Questions & Answers

How to replace a string in multiple files in linux command line

I need to replace a string in a lot of files in a folder, with only ssh access to the server. How can I do this?

2023-01-22 23:52:07
If you'd prefer to use Notepad++ instead of command line, I found this really helpful:
Answers(28) :

To replace a path within files (avoiding escape characters) you may use the following command:

sed -i 's@old_path@new_path@g'

The @ sign means that all of the special characters should be ignored in a following string.

2023-01-22 23:52:07
Exactly what I was looking for in all the other answers. Namely, how to deal with special characters, such as when changing a string that is a path. Thank you. Seemed like a big oversight in the other answers.
cd /path/to/your/folder
sed -i 's/foo/bar/g' *

Occurrences of "foo" will be replaced with "bar".

On BSD systems like macOS, you need to provide a backup extension like -i '.bak' or else "risk corruption or partial content" per the manpage.

cd /path/to/your/folder
sed -i '.bak' 's/foo/bar/g' *
2023-01-22 23:52:07
This doesn't seem to work for me if the string has whitespaces or special characters in it. Any idea why that might be, or do I need to escape them some how? Thanks!
2023-01-22 23:52:07
At least for my version of sed, -i requires a string after it, which is appended to the old file names. So sed -i .bak 's/foo/bar/g' *.xx moves all .xx files to the equivalent .xx.bak name and then generates the .xx files with the foo→bar substitution.
2023-01-22 23:52:07
If anybody want to look for more options, there is an answer on unix stack exchange which covers more use cases site
2023-01-22 23:52:07
@MatthewHerbst to escape spaces, use \ like sed -i 's/foo\ with\ spaces/bar/g' *. I suppose you've found out after so long... but this way it stays for others finding the same issue.
2023-01-22 23:52:07
I am getting this error: sed -i 's/[./(/g' * sed: -e expression #1, char 1: unknown command: `''
2023-01-22 23:52:07
@Anaphory You can also pass in an empty string for no backups (at least on OS X). Like: sed -i "" "s/foo/bar/g" *
2023-01-22 23:52:07
For something recursive you could try the following. Note that it doesn't work if the list of files is huge. sed -i -e 's/foo/bar/g' $(find /home/user/base/dir)
2023-01-22 23:52:07
@Scot that doesn't work. sed -i -e 's/foo/bar/g' $(find /home/user/dir) sed: couldn't edit /home/user/dir: not a regular file.
2023-01-22 23:52:07
To get around the "not a regular file" problem, just find only files: sed -i 's/foo/bar/g' $(find /path/to/your/folder -type f)
2023-01-22 23:52:07
its not working for me to replace a string like "my-syste" to ip address "" getting error like "sed: 1: "filename.txt": invalid command code .". command i used is: find ./ -type f -exec sed -i 's/"dm-system7"/""/g' {} \; Any help is appreciated.
2023-01-22 23:52:07
@FearlessFuture Use **/*, i.e. sed -i -e 's/foo/bar/g' **/*, where you can change it into a better match, such as **/*.tex for example.
2023-01-22 23:52:07
not one month passes without telling myself: one day I will sit down for 2 hours and learn how to use sed... years ago I bit the bullet for vim and I never regretted it.
2023-01-22 23:52:07
You can also try rpl (sudo apt install rpl)
2023-01-22 23:52:07
I used sed -i '.bak' 's/A.h/B.h/g' * with the goal to change all ocurrences of A.h to B.h, and now there is copies of all my files with .bkk
2023-01-22 23:52:07
This blog,…, answers this extensively as well.

If you have list of files you can use

replace "old_string" "new_string" -- file_name1 file_name2 file_name3

If you have all files you can use

replace "old_string" "new_string" -- *

If you have list of files with extension, you can use

replace "old_string" "new_string" -- *.extension
2023-01-22 23:52:07
actually, "--file" should just be "--", at least in my version
2023-01-22 23:52:07
This utility is distributed within MySQL packages.
2023-01-22 23:52:07
Though I would like to appreciate solution, which works without quoting regular line to be regular expression.
2023-01-22 23:52:07
This works fine - and if you want to replace all strings in multiple files, which end in e.g. ".txt" , you can just do replace "old_string" "new_string" -- *.txt
2023-01-22 23:52:07
How do you include subdirectories of one specific directory?
2023-01-22 23:52:07
It's better to add from where to get this replace utility. Without this information, this answer is incomplete.
2023-01-22 23:52:07
They would have done better to distribute this utility separately. Anyway, I’ve found the rpl utility, which can be installed separately on Debian-based distros, to be an adequate alternative.

I'd just like to add a note to do two things at once - find a file that contains a string and then do a replace, using the find 'chaining' method:

find  . -type f -iname \*.php -exec fgrep -l "www." {} \; -exec sed -i "s|www||g" {} \;      
  • In this real case, remove the anachronistic 'www' from urls found in PHP files.

  • The 'fgrep -l' only triggers if it finds at least one match in a file, it produces no other output. Don't forget the '\;' separators!

Really lame, but I couldn't get any of the sed commands to work right on OSX, so I did this dumb thing instead:


^- copy these three lines into my clipboard (yes, include the ending newline), then:

vi *

and hold down command-v until it says there's no files left.


This worked for me:

find ./ -type f -exec sed -i 's/string1/string2/' {} \;

However, this did not: sed -i 's/string1/string2/g' *. Maybe "foo" was not meant to be string1 and "bar" not string2.

2023-01-22 23:52:07
It is because sed treats the wildcard * differently. [abc]* means an arbitrary number of characters of the set {a, b, c}. [a-z0-9]* works similar to the wildcard *.
2023-01-22 23:52:07
On OSX use: find ./ -type f -exec sed -i '' -e 's/string1/string2/' {} \;

The stream editor does modify multiple files “inplace” when invoked with the -i switch, which takes a backup file ending as argument. So

sed -i.bak 's/foo/bar/g' *

replaces foo with bar in all files in this folder, but does not descend into subfolders. This will however generate a new .bak file for every file in your directory. To do this recursively for all files in this directory and all its subdirectories, you need a helper, like find, to traverse the directory tree.

find ./ -print0 | xargs -0 sed -i.bak 's/foo/bar/g' *

find allows you further restrictions on what files to modify, by specifying further arguments like find ./ -name '*.php' -or -name '*.html' -print0, if necessary.

Note: GNU sed does not require a file ending, sed -i 's/foo/bar/g' * will work, as well; FreeBSD sed demands an extension, but allows a space in between, so sed -i .bak s/foo/bar/g * works.

"You could also use find and sed, but I find that this little line of perl works nicely.

perl -pi -w -e 's/search/replace/g;' *.php
  • -e means execute the following line of code.
  • -i means edit in-place
  • -w write warnings
  • -p loop

" (Extracted from

My best results come from using perl and grep (to ensure that file have the search expression )

perl -pi -w -e 's/search/replace/g;' $( grep -rl 'search' )
2023-01-22 23:52:08
I prefer this as its cross-platform with no effort between Mac/Linux

To maintain my personal English node, I wrote an utility script that help to replace multiple pair of old/new string, for all files under a directory recursively.

The multiple pair of old / new string are managed in a hash map.

The dir can be set via command line or environment variable, the map is hard coded in the script, but you can modify the code to load from a file, if necessary.

It requires bash 4.2, due to some new feature.

#! /bin/bash
# (need bash 4.2+,)
# Standardize phonetic symbol of English.
# format:
# [<dir>]
# params:
# * dir
#   target dir, optional,
#   if not specified then use environment variable "$node_dir_en",
#   if both not provided, then will not execute,
# * 


# figure target dir,
if [ $paramCount -ge 1 ]; then # dir specified
    echo -e "dir specified (in command):\n\t$1\n"
elif [[ -v node_dir_en ]]; then # environable set,
    echo -e "dir specified (in environment vairable):\n\t$node_dir_en\n"
else # environable not set,
    echo "dir not specified, won't execute"

# check whether dir exists,
if [ -d $targetDir ]; then
    cd $targetDir
    echo -e "invalid dir location:\n\t$targetDir\n"

# initial map,
declare -A itemMap
itemMap=( ["ɪ"]="i" ["ː"]=":" ["ɜ"]="ə" ["ɒ"]="ɔ" ["ʊ"]="u" ["ɛ"]="e")

# print item maps,
echo 'maps:'
for key in "${!itemMap[@]}"; do
    echo -e "\t$key\t->\t${itemMap[$key]}"
echo -e '\n'

# do replace,
for key in "${!itemMap[@]}"; do
    grep -rli "$key" * | xargs -i@ sed -i "s/$key/${itemMap[$key]}/g" @

echo -e "\nDone."

