[python] Python glob multiple filetypes

Is there a better way to use glob.glob in python to get a list of multiple file types such as .txt, .mdown, and .markdown? Right now I have something like this:

projectFiles1 = glob.glob( os.path.join(projectDir, '*.txt') )
projectFiles2 = glob.glob( os.path.join(projectDir, '*.mdown') )
projectFiles3 = glob.glob( os.path.join(projectDir, '*.markdown') )

One glob, many extensions... but imperfect solution (might match other files).

filetypes = ['tif', 'jpg']

filetypes = zip(*[list(ft) for ft in filetypes])
filetypes = ["".join(ch) for ch in filetypes]
filetypes = ["[%s]" % ch for ch in filetypes]
filetypes = "".join(filetypes) + "*"
print(filetypes)
# => [tj][ip][fg]*

glob.glob("/path/to/*.%s" % filetypes)

By the results I've obtained from empirical tests, it turned out that glob.glob isn't the better way to filter out files by their extensions. Some of the reason are:

• The globbing "language" does not allows perfect specification of multiple extension.
• The former point results in obtaining incorrect results depending on file extensions.
• The globbing method is empirically proven to be slower than most other methods.
• Even if it's strange even other filesystems objects can have "extensions", folders too.

I've tested (for correcteness and efficiency in time) the following 4 different methods to filter out files by extensions and puts them in a list:

from glob import glob, iglob
from re import compile, findall
from os import walk


def glob_with_storage(args):

    elements = ''.join([f'[{i}]' for i in args.extensions])
    globs = f'{args.target}/**/*{elements}'
    results = glob(globs, recursive=True)

    return results


def glob_with_iteration(args):

    elements = ''.join([f'[{i}]' for i in args.extensions])
    globs = f'{args.target}/**/*{elements}'
    results = [i for i in iglob(globs, recursive=True)]

    return results


def walk_with_suffixes(args):

    results = []
    for r, d, f in walk(args.target):
        for ff in f:
            for e in args.extensions:
                if ff.endswith(e):
                    results.append(path_join(r,ff))
                    break
    return results


def walk_with_regs(args):

    reg = compile('|'.join([f'{i}$' for i in args.extensions]))

    results = []
    for r, d, f in walk(args.target):
        for ff in f:
            if len(findall(reg,ff)):
                results.append(path_join(r, ff))

    return results

By running the code above on my laptop I obtained the following auto-explicative results.

Elapsed time for '7 times glob_with_storage()':  0.365023 seconds.
mean   : 0.05214614
median : 0.051861
stdev  : 0.001492152
min    : 0.050864
max    : 0.054853

Elapsed time for '7 times glob_with_iteration()':  0.360037 seconds.
mean   : 0.05143386
median : 0.050864
stdev  : 0.0007847381
min    : 0.050864
max    : 0.052859

Elapsed time for '7 times walk_with_suffixes()':  0.26529 seconds.
mean   : 0.03789857
median : 0.037899
stdev  : 0.0005759071
min    : 0.036901
max    : 0.038896

Elapsed time for '7 times walk_with_regs()':  0.290223 seconds.
mean   : 0.04146043
median : 0.040891
stdev  : 0.0007846776
min    : 0.04089
max    : 0.042885

Results sizes:
0 2451
1 2451
2 2446
3 2446

Differences between glob() and walk():
0 E:\x\y\z\venv\lib\python3.7\site-packages\Cython\Includes\numpy
1 E:\x\y\z\venv\lib\python3.7\site-packages\Cython\Utility\CppSupport.cpp
2 E:\x\y\z\venv\lib\python3.7\site-packages\future\moves\xmlrpc
3 E:\x\y\z\venv\lib\python3.7\site-packages\Cython\Includes\libcpp
4 E:\x\y\z\venv\lib\python3.7\site-packages\future\backports\xmlrpc

Elapsed time for 'main':  1.317424 seconds.

The fastest way to filter out files by extensions, happens even to be the ugliest one. Which is, nested for loops and string comparison using the endswith() method.

Moreover, as you can see, the globbing algorithms (with the pattern E:\x\y\z\**/*[py][pyc]) even with only 2 extension given (py and pyc) returns also incorrect results.

