Speed up loop with ForEachAsync

I’m reposting an useful old post by Stephen Toub.

ForEach loop doesn’t supports Async/Await operators. If you like the possibility to have a foreach loop that can run in parallel, and can also leverage async/await operators to improve the performances of your loop, you can use the following, very useful, extension method.

public static class Extensions 
{
    public static Task ForEachAsync<TSource, TResult>(
        this IEnumerable<TSource> source,
        Func<TSource, Task<TResult>> taskSelector, Action<TSource, TResult> resultProcessor)
    {
        var oneAtATime = new SemaphoreSlim(initialCount:1, maxCount:1);
        return Task.WhenAll(
            from item in source
            select ProcessAsync(item, taskSelector, resultProcessor, oneAtATime));
    }

    private static async Task ProcessAsync<TSource, TResult>(
        TSource item,
        Func<TSource, Task<TResult>> taskSelector, Action<TSource, TResult> resultProcessor,
        SemaphoreSlim oneAtATime)
    {
        TResult result = await taskSelector(item);
        await oneAtATime.WaitAsync();
        try { resultProcessor(item, result); }
        finally { oneAtATime.Release(); }
    }
}

Usage:

var collection = new List<object>();

await collection.ForEachAsync(
    async obj => { ... loop code, containing the await operators and a return statement ... },
    (obj, result) => { ...action to perform with the result of the operation... });

Example:

var client = new MyAsyncServiceClient();
var requestUrls = new List<string>();
var responseList = new List<Response>();

await requestUrls.ForEachAsync(
    async requestUrl => {
        var response = await client.ExecuteRequest(requestUrl);
        return response;
    },
    (requestUrl, response) => { 
        responseList.Add(response);
    }
);

You can find here the original article, and here an useful post that explains the reasons behind the architectural choices.

Thanks Stephen!