Using the ack command would be alot faster like this:

ack '25 Essex' -l | xargs sed -i 's/The\ fox \jump/abc 321/g'

Also if you have a white space in the search result. You need to escape it.

In case your string has a forward slash(/) in it, you could change the delimiter to '+'.

find . -type f -exec sed -i 's+' {} +

This command would run recursively in the current directory.

2023-01-22 23:52:08
Thanks! Helped change a load of references of ../ to

Similar to Kaspar's answer but with the g flag to replace all the occurrences on a line.

find ./ -type f -exec sed -i 's/old_string/new_string/g' {} \;

For global case insensitive:

find ./ -type f -exec sed -i 's/old_string/new_string/gI' {} \;
2023-01-22 23:52:08
If you are on OSX and your patterns might contain dots and you want in-place replacement (no backup file) you should use LC_ALL=C find ./ -type f -exec sed -i '' -e 's/proc.priv/priv/g' {} \; (see this post and this one)
2023-01-22 23:52:08
This definitely worked for me, as it enables filtering with -name "*.js"'
2023-01-22 23:52:08
A verbose option would be cool, but you can just re-grep to see if the changes were made. Note: For wildcards, try '-name "*.php"' and grep is bad with recursion and wildcards, you need to add --include=*.whatever with -r
2023-01-22 23:52:08
Don't do this from the root of a checked-out Git repo. You might accidentally corrupt its database in .git/. Be sure to cd down to a lower level.
2023-01-22 23:52:08
This should really be the correct answer, but make sure to have the space after {} or you'll get an error.
2023-01-22 23:52:08
I just did this in my git repo and now git status returns: error: bad index file sha1 signature.... fatal: index file corrupt. What gives?
2023-01-22 23:52:08
I fixed the error by doing an inverse sed regex. Now repo's back to normal.
2023-01-22 23:52:08
This works, but not if you have special characters in the string, such as when changing a path. To get around this, us the following in the respective part of the command in this answer, sed -i 's@old_path@new_path@g'
2023-01-22 23:52:08
It would be good if there was a way to ignore the .git/ folder (as well as exclude other folders e.g. the ones specified in .gitignore) instead of having to cd to a lower level, which isn't convenient when there are lots of folders. Otherwise you could move them all into one folder and cd to that folder.
2023-01-22 23:52:08
Works! But how do we see verbose output or progress?
2023-01-22 23:52:08
@Shelljohn's comment above with LC_ALL=C find ./ -type f -exec sed -i '' -e 's/proc.priv/priv/g' {} \; was the only thing that worked for me on OSX
2023-01-22 23:52:08
Does ./ start searching from the current directory?
2023-01-22 23:52:08
@Black yes, it does
2023-01-22 23:52:08
This seems to me like a better answer than the currently most upvoted one, as it is simple and clear, less prone to bugs and works recursively as it should.
2023-01-22 23:52:08
Has anyone figured out how to combine the above command to something like this to ignore directories (.git/, node_modules/, etc)?
2023-01-22 23:52:08
This blog,…, answers this extensively with solid info.

@kev's answer is good, but only affects files in the immediate directory.The example below uses grep to recursively find files. It works for me everytime.

grep -rli 'old-word' * | xargs -i@ sed -i 's/old-word/new-word/g' @

Command breakdown

grep -r: --recursive, recursively read all files under each directory.
grep -l: --print-with-matches, prints the name of each file that has a match, instead of printing matching lines.
grep -i: --ignore-case.

xargs: transform the STDIN to arguments, follow this answer.
xargs -i@ ~command contains @~: a placeholder for the argument to be used in a specific position in the ~command~, the @ sign is a placeholder which could replaced by any string.

sed -i: edit files in place, without backups.
sed s/regexp/replacement/: substitute string matching regexp with replacement.
sed s/regexp/replacement/g: global, make the substitution for each match instead of only the first match.