For example:

import glob
lst_img = []
base_dir = '/home/xy/img/'

# get all the jpg file in base_dir 
lst_img += glob.glob(base_dir + '*.jpg')
print lst_img
# ['/home/xy/img/2.jpg', '/home/xy/img/1.jpg']

# append all the png file in base_dir to lst_img
lst_img += glob.glob(base_dir + '*.png')
print lst_img
# ['/home/xy/img/2.jpg', '/home/xy/img/1.jpg', '/home/xy/img/3.png']

A function:

import glob
def get_files(base_dir='/home/xy/img/', lst_extension=['*.jpg', '*.png']):
    """
    :param base_dir:base directory
    :param lst_extension:lst_extension: list like ['*.jpg', '*.png', ...]
    :return:file lists like ['/home/xy/img/2.jpg','/home/xy/img/3.png']
    """
    lst_files = []
    for ext in lst_extension:
        lst_files += glob.glob(base_dir+ext)
    return lst_files

import os    
import glob
import operator
from functools import reduce

types = ('*.jpg', '*.png', '*.jpeg')
lazy_paths = (glob.glob(os.path.join('my_path', t)) for t in types)
paths = reduce(operator.add, lazy_paths, [])

https://docs.python.org/3.5/library/functools.html#functools.reduce https://docs.python.org/3.5/library/operator.html#operator.add

To glob multiple file types, you need to call glob() function several times in a loop. Since this function returns a list, you need to concatenate the lists.

For instance, this function do the job:

import glob
import os


def glob_filetypes(root_dir, *patterns):
    return [path
            for pattern in patterns
            for path in glob.glob(os.path.join(root_dir, pattern))]

Simple usage:

project_dir = "path/to/project/dir"
for path in sorted(glob_filetypes(project_dir, '*.txt', '*.mdown', '*.markdown')):
    print(path)

You can also use glob.iglob() to have an iterator:

Return an iterator which yields the same values as glob() without actually storing them all simultaneously.

def iglob_filetypes(root_dir, *patterns):
    return (path
            for pattern in patterns
            for path in glob.iglob(os.path.join(root_dir, pattern)))

You can try to make a manual list comparing the extension of existing with those you require.

ext_list = ['gif','jpg','jpeg','png'];
file_list = []
for file in glob.glob('*.*'):
  if file.rsplit('.',1)[1] in ext_list :
    file_list.append(file)

This worked for me!

split('.')[-1]

Will separate the filename suffix (*.xxx) so using if help to check it

    for filename in glob.glob(folder + '*.*'):
        print(folder+filename)
        if  filename.split('.')[-1] != 'tif' and \
            filename.split('.')[-1] != 'tiff' and \
            filename.split('.')[-1] != 'bmp' and \
            filename.split('.')[-1] != 'jpg' and \
            filename.split('.')[-1] != 'jpeg' and \
            filename.split('.')[-1] != 'png':
                continue
        # Your code

You could also use reduce() like so:

import glob
file_types = ['*.txt', '*.mdown', '*.markdown']
project_files = reduce(lambda list1, list2: list1 + list2, (glob.glob(t) for t in file_types))

this creates a list from glob.glob() for each pattern and reduces them to a single list.

files = glob.glob('*.txt')
files.extend(glob.glob('*.dat'))

Not glob, but here's another way using a list comprehension:

extensions = 'txt mdown markdown'.split()
projectFiles = [f for f in os.listdir(projectDir) 
                  if os.path.splitext(f)[1][1:] in extensions]

After coming here for help, I made my own solution and wanted to share it. It's based on user2363986's answer, but I think this is more scalable. Meaning, that if you have 1000 extensions, the code will still look somewhat elegant.

from glob import glob

directoryPath  = "C:\\temp\\*." 
fileExtensions = [ "jpg", "jpeg", "png", "bmp", "gif" ]
listOfFiles    = []

for extension in fileExtensions:
    listOfFiles.extend( glob( directoryPath + extension ))

for file in listOfFiles:
    print(file)   # Or do other stuff

From previous answer

glob('*.jpg') + glob('*.png')

Here is a shorter one,

from glob import glob
extensions = ['jpg', 'png'] # to find these filename extensions

# Method 1: loop one by one and extend to the output list
output = []
[output.extend(glob(f'*.{name}')) for name in extensions]
print(output)

# Method 2: even shorter
# loop filename extension to glob() it and flatten it to a list
output = [p for p2 in [glob(f'*.{name}') for name in extensions] for p in p2]
print(output)

