How to Async Files.ReadAllLines and await for results?

I have the following code,

    private void button1_Click(object sender, RoutedEventArgs e)
{
button1.IsEnabled = false;


var s = File.ReadAllLines("Words.txt").ToList(); // my WPF app hangs here
// do something with s


button1.IsEnabled = true;
}

Words.txt has a ton of words which i read into the s variable, I am trying to make use of async and await keywords in C# 5 using Async CTP Library so the WPF app doesn't hang. So far I have the following code,

    private async void button1_Click(object sender, RoutedEventArgs e)
{
button1.IsEnabled = false;


Task<string[]> ws = Task.Factory.FromAsync<string[]>(
// What do i have here? there are so many overloads
); // is this the right way to do?


var s = await File.ReadAllLines("Words.txt").ToList();  // what more do i do here apart from having the await keyword?
// do something with s


button1.IsEnabled = true;
}

The goal is to read the file in async rather than sync, to avoid freezing of WPF app.

Any help is appreciated, Thanks!

68466 次浏览

UPDATE: Async versions of File.ReadAll[Lines|Bytes|Text], File.AppendAll[Lines|Text] and File.WriteAll[Lines|Bytes|Text] have now been merged into .NET Core and shipped with .NET Core 2.0. They are also included in .NET Standard 2.1.

Using Task.Run, which essentially is a wrapper for Task.Factory.StartNew, for asynchronous wrappers is a code smell.

If you don't want to waste a CPU thread by using a blocking function, you should await a truly asynchronous IO method, StreamReader.ReadToEndAsync, like this:

using (var reader = File.OpenText("Words.txt"))
{
var fileText = await reader.ReadToEndAsync();
// Do something with fileText...
}

This will get the whole file as a string instead of a List<string>. If you need lines instead, you could easily split the string afterwards, like this:

using (var reader = File.OpenText("Words.txt"))
{
var fileText = await reader.ReadToEndAsync();
return fileText.Split(new[] { Environment.NewLine }, StringSplitOptions.None);
}

EDIT: Here are some methods to achieve the same code as File.ReadAllLines, but in a truly asynchronous manner. The code is based on the implementation of File.ReadAllLines itself:

using System.Collections.Generic;
using System.IO;
using System.Text;
using System.Threading.Tasks;


public static class FileEx
{
/// <summary>
/// This is the same default buffer size as
/// <see cref="StreamReader"/> and <see cref="FileStream"/>.
/// </summary>
private const int DefaultBufferSize = 4096;


/// <summary>
/// Indicates that
/// 1. The file is to be used for asynchronous reading.
/// 2. The file is to be accessed sequentially from beginning to end.
/// </summary>
private const FileOptions DefaultOptions = FileOptions.Asynchronous | FileOptions.SequentialScan;


public static Task<string[]> ReadAllLinesAsync(string path)
{
return ReadAllLinesAsync(path, Encoding.UTF8);
}


public static async Task<string[]> ReadAllLinesAsync(string path, Encoding encoding)
{
var lines = new List<string>();


// Open the FileStream with the same FileMode, FileAccess
// and FileShare as a call to File.OpenText would've done.
using (var stream = new FileStream(path, FileMode.Open, FileAccess.Read, FileShare.Read, DefaultBufferSize, DefaultOptions))
using (var reader = new StreamReader(stream, encoding))
{
string line;
while ((line = await reader.ReadLineAsync()) != null)
{
lines.Add(line);
}
}


return lines.ToArray();
}
}

Try this:

private async void button1_Click(object sender, RoutedEventArgs e)
{
button1.IsEnabled = false;
try
{
var s = await Task.Run(() => File.ReadAllLines("Words.txt").ToList());
// do something with s
}
finally
{
button1.IsEnabled = true;
}
}

Edit:

You don't need the try-finally for this to work. It's really only the one line that you need to change. To explain how it works: This spawns another thread (actually gets one from the thread pool) and gets that thread to read the file. When the file is finished reading then the remainder of the button1_Click method is called (from the GUI thread) with the result. Note that this is probably not the most efficient solution, but it is probably the simplest change to your code which doesn't block the the GUI.

