Bash sed in loop

Question

I'm trying to use the follow to substitute the tab with comma in several file:

#!/bin/sh 
for i in *output_*.txt  
do   
sed 's/ /;/g' $i > $i 
done

But it is not working because in the output file I still have the tab delimiter. It just work when I'm using it on a single file without the for loop.

Any help?

Thanks.

How are you inserting the tab character? bash takes tab as a completion, which can cause issues. Most sed's support \t for a tab, else tab=$'\t' then sed "s/$tab/;/g" $i > $i — cdarke
– cdarke, Commented Jun 25, 2012 at 8:14
it looks like you're substituting space for semicolon. try using sed -i 's/\t/,/g' $i. -i option means in place. — Grzegorz Grzybek
– Grzegorz Grzybek, Commented Jun 25, 2012 at 8:15
Try using the -i flag of sed that enables in-place substitution. — jpe
– jpe, Commented Jun 25, 2012 at 8:18
Be careful; I hope you've got copies of your data, because 'sed ... $i > $i' has the shell create an empty output file called $i before sed gets to read it. You cannot edit a file in situ like that. If you have GNU sed, there's a -i option. Otherwise, write to a temporary file and then copy the temporary over the original when done (or move the temporary). — Jonathan Leffler
– Jonathan Leffler, Commented Jun 25, 2012 at 8:18
You seem to be replacing a space (or tab) with a semi-colon, too, but that could just be a confusing layout. — Jonathan Leffler
– Jonathan Leffler, Commented Jun 25, 2012 at 8:20

tvm · Accepted Answer · 2012-06-25 08:15:03Z

3

Several things are wrong. Unqouted variables and output redirection into same file. Loop is also not needed.

Try: sed -i 's/ /;/g' *output_*.txt

answered Jun 25, 2012 at 8:15

tvm

3,45929 silver badges39 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

user1256539 Over a year ago

I have different file with different name so just sed -i 's/ /;/g' output_.txt will not work.

user1256539 Over a year ago

Yes but the problem is that I need to write to file. I was doing this by overwriting the original file. But has someone point out sed first create an empty file.

aps2012 · Accepted Answer · 2012-06-26 06:21:01Z

The correct script you need is as follows:

find . -name '*output_*.txt' | while read FILENAME; do
    (sed -e "s/\\t/;/g" <${FILENAME} >${FILENAME%.txt}.tmp) && (mv ${FILENAME%.txt}.tmp ${FILENAME});
done

This script has several important features:

It finds all files called *output_*.txt in the current directory and all subdirectories. If you do not want to recurse into subdirectories, then use:
```
find . -maxdepth 1 -name '*output_*.txt' | while read FILENAME; do
```
as the first line.
It does not overwrite your original input file if sed encounters an error. sed generates its output to a temporary file (<filename>.tmp) and it only replaces the original file if it is successful.
As pointed out by other posters, the tab character is represented by \t in sed scripts.

An example transformation performed by this script is as follows (the sequence <tab> represents a tab character):

Input:

    <tab><tab><tab><tab><tab>line 1<tab><tab>
    <tab><tab><tab>line 2<tab><tab>
    <tab><tab>line 3<tab><tab>
    <tab><tab><tab>line 4<tab><tab>
    <tab><tab><tab><tab><tab>line<tab><tab> 5

Output:

    ;;;;line 1;;
    ;;;line 2;;
    ;;line 3;;
    ;;;line 4;;
    ;;;;;line;; 5

Collectives™ on Stack Overflow

Bash sed in loop

2 Answers 2

2 Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related