In SignalR Core using ChannelWriter: Do I need to call TryComplete twice if there's an exception?

Question

Here is an excerpt from the Use streaming in ASP.NET Core SignalR article by Microsoft:

private async Task WriteItemsAsync(
    ChannelWriter<int> writer,
    int count,
    int delay,
    CancellationToken cancellationToken)
{
    try
    {
        for (var i = 0; i < count; i++)
        {
            // Check the cancellation token regularly so that the server will stop
            // producing items if the client disconnects.
            cancellationToken.ThrowIfCancellationRequested();
            await writer.WriteAsync(i);

            // Use the cancellationToken in other APIs that accept cancellation
            // tokens so the cancellation can flow down to them.
            await Task.Delay(delay, cancellationToken);
        }
    }
    catch (Exception ex)
    {
        writer.TryComplete(ex);
    }

    writer.TryComplete();
}

If there's an exception, it will first call writer.TryComplete(ex), and then writer.TryComplete(). In other words, it calls TryComplete (albeit different overloads) twice.

Is this necessary? Should I add a return statement after writer.TryComplete(ex) to avoid calling it twice? Or does the second writer.TryComplete() serve some meaningful purpose after calling the former?

score 1 · Accepted Answer · answered Sep 06 '19 at 15:27

It's not necessary. This isn't the best example of a channel producer. writer.TryComplete() should be the last call of the try{} block :

private async Task WriteItemsAsync(
    ChannelWriter<int> writer,
    int count,
    int delay,
    CancellationToken cancellationToken)
{
    try
    {
        for (var i = 0; i < count; i++)
        {
            // Check the cancellation token regularly so that the server will stop
            // producing items if the client disconnects.
            cancellationToken.ThrowIfCancellationRequested();
            await writer.WriteAsync(i);

            // Use the cancellationToken in other APIs that accept cancellation
            // tokens so the cancellation can flow down to them.
            await Task.Delay(delay, cancellationToken);
        }
        writer.TryComplete();
    }
    catch (Exception ex)
    {
        writer.TryComplete(ex);
    }   
}

This way, it's only called once, either when the loop terminates succesfully or when an exception is thrown for any reason.

Instead of throwing on cancellation with ThrowIfCancellationRequested you could simply break out of the loop :

for (var i = 0; i < count; i++)
{
    if (cancellationToken.IsCancellationRequested)
    {
        break;
    }
    await writer.WriteAsync(i);
    await Task.Delay(delay, cancellationToken);
}
writer.TryComplete();

In case of bounded channels, WriteAsync should also receive the cancellation token, otherwise it may get stuck if the channel is full:

await writer.WriteAsync(i,cancellationToken);

In SignalR Core using ChannelWriter: Do I need to call TryComplete twice if there's an exception?

1 Answers1