Question

Parallel.ForEach继续运行，我的程序没有结束。我无法追踪第一次迭代后的位置。我的猜测是陷入僵局并继续进行上下文切换。

private void ReadInputFile()
{
    var collection = new ConcurrentBag<PropertyRecord>();
    var lines = System.IO.File.ReadLines(InputFileName);
    int i = 0;
    int RecordsCount = lines.Count();
    Parallel.ForEach(lines, line =>
    {
        if (string.IsNullOrWhiteSpace(line))
        {
            return;                    
        }

        var tokens = line.Split(',');
        var postalCode = tokens[0];
        var country = tokens.Length > 1 ? tokens[1] : "england";

        SetLabelNotifyTwoText(
            string.Format(
                "Reading PostCode {0} out of {1}"
                i,
                lines.Length));

        var tempRecord = GetAllAddesses(postalCode, country);
        if (tempRecord != null)
        {
            foreach (PropertyRecord r in tempRecord)
            {
                collection.Add(r);
            }
        }    
    });
}

private List<PropertyRecord> GetAllAddesses(
        string postalCode,
        string country = "england")
{
    SetLabelNotifyText("");
    progressBar1.Value = 0;
    progressBar1.Update();

    var records = new List<PropertyRecord>();
    using (WebClient w = new WebClient())
    {
        var url = CreateUrl(postalCode, country);
        var document = w.DownloadString(url);
        var pagesCount = GetPagesCount(document);
        if (pagesCount == null)
        {
            return null;
        }

        for (int i = 0; i < pagesCount; i++)
        {
            SetLabelNotifyText(
                string.Format(
                    "Reading Page {0} out of {1}",
                    i,
                    pagesCount - 1));

            url = CreateUrl(postalcode,country, i);
            document = w.DownloadString(url);
            var collection = Regex.Matches(
                document,
                "<div class=\"soldDetails\">(.|\\n|\\r)*?class=" +
                "\"soldAddress\".*?>(?<address>.*?)(</a>|</div>)" +
                "(.|\\n|\\r)*?class=\\\"noBed\\\">(?<noBed>.*?)" +
                "</td>|</tbody>");

            foreach (var match in collection)
            {
                var r = new PropertyRecord();

                var bedroomCount = match.Groups["noBed"].Value;
                if(!string.IsNullOrEmpty(bedroomCount))
                {
                    r.BedroomCount = bedroomCount;             
                }
                else
                {
                    r.BedroomCount = "-1";
                }

                r.address = match.Groups["address"].Value;

                var line = string.Format(
                    "\"{0}\",{1}",
                    r.address
                    r.BedroomCount);
                OutputLines.Add(line);

                Records.Add(r);
            }
        }
    }

    return Records;
}

在没有Parallel.ForEach的情况下运行良好，但在要求中使用Parallel.ForEach。

我已经调试了它并且在从第GetAllAdresses方法返回后 - 第一次 Step Next 按钮暂停，它只是在后台继续调试。它没有回到我放置的任何书签上。

Answer 1

正如您在评论中所说，您的SetLabelNotifyText和SetLabelNotifyTwoText方法会调用Control.Invoke。

要使Control.Invoke正常工作，主线程必须是免费的，但在你的情况下，你似乎通过在其中调用Parallel.ForEach来阻止主线程。

这是一个极小的复制品：

private void button1_Click(object sender, EventArgs e)
{
    Parallel.ForEach(Enumerable.Range(1, 100), (i) =>
    {
        Thread.Sleep(10);//Simulate some work
        this.Invoke(new Action(() => SetText(i)));
    });
}

private void SetText(int i)
{
    textBox1.Text = i.ToString();
}

主线程等待Parallel.ForEach，工作线程等待主线程，从而导致死锁。

如何修复：不要使用Invoke只使用BeginInvoke或不要阻止MainThread。

_{如果这不是sscce的案例，那对我们有帮助}

Answer 2

更改您的代码，使用async and await。这是使用BeginInvoke和其他异步代码模型的现代替代方法。

private async Task ReadInputFile()
{
    var collection = new ConcurrentBag<PropertyRecord>();
    var lines = System.IO.File.ReadLines(InputFileName);
    int i = 0;
    int RecordsCount = lines.Count();
    Parallel.ForEach(lines, line =>
    {
        if (string.IsNullOrWhiteSpace(line))
        {
            return;                    
        }

        var tokens = line.Split(',');
        var postalCode = tokens[0];
        var country = tokens.Length > 1 ? tokens[1] : "england";

        SetLabelNotifyTwoText(
            string.Format(
                "Reading PostCode {0} out of {1}"
                i,
                lines.Length));

        var tempRecord = await GetAllAddesses(postalCode, country);
        if (tempRecord != null)
        {
            foreach (PropertyRecord r in tempRecord)
            {
                collection.Add(r);
            }
        }    
    });
}

private async Task<List<PropertyRecord>> GetAllAddesses(
        string postalCode,
        string country = "england")
{
    SetLabelNotifyText("");
    progressBar1.Value = 0;
    progressBar1.Update();

    var records = new List<PropertyRecord>();
    using (WebClient w = new WebClient())
    {
        var url = CreateUrl(postalCode, country);
        var document = await w.DownloadStringTaskAsync(url);
        var pagesCount = GetPagesCount(document);
        if (pagesCount == null)
        {
            return null;
        }

        for (int i = 0; i < pagesCount; i++)
        {
            SetLabelNotifyText(
                string.Format(
                    "Reading Page {0} out of {1}",
                    i,
                    pagesCount - 1));

            url = CreateUrl(postalcode,country, i);
            document = await w.DownloadStringTaskAsync(url);
            var collection = Regex.Matches(
                document,
                "<div class=\"soldDetails\">(.|\\n|\\r)*?class=" +
                "\"soldAddress\".*?>(?<address>.*?)(</a>|</div>)" +
                "(.|\\n|\\r)*?class=\\\"noBed\\\">(?<noBed>.*?)" +
                "</td>|</tbody>");

            foreach (var match in collection)
            {
                var r = new PropertyRecord();

                var bedroomCount = match.Groups["noBed"].Value;
                if(!string.IsNullOrEmpty(bedroomCount))
                {
                    r.BedroomCount = bedroomCount;             
                }
                else
                {
                    r.BedroomCount = "-1";
                }

                r.address = match.Groups["address"].Value;

                var line = string.Format(
                    "\"{0}\",{1}",
                    r.address
                    r.BedroomCount);
                OutputLines.Add(line);

                Records.Add(r);
            }
        }
    }

    return Records;
}

然后像这样称呼它

ReadInputFile.Wait();

或者更好的是，来电者是async，

await ReadInputFile();

多线程问题，也许是使用Foreach的DeadLock

2 个答案: