Wednesday, April 24, 2013

Sasa.IO.FilePath - Easy and Safe Path Manipulations

This is the fifteenth post in my ongoing series covering the abstractions in Sasa. Previous posts:

One persistent difficulty in dealing with IO in .NET is path handling. .NET exposes a platform's directory separator characters, but really this sort of thing should be automated. Furthermore, paths are considered simple strings and so concatenating fragments could leave you with a path string that goes up and down directories with no clear final result, ie. resolving the final path string is left to the OS.

This means that you can't easily reason about the constructed paths without consulting the OS, which is a relatively expensive operation. Furthermore, Path.Combine has a number of corner cases that make constructing paths non-compositional, requiring numerous argument validations to ensure a correct result.

Enter Sasa.IO.FilePath. It's a simple struct that encapsulates an underlying path string, so FilePath operations are just as efficient as your current path handling. FilePath fully resolves directory change operations where possible, so the final path string designates the path the OS will actually look up. Null or empty strings are considered references to the current directory, ie. ".". There is also a "jail" operation which ensures that a provided path cannot escape a particular sub-directory, just like the OS-level chroot jail.

I first posted about this abstraction back in 2009, but the name and implementation has changed a little since then.

Sasa.IO.FilePath Constructor

The constructor takes an arbitrary path string, resolves all the inner directory change operations, and returns a final path:

var foo = "foo/../..";
var path1 = new FilePath(foo);
var path2 = new FilePath("bar/" + foo);
Console.WriteLine(path1);
Console.WriteLine(path2);
Console.WriteLine("root" / path2); // FilePath composition using /
// output:
// ..
// .
// root

Sasa.IO.FilePath.Combine

The Combine method exposes the same semantics as System.IO.Path.Combine, but on FilePath instances:

var foo = new FilePath("foo/../..");
var bar = new FilePath("/bar");
Console.WriteLine(FilePath.Combine(foo, bar));
Console.WriteLine(FilePath.Combine(bar, foo));
// output:
// ..\bar
// .

Sasa.IO.FilePath.IsParentOf

The IsParentOf method compares two paths to see if one is a subset of the other:

var foobar = new FilePath("foo/bar");
var bar = new FilePath("bar");
var foo = new FilePath("foo");
Console.WriteLine(bar.IsParentOf(foobar));
Console.WriteLine(foo.IsParentOf(foobar));
Console.WriteLine(foobar.IsParentOf(foo));
// output:
// false
// true
// false

Sasa.IO.FilePath.Jail

The Jail methods provide a chroot-jail operation on file paths, to ensure a provided path string cannot escape a certain sub-directory:

var escape = new FilePath("../..");
var bar = new FilePath("/bar");
Console.WriteLine(FilePath.Jail(bar, escape));
// output:
// bar

Sasa.IO.FilePath.Resolve

The real workhorse behind FilePath, Sasa.IO.FilePath.Resolve resolves all path change operations in a string and returns a simplified path string. In case you want to stick with raw string paths, you can use this method to perform path simplification:

var path1 = "foo/../..";
var path2 = "bar/" + path1;
Console.WriteLine(FilePath.Resolve(path1));
Console.WriteLine(FilePath.Resolve(path2));
Console.WriteLine(FilePath.Resolve("root/" + path2));
// output:
// ..
// 
// root

Sasa.IO.FilePath./ Operator

The division operator on file paths is a convenient shorthand for composing paths, and a shorthand for FilePath.Combine:

var foo = "foo/../..";
var path1 = new FilePath(foo);
var path2 = "bar" / path1;
Console.WriteLine(path1);
Console.WriteLine(path2);
Console.WriteLine("root" / path2);
// output:
// ..
// .
// root

Sasa.IO.FilePath's Interfaces

  • IEnumerable<string>: the sequence of path components making up a full path.
  • IEquatable<FilePath>: compare two paths for equality
  • IComparable<FilePath>: order two paths alphanumerically

No comments: