scandir is a module which provides a generator version of os.listdir() that also exposes the extra file information the operating system returns when you iterate a directory. scandir also provides a much faster version of os.walk(), because it can use the extra file information exposed by the scandir() function.

scandir is intended to work on Python 2.6+ and Python 3.2+ (and it has been tested on those versions).

Note that this module is currently beta software -- working and used to some extent, and more than a proof-of-concept, but not battle-tested or extremely widely used. It's my hope that scandir will be included in Python 3.5 as os.scandir().

Python's built-in os.walk() is significantly slower than it needs to be, because -- in addition to calling listdir() on each directory -- it calls stat() on each file to determine whether the filename is a directory or not. But both FindFirstFile / FindNextFile on Windows and readdir on Linux/OS X already tell you whether the files returned are directories or not, so no further stat system calls are needed. In short, you can reduce the number of system calls from about 2N to N, where N is the total number of files and directories in the tree.

In practice, removing all those extra system calls makes os.walk() about 8-9 times as fast on Windows, and about 2-3 times as fast on Linux and Mac OS X. So we're not talking about micro-optimizations. See more benchmarks below.

Somewhat relatedly, many people have also asked for a version of os.listdir() that yields filenames as it iterates instead of returning them as one big list. This improves memory efficiency for iterating very large directories.

So as well as a faster walk(), scandir adds a new scandir() function. They're pretty easy to use, but see below for the full API docs.

I'd love for these incremental (but significant!) improvements to be added to the Python standard library. This scandir module was released to help test the concept and get it in shape for inclusion in the standard os module.

There are various third-party "path" and "walk directory" libraries available, but Python's os module isn't going away anytime soon. So we might as well speed it up and add small improvements where possible.

So I'd love it if you could help test scandir, report bugs, suggest improvements, or comment on the API.

Below are results showing how many times as fast scandir.walk() is than os.walk() on various systems, found by running benchmark.py with no arguments as well as with the -s argument (which totals the directory size).

System version          Python version  Speed ratio    With -s
--------------------------------------------------------------
Windows 7 64-bit        2.7.5 64-bit    7.5            14.2
Windows 7 64-bit SSD    2.7.6 64-bit    10.0           18.5
Windows 7 64-bit NFS    2.7.6 64-bit    23.2           46.4
Windows 7 64-bit        3.4.1 64-bit    TODO

CentOS 6.5 64-bit       2.7.6 64-bit    5.5            2.3
Ubuntu 12.04 32-bit     2.7.3 32-bit    4.3            2.2

Mac OS X 10.9.3         2.7.5 64-bit    5.3            2.1

All of the above tests were done using the version of scandir with the fast C scandir_helper() function.

Note that the gains are less than the above on smaller directories and greater on larger directories. This is why benchmark.py creates a test directory tree with a standardized size.

Another quick benchmark I've done (on Windows 7 64-bit) is running Eli Bendersky's pss source code searching tool across a fairly large code tree (4938 files, 598 dirs, 200 MB). Using pss out of the box with os.walk() on a not-found string takes 0.91 seconds. But after monkey-patching in scandir.walk() it takes only 0.34 seconds -- 2.7 times as fast.

The API for scandir.walk() is exactly the same as os.walk(), so just read the Python docs.

The scandir() function is the scandir module's main workhorse. It's defined as follows:

scandir(path='.', windows_wildcard='*.*') -> iterator of DirEntry objects

It yields a DirEntry for each file and directory in path. Like os.listdir(), . and .. are skipped, and the entries are yielded in system-dependent order. Each DirEntry object has the following attributes and methods:

• name: filename, relative to path (like that returned by os.listdir)
• is_dir(): like os.path.isdir(), but requires no OS calls on most systems (Linux, Windows, OS X)
• is_file(): like os.path.isfile(), but requires no OS calls on most systems (Linux, Windows, OS X)
• is_symlink(): like os.path.islink(), but requires no OS calls on most systems (Linux, Windows, OS X)
• lstat(): like os.lstat(), but requires no OS calls on Windows

Obviously windows_wildcard is only available on Windows. It allows Windows power users to pass a custom wildcard to FindFirstFile, which may avoid the need to use fnmatch on the resulting names.

Here's a good usage pattern for scandir. This is in fact almost exactly how the faster os.walk() implementation uses it:

dirs = []
nondirs = []
for entry in scandir(path):
    if entry.is_dir():
        dirs.append(entry)
    else:
        nondirs.append(entry)

• Thread I started on the python-ideas list about speeding up os.walk()
• Python Issue 11406, original proposal for scandir(), a generator without the dirent/stat info
• Further thread I started on python-dev that refined the scandir() API
• Question on StackOverflow about why os.walk() is slow and pointers to fix it
• Question on StackOverflow asking about iterating over a directory
• BetterWalk, my previous attempt at this, on which this code is based
• Info about Win32 reparse points / symbolic links

• Finish the C extension version (_scandir.c)
• Get scandir() included in the Python 3.5 standard library! :-)

Please send flames, comments, and questions about scandir to Ben Hoyt:

http://benhoyt.com/

File bug reports or feature requests at the GitHub project page:

https://github.com/benhoyt/scandir