Delete all files/directories except two specific directories

Question

So, there seems to be a few questions asking about removing files/directories matching certain cases, but I'm looking for the exact opposite: Delete EVERYTHING in a folder that DOESN'T match my provided examples.

For example, here is an example directory tree:

.
|-- coke
|   |-- diet
|   |-- regular
|   `-- vanilla
|-- icecream
|   |-- chocolate
|   |-- cookiedough
|   |-- cupcake
|   |   |-- file1.txt
|   |   |-- file2.txt
|   |   |-- file3.txt
|   |   |-- file4.txt
|   |   `-- file5.txt
|   `-- vanilla
|-- lol.txt
|-- mtndew
|   |-- classic
|   |-- codered
|   |-- livewire
|   |   |-- file1.txt
|   |   |-- file2.txt
|   |   |-- file3.txt
|   |   |-- file4.txt
|   |   `-- file5.txt
|   `-- throwback
`-- pepsi
    |-- blue
    |-- classic
    |-- diet
    `-- throwback

I want to delete everything but the files in test/icecream/cupcake/ and test/mtndew/livewire/. Everything else can go, including the directory structure. So, how can I achieve this? Languages I wouldn't mind this being in: bash or python.

Solution 1

find's -prune comes to mind, but it's a pain to get it to work for specific paths (icecream/cupcake/) rather than specific directories (cupcake/).

Personally, I'd just use cpio and hard-link (to avoid having to copy them) the files in the directories you want to keep to a new tree and then remove the old one:

find test -path 'test/icecream/cupcake/*' -o -path 'test/mtndew/livewire/*' | cpio -padluv test-keep
rm -rf test

That'll also keep your existing directory structure for the directories you intend to keep.

Solution 2

This command will leave only the desired files in their original directories:

find test \( ! -path "test/mtndew/livewire/*" ! -path "test/icecream/cupcake/*" \) -delete

No need for cpio. It works on Ubuntu, Debian 5, and Mac OS X.

On Linux, it will report that it cannot delete non-empty directories, which is exactly the desired result. On Mac OS X, it will quietly do the right thing.

Solution 3

Everything "except" is why we have if-statements; and why os.walk's list of directories is a mutable list.

for path, dirs, files in os.walk( 'root' ):
    if 'coke' in dirs:
        dirs.remove('coke')
        dirs.remove('pepsi')

Solution 4

You could do something based on Python's os.walk function:

import os
for root, dirs, files in os.walk(top, topdown=False):
    for name in files:
        os.remove(os.path.join(root, name))
    for name in dirs:
        os.rmdir(os.path.join(root, name))

...just add something to ignore the paths you're interested in.

Solution 5

find /path/to/test/ -depth -mindepth 1 \
! -path "/path/to/test/icecream/cupcake/*" \
! -path "/path/to/test/icecream/cupcake" \
! -path "/path/to/test/icecream" \
! -path "/path/to/test/mtndew/livewire/*" \
! -path "/path/to/test/mtndew/livewire" \
! -path "/path/to/test/mtndew"
 -delete -print

It's a bit tedious to write all the paths to preserve but thi is the only way to use find alone.

View more solutions

Comments

[phuzion]

Tried this, and it says the following: find /home/phuzion/test/ \( -prune 'icecream/cupcake/' \) -delete find: paths must precede expression

[SourceSeeker]

From the man page for find: "Because -delete implies -depth, you cannot usefully use -prune and -delete together."

[SourceSeeker]

The correct syntax would be more like "find $directory -path ./pepsi/diet -prune -o -exec some-command '{}' \;" anyway to eliminate that error message.

[SourceSeeker]

In your first example, it will rm -r test/icecream which will include cupcake (for example). So, even though you've grep -v the cupcakes, they still get "eaten". The second example has a couple of typos and unbalanced single quotes. There should be a space after sed and the zero should be an ampersand.

[SourceSeeker]

This finds test and deletes it, thus deleting the directories that you're trying to save.

[Patrick Webster]

This finds test and does NOT delete it, because it is non-empty. Are there any modern flavors of Linux or Unix where this command fails to work?

[SourceSeeker]

Sorry, I made a mistake in setting up the test on my system. Your example works correctly for me.

[SourceSeeker]

You still have unbalanced single quotes around the paths. For example, it should be "rm './test/icecream/cupcake/'"

[kevinmicke]

Though it wasn't the original question, some may find this useful: you can preserve the entire directory structure and only delete files not excluded with test by adding the -type f flag before -delete.