ChatGPT解决这个技术问题 Extra ChatGPT

How to use 'cp' command to exclude a specific directory?

I want to copy all files in a directory except some files in a specific sub-directory. I have noticed that cp command didn't have the --exclude option. So, how can I achieve this?

tar -c | tar -x ?
@mvds, agree with you, using tar with '--exclude' is a good idea.

s
syntagma

rsync is fast and easy:

rsync -av --progress sourcefolder /destinationfolder --exclude thefoldertoexclude

You can use --exclude multiples times.

rsync -av --progress sourcefolder /destinationfolder --exclude thefoldertoexclude --exclude anotherfoldertoexclude

Note that the dir thefoldertoexclude after --exclude option is relative to the sourcefolder, i.e., sourcefolder/thefoldertoexclude.

Also you can add -n for dry run to see what will be copied before performing real operation, and if everything is ok, remove -n from command line.


Agreed, you can't beat the simplicity and power of --exclude
is the thefoldertoexclude relative to the sourcefolder or the current working dir? thanks
It's relative to the source folder. This will exclude the folder source/.git from being copied. rsync -r --exclude '.git' source target
Maybe I'm wrong but I think it's a good practice to add "switches" before "parameters". Also the man page of rsync reports --exclude usable as with the "=" syntax or without. So to standarize across operating systems, I'd use rsync -av --progress --exclude="thefoldertoexclude" sourcefolder /destinationfolder - anyway upvote for the rsync instead of the find, as you can easily use absolute paths for the source while in the find it's trickier as it uses the {} in the dst.
I think slash at wrong place,rsync -av --progress sourcefolder/ destinationfolder --exclude thefoldertoexclude
C
Community

Well, if exclusion of certain filename patterns had to be performed by every unix-ish file utility (like cp, mv, rm, tar, rsync, scp, ...), an immense duplication of effort would occur. Instead, such things can be done as part of globbing, i.e. by your shell.

bash

man 1 bash, / extglob.

Example:

$ shopt -s extglob
$ echo images/*
images/004.bmp images/033.jpg images/1276338351183.jpg images/2252.png
$ echo images/!(*.jpg)
images/004.bmp images/2252.png

So you just put a pattern inside !(), and it negates the match. The pattern can be arbitrarily complex, starting from enumeration of individual paths (as Vanwaril shows in another answer): !(filename1|path2|etc3), to regex-like things with stars and character classes. Refer to the manpage for details.

zsh

man 1 zshexpn, / filename generation.

You can do setopt KSH_GLOB and use bash-like patterns. Or,

% setopt EXTENDED_GLOB
% echo images/*
images/004.bmp images/033.jpg images/1276338351183.jpg images/2252.png
% echo images/*~*.jpg
images/004.bmp images/2252.png

So x~y matches pattern x, but excludes pattern y. Once again, for full details refer to manpage.

fishnew!

The fish shell has a much prettier answer to this:

🐟 cp (string match -v '*.excluded.names' -- srcdir/*) destdir

Bonus pro-tip

Type cp *, hit CtrlX* and just see what happens. it's not harmful I promise


@MikhailGolubtsov perhaps that's because globbing is not recursive and works one level at a time. Edited out. P.S: it works in zsh though.
Nice pro-tip! This way you can remove single items easily. Thanks a lot!
BTW, to turn off extended pattern matching features in Bash, run setopt -u extglob.
"... an immense duplication of effort..." shouldn't it be just a one-liner: exclude paths from the list that match a regex? How the file manipulating utilities like cp don't support this most simple and straightforward use case out of the box is beyond me. Thanks for the tip though!
@ayorgo well yes it "should" — but in C, a oneliner can't do much: multiply some ints and maybe move a pointer, that's it. Even ignoring the source level, regex matching in C involves additional library dependency and additional machine code output — now multiply this by the number of commands, and you've got nontrivial (unbounded?..) overhead. At least that's my understanding why it was "refactored" to the shell; I can totally relate to the subpar UI aspect of it, but hopefully you can also see the technical justification now. Best wishes!
L
Linus Kleen

Why use rsync when you can do:

find . -type f -not -iname '*/not-from-here/*' -exec cp '{}' '/dest/{}' ';'

This assumes the target directory structure being the same as the source's.