2023-01-22 23:52:08
this did not work for me, but this did: grep --include={*.php,*.html,*.js} -rnl './' -e "old-word" | xargs -i@ sed -i 's/old-word/new-word/g' @
2023-01-22 23:52:08
@pymarco Can you please explain this command? I don't know why you had to use xargs instead of just using sed after |, also, why -i in xargs command? I read in the manual it is deprecated and should use -I instead. And is @ used as a delimiter for beginning and end for pattern?
2023-01-22 23:52:08
the question is specifically for linux though.
2023-01-22 23:52:08
On osx this is ok : grep -rli 'old-word' * | xargs -I@ sed -i '' 's/2.0.0/latest/g' @
2023-01-22 23:52:08
It would be nice if you broke down some of the options (rli) and the @ sign for example
2023-01-22 23:52:08
I also get sed no input files
2023-01-22 23:52:08
Works great, but be careful doing this within Git repositories. If you do this at the root, it changes files within .git, and messes up the repository (it doesn't track changes correctly anymore). So, be sure to do it in a subdirectory (e.g. /src).
2023-01-22 23:52:08
I am using @PhilippeSultan suggestion (osx) but I have a No such file or directory error. Does anyone can help please? I am seaching for "5.9.0", one is found in README but replacing does not work : sed: ../ current version is 5.9.0.: No such file or directory
2023-01-22 23:52:08
Worked like a charm
2023-01-22 23:52:08
The explanation of the command can be found here.
2023-01-22 23:52:08
On macOS (zsh 5.8.1), grep -rli 'd<C-A>' * | xargs -I@ sed -i '.bak' 's/d<C-A>/z<C-A>/g' @ works. As @Frungi said, -i '.bak' is usually necessary on macOS (and other BSD) systems.

There are a few standard answers to this already listed. Generally, you can use find to recursively list the files and then do the operations with sed or perl.


For most quick uses, you may find the command rpl is much easier to remember.

Replace foo with bar on all .txt files:

rpl -v foo bar '*.txt' 

Simulate replacing the regex foo.* with bar in all .txt files recursively:

rpl --dry-run 'foo.*' bar '**/*.txt'

You'll probably need to install it (apt-get install rpl or similar).


However, for tougher jobs that involve regular expressions and back substitution, or file renames as well as search-and-replace, the most general and powerful tool I'm aware of is repren, a small Python script I wrote a while back for some thornier renaming and refactoring tasks. The reasons you might prefer it are:

  • Support renaming of files as well as search-and-replace on file contents.
  • See changes before you commit to performing the search and replace.
  • Support regular expressions with back substitution, whole words, case insensitive, and case preserving (replace foo -> bar, Foo -> Bar, FOO -> BAR) modes.
  • Works with multiple replacements, including swaps (foo -> bar and bar -> foo) or sets of non-unique replacements (foo -> bar, f -> x).

To use it, pip install repren. Check the README for examples.

2023-01-22 23:52:08
Wow, repren is excellent! Just used it to change part of a word inside of class names, methods and variables while renaming files to match across 1,000+ C++ header and source files and it worked perfectly the first time with one command. Thank you!
2023-01-22 23:52:08
Create tools, thanks for sharing!
grep --include={*.php,*.html} -rnl './' -e "old" | xargs -i@ sed -i 's/old/new/g' @

script for multiedit command


From Kaspar's answer I made a bash script to accept command line arguments and optionally limit the filenames matching a pattern. Save in your $PATH and make executable, then just use the command above.

