The following describes some core concepts when working with PyFilesystem. If you are skimming this documentation, pay particular attention to the first section on paths.


With the possible exception of the constructor, all paths in a filesystem are PyFilesystem paths, which have the following properties:

  • Paths are str type in Python3, and unicode in Python2
  • Path components are separated by a forward slash (/)
  • Paths beginning with a / are absolute
  • Paths not beginning with a forward slash are relative
  • A single dot (.) means ‘current directory’
  • A double dot (..) means ‘previous directory’

Note that paths used by the FS interface will use this format, but the constructor may not. Notably the OSFS constructor which requires an OS path – the format of which is platform-dependent.


There are many helpful functions for working with paths in the path module.

PyFilesystem paths are platform-independent, and will be automatically converted to the format expected by your operating system – so you won’t need to make any modifications to your filesystem code to make it run on other platforms.

System Paths

Not all Python modules can use file-like objects, especially those which interface with C libraries. For these situations you will need to retrieve the system path. You can do this with the getsyspath() method which converts a valid path in the context of the FS object to an absolute path that would be understood by your OS.

For example:

>>> from fs.osfs import OSFS
>>> home_fs = OSFS('~/')
>>> home_fs.getsyspath('test.txt')

Not all filesystems map to a system path (for example, files in a MemoryFS will only ever exists in memory).

If you call getsyspath on a filesystem which doesn’t map to a system path, it will raise a NoSysPath exception. If you prefer a look before you leap approach, you can check if a resource has a system path by calling hassyspath()


FS objects are not permitted to work with any files outside of their root. If you attempt to open a file or directory outside the filesystem instance (with a backref such as "../foo.txt"), a IllegalBackReference exception will be thrown. This ensures that any code using a FS object won’t be able to read or modify anything you didn’t intend it to, thus limiting the scope of any bugs.

Unlike your OS, there is no concept of a current working directory in PyFilesystem. If you want to work with a sub-directory of an FS object, you can use the opendir() method which returns another FS object representing the contents of that sub-directory.

For example, consider the following directory structure. The directory foo contains two sub-directories; bar and baz:

  |  |--readme.txt
  |  `--photo.jpg

We can open the foo directory with the following code:

from fs.osfs import OSFS
foo_fs = OSFS('foo')

The foo_fs object can work with any of the contents of bar and baz, which may not be desirable if we are passing foo_fs to a function that has the potential to delete files. Fortunately we can isolate a single sub-directory with the opendir() method:

bar_fs = foo_fs.opendir('bar')

This creates a completely new FS object that represents everything in the foo/bar directory. The root directory of bar_fs has been re- position, so that from bar_fs’s point of view, the readme.txt and photo.jpg files are in the root:



This sandboxing only works if your code uses the filesystem interface exclusively. It won’t prevent code using standard OS level file manipulation.


PyFilesystem converts errors in to a common exception hierarchy. This ensures that error handling code can be written once, regardless of the filesystem being used. See errors for details.