具有异步Task函数C＃/。NET的线程/等待未编译

Question

I have a project where I need multiple data for a list of projects. For each project I call an api to get me this information. The loop works, although it takes 4 to 5 minutes to finish(Which is A LOT).

The code used to look like this :

foreach (var project in projects)
{
    string url = urlOneProject + project.name + secondPartUrl + "?authtoken=" + authToken;
    HttpWebRequest request;
    request = (HttpWebRequest)WebRequest.Create(url);
    request.Accept = "application/json";
    request.ContentType = "application/json";
    var executions = new Execs();

    var response = (HttpWebResponse)(await request.GetResponseAsync());

    using (response)
    {
        using (var reader = new StreamReader(response.GetResponseStream()))
        {
            JavaScriptSerializer js = new JavaScriptSerializer();
            var objText = reader.ReadToEnd();
            executions = (Execs)js.Deserialize(objText, typeof(Execs));
        }
    }


    execs.AddRange(executions.executions);
} 

To improve performance I thought that using threads might be a good idea. So, I came up with something like this:

ManualResetEvent resetEvent = new ManualResetEvent(false);
int toProcess = projects.Count;
foreach (var project in projects)
{
    new Thread(() =>
    {

        string url = urlOneProject + project.name + secondPartUrl + "?authtoken=" + authToken;
        HttpWebRequest request;
        request = (HttpWebRequest)WebRequest.Create(url);
        request.Accept = "application/json";
        request.ContentType = "application/json";
        var executions = new Execs();

        var response = (HttpWebResponse)(await request.GetResponseAsync());

        using (response)
        {
            using (var reader = new StreamReader(response.GetResponseStream()))
            {
                JavaScriptSerializer js = new JavaScriptSerializer();
                var objText = reader.ReadToEnd();
                executions = (Execs)js.Deserialize(objText, typeof(Execs));
            }
        }

        lock (execs)
        {
            execs.AddRange(executions.executions);
        }

        if (Interlocked.Decrement(ref toProcess) == 0)
            resetEvent.Set();
    }).Start();

}

The problem with this code is that the line:

var response = (HttpWebResponse)(await request.GetResponseAsync());

doesn't compile anymore from the moment I added Thread . And the error I get is

"The 'await' operator can only be used within an async lambda expression "

That wasn't a problem when I didn't use threads. The GetResponseAsync is an async function and the use of the await is compulsory. I tried deleting it (which wasn't logical I agree but I ran out of options) but the compiler tells me I need an await for an async function. I don't quite understand what changes with the implementation of the Thread .

Didn't I use the threads mechanically correctly? What should I do to correct this or to implement what I want to do correctly?

Answer 1

You are mixing multiple paradigms of coding, viz async / await , and oldschool Thread starting and synchronization, which is likely to lead to trouble.

As per the above comments

• The reason your code doesn't compile is because you are attempting to use await in otherwise synchronous code passed to the thread. You can qualify lambdas as async as well.
• Task is a much safer paradigm than Thread , and TPL provides rich and expressive tools to assist in asynchrony and parallelism.
• If you process each parallel task in isolation, without sharing any data (such as the collections that you are locking), but instead return the resultant data from each Task you can then use LINQ to collate the results in a thread-safe manner.

var myTasks = projects.Select(async project =>
{
    var url = $"urlOneProject{project.name}{secondPartUrl}?authtoken={authToken}";
    var request = (HttpWebRequest) WebRequest.Create(url);
    request.Accept = "application/json";
    request.ContentType = "application/json";

    using (var response = (HttpWebResponse) (await request.GetResponseAsync()))
    using (var reader = new StreamReader(response.GetResponseStream()))
    {
        var objText = await reader.ReadToEndAsync();
        return JsonConvert.DeserializeObject<Execs>(objText);
    }
});

var execs = (await Task.WhenAll(myTasks))
    .SelectMany(result => result.executions);

Other notes

• Don't use JavaScriptSerializer - even the MSDN docco says to use NewtonSoft Json
• There's an async version of reader.ReadToEndAsync which I've included.
• You can drop the locks and the ManualResetEvent - since each Task returns it's result, we'll leave it to Task.WhenAll to collate the data.
• You can flatten the children of multiple executions with a SelectMany
• Adjacent using clauses are stackable - it saves a bit of eyestrain on the indentation.