A FileSystem Abstraction System for Go

Afero is an filesystem framework providing a simple, uniform and universal API interacting with any filesystem, as an abstraction layer providing interfaces, types and methods. Afero has an exceptionally clean interface and simple design without needless constructors or initialization methods.

Afero is also a library providing a base set of interoperable backend filesystems that make it easy to work with afero while retaining all the power and benefit of the os and ioutil packages.

Afero provides significant improvements over using the os package alone, most notably the ability to create mock and testing filesystems without relying on the disk.

It is suitable for use in a any situation where you would consider using the OS package as it provides an additional abstraction that makes it easy to use a memory backed file system during testing. It also adds support for the http filesystem for full interoperability.

• A single consistent API for accessing a variety of filesystems
• Interoperation between a variety of file system types
• A set of interfaces to encourage and enforce interoperability between backends
• An atomic cross platform memory backed file system
• Support for compositional file systems by joining various different file systems (see httpFs)

Afero is easy to use and easier to adopt.

A few different ways you could use Afero:

• Use the interfaces alone to define you own file system.
• Wrap for the OS packages.
• Define different filesystems for different parts of your application.
• Use Afero for mock filesystems while testing

First use go get to install the latest version of the library.

$ go get github.com/spf13/afero

Next include Afero in your application.

import "github.com/spf13/afero"

First define a package variable and set it to a pointer to a filesystem.

var AppFs afero.Fs = &afero.MemMapFs{}

or

var AppFs afero.Fs = &afero.OsFs{}

It is important to note that if you repeat the composite literal you will be using a completely new and isolated filesystem. In the case of OsFs it will still use the same underlying filesystem but will reduce the ability to drop in other filesystems as desired.

Throughout your application use any function and method like you normally would.

So if my application before had:

os.Open('/tmp/foo')

We would replace it with a call to AppFs.Open('/tmp/foo').

AppFs being the variable we defined above.

File System Methods Available:

Chmod(name string, mode os.FileMode) : error
Chtimes(name string, atime time.Time, mtime time.Time) : error
Create(name string) : File, error
Mkdir(name string, perm os.FileMode) : error
MkdirAll(path string, perm os.FileMode) : error
Name() : string
Open(name string) : File, error
OpenFile(name string, flag int, perm os.FileMode) : File, error
Remove(name string) : error
RemoveAll(path string) : error
Rename(oldname, newname string) : error
Stat(name string) : os.FileInfo, error

File Interfaces and Methods Available:

io.Closer
io.Reader
io.ReaderAt
io.Seeker
io.Writer
io.WriterAt

Name() : string
Readdir(count int) : []os.FileInfo, error
Readdirnames(n int) : []string, error
Stat() : os.FileInfo, error
Sync() : error
Truncate(size int64) : error
WriteString(s string) : ret int, err error

In some applications it may make sense to define a new package that simply exports the file system variable for easy access from anywhere.

There is a large benefit to using a mock filesystem for testing. It has a completely blank state every time it is initialized and can be easily reproducible regardless of OS. You could create files to your heart’s content and the file access would be fast while also saving you from all the annoying issues with deleting temporary files, Windows file locking, etc. The MemMapFs backend is perfect for testing.

• Much faster than performing I/O operations on disk
• Avoid security issues and permissions
• Far more control. 'rm -rf /' with confidence
• Test setup is far more easier to do
• No test cleanup needed

One way to accomplish this is to define a variable as mentioned above. In your application this will be set to &afero.OsFs{} during testing you can set it to &afero.MemMapFs{}.

It wouldn't be uncommon to have each test initialize a blank slate memory backend. To do this I would define my appFS = &afero.OsFs{} somewhere appropriate in my application code. This approach ensures that Tests are order independent, with no test relying on the state left by an earlier test.

Then in my tests I would initialize a new MemMapFs for each test:

func TestExist(t *testing.T) {
    appFS = &afero.MemMapFs{}
    // create test files and directories
    appFS.MkdirAll("src/a", 0755))
    appFS.WriteFile("src/a/b", []byte("file b"), 0644)
    appFS.WriteFile("src/c", []byte("file c"), 0644)
    testExistence("src/c", true, t)
}

func testExistence(name string, e bool, t *testing.T) {
    _, err := appFS.Stat(name)
    if os.IsNotExist(err) {
        if e {
            t.Errorf("file \"%s\" does not exist.\n", name)
        }
    } else if err != nil {
        panic(err)
    } else {
        if !e {
            t.Errorf("file \"%s\" exists.\n", name)
        }
    }
}

Afero provides an http compatible backend which can wrap any of the existing backends.

The Http package requires a slightly specific version of Open which returns an http.File type.

Afero provides an httpFs file system which satisfies this requirement. Any Afero FileSystem can be used as an httpFs.

httpFs := &afero.HttpFs{SourceFs: <ExistingFS>}
fileserver := http.FileServer(httpFs.Dir(<PATH>)))
http.Handle("/", fileserver)

The first is simply a wrapper around the native OS calls. This makes it very easy to use as all of the calls are the same as the existing OS calls. It also makes it trivial to have your code use the OS during operation and a mock filesystem during testing or as needed.

Afero also provides a fully atomic memory backed filesystem perfect for use in mocking and to speed up unnecessary disk io when persistence isn’t necessary. It is fully concurrent and will work within go routines safely.

As part of MemMapFs, Afero also provides an atomic, fully concurrent memory backed file implementation. This can be used in other memory backed file systems with ease. Plans are to add a radix tree memory stored file system using InMemoryFile.

The following is a short list of possible backends we hope someone will implement:

• SSH/SCP
• ZIP
• TAR
• S3
• Mem buffering to disk/network
• BasePath (where all paths are relative to a fixed basepath)

Initially this project was called fs. Unfortunately as I used it, the name proved confusing, there were too many fs’. In looking for alternatives I looked up the word 'abstract' in a variety of different languages. Afero is the Greek word for abstract and it seemed to be a fitting name for the project. It also means ‘to do’ or ‘thing’ in Esperanto which is also fitting.

• 0.9.0 2015.11.05
  • New Walk function similar to filepath.Walk
  • MemMapFs.OpenFile handles O_CREATE, O_APPEND, O_TRUNC
  • MemMapFs.Remove now really deletes the file
  • InMemoryFile.Readdir and Readdirnames work correctly
  • InMemoryFile functions lock it for concurrent access
  • Test suite improvements
• 0.8.0 2014.10.28
  • First public version
  • Interfaces feel ready for people to build using
  • Interfaces satisfy all known uses
  • MemMapFs passes the majority of the OS test suite
  • OsFs passes the majority of the OS test suite

1. Fork it
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create new Pull Request

Names in no particular order:

• spf13
• jaqx0r
• mbertschler

Afero is released under the Apache 2.0 license. See LICENSE.txt