I also encountered a problem described in your question. I've solved it just simplier that in previous answers:

string[] values;
StorageFolder folder = ApplicationData.Current.LocalFolder; // Put your location here.
IList<string> lines = await FileIO.ReadLinesAsync(await folder.GetFileAsync("Words.txt"););
lines.CopyTo(values, 0);
private async Task<string> readFile(string sourceFilePath)
{
using (var fileStream = new FileStream(sourceFilePath, FileMode.Open, FileAccess.ReadWrite, FileShare.ReadWrite))
{
using (var streamReader = new StreamReader(fileStream))
{
string data = await streamReader.ReadToEndAsync().ConfigureAwait(false);
streamReader.Close();
fileStream.Close();
return data;
}


}
}

Here are the helper methods I've created for a NetStandart 2.0 class library, that was used both in NetCore 3.1 and NetFramework 4.7.2 projects.

These implementations have matched exactly the names and signatures of the net core 3.1 / net standard 2.1 File class methods, so you only need to put them in any public class. (FileHelper for example...):

Also, this should be most efficient and similar to the source code of .net implementation.

    private const int DefaultBufferSize = 4096;
// File accessed asynchronous reading and sequentially from beginning to end.
private const FileOptions DefaultOptions = FileOptions.Asynchronous | FileOptions.SequentialScan;


public static async Task WriteAllTextAsync(string filePath, string text)
{
byte[] encodedText = Encoding.Unicode.GetBytes(text);


using FileStream sourceStream = new FileStream(filePath, FileMode.Append, FileAccess.Write, FileShare.None,
DefaultBufferSize, true);
await sourceStream.WriteAsync(encodedText, 0, encodedText.Length);
}


public static async Task<IEnumerable<string>> ReadAllLinesAsync(string filePath)
{
var lines = new List<string>();


using var sourceStream = new FileStream(filePath, FileMode.Open, FileAccess.Read, FileShare.Read,
DefaultBufferSize, DefaultOptions);
using var reader = new StreamReader(sourceStream, Encoding.Unicode);
string line;
while ((line = await reader.ReadLineAsync()) != null) lines.Add(line);


return lines;
}


public static async Task<string> ReadAllTextAsync(string filePath)
{
using var sourceStream = new FileStream(filePath, FileMode.Open, FileAccess.Read, FileShare.Read,
DefaultBufferSize, DefaultOptions);
using var reader = new StreamReader(sourceStream, Encoding.Unicode);
return await reader.ReadToEndAsync();
}

Edit

Apparently that the StreamReader "async" methods block the current thread for a considerable amount of time before returning an incomplete Task.

(Even the netcore 3.1 File.ReadAllLinesAsyn,File.ReadAllTextAsync currently aren't seems to be fully async. as you can check in source code, they are based on the StreamReader "async" methods).

So, I'm sharing an implementation that seems like the most efficient way currently. \

It's better than options like to run the sync methods in Task.Run(()=>File.ReadAllLines(...)), since its a very bad practice to wrap your sync code with Task.Run and expect this to be full async flow. \ Actually, it breaks the internal queues mechanism of the real asynchronous dotnet structure.

public static async Task<string> ReadAllTextAsync(string filePath)
{
using (var sourceStream = new FileStream(filePath, FileMode.Open, FileAccess.Read, FileShare.Read,
DefaultBufferSize, DefaultOptions))
{
var sb = new StringBuilder();
var buffer = new byte[0x1000];
var numRead = 0;


while ((numRead = await sourceStream.ReadAsync(buffer, 0, buffer.Length)) != 0)
sb.Append(Encoding.Unicode.GetString(buffer, 0, numRead));
return sb.ToString();
}
}

Testing Time

Here is my test and its output that displays clearly that the actual run is async:

        var stopwatch = Stopwatch.StartNew();
var fileTask = FileHelper.ReadAllTextAsync("48MB_file.txt");
var duration1 = stopwatch.ElapsedMilliseconds;
var isCompleted = fileTask.IsCompleted;


stopwatch.Restart();
await fileTask;
var duration2 = stopwatch.ElapsedMilliseconds;


Console.WriteLine($"Creation took: {duration1:#,0} ms, Task.IsCompleted: {isCompleted}");
Console.WriteLine($"Calling await took:  {duration2:#,0} ms, Task.IsCompleted: {fileTask.IsCompleted}");

Creation took: 43 ms, Task.IsCompleted: False
Calling await took: 508 ms, Task.IsCompleted: True

You can find more in the comments, and in this question: File.ReadAllLinesAsync() blocks the UI thread