Path manipulation library for .Net
We're on NuGet!
- Added support for
PosixPath.Resolve()
(thanks @asasine)
Paths are commonly used in programming, from opening files to storage directories. They're integral to any program, yet unlike their siblings URLs and URIs very few programming languages (with strong typing) have a strongly typed solution for storing and manipulating paths.
Instead, programmers are forced to store these paths as strings and use a host
of static methods to pinch and twist one path into another. In .Net, these are
found in the System.IO.Path
and System.IO.Directory
namespaces. Common
operations include combining paths (Path.Combine(path1, path2)
) and
extracting the filename (Path.GetFileName("C:\file.txt")
). Since these are
only valid for a particular subset of strings, I'm surprised that more
languages do not have objects corresponding to a path so that methods and
libraries can accept a "path object" and be confident that the input data
conforms to at least a rudimentary set of validation criteria (even if the
path itself doesn't exist on disk).
Despite being available in Ruby, Python, C++, and Java, I've never seen real-life code that uses any of these libraries, nor have I read a single blog post extolling their virtues (and if my experience dogfooding PathLib is any indication, there are many). The best thing about having a class dedicated to paths, in my opinion, is setting expectations for methods that use paths. Which of these more clearly explains what value to pass into the method?
public Database OpenDatabase(string dbLocation) {}
public Database OpenDatabase(Path dbLocation) {}
I can't speak for anyone else, but I'm certainly less tempted to pass
"http://localhost:4000"
to the latter than the former.
To give you a small taste of the power of PathLib, compare implementations of
this (real) scenario: list all files within the user's "myapp" directory and
copy all alphanumeric characters into a new file with an extra ".clean"
extension (so file.txt
becomes file.clean.txt
). I've avoided using var
to show how many "stringly typed"
objects are created in the non-PathLib version compared to using PathLib.
IPath appDir = new WindowsPath("~/myapp").ExpandUser();
foreach(IPath file in appDir.ListDir())
{
if(!file.IsFile()) continue;
string text = file.ReadAsText();
text = Regex.Replace(text, @"\W+", "");
IPath newFile = file.WithFilename(
file.Basename + ".clean" + file.Extension);
using(var output = new StreamWriter(newFile.Open(FileMode.Create)))
{
output.Write(text);
}
}
string userDir = Environment.GetFolderPath(
System.Environment.SpecialFolder.UserProfile); // Only in .Net 4.0
string appDir = Path.Combine(userDir, "myapp");
foreach(string file in Directory.EnumerateFiles(appDir))
{
string text = File.ReadAllText(file);
text = Regex.Replace(text, @"\W+", "");
string newFile = Path.Combine(appDir,
Path.GetFileNameWithoutExtension(file) +
".clean" +
Path.GetExtension(file));
using(var output = new StreamWriter(File.Open(newFile, FileMode.Create)))
{
output.Write(text);
}
}
The goal of PathLib is to extend the feature set of System.IO.Path
and bundle
it all into a strongly typed path object. It borrows some terminology from the
similarly named Python library mentioned above.
There are four main classes and two main interfaces in the library:
- IPurePath: A platform-agnostic interface for "pure paths", or those that do not touch the filesystem. All operations are guaranteed to be supported on any platform so, for instance, your application can create and use Windows-style paths on a Linux machine (perfect for remote applications or web apps that manipulate client paths on a server).
- IPath: A platform-agnostic interface for concrete paths. These can perform operations that touch the filesystem such as file/directory exists, resolving symbolic links, and reading the contents of files. All IPaths inherit the IPurePath interface which works as a form of multiple inheritance even though the language itself doesn't support it.
- PureWindowsPath: A pure path using Windows validation and styling rules. For example, absolute (non-UNC) paths begin with a drive letter and use backslashes as separators and comparisons are case insensitive.
- PurePosixPath: A pure path using POSIX validation and styling rules. POSIX-compliant systems include Linux, UNIX, and Mac OSX. These paths use forward slashes as component separators and have case sensitive comparisons among other differences.
- WindowsPath: A concrete path using Windows validation and styling rules. Additionally, this class has methods that touch the filesystem. Due to this, the class can only be used on Windows systems.
- PosixPath: A concrete path using POSIX validation and styling rules. Additionally, this class has methods that touch the filesystem. Due to this, the class can only be used on POSIX-compliant systems.
Since application and library developers usually want to be as cross-platform-compatible as possible, it doesn't make much sense to explicitly create instances of a "windows path" or "posix path". To that end, the library provides a couple of path factories that automatically detect the user's operating system and create the appropriate path on command.
This factory builds a pure path for the current operating system. You may also
provide a set of PurePathFactoryOptions
to the builder:
- AutoNormalizeCase: Always normalize a path's case in the created object (this has no effect on case sensitive platforms).
For convenience, a static, global PurePathFactory instance can be accessed from
the PurePath
class (no generic arguments).
This factory builds a concrete path for the current operating system. You may
also provide a set of PathFactoryOptions
to the builder:
- AutoNormalizeCase: Always normalize a path's case in the created object (this has no effect on case sensitive platforms).
- AutoExpandEnvironmentVariables: Always replace the value of environment variables present in the created object.
- AutoExpandUserDirectory: If the path begins with a tilde (
~
), replace it with the current user's directory. If theUserDirectory
property on the options class is non-null, use that path as the user directory.
For convenience, a static, global PathFactory instance can be accessed from
the Paths
class (no generic arguments).
One of the key behaviors of a new data type is the ability to serialize and deserialize to/from the various data storage and transport formats out there (namely, XML and JSON). Due to the magic of TypeConverters, that support is built in! Here's an example of deserializing a path in JSON.Net:
class Data
{
public IPath Path { get; set; }
}
public static void Main()
{
var json = @"{ ""path"": ""C:/users/me/file.txt"" }";
var data = JsonConvert.DeserializeObject<Data>(json);
Console.WriteLine(data.Path.Directory);
// C:\users\me
}
Note that I used an IPath interface rather than WindowsPath or PosixPath.
While each of those have their own TypeConverters, both IPath
and
IPurePath
have special converters that use a PathFactory to choose the
correct interface type. This makes it simple to support users on any platform.