.com/WolfgangFahl/JPAJava8Sorting中获取persistence.xml和pom.xml。
该测试可以使用MYSQL数据库运行，也可以使用DERBY在内存中运行（默认）

package com.bitplan.java8sorting;

import static org.junit.Assert.assertEquals;

import java.util.ArrayList;
import java.util.Collections;
import java.util.Comparator;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.logging.Level;
import java.util.logging.Logger;

import javax.persistence.Access;
import javax.persistence.AccessType;
import javax.persistence.CascadeType;
import javax.persistence.Entity;
import javax.persistence.EntityManager;
import javax.persistence.EntityManagerFactory;
import javax.persistence.FetchType;
import javax.persistence.Id;
import javax.persistence.ManyToOne;
import javax.persistence.OneToMany;
import javax.persistence.Persistence;
import javax.persistence.Query;
import javax.persistence.Table;

import org.eclipse.persistence.indirection.IndirectList;
import org.junit.Test;

/**
 * Testcase for 
 * https://stackoverflow.com/questions/26816650/java8-collections-sort-sometimes-does-not-sort-jpa-returned-lists
 * @author wf
 *
 */
public class TestJPASorting {

  // the number of documents we want to sort
  public static final int NUM_DOCUMENTS = 3;

  // Logger for debug outputs
  protected static Logger LOGGER = Logger.getLogger("com.bitplan.java8sorting");

  /**
   * a classic comparator
   * @author wf
   *
   */
  public static class ByNameComparator implements Comparator<Document> {

    // @Override
    public int compare(Document d1, Document d2) {
      LOGGER.log(Level.INFO,"comparing " + d1.getName() + "<=>" + d2.getName());
      return d1.getName().compareTo(d2.getName());
    }
  }

  // Document Entity - the sort target
  @Entity(name = "Document")
  @Table(name = "document")
  @Access(AccessType.FIELD)
  public static class Document {
    @Id
    String name;

    @ManyToOne
    Folder parentFolder;

    /**
     * @return the name
     */
    public String getName() {
      return name;
    }
    /**
     * @param name the name to set
     */
    public void setName(String name) {
      this.name = name;
    }
    /**
     * @return the parentFolder
     */
    public Folder getParentFolder() {
      return parentFolder;
    }
    /**
     * @param parentFolder the parentFolder to set
     */
    public void setParentFolder(Folder parentFolder) {
      this.parentFolder = parentFolder;
    }
  }

  // Folder entity - owning entity for documents to be sorted
  @Entity(name = "Folder")
  @Table(name = "folder")
  @Access(AccessType.FIELD)
  public static class Folder {
    @Id
    String name;

    // https://stackoverflow.com/questions/8301820/onetomany-relationship-is-not-working
    @OneToMany(cascade = CascadeType.ALL, mappedBy = "parentFolder", fetch=FetchType.EAGER)
    List<Document> documents;

    /**
     * @return the name
     */
    public String getName() {
      return name;
    }

    /**
     * @param name the name to set
     */
    public void setName(String name) {
      this.name = name;
    }

    /**
     * @return the documents
     */
    public List<Document> getDocuments() {
      return documents;
    }

    /**
     * @param documents the documents to set
     */
    public void setDocuments(List<Document> documents) {
      this.documents = documents;
    }

    /**
     * get the documents of this folder by name
     * 
     * @return a sorted list of documents
     */
    public List<Document> getDocumentsByName() {
      List<Document> docs = this.getDocuments();
      LOGGER.log(Level.INFO, "sorting " + docs.size() + " documents by name");
      if (docs instanceof IndirectList) {
        LOGGER.log(Level.INFO, "The document list is an IndirectList");
      }
      Comparator<Document> comparator = new ByNameComparator();
      // here is the culprit - do or don't we sort correctly here?
      Collections.sort(docs, comparator);
      return docs;
    }

    /**
     * get a folder example (for testing)
     * @return - a test folder with NUM_DOCUMENTS documents
     */
    public static Folder getFolderExample() {
      Folder folder = new Folder();
      folder.setName("testFolder");
      folder.setDocuments(new ArrayList<Document>());
      for (int i=NUM_DOCUMENTS;i>0;i--) {
        Document document=new Document();
        document.setName("test"+i);
        document.setParentFolder(folder);
        folder.getDocuments().add(document);
      }
      return folder;
    }
  }

  /** possible Database configurations
  using generic persistence.xml:
    <?xml version="1.0" encoding="UTF-8"?>
    <!-- generic persistence.xml which only specifies a persistence unit name -->
    <persistence xmlns="http://java.sun.com/xml/ns/persistence"
      version="2.0">
      <persistence-unit name="com.bitplan.java8sorting" transaction-type="RESOURCE_LOCAL">
        <description>sorting test</description>
        <provider>org.eclipse.persistence.jpa.PersistenceProvider</provider>
        <exclude-unlisted-classes>false</exclude-unlisted-classes> 
        <properties>
        <!--  set programmatically -->
         </properties>
      </persistence-unit>
    </persistence>
  */
  // in MEMORY database
  public static final JPASettings JPA_DERBY=new JPASettings("Derby","org.apache.derby.jdbc.EmbeddedDriver","jdbc:derby:memory:test-jpa;create=true","APP","APP");
  // MYSQL Database
  //  needs preparation:
  //    create database testsqlstorage;
  //    grant all privileges on testsqlstorage to cm@localhost identified by 'secret';
  public static final JPASettings JPA_MYSQL=new JPASettings("MYSQL","com.mysql.jdbc.Driver","jdbc:mysql://localhost:3306/testsqlstorage","cm","secret");