Here's the script:

Replace OLDSTRING with NEWSTRING recursively starting from current directory\n

[-n PATTERN] option limits to filenames matching PATTERN\n
Note: backslash escape special characters\n
Note: enclose STRINGS with spaces in double quotes\n
Example to limit the edit to python files:\n
multiedit -n \*.py \"OLD STRING\" NEWSTRING\n"

# ensure correct number of arguments, otherwise display help...
if [ $# -lt 2 ] || [ $# -gt 4 ]; then echo -e $_help ; exit ; fi
if [ $1 == "-n" ]; then  # if -n option is given:
        # replace OLDSTRING with NEWSTRING recursively in files matching PATTERN
        find ./ -type f -name "$2" -exec sed -i "s/$3/$4/g" {} \;
        # replace OLDSTRING with NEWSTRING recursively in all files
        find ./ -type f -exec sed -i "s/$1/$2/" {} \;

I am giving an example for fixing a common shebang error in python sources.

You can try the grep/sed approach. Here is one that works with GNU sed and won't break a git repo:

$ grep -rli --exclude '*.git*' '#!/usr/bin/python' . | xargs -I {} \
gsed -i '' -e 's/#!\/usr\/bin\/python/#!\/usr\/bin\/env python/' {}

Or you can use greptile :)

$ greptile -x .py -l -i -g '#!/usr/bin/env python' -r '#!/usr/bin/python' .

I just tested the first script, and the second should work as well. Be careful with escape characters, I think it should be easier to use greptile in most cases. Of course, you can do many interesting things with sed, and for that it may be preferable to master using it with xargs.

To replace a string in multiple files you can use:

grep -rl string1 somedir/ | xargs sed -i 's/string1/string2/g'


grep -rl 'windows' ./ | xargs sed -i 's/windows/linux/g'

Source blog

Given you want to search for the string search and replace it with replace across multiple files, this is my battle-tested, one-line formula:

grep -RiIl 'search' | xargs sed -i 's/search/replace/g'

Quick grep explanation:

  • -R - recursive search
  • -i - case-insensitive
  • -I - skip binary files (you want text, right?)
  • -l - print a simple list as output. Needed for the other commands

The grep output is then piped to sed (through xargs) which is used to actually replace text. The -i flag will alter the file directly. Remove it for a kind of "dry run" mode.

2023-01-22 23:52:08
If the text happens to be a url, remember you use different delimiters with sed ike this: 's#search#replace#g'
2023-01-22 23:52:09
When using grep -i for case-insensitive search, you also need to let sed do its own search in a case-insensitive way. For that, replace the sed command with sed -i 's/search/replace/gI'.

The first line occurrences of "foo" will be replaced with "bar". And you can using the second line to check.

grep -rl 'foo' . | xargs sed -i 's/foo/bar/g'
grep 'foo' -r * | awk -F: {'print $1'} | sort -n | uniq -c
2023-01-22 23:52:09
Why are you linking to to your blog? It contains exactly the same text as your answer.
2023-01-22 23:52:09
i like grep -rl 'foo' . | xargs sed -i 's/foo/bar/g'

If the file contains backslashes (paths usually) you can try something like this:

sed -i -- 's,<path1>,<path2>,g' *


sed -i -- 's,/foo/bar,/new/foo/bar,g' *.sh (in all shell scripts available)

Below command can be used to first search the files and replace the files:

find . | xargs grep 'search string' | sed 's/search string/new string/g'

For example

find . | xargs grep abc | sed 's/abc/xyz/g'

I did concoct my own solution before I found this question (and answers). I searched for different combinations of "replace" "several" and "xml," because that was my application, but did not find this particular one.

My problem: I had spring xml files with data for test cases, containing complex objects. A refactor on the java source code changed a lot of classes and did not apply to the xml data files. In order to save the test cases data, I needed to change all the class names in all the xml files, distributed across several directories. All while saving backup copies of the original xml files (although this was not a must, since version control would save me here).

I was looking for some combination of find + sed, because it worked for me in other situations, but not with several replacements at once.

Then I found ask ubuntu response and it helped me build my command line:

find -name "*.xml" -exec sed -s --in-place=.bak -e 's/firstWord/newFirstWord/g;s/secondWord/newSecondWord/g;s/thirdWord/newThirdWord/g' {} \;

And it worked perfectly (well, my case had six different replacements). But please note that it will touch all *.xml files under current directory. Because of that, and if you are accountable to a version control system, you might want to filter first and only pass on to sed those actually having the strings you want; like:

find -name "*.xml" -exec grep -e "firstWord" -e "secondWord" -e "thirdWord" {} \; -exec sed -s --in-place=.bak -e 's/firstWord/newFirstWord/g;s/secondWord/newSecondWord/g;s/thirdWord/newThirdWord/g' {} \;
2023-01-22 23:52:09
and for windows, I just found out that there is a way to find the string --didn't check yet how to replace it-- in one command: findstr /spin /c:"quéquieresbuscar" *.xml It will be handy.

