Using Python's os.path, how do I go up one directory?

To get the folder of a file just use:

os.path.dirname(path) 

To get a folder up just use os.path.dirname again

os.path.dirname(os.path.dirname(path))

You might want to check if __file__ is a symlink:

if os.path.islink(__file__): path = os.readlink (__file__)

You want exactly this:

BASE_DIR = os.path.join( os.path.dirname( __file__ ), '..' )

from os.path import dirname, realpath, join
join(dirname(realpath(dirname(__file__))), 'templates')

Update:

If you happen to “copy” settings.py through symlinking, @forivall’s answer is better:

~user/
    project1/  
        mysite/
            settings.py
        templates/
            wrong.html

    project2/
        mysite/
            settings.py -> ~user/project1/settings.py
        templates/
            right.html

The method above will ‘see’ wrong.html while @forivall’s method will see right.html

In the absense of symlinks the two answers are identical.

For a paranoid like me, I’d prefer this one

TEMPLATE_DIRS = (
    __file__.rsplit('/', 2)[0] + '/templates',
)

Personally, I’d go for the function approach

def get_parent_dir(directory):
    import os
    return os.path.dirname(directory)

current_dirs_parent = get_parent_dir(os.getcwd())

I think the easiest thing to do is just to reuse dirname()
So you can call

os.path.dirname(os.path.dirname( __file__ ))

if you file is at /Users/hobbes3/Sites/mysite/templates/method.py

This will return “/Users/hobbes3/Sites/mysite”

This might be useful for other cases where you want to go x folders up. Just run walk_up_folder(path, 6) to go up 6 folders.

def walk_up_folder(path, depth=1):
    _cur_depth = 1        
    while _cur_depth < depth:
        path = os.path.dirname(path)
        _cur_depth += 1
    return path   

If you are using Python 3.4 or newer, a convenient way to move up multiple directories is pathlib:

from pathlib import Path

full_path = "path/to/directory"
str(Path(full_path).parents[0])  # "path/to"
str(Path(full_path).parents[1])  # "path"
str(Path(full_path).parents[2])  # "."

To go n folders up… run up(n)

import os

def up(n, nth_dir=os.getcwd()):
    while n != 0:
        nth_dir = os.path.dirname(nth_dir)
        n -= 1
    return nth_dir

Of course: simply use os.chdir(..).

With using os.path we can go one directory up like that

one_directory_up_path = os.path.dirname('.')

also after finding the directory you want you can join with other file/directory path

other_image_path = os.path.join(one_directory_up_path, 'other.jpg')

Go up a level from the work directory

import os
os.path.dirname(os.getcwd())

or from the current directory

import os
os.path.dirname('current path')

From the current file path you could use:

os.path.join(os.path.dirname(__file__),'..','img','banner.png')

I’m surprised handling for an arbitrary number of ".." parent directory tokens in a path string isn’t directly handled for by the os library. Here’s a quick and dirty function that’ll give you an absolute path string from a relative one:

def get_abs_from_relpath(relpath:str) -> str:
    ap = os.path.abspath(__file__).split("/")[:-1]
    sp = relpath.split("/")
    sp_start_index = 0
    for slug in sp:
        if slug == "..":
            ap.pop(-1)
            sp_start_index += 1
        else:
            return "/".join(ap+sp[sp_start_index:])

You can call it with open() like this:

with open(get_abs_from_relpath('../../../somedir/myfile.txt')) as f:
    foo = f.read()

If you prefer a one-liner for getting the parent directory, I’d suggest this:

import os
    
parent_dir = os.path.split(os.getcwd())[0]

os.path.split() method returns a tuple (head, tail) where tail is everything after the final slash. So the first index is the parent of your absolute path.

Not an answer but a long tangential comment that probably should be made as someone may be led astray…

The syntax os.path.join( os.path.dirname( __file__ ), 'foo.txt') to get a file within the same folder as the python file getting run is not the "advised" solution for packages, instead package data is preferred for a couple of reasons, for example in a zip packaged package or a more complicated filesystem.

pkg_resources.read_text(__package__, 'foo.txt') was the formerly recommended solution, but will be removed at some point and importlib.resources.read_text(__package__, 'foo.txt') is the recommended way —see https://docs.python.org/3/library/importlib.html#module-importlib.resources for the many options.
However, this

• requires include_package_data=True and package_data with a Dict[str, List[str] in the setup.py file
• requires a MANIFEST.in if pip distributed as sdist (but not a built wheel)
• will not work for relative imports (i.e. not installed)
• is wisely and generally ignored for sake of sanity in webapps due to the way they run