  /**
   * Wrapper class for JPASettings
   * @author wf
   *
   */
  public static class JPASettings {
    String driver;
    String url;
    String user;
    String password;
    String targetDatabase;

    EntityManager entityManager;
    /**
     * @param driver
     * @param url
     * @param user
     * @param password
     * @param targetDatabase
     */
    public JPASettings(String targetDatabase,String driver, String url, String user, String password) {
      this.driver = driver;
      this.url = url;
      this.user = user;
      this.password = password;
      this.targetDatabase = targetDatabase;
    }

    /**
     * get an entitymanager based on my settings
     * @return the EntityManager
     */
    public EntityManager getEntityManager() {
      if (entityManager == null) {
        Map<String, String> jpaProperties = new HashMap<String, String>();
        jpaProperties.put("eclipselink.ddl-generation.output-mode", "both");
        jpaProperties.put("eclipselink.ddl-generation", "drop-and-create-tables");
        jpaProperties.put("eclipselink.target-database", targetDatabase);
        jpaProperties.put("eclipselink.logging.level", "FINE");

        jpaProperties.put("javax.persistence.jdbc.user", user);
        jpaProperties.put("javax.persistence.jdbc.password", password);
        jpaProperties.put("javax.persistence.jdbc.url",url);
        jpaProperties.put("javax.persistence.jdbc.driver",driver);

        EntityManagerFactory emf = Persistence.createEntityManagerFactory(
            "com.bitplan.java8sorting", jpaProperties);
        entityManager = emf.createEntityManager();
      }
      return entityManager;
    }
  }

  /**
   * persist the given Folder with the given entityManager
   * @param em - the entityManager
   * @param folderJpa - the folder to persist
   */
  public void persist(EntityManager em, Folder folder) {
    em.getTransaction().begin();
    em.persist(folder);
    em.getTransaction().commit();    
  }

  /**
   * check the sorting - assert that the list has the correct size NUM_DOCUMENTS and that documents
   * are sorted by name assuming test# to be the name of the documents
   * @param sortedDocuments - the documents which should be sorted by name
   */
  public void checkSorting(List<Document> sortedDocuments) {
    assertEquals(NUM_DOCUMENTS,sortedDocuments.size());
    for (int i=1;i<=NUM_DOCUMENTS;i++) {
      Document document=sortedDocuments.get(i-1);
      assertEquals("test"+i,document.getName());
    }
  }

  /**
   * this test case shows that the list of documents retrieved will not be sorted if 
   * JDK8 and lazy fetching is used
   */
  @Test
  public void testSorting() {
    // get a folder with a few documents
    Folder folder=Folder.getFolderExample();
    // get an entitymanager JPA_DERBY=inMemory JPA_MYSQL=Mysql disk database
    EntityManager em=JPA_DERBY.getEntityManager();
    // persist the folder
    persist(em,folder);
    // sort list directly created from memory
    checkSorting(folder.getDocumentsByName());

    // detach entities;
    em.clear();
    // get all folders from database
    String sql="select f from Folder f";
    Query query = em.createQuery(sql);
    @SuppressWarnings("unchecked")
    List<Folder> folders = query.getResultList();
    // there should be exactly one
    assertEquals(1,folders.size());
    // get the first folder
    Folder folderJPA=folders.get(0);
    // sort the documents retrieved
    checkSorting(folderJPA.getDocumentsByName());
  }
}

问题答案：

好吧，这是一个完美的教学法，告诉您程序员为什么不应该扩展不是为子类设计的类。诸如“ Effective
Java”之类的书告诉您原因：当超类演化时，尝试拦截每种方法以改变其行为的尝试将失败。

在这里，IndirectList扩展Vector并覆盖几乎所有修改其行为的方法，这是一种清晰的反模式。现在，随着Java 8基类的发展。

从Java
8开始，接口可以具有default方法，因此sort添加了类似的方法，这些方法的优点在于，与Collections.sort实现不同，实现可以覆盖该方法，并提供更适合特定interface实现的实现。Vector这样做有两个原因：现在所有协定都synchronized扩展为排序，并且优化的实现可以将其内部数组传递给该Arrays.sort方法，从而跳过先前实现中已知的复制操作（ArrayList执行相同操作）。

为了立即获得这种好处，即使对于现有代码，Collections.sort也已进行了改进。它委托List.sort默认情况下将委托给另一个方法，该方法实现通过toArray和使用复制的旧行为TimSort。但是，如果List实现重写，List.sort则也会影响行为Collections.sort。

                  interface method              using internal
                  List.sort                     array w/o copying
Collections.sort ─────────────────> Vector.sort ─────────────────> Arrays.sort

Java8 Collections.sort（有时）不对JPA返回的列表进行排序

相关阅读

相关文章

相关问答

相关工具

相关文档