How Do I Replace A Text String In Many Files At Once?

by on November 12, 2005 · 21 comments· LAST UPDATED July 10, 2011

in , ,

The replace command is a string-replacement utility. It changes strings in place in files or on the standard input. This command uses a finite state machine to match longer strings first. It can be used to swap strings. This command is similar to the Perl -pie syntax or sed (stream editor) command.

Please note that the replace command is part of is MySQL database system. If you don't have MySQL installed, you don’t have replace command.

Syntax

 
replace OLD-STRING NEW-STRING < INPUT-FILE > OUTPUT-FILE
 

Examples

To replace all occurrences of word UNIX with Linux, enter:
$ replace UNIX Linux < oldfile > newfile

The replace command can be used in a pipeline, run:
$ cat /etc/passwd | replace : '|'

You can skip the cat command, enter:
$ replace : '|' < /etc/passwd

It also supports few special characters in string replacement:

  • \^ : Match start of line.
  • $ : Match end of line.

How Do I Update All *.txt Files At Once?

You use bash for loop as follows:

#!/bin/bash
for f in /path/to/*.txt
do
   replace UNIX Linux < "$f" > "$f.new"
done
 

The replace command does not understand regular expression. To use regular expression try the sed command or Perl.

Sed Command Example

To replace all occurrences of word UNIX with Linux using the sed command, enter:

 
sed 's/UNIX/Linux/g' < input.file > output.file
 

OR

 
sed -i 's/UNIX/Linux/g' input.file
 

OR use bash shell for loop as follows to update all *.doc files at once:

#!/bin/bash
for f in /path/to/*.doc
do
   sed -i  's/UNIX/Linux/g' "$f"
done
 

Updated for accuracy!

TwitterFacebookGoogle+PDF versionFound an error/typo on this page? Help us!

{ 21 comments… read them below or add one }

1 Joe November 12, 2005 at 11:30 pm

To use perl, just remember Perl Pie!

perl -p -i -e ‘s/hello/goodbye/g’ textfile.txt

Reply

2 LinuxTitli November 12, 2005 at 11:47 pm

Joe, very nice. Your tip sound yummy :) thanks for sharing with us

Reply

3 Anonymous November 30, 2005 at 9:34 am

http://www.debian-administration.org/articles/298 has a fine article and discussion on Perl Pie.

Reply

4 Anonymous December 12, 2005 at 12:13 am

what about this line:

perl -p -i -e ‘s/|00000000.00|/||/g’ myfile.txt

I want to replace |00000000.00| with ||

I get a compilation error.

Reply

5 nixcraft December 12, 2005 at 1:41 am

You need to write it as follows:

perl -p -i -e ‘s/|00000000.00|/||/g’ myfile.txt

|| got some special meaning (regex) | will disable it

Reply

6 Anonymous December 12, 2005 at 9:02 pm

This does the trick perfectly! Thanks for sharing this special character technique nixcraft!

Reply

7 walter August 9, 2007 at 6:59 am

or try this…
first make a bash script, ‘fixer.sh’


#!/bin/bash
replace CHANGEFROM CHANGETO $1.tmp
rm $1
mv $1.tmp $1

now run this command line…

$ grep CHANGEFROM |cut -d':' -f1 |xargs -n 1 fixer.sh

the results is that all files in the directory (or whatever you grep for) will be changed automagically.

just make sure the grep doesn’t include the fixer script itself, or it will die half-way through changing when execute permissions are reset!

;)

Reply

8 Raj April 10, 2008 at 1:25 pm

thanks for sharing this info

Reply

9 lefty.crupps June 16, 2008 at 4:28 pm

Am I the only one who sees no difference in these?

>> perl -p -i -e ’s/|00000000.00|/||/g’ myfile.txt
>> I want to replace |00000000.00| with ||
>> I get a compilation error.

— —
>> You need to write it as follows:
>> perl -p -i -e ’s/|00000000.00|/||/g’ myfile.txt

Reply

10 Eden February 11, 2009 at 12:21 am

Cant see the diference… :(

Reply

11 2046 March 31, 2009 at 6:09 pm

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

Reply

12 Matthew Scott July 18, 2010 at 3:43 am

`replace`

Reply

13 Matthew Scott July 18, 2010 at 3:46 am

`replace` can work with files, which may be simpler than writing a shell script as mentioned above. It will even convert files in place. Say you want to change an instance of ‘foo’ to ‘bar’ in all files in a certain directory, recursively. In bash,

for i in `grep -lR foo dir/to/files`; do replace foo bar — $i; done

Simply put, use two dashes to separate filenames from the from/to strings, and it’ll convert the files.

Reply

14 DougTheBug August 17, 2010 at 10:07 pm

I couldn’t even find this command in Ubunto 10.04 (or in repositories). I ended up using rpl instead. Is really fun for databending images.

Reply

15 Ben March 7, 2011 at 4:23 am

‘Replace’ is found in mySQL. If you don’t have mySQL, you don’t have `replace`

Reply

16 saravanan January 4, 2012 at 2:35 pm

I want to replace a string in sub directories also(in Linux), is there any specific command for that?

Reply

17 Jay March 23, 2012 at 5:54 pm

hi,
thanks for the info..nice script. just wanted to add that when i used to ‘for’ loop version w/ ‘sed’ command (in Linux) it would read all the “*.sh” files that i wanted to substitue a string but did not actually make the change. just an fyi, i removed the single quote (‘) in the sed line and it works.

See below:
#!/bin/bash
for f in /path/to/*.doc
do
sed -i s/UNIX/Linux/g “$f”
done

Thanks

Reply

18 SAURABH SHARMA April 1, 2012 at 7:04 am

for i in `find . -type f`
do
j=`echo $i|sed ‘s/$1/$2/g’`
mv $i $j
done;

Reply

19 Nathan April 20, 2012 at 12:28 am

I noticed in my bash shell I need to do the following on my Mac:

sed -i ” ‘s/UNIX/Linux/g’ input.file

Otherwise I get stdin error.

Reply

20 drankinatty April 5, 2013 at 5:37 pm

Your example below will fail on filenames containing spaces due to the default internal field separator (IFS) breaking on a space character (default=$’ \t\n’):

#!/bin/bash
for f in /path/to/*.txt
do
   replace UNIX Linux  "$f.new"
done

To properly handle filenames with spaces, you should set IFS to break on new lines only:

#!/bin/bash
oldIFS=IFS
IFS=$'\n'
for f in /path/to/*.txt
do
   replace UNIX Linux  "$f.new"
done
IFS=oldIFS

You can also simply run the command in a subshell by enclosing it in parenthesis and not have to worry about resetting IFS at the completion of the loop. Eg.:

(IFS=$'\n'; for f in /path/to/*.txt; do replace UNIX Linux  "$f.new"; done)

However, while replace is a valid solution, I have found sed to be the most robust and much faster than perl for large search and replace tasks.

Reply

21 asdds June 17, 2013 at 1:11 pm

Thanks!

Reply

Leave a Comment

Tagged as: , , , , , , , , , , , , , ,

Previous post:

Next post: