Ruby’s Pathname API
Ruby’s standard library is a treasure trove of useful classes, and the Pathname
class is certainly one of them. Introduced in Ruby 1.9, Pathname
represents a path on the file system, providing convenient access to functionality that is otherwise scattered across a handful of other classes like File
, FileTest
, and Dir
.
Now that the life of Ruby 1.8 has officially come to an end, and 1.9 is nearing its fifth birthday, it’s high time to drop the legacy support and embrace the elegance and ease of use Pathname
has to offer.
To show why Pathname
was a much needed addition, let’s consider the more traditional way of manipulating path information. This is how we did things in the 1.8 days:
CONFIG_PATH = File.join(File.dirname(__FILE__), '..', 'config.yml')
Notice the use of class methods, a rare sight in a language where everything is an object. But what’s the alternative? The object-oriented thing-to-do would be to move the functionality closer to the data on which it operates. In this case, we’re simply manipulating strings, but it doesn’t really make sense to add join
and dirname
to String
. A string is a far too generic a data type for such specific functionality.
Enter Pathname
:
CONFIG_PATH = Pathname.new(__FILE__).dirname.join('..', 'config.yml')
By wrapping the string in a Pathname
object, we get a clean API for manipulating the path. This code already looks a lot more like idiomatic Ruby. An important difference from Ruby’s vanilla strings is that Pathname
instances are immutable. There are no methods that change the Pathname
instance itself. Instead, all methods like dirname
, join
or basename
return a new Pathname
instance, leaving the existing instance unchanged.
The Pathname API
Methods on Pathname
fall roughly into one of two categories: Either they simply manipulate the path string without accessing the file system, or they call out to the operating system to perform checks and operations on the file or directory itself. For the full list of eighty-something methods, you can check out the official Pathname API documentation. It’s worth noting that the docs are also available from a terminal through ri Pathname
, or in irb with help 'Pathname'
.
Let’s look at some methods you might find useful.
File Type and Permission Checks
pn = Pathname.new('/usr/bin/ruby')
pn.file? # => true
pn.directory? # => false
pn.absolute? # => true
pn.relative? # => false
pn.executable? # => true
pn.readable? # => true
pn.writable? # => false
pn.root? # => false
Most of these correspond with methods on FileTest
File System Navigation and Querying
pn = Pathname.getwd # current working directory
pn.children # => [ #<Pathname...>, ... ]
pn.each_child {|ch| ... }
pn = pn.parent
Pathname.glob('**/*.rb') # like Dir['**/*.rb'] but returns Pathnames
Pathname Manipulation
pn = Pathname.new('lib/mylib/awesome.rb')
pn.dirname # => #<Pathname:lib/mylib>
pn.basename # => #<Pathname:awesome.rb>
pn.extname # => ".rb"
pn.expand_path('/home/arne') # => #<Pathname:/home/arne/lib/mylib/awesome.rb>
Working with the Actual File
pn = Pathname.new(ENV['HOME']).join('.bashrc')
pn.size
pn.read
pn.open {|io| ... }
pn.each_line {|line| ... }
pn.rename('/tmp/foo')
pn.delete
Pathname()
and to_path
There are two ways to create Pathname
instances: through explicit construction as above, or through the Pathname()
conversion method. While capitalized identifiers are typically reserved for classes and constants, Ruby has a number of built-in methods for converting arbitrary values to a specific type, named after the classes they convert to.
Array(1) # => [1]
Array(nil) # => []
Array([1,2]) # => [1,2]
String(7) # => "7"
URI("http://example.com") # => #<URI::HTTP:0x1c0e URL:http://example.com>
These go hand in hand with a number of conversion hooks that objects can implement, like to_ary
or to_str
. Likewise, objects can implement to_path
to indicate how they can be converted to a pathname. The pathname constructor is implemented in C for efficiency, but it would roughly translate to Ruby like this:
class Pathname
def initialize(path)
path = path.to_path if path.respond_to? :to_path
@path = String(path)
end
end
Say you are writing some scripts to help you manage your music collection. An AudioFile
might look like this:
class AudioFile
def initialize(path)
@path = path
end
def to_path
@path
end
# ...
end
Many built-in methods will wrap their file name arguments in a call to Pathname()
, so you can pass in your AudioFile
instance just the same.
af = AudioFile.new('...')
File.open(af) do |io|
#...
end
When writing library code, reuse this pattern for maximum flexibility:
def read_configuration(config_file)
parse_configuration(Pathname(config_file).read)
end
Now the caller of this code can pass in as config_file
- a string
- a Pathname
- an object that responds to
to_path
- any object that can be converted with
String()
Building a Music Collection Browser
To show how powerful and intuitive working with Pathname
is, we’ll write a little API for browsing our music collection. The files are organized with songs grouped in directories per album, which are again grouped per artist. For example: ~/Music/Arsenal/Oyebo Soul/09_How Come.flacc
.
We’re aiming for an easy to use API that looks like this:
mc = MusicCollection.new('~/Music') # => #<MusicCollection 'Music'>
mc.artists.first # => #<Artist 'Arsenal'>
mc.artists.first.albums.first # => #<Album 'Oyebo Soul'>
mc.songs.count # => 120
mc.songs.first.path # => #<Pathname: ...>
mc.songs.first.play
The Song
class is essentially our AudioFile
class from above:
class Song
attr_reader :path
alias to_path path
def initialize(path)
@path = Pathname(path).expand_path
end
def name
String(path.basename(path.extname))
end
def play(options = {})
exec(options.fetch(:player, 'mplayer'), String(path))
# OS X users can use player: '/usr/bin/afplay'
end
end
We immediately call expand_path
to normalize our input. This will expand ~
into the user’s home directory, and will resolve any relative file names (like .
and ..
). This way, we are sure to have an absolute file name.
Now we need to add classes for Album
, Artist
, and, finally, the MusicCollection
itself. Since all of these are wrapper classes around a Pathname
, we can pull the common plumbing into a base class.
class PathnameWrapper
attr_reader :path
alias to_path path
def initialize(path)
@path = Pathname(path).expand_path
end
def name
# ...
end
def to_s
# ...
end
end
class Song < PathnameWrapper
def play
# ...
end
end
Now we can add an Album
and provide navigation methods back and forth. We will also add files
and directories
helpers to our base class, along with a to_proc
class method. When passing in an object as a block with &
, Ruby will use to_proc
to find an implementation for the block. This way we get shorthand for typecasting all elements of a collection.
class PathnameWrapper
# ...
def directories
path.children.select(&:directory?)
end
def files
path.children.select(&:file?)
end
def self.to_proc
->(path) { new(path) }
end
end
class Album < PathnameWrapper
EXTENSIONS = %w[.flacc .m4a .mp3 .ogg .wav]
def songs
files.select do |file|
EXTENSIONS.any? {|ext| ext == file.extname }
end.map(&Song)
end
end
class Song < PathnameWrapper
# ...
def album
Album.new(path.dirname)
end
end
To wrap things up we introduce Artist
and MusicCollection
.
class MusicCollection < PathnameWrapper
def artists
directories.map(&Artist)
end
def albums
artists.flat_map(&:albums)
end
def songs
albums.flat_map(&:songs)
end
end
class Artist < PathnameWrapper
def albums
directories.map(&Album)
end
end
class Album < PathnameWrapper
# ...
def artist
Artist.new(path.dirname)
end
end
class Song < PathnameWrapper
# ...
def artist
album.artist
end
end
Now we can browse back and forth between songs, albums, and artists.
This concludes the basics of our MusicCollection
API, as well as our introduction to the Pathname
class. Having a good grasp on Ruby’s standard library can help to make your code more elegant, correct, and concise. Here we’ve zoomed in on a single small component, highlighting its many uses. Go forward and use Pathname
wisely.