I think you need the -path argument to test path hierarchies, not -iname
And you'll also need a semi-colon at the end: find . -type f -not -path '*/not-from-here/*' -exec cp '{}' '/dest/{}' \;
Wow, it won't let me: "Edits must be at least 6 characters" !
@MatthewWilcoxson Meh. Those restrictions will be lifted, as soon as you gain a little more rep. I edited the answer accordingly. Thanks again!
@Henning why not rsync? Coz it may be not present in the system! while find, cp is always on their places. Or you from kind of guys, who installed 2gigs of stuff to do simple things?
p
pts
cp -r `ls -A | grep -v "c"` $HOME/

Worked for me in Windows 10 .sh
Made a shell function which simplifies the usage for custom source path and exclusion of just one file or directory: # $1 = source path # $2 = destination path # $3 = filter copy_from_source_to_destination_except_filter() { cp -r $(ls -A $1 | grep -v -w $3 | awk -v path=$1 '{printf "%s/%s ", path, $1}') $2 }
fails with directories with spaces
@Sérgio I haven't tested it but cp -r "$(ls -A | grep -v "c")" $HOME/ should work. The command in the answer fails because there cp operates on the output of ls -A | grep -v "c", which is unquoted and therefore breaks on spaces. "$(…)" is the same as "`…`" but easier on the eyes.
o
ostergaard

The easiest way I found, where you can copy all the files excluding files and folders just by adding their names in the parentheses:

shopt -s extglob
cp -r !(Filename1 | FoldernameX | Filename2) Dest/

Doesn't work for me. I get -bash: !: event not found
shopt -s extglob (execute this to enable ! in cp, rm and others)
@geneorama This happens if history substitution is enabled. serverfault.com/a/208414/352016
Nice tip, but it does not work in sh.
S
Skandix

It's relative to the source directory.
This will exclude the directory source/.git from being copied.

rsync -r --exclude '.git' source target

What is the difference/improvement compared to the top answer?
@reducingactivity less obsolete flags
I feel like the '-a' in the first answer is better than plain old -r: explainshell.com/explain?cmd=rsync+-a
@reducingactivity nothing much but simple to digest as the expression is shorter, just my personal preference
C
Community

Expanding on mvds’s comment, this works for me

cd dotfiles
tar -c --exclude .git --exclude README . | tar -x -C ~/dotfiles2

The nice thing about tar, you can use exclude.tag files to ignore directories stackoverflow.com/a/13280610/722796 also gnu.org/software/tar/manual/html_node/exclude.html
L
LeOn - Han Li

rsync is actually quite tricky. have to do multiple tests to make it work.

Let's say you want to copy /var/www/html to /var/www/dev but need to exclude /var/www/html/site/video/ directory maybe due to its size. The command would be:

rsync -av --exclude 'sites/video' /var/www/html/ /var/www/dev

Some caveat:

The last slash / in the source is needed, otherwise it will also copy the source directory rather than its content and becomes /var/www/dev/html/xxxx, which maybe is not what you want. The the --exclude path is relative to the source directly. Even if you put full absolute path, it will not work. -v is for verbose, -a is for archive mode which means you want recursion and want to preserve almost everything.


a simple solution that take cares of special characters and white spaces
Thanks for explaining parameters, unlike the current top answer!
how about multiple folder to exclude?
@ddzzbbwwmm You probably figured it out by now, but for posterity's sake: you can add multiple --exclude flags, like: --exclude 'foo' --exclude 'bar'
s
sudo work
cp -rv `ls -A | grep -vE "dirToExclude|targetDir"` targetDir

Edit: forgot to exclude the target path as well (otherwise it would recursively copy).


watch out for directory entries containing spaces.
d
dzon

rsync

rsync -r --verbose --exclude 'exclude_pattern' ./* /to/where/

and first try it with -n option to see what is going to be copied


What is the difference/improvement compared to the top answer?
V
Vanwaril

I assume you're using bash or dash. Would this work?

shopt -s extglob  # sets extended pattern matching options in the bash shell
cp $(ls -laR !(subdir/file1|file2|subdir2/file3)) destination

Doing an ls excluding the files you don't want, and using that as the first argument for cp


You can skip the extra ls and simply do cp !(file1|file1) dest.
Do not use -laR. it add string that interfere with cp. cp $(ls folder/!exclude_folder0|exclude_folder1)) dest
e
ehudokai

Another simpler option is to install and use rsync which has an --exclude-dir option, and can be used for both local and remote files.


k
kungfooman

Just move it temporally into a hidden directory (and rename it after, if wanted).