The following function _glob globs for multiple file extensions.

import glob
import os
def _glob(path, *exts):
    """Glob for multiple file extensions

    Parameters
    ----------
    path : str
        A file name without extension, or directory name
    exts : tuple
        File extensions to glob for

    Returns
    -------
    files : list
        list of files matching extensions in exts in path

    """
    path = os.path.join(path, "*") if os.path.isdir(path) else path + "*"
    return [f for files in [glob.glob(path + ext) for ext in exts] for f in files]

files = _glob(projectDir, ".txt", ".mdown", ".markdown")

Use a list of extension and iterate through

from os.path import join
from glob import glob

files = []
extensions = ['*.gif', '*.png', '*.jpg']
for ext in extensions:
   files.extend(glob(join("path/to/dir", ext)))

print(files)

Python 3

We can use pathlib; .glob still doesn't support globbing multiple arguments or within braces (as in POSIX shells) but we can easily filter the result.

For example, where you might ideally like to do:

# NOT VALID
Path(config_dir).glob("*.{ini,toml}")
# NOR IS
Path(config_dir).glob("*.ini", "*.toml")

you can do:

filter(lambda p: p.suffix in {".ini", ".toml"}, Path(config_dir).glob("*"))

which isn't too much worse.

I have released Formic which implements multiple includes in a similar way to Apache Ant's FileSet and Globs.

The search can be implemented:

import formic
patterns = ["*.txt", "*.markdown", "*.mdown"]
fileset = formic.FileSet(directory=projectDir, include=patterns)
for file_name in fileset.qualified_files():
    # Do something with file_name

Because the full Ant glob is implemented, you can include different directories with each pattern, so you could choose only those .txt files in one subdirectory, and the .markdown in another, for example:

patterns = [ "/unformatted/**/*.txt", "/formatted/**/*.mdown" ]

I hope this helps.

with glob it is not possible. you can use only:
* matches everything
? matches any single character
[seq] matches any character in seq
[!seq] matches any character not in seq

use os.listdir and a regexp to check patterns:

for x in os.listdir('.'):
  if re.match('.*\.txt|.*\.sql', x):
    print x

import glob
import pandas as pd

df1 = pd.DataFrame(columns=['A'])
for i in glob.glob('C:\dir\path\*.txt'):
    df1 = df1.append({'A': i}, ignore_index=True)
for i in glob.glob('C:\dir\path\*.mdown'):
    df1 = df1.append({'A': i}, ignore_index=True)
for i in glob.glob('C:\dir\path\*.markdown):
    df1 = df1.append({'A': i}, ignore_index=True)

this worked for me:

import glob
images = glob.glob('*.JPG' or '*.jpg' or '*.png')

I had the same issue and this is what I came up with

import os, sys, re

#without glob

src_dir = '/mnt/mypics/'
src_pics = []
ext = re.compile('.*\.(|{}|)$'.format('|'.join(['png', 'jpeg', 'jpg']).encode('utf-8')))
for root, dirnames, filenames in os.walk(src_dir):
  for filename in filter(lambda name:ext.search(name),filenames):
    src_pics.append(os.path.join(root, filename))

Yet another solution (use glob to get paths using multiple match patterns and combine all paths into a single list using reduce and add):

import functools, glob, operator
paths = functools.reduce(operator.add, [glob.glob(pattern) for pattern in [
    "path1/*.ext1",
    "path2/*.ext2"]])

A one-liner, Just for the hell of it..

folder = "C:\\multi_pattern_glob_one_liner"
files = [item for sublist in [glob.glob(folder + ext) for ext in ["/*.txt", "/*.bat"]] for item in sublist]

output:

['C:\\multi_pattern_glob_one_liner\\dummy_txt.txt', 'C:\\multi_pattern_glob_one_liner\\dummy_bat.bat']

glob returns a list: why not just run it multiple times and concatenate the results?

from glob import glob
project_files = glob('*.txt') + glob('*.mdown') + glob('*.markdown')

So many answers that suggest globbing as many times as number of extensions, I'd prefer globbing just once instead:

from pathlib import Path

files = {p.resolve() for p in Path(path).glob("**/*") if p.suffix in [".c", ".cc", ".cpp", ".hxx", ".h"]}

from glob import glob

files = glob('*.gif')
files.extend(glob('*.png'))
files.extend(glob('*.jpg'))

print(files)

