{"id":48604,"date":"2023-11-06T10:05:00","date_gmt":"2023-11-06T18:05:00","guid":{"rendered":"https:\/\/devblogs.microsoft.com\/dotnet\/?p=48604"},"modified":"2023-11-06T19:24:11","modified_gmt":"2023-11-07T03:24:11","slug":"the-convenience-of-system-io","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/dotnet\/the-convenience-of-system-io\/","title":{"rendered":"The convenience of System.IO"},"content":{"rendered":"

Reading and writing files is very common, just like other forms of I\/O. File APIs are needed for reading application configuration, caching content, and loading data (from disk) into memory to do some computation like (today’s topic) word counting. File<\/code>, FileInfo<\/code>, FileStream<\/code>, and related types do a lot of the heavy lifting for .NET developers needing to access files. In this post, we\u2019re going to look at the convenience and performance of reading text files with System.IO<\/a>, with some help from System.Text<\/a> APIs.<\/p>\n

We recently kicked off a series on the Convenience of .NET<\/a> that describes our approach for providing convenient solutions to common tasks. The convenience of System.Text.Json<\/a> is another post in the series, about reading and writing JSON documents. Why .NET?<\/a> describes the architectural choices that enable the solutions covered in these posts.<\/p>\n

This post analyzes the convenience and performance of file I\/O and text APIs being put to the task of counting lines, words, and bytes in a large novel. The results show that the high-level APIs are straightforward to use and deliver great performance, while the lower-level APIs require a little more effort and deliver excellent results. You’ll also see how native AOT<\/a> shifts .NET into a new class of performance for application startup.<\/p>\n

The APIs<\/h2>\n

The following File<\/code> APIs (with their companions) are used in the benchmarks.<\/p>\n

    \n
  1. File.OpenHandle<\/code><\/a> with RandomAccess.Read<\/code><\/a><\/li>\n
  2. File.Open<\/code><\/a> with FileStream.Read<\/code><\/a><\/li>\n
  3. File.OpenText<\/code><\/a> with StreamReader.Read<\/code><\/a> and StreamReader.ReadLine<\/code><\/a><\/li>\n
  4. File.ReadLines<\/code><\/a> with IEnumerable<string><\/code><\/a><\/li>\n
  5. File.ReadAllLines<\/code><\/a> with string[]<\/code><\/a><\/li>\n<\/ol>\n

    The APIs are listed from highest-control to most convenient. It’s OK if they are new to you. It should still be an interesting read.<\/p>\n

    The lower-level benchmarks rely on the following System.Text<\/code> types:<\/p>\n