I found this one from another post (can't remember which) and while not the most elegant, it's simple and as a novice Linux user has given me no trouble

for i in *old_str* ; do mv -v "$i" "${i/\old_str/new_str}" ; done

if you have spaces or other special characters use a \

for i in *old_str\ * ; do mv -v "$i" "${i/\old_str\ /new_str}" ; done

for strings in sub-directories use **

for i in *\*old_str\ * ; do mv -v "$i" "${i/\old_str\ /new_str}" ; done

On a MacBook Pro, I used the following (inspired by

sed -i '' -e 's/<STR_TO_REPLACE>/<REPLACEMENT_STR>/g' *

-i '' will ensure you are taking no backups.

-e for modern regex.

2023-01-22 23:52:09
-e just tells sed that the next token is a command. For extended regular expressions, use -E instead.

There is an easier way by using a simple script file:

   # sudo chmod +x /bin/replace_string_files_present_dir

open the file in gedit or an editor of your choice, I use gedit here.

   # sudo gedit /bin/replace_string_files_present_dir

Then in the editor paste the following in the file

   replace "oldstring" "newstring" -- *
   replace "oldstring1" "newstring2" -- *
   #add as many lines of replace as there are your strings to be replaced for 
   #example here i have two sets of strings to replace which are oldstring and 
   #oldstring1 so I use two replace lines.

Save the file, close gedit, then exit your terminal or just close it and then start it to be able load the new script you added.

Navigate to the directory where you have multiple files you want to edit. Then run:


Press enter and this will automatically replace the oldstring and oldstring1 in all the files that contain them with the correct newstring and newstring1 respectively.

It will skip all the directories and files that don't contain the old strings.

This might help in eliminating the tedious work of typing in case you have multiple directories with files that need string replacement. All you have to do is navigate to each of those directories, then run:


All you have to do is to ensure you've included or added all the replacement strings as I showed you above:

replace "oldstring" "newstring" -- *

at the end of the file /bin/replace_string_files_present_dir.

To add a new replace string just open the script we created by typing the following in the terminal:

sudo gedit /bin/replace_string_files_present_dir

Don't worry about the number of replace strings you add, they will have no effect if the oldstring is not found.

2023-01-22 23:52:09
Usually, when people ask "how to ${whatever}" in bash, they are asking for a compact version to include in a script or CI jobs instruction (say a .gitlab-ci.yml or a travis.yml). Every turn around consisting in writing a script to execute it later is an anti-pattern, for you'll then need to script the script creation (and I goes messy, most of the time)
2023-01-22 23:52:09
@zar3bski wut? You really should be putting your individual CI steps in separate shell files so that they are portable to other CI systems. It also makes them testable locally.

I used ag, the_silver_searcher:

ag -0 -l 'old' | xargs -0 sed -ri.bak -e 's/old/new/g';

Then git clean to remove .bak files (rm was being buggy when running inside a git rebase exec)

git clean -f '**/*.bak';

My problem with many of the answers was that I needed to replace a filepath inside of many files. Though one answer provided mentioned this, it did not work for me. My solution:

First, generate a list of the filenames that you want to be changed.

filelist=($(find /path/to/your/folder | xargs grep '/path/to/fix' | cut -d : -f 1 | tr '\n' ' '))

What the commands above do is that the find piped to grep generates the names of the files with the /path/to/fix inside. However, grep also prints out the line that the string was found on, so the cut command gets rid of this and just keeps the filename. tr replaces newline characters with spaces, which allows for filelist to be stored as an array.

for file in "${filelist[@]}"; do sed -i.bak 's+/path/to/fix+/new/path/for/my/file+g' $file; done

This sed command draws on other answers to this question, and uses + as delimiters rather than the normal /, since the / characters are being used in the filepath.