mkdir .hiddendir
cp * .hiddendir -R
mv .hiddendir realdirname

Not pretty maybe – but this is the only option I’ve found here which works with cp and a standard POSIX shell like sh.
This answer is dramatically underrated. This is the most compatible, easiest to read and easy to understand answer. Kudos, I don't know why I didn't think of it.
Thanks @RobertTalada, how far will the answer go from now? ‎️‍🌈
Obvious drawback is that you might be avoiding copying something because it's too big.
M
Milan Simek

This is a modification of Linus Kleen's answer. His answer didn't work for me because there would be a . added in front of the file path which cp doesn't like (the path would look like source/.destination/file).

This command worked for me:

find . -type f -not -path '*/exlude-path/*' -exec cp --parents '{}' '/destination/' \;

the --parents command preserves the directory structure.


z
zyfyy
cp -r `ls -A | grep -v "Excluded_File_or_folder"` ../$target_location -v

P
Panduka

rsync went unavailable for us. Below is an alternative that works.

tar -cf - --exclude='./folder' --exclude='./file.tar' ./source_directory | tar -xf - -C ./destination_directory

p
pts
mv tobecopied/tobeexcluded .
cp -r tobecopied dest/
mv tobeexcluded tobecopied/

L
LouisXW
ls -I "filename1" -I "filename2" | xargs cp -rf -t destdir 

The first part ls all the files but hidden specific files with flag -I. The output of ls is used as standard input for the second part. xargs build and execute command cp -rf -t destdir from standard input. the flag -r means copy directories recursively, -f means copy files forcibly which will overwrite the files in the destdir, -t specify the destination directory copy to.


q
qräbnö

10 years late. Credits to Linus Kleen.

I hate rsync! ;) So why not use find and cp? And with this answer also mkdir to create a non-existent folder structure.

cd /source_folder/ && find . -type d -not -path '*/not-from-here/*' -print -exec mkdir -p '/destination_folder/{}' \;

cd /source_folder/ && find . -type f -not -path '*/not-from-here/*' -print -exec cp -au '{}' '/destination_folder/{}' \;

It looks like cd ìs necessary to concat relative paths with find.

mkdir -p will create all subfolders and will not complain when a folder already exists.

Housten we have the next problem. What happens when someone creates a new folder with a new file in the middle of it? Exactly: it will fail for these new files. (Solution: just run it again! :)) The solution to put everything into one find command seems difficult.

For clean-up: https://unix.stackexchange.com/q/627218/239596


Gave it a try. Current version of your answer creates in the first step directories. Because this step currently contains -not -path '*/not-from-here/*', it will create directory ./not-from-here. Probably, this is not intended. Therefore, for the first step (directory creation), you probably want -not -path '*/log' instead.
y
yanana

I use a "do while" loop to read the output of the find command. In this example, I am matching (rather than excluding) certain patterns since there are a more limited number of pattern matches that I want than that I don't want. You could reverse the logic with a -not in front of the -iname flags:

find . -type f -iname "*.flac" -o -print0 -iname "*.mp3" -print0 -o -iname "*.wav" -print0 -o -iname "*.aac" -print0 -o -iname "*.wma" -print0 | while read -d $'\0' file; do cp -ruv "$file" "/media/wd/network_sync/music/$file"; done

I use the above to copy all music type files that are newer on my server than the files on a Western Digital TV Live Hub that I have mounted at /media/wd. I use the above because I have a lot of DVD files, mpegs, etc. that I want to exclude AND because for some reason rsync looks like it is copying, but after I look at the wd device, the files are not there despite no errors during the rsync with this command:

rsync -av --progress --exclude=*.VOB --exclude=*.avi --exclude=*.mkv --exclude=*.ts --exclude=*.mpg --exclude=*.iso --exclude=*ar --exclude=*.vob --exclude=*.BUP --exclude=*.cdi --exclude=*.ISO --exclude=*.shn --exclude=*.MPG --exclude=*.AVI --exclude=*.DAT --exclude=*.img --exclude=*.nrg --exclude=*.cdr --exclude=*.bin --exclude=*.MOV --exclude=*.goutputs* --exclude=*.flv --exclude=*.mov --exclude=*.m2ts --exclude=*.cdg --exclude=*.IFO --exclude=*.asf --exclude=*.ite /media/2TB\ Data/data/music/* /media/wd/network_sync/music/