C# Language
Task Parallel Library (TPL) Dataflow Constructs
TransformManyBlock
Example
(SelectMany, 1-m: The results of this mapping are “flattened”, just like LINQ’s SelectMany)
TransformManyBlock<TInput, TOutput> is very similar to TransformBlock<TInput, TOutput>.
The key difference is that whereas a TransformBlock<TInput, TOutput> produces one and only one output for each input, TransformManyBlock<TInput, TOutput> produces any number (zero or more) outputs for each input. As with ActionBlock and TransformBlock<TInput, TOutput>, this processing may be specified using delegates, both for synchronous and asynchronous processing.
A Func<TInput, IEnumerable> is used for synchronous, and a Func<TInput, Task<IEnumerable>> is used for asynchronous. As with both ActionBlock and TransformBlock<TInput, TOutput>, TransformManyBlock<TInput, TOutput> defaults to sequential processing, but may be configured otherwise.
The mapping delegate retuns a collection of items, which are inserted individually into the output buffer.
Asynchronous Web Crawler
var downloader = new TransformManyBlock<string, string>(async url =>
{
Console.WriteLine(“Downloading “ + url);
try
{
return ParseLinks(await DownloadContents(url));
}
catch{}
return Enumerable.Empty<string>();
});
downloader.LinkTo(downloader);
Expanding an Enumerable Into Its Constituent Elements
var expanded = new TransformManyBlock<T[], T>(array => array);
Filtering by going from 1 to 0 or 1 elements
public IPropagatorBlock<T> CreateFilteredBuffer<T>(Predicate<T> filter)
{
return new TransformManyBlock<T, T>(item =>
filter(item) ? new [] { item } : Enumerable.Empty<T>());
}
Introduction to TPL Dataflow by Stephen Toub
