0

以下代码（部分显示）创建类 DescCalculator 的对象并计算描述符并将它们作为字符串数组返回。传入一个分子和一个描述符对象的 ArrayList。 

    private void calcDesc()
    {
        try
        {
        StatusPanel.setStatus("Calculating Molecular Descriptors Using CDK...\n");
        File df = new File(Settings.getCurrentDirectory() + sep + "molDesc.csv");
        FileWriter dfw = new FileWriter(df);
        LoadSDF lsdf1 = new LoadSDF(Settings.getCurrentDirectory() + sep + "marvin3D.sdf");
        List<IAtomContainer> mols3D = lsdf1.getCompounds();
        DescriptorEngine engine = new DescriptorEngine(DescriptorEngine.MOLECULAR);
        List<String> classNames = engine.getDescriptorClassNames();
        List<String> removeList = new ArrayList();
        removeList.add("org.openscience.cdk.qsar.descriptors.molecular.IPMolecularLearningDescriptor");
        classNames.removeAll(removeList);
        List<IDescriptor> instances = engine.instantiateDescriptors(classNames);
        engine.setDescriptorInstances(instances);

        List<String> headerItems = new ArrayList<String>();
        headerItems.add("CID");
        headerItems.add("MobCSA");
        for (IDescriptor descriptor : instances) {
            String[] names = descriptor.getDescriptorNames();
            headerItems.addAll(Arrays.asList(names));
        }
        ArrayList<IMolecularDescriptor> descriptors = new ArrayList();

        for (Object object : instances)
        {
        IMolecularDescriptor descriptor = (IMolecularDescriptor) object;
        String[] comps = descriptor.getSpecification().getSpecificationReference().split("#");
        descriptors.add(descriptor);
        }
        String headerLine = "";
        for (String header : headerItems) {
            headerLine = headerLine + header + ",";
        }

        dfw.append(headerLine+"\n");
        ExecutorService eservice = Executors.newFixedThreadPool(threads);
        CompletionService <List<String>> cservice = new ExecutorCompletionService <List<String>> (eservice);
        int k=0;
        for (IAtomContainer mol : mols3D)
        {
            DescCalculator dc = new DescCalculator(mol,descriptors);
            cservice.submit(dc);
            k=k+1;

        }
        for (int j=1 ; j<=k; j++)
        {
            StatusPanel.setStatus("Calculating Descriptors for Molecule "+j+"/"+compounds.size()+" Using "+threads+" Processors\n");
            List<String> dataItems = cservice.take().get();
                for (int i = 0; i < dataItems.size(); i++) {
                if (dataItems.get(i).equals("NaN")) {
                    dataItems.set(i, "NA");
                }
            }

            try {
                String dataLine = "";
                for (String data : dataItems) {
                    dataLine = dataLine + data + ",";
                }
                dfw.append(dataLine+"\n");
            } catch (Exception e) {
                System.out.println(e.toString());
            }
        }
 dfw.close();
        }
        catch (Exception e)
        {
            e.printStackTrace();
        }
    }

在类内部有一个 for 循环遍历描述符列表，如下所示（部分显示）。此代码遇到并发修改异常。如果我在 synchronized{} 块中使用线程 = 1 或描述符迭代，则代码运行良好，但我没有获得所需的并行化。如何迭代 DesCalculator 类中的列表？

    public class DescCalculator implements Callable<List<String>>{

    private IAtomContainer mol = new Molecule();
    private ArrayList<IMolecularDescriptor> molDesc;

    DescCalculator(IAtomContainer mol_, ArrayList<IMolecularDescriptor> molDesc_)
    {
        this.mol = mol_;
        this.molDesc = molDesc_;
    }

    @Override
    public List<String> call() {
        List<String> dataItems = new ArrayList<String>();
        try
        {
            String title = (String) mol.getProperty("PUBCHEM_COMPOUND_CID");
            dataItems.add(title);
            //String csa = Double.toString(mobcalCSA.get(ind));
            String csa = "NA";
            dataItems.add(csa);
            int ndesc = 0;
            for (IMolecularDescriptor descriptor : molDesc) {
                descriptor.calculate(mol);
                DescriptorValue value = descriptor.calculate(mol);
                if (value.getException() != null) {
                    for (int i = 0; i < value.getNames().length; i++) {
                        dataItems.add("NA");
                    }
                    continue;
                }

                IDescriptorResult result = value.getValue();
                if (result instanceof DoubleResult) {
                    dataItems.add(String.valueOf(((DoubleResult) result).doubleValue()));
                } else if (result instanceof IntegerResult) {
                    dataItems.add(String.valueOf(((IntegerResult) result).intValue()));
                } else if (result instanceof DoubleArrayResult) {
                    for (int i = 0; i < ((DoubleArrayResult) result).length(); i++) {
                        dataItems.add(String.valueOf(((DoubleArrayResult) result).get(i)));
                    }
                } else if (result instanceof IntegerArrayResult) {
                    for (int i = 0; i < ((IntegerArrayResult) result).length(); i++) {
                        dataItems.add(String.valueOf(((IntegerArrayResult) result).get(i)));
                    }
                }

                ndesc++;
            } 

        }
        catch(Exception e)
        {
            e.printStackTrace();
        }
        return dataItems;
    }

