Task.WaitAll(taskList.ToArray()) never fisnishes

[Dash83UPV]

Dear all
As always, thank you for your hard work

Maybe I'm doing something wrong, but I think I have found an issue.

Launching several tasks to Create several tweets in parallel and using Task.WaiAll() does not work correctly

I have been using the latest 0.40 version with Core NET 8.0
SurrealDB (latest version 1.3.1) is the table tweets with no records.

This is the test code I have prepared.

private static async Task Main(string[] args)
{
    DataBaseParams dbParams = new DataBaseParams()
    {
        ipAddress = "192.168.0.1",
        port = 8000,
        username = "root",
        password = "root",
        database = "testdb",
        dataBaseEngine = DataBaseEngine.SurrealDB
    };

    var surrealDBClient = new SurrealDbClient(
                    SurrealDbOptions
                        .Create()
                        .WithEndpoint($"ws://{dbParams.ipAddress}:{dbParams.port}/rpc")
                        .WithUsername(dbParams.username)
                        .WithPassword(dbParams.password)
                        .WithNamespace(dbParams.database)
                        .WithDatabase(dbParams.database)
                        .Build(),
                    null,
                    (_options) =>
                    {
                        _options.IncludeFields = true;
                    }
    );

    List<Tweet> tweets = new List<Tweet>();
    for (int iCount = 0; iCount < 5; iCount++)
    {
        tweets.Add(new Tweet()
        {
            id = $"ID{iCount}",
            username = $"user{iCount}",
        });
    }

    List<Task<Tweet>> list = new List<Task<Tweet>>();
    tweets.ForEach(tweet =>
    {
        var task = surrealDBClient.Create("tweets2", tweet);
        //var task = dummyTask();
        list.Add(task);
    });

    Task.WaitAll(list.ToArray());
}

Please notice there are 2 methods. One is the basic SurrealDB Create method

var task = surrealDBClient.Create("tweets2", tweet);

The other is a dummy method used to compare the execution.

var task = dummyTask();

private static async Task<Tweet> dummyTask()
{
   await Task.Delay(2000);
   return new Tweet()
   {
       id = "dummy",
       username = "pepito"
   };
}

When I use the code with the Create the execution is stuck on the Task.WaitAll() forever, but checking the DB with Surrealist I can see that the 5 records were inserted.

When I use the code with the DummyTask the execution continues and the program finishes.

I have not tested the code with HTTP instead of WS, but it could have the same behaviour.

[Odonno]

Hello Alfonso,

I can reproduce this issue. It stops responding on WS calls but works fine on HTTP calls. The root cause it that you are using Task.WaitAll instead of Task.WhenAll.

I am not sure of what you want to do but one major difference between the two methods is that WaitAll is synchronous hence blocking the current (main/UI) thread. WhenAll is asynchronous and return a Task, you can call it anytime you want without blocking the main thread.

In general, it is a best practice to use WhenAll instead of WaitAll unless you really want to use it.

Like you, I wasn't expecting this behavior because I have to consume an Observable that seems to be blocked by another task that runs on the current thread. This can lead to some issue. I am curious to see if a migration to a Channel fixes this (among other things). I do not think this is much important. Can you confirm that WhenAll is a valid usage for you?

Thanks for raising this issue.

[Dash83UPV]

Hello Odonno

As you said, by modifying my code from

Task.WaitAll(list.ToArray());

to

await Task.WhenAll(list.ToArray());

it works. For now I think I can work with it.

Unfortunately, I just tested the parallelization of several Create method calls, i.e.:

var tasks = tweets.Select(tweet => surrealDBClient.Create("tweets2", tweet));
await Task.WhenAll(tasks);

against

tweets.ForEach(async tweet => await surrealDBClient.Create("tweets2", tweet));

And the ForEach one is faster for Lists below 5000 records.

Really thank you for your help and fast answer.
I will continue bench-working and testing the API.

Best Regards

[Odonno]

Yes. It depends on what you need once again. :)

Small article that explains the difference: https://blog.nimblepros.com/blogs/task-whenall-vs-parallel-foreach/