tags:

views:

1283

answers:

2

I don't know if it's legit at StackOverflow to post your own answer to a question, but I saw nobody had asked this already. I went looking for a C# Glob and didn't find one, so I wrote one that others might find useful.

+6  A: 
    /// <summary>
    /// return a list of files that matches some wildcard pattern, e.g. 
    /// C:\p4\software\dotnet\tools\*\*.sln to get all tool solution files
    /// </summary>
    /// <param name="glob">pattern to match</param>
    /// <returns>all matching paths</returns>
    public static IEnumerable<string> Glob(string glob)
    {
        foreach (string path in Glob(PathHead(glob) + DirSep, PathTail(glob)))
            yield return path;
    }

    /// <summary>
    /// uses 'head' and 'tail' -- 'head' has already been pattern-expanded
    /// and 'tail' has not.
    /// </summary>
    /// <param name="head">wildcard-expanded</param>
    /// <param name="tail">not yet wildcard-expanded</param>
    /// <returns></returns>
    public static IEnumerable<string> Glob(string head, string tail)
    {
        if (PathTail(tail) == tail)
            foreach (string path in Directory.GetFiles(head, tail).OrderBy(s => s))
                yield return path;
        else
            foreach (string dir in Directory.GetDirectories(head, PathHead(tail)).OrderBy(s => s))
                foreach (string path in Glob(Path.Combine(head, dir), PathTail(tail)))
                    yield return path;
    }

    /// <summary>
    /// shortcut
    /// </summary>
    static char DirSep = Path.DirectorySeparatorChar;

    /// <summary>
    /// return the first element of a file path
    /// </summary>
    /// <param name="path">file path</param>
    /// <returns>first logical unit</returns>
    static string PathHead(string path)
    {
        // handle case of \\share\vol\foo\bar -- return \\share\vol as 'head'
        // because the dir stuff won't let you interrogate a server for its share list
        // FIXME check behavior on Linux to see if this blows up -- I don't think so
        if (path.StartsWith("" + DirSep + DirSep))
            return path.Substring(0, 2) + path.Substring(2).Split(DirSep)[0] + DirSep + path.Substring(2).Split(DirSep)[1];

        return path.Split(DirSep)[0];
    }

    /// <summary>
    /// return everything but the first element of a file path
    /// e.g. PathTail("C:\TEMP\foo.txt") = "TEMP\foo.txt"
    /// </summary>
    /// <param name="path">file path</param>
    /// <returns>all but the first logical unit</returns>
    static string PathTail(string path)
    {
        if (!path.Contains(DirSep))
            return path;

        return path.Substring(1 + PathHead(path).Length);
    }
Mark Maxham
Bug? I had to replace "Path.Combine(head, dir)" with "dir" since Directory.GetDirectories already returns the full path. This caused a bug with paths like "..\SomeDir\*.dll" since "..\" were duplicated by Combine
total
A: 

I stumbled upon the source to iron ruby that contains a pretty neat Glob class. It's fairly easy extract it from the related code.

http://github.com/ironruby/ironruby/blob/master/Merlin/Main/Languages/Ruby/Ruby/Builtins/Glob.cs

Andreas