}

打印堆栈跟踪

java.util.ConcurrentModificationException
    at java.util.AbstractList$Itr.checkForComodification(AbstractList.java:372)
    at java.util.AbstractList$Itr.next(AbstractList.java:343)
    at org.openscience.cdk.ChemObject.notifyChanged(ChemObject.java:187)
    at org.openscience.cdk.ChemObject.setFlag(ChemObject.java:375)
    at org.openscience.cdk.graph.PathTools.depthFirstTargetSearch(PathTools.java:168)
    at org.openscience.cdk.graph.PathTools.depthFirstTargetSearch(PathTools.java:177)
    at org.openscience.cdk.graph.PathTools.depthFirstTargetSearch(PathTools.java:177)
    at org.openscience.cdk.graph.SpanningTree.getRing(SpanningTree.java:185)
    at org.openscience.cdk.graph.SpanningTree.getCyclicFragmentsContainer(SpanningTree.java:221)
    at org.openscience.cdk.atomtype.CDKAtomTypeMatcher.getRing(CDKAtomTypeMatcher.java:912)
    at org.openscience.cdk.atomtype.CDKAtomTypeMatcher.perceiveNitrogens(CDKAtomTypeMatcher.java:730)
    at org.openscience.cdk.atomtype.CDKAtomTypeMatcher.findMatchingAtomType(CDKAtomTypeMatcher.java:117)
    at org.openscience.cdk.tools.manipulator.AtomContainerManipulator.percieveAtomTypesAndConfigureAtoms(AtomContainerManipulator.java:719)
    at org.openscience.cdk.smiles.smarts.SMARTSQueryTool.initializeMolecule(SMARTSQueryTool.java:435)
    at org.openscience.cdk.smiles.smarts.SMARTSQueryTool.matches(SMARTSQueryTool.java:214)
    at org.openscience.cdk.smiles.smarts.SMARTSQueryTool.matches(SMARTSQueryTool.java:189)
    at org.openscience.cdk.qsar.descriptors.molecular.AcidicGroupCountDescriptor.calculate(AcidicGroupCountDescriptor.java:135)
    at edu.uconn.pharmacy.molfind.DescCalculator.call(DescCalculator.java:48)
    at edu.uconn.pharmacy.molfind.DescCalculator.call(DescCalculator.java:25)
    at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:303)
    at java.util.concurrent.FutureTask.run(FutureTask.java:138)
    at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:441)
    at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:303)
    at java.util.concurrent.FutureTask.run(FutureTask.java:138)
    at java.util.concurrent.ThreadPoolExecutor$Worker.runTask(ThreadPoolExecutor.java:886)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:908)
    at java.lang.Thread.run(Thread.java:680)
4

3 回答 3

1

除非您在走过它的同时实际修改a ，否则您不应该得到异常。Collection通常，这可以通过从Collection您正在迭代的 a 中删除来在 1 个线程中完成 -for例如在 a 中。

但是在您的情况下，我猜您正在从molDesc某个地方的列表中删除，尽管我在您提供的代码示例中看不到这一点。如果您需要从列表中删除条目，那么您将不得不使用其他一些机制来进行删除。您不能在多个线程中更改同一个集合，除非它是某种方式synchronized。

结合其他想法：

  • 现在确定它是否需要完全相同的列表。您可以让每个线程使用列表的副本。

    DescCalculator dc =
   new DescCalculator(mol, new ArrayList<IMolecularDescriptor>(descriptors));

  • 尽管我不确定这是否是您想要的，但您可以只传递一份Collections.synchronizedList副本。molDesc

    List<IMolecularDescriptor> syncedList =
    Collections.synchronizedList(descriptors);
...
DescCalculator dc = new DescCalculator(mol, syncedList);

  • 每个线程都可以保留需要删除的项目列表。最后，您可以使用Futurewhen the thread is reaped 来获取已删除的项目列表并将它们从最后的列表中删除。

于 2012-03-20T20:07:02.197 回答
0

事实证明，问题出在非线程安全的描述符对象上。我感谢 Affe 指出！在 DescCalculator 类中生成它们解决了这个问题。我感谢大家的意见！

于 2012-03-21T13:25:16.283 回答
0

您的 molDesc 字段是访问 call() 方法的 2 个或多个线程之间的共享状态。正如您所说，避免 CME 的唯一方法是将迭代包装在 synchronized() 块中（这将阻止您进行并行化），或者使用像 Victor 提到的 CopyOnWriteArrayList 这样的线程安全结构。

于 2012-03-20T20:18:59.987 回答