If you need to specify a path, loop over match patterns and keep the join inside the loop for simplicity:

from os.path import join
from glob import glob

files = []
for ext in ('*.gif', '*.png', '*.jpg'):
   files.extend(glob(join("path/to/dir", ext)))

print(files)

You could use filter:

import os
import glob

projectFiles = filter(
    lambda x: os.path.splitext(x)[1] in [".txt", ".mdown", ".markdown"]
    glob.glob(os.path.join(projectDir, "*"))
)

This is a Python 3.4+ pathlib solution:

exts = ".pdf", ".doc", ".xls", ".csv", ".ppt"
filelist = (str(i) for i in map(pathlib.Path, os.listdir(src)) if i.suffix.lower() in exts and not i.stem.startswith("~"))

Also it ignores all file names starting with ~.

While Python's default glob doesn't really follow after Bash's glob, you can do this with other libraries. We can enable braces in wcmatch's glob.

>>> from wcmatch import glob
>>> glob.glob('*.{md,ini}', flags=glob.BRACE)
['LICENSE.md', 'README.md', 'tox.ini']

You can even use extended glob patterns if that is your preference:

from wcmatch import glob
>>> glob.glob('*.@(md|ini)', flags=glob.EXTGLOB)
['LICENSE.md', 'README.md', 'tox.ini']

This Should Work:

import glob
extensions = ('*.txt', '*.mdown', '*.markdown')
for i in extensions:
    for files in glob.glob(i):
        print (files)

For example, for *.mp3 and *.flac on multiple folders, you can do:

mask = r'music/*/*.[mf][pl][3a]*'
glob.glob(mask)

The idea can be extended to more file extensions, but you have to check that the combinations won't match any other unwanted file extension you may have on those folders. So, be careful with this.

To automatically combine an arbitrary list of extensions into a single glob pattern, you can do the following:

def multi_extension_glob_mask(mask_base, *extensions):
    mask_ext = ['[{}]'.format(''.join(set(c))) for c in zip(*extensions)]
    if not mask_ext or len(set(len(e) for e in extensions)) > 1:
        mask_ext.append('*')
    return mask_base + ''.join(mask_ext)

mask = multi_extension_glob_mask('music/*/*.', 'mp3', 'flac', 'wma')
print(mask)  # music/*/*.[mfw][pml][a3]*

If you use pathlib try this:

import pathlib

extensions = ['.py', '.txt']
root_dir = './test/'

files = filter(lambda p: p.suffix in extensions, pathlib.Path(root_dir).glob('**/*'))

print(list(files))

Chain the results:

import itertools as it, glob

def multiple_file_types(*patterns):
    return it.chain.from_iterable(glob.iglob(pattern) for pattern in patterns)

Then:

for filename in multiple_file_types("*.txt", "*.sql", "*.log"):
    # do stuff

Here is one-line list-comprehension variant of Pat's answer (which also includes that you wanted to glob in a specific project directory):

import os, glob
exts = ['*.txt', '*.mdown', '*.markdown']
files = [f for ext in exts for f in glob.glob(os.path.join(project_dir, ext))]

You loop over the extensions (for ext in exts), and then for each extension you take each file matching the glob pattern (for f in glob.glob(os.path.join(project_dir, ext)).

This solution is short, and without any unnecessary for-loops, nested list-comprehensions, or functions to clutter the code. Just pure, expressive, pythonic Zen.

This solution allows you to have a custom list of exts that can be changed without having to update your code. (This is always a good practice!)

The list-comprehension is the same used in Laurent's solution (which I've voted for). But I would argue that it is usually unnecessary to factor out a single line to a separate function, which is why I'm providing this as an alternative solution.

Bonus:

If you need to search not just a single directory, but also all sub-directories, you can pass recursive=True and use the multi-directory glob symbol ** ¹:

files = [f for ext in exts 
         for f in glob.glob(os.path.join(project_dir, '**', ext), recursive=True)]

This will invoke glob.glob('<project_dir>/**/*.txt', recursive=True) and so on for each extension.

¹ Technically, the ** glob symbol simply matches one or more characters including forward-slash / (unlike the singular * glob symbol). In practice, you just need to remember that as long as you surround ** with forward slashes (path separators), it matches zero or more directories.