How to replace the leading tab (including tabs after some space characters) only?

Question

I have some files that are indented by tabs or spaces or both. I want to convert leading tabs to spaces (one tab to 4 spaces). This includes tabs after some leading spaces. Below are some examples of some input line and the expected result.

+-----+---------------+-----------------+
| No. | Original line | Expected result |
+-----+---------------+-----------------+
|   1 | \t␣xxx        | ␣␣␣␣␣xxx       |
|   2 | ␣␣␣\txxx      | ␣␣␣␣␣␣␣xxx     |
|   3 | \t␣\txxx      | ␣␣␣␣␣␣␣␣␣xxx   |
|   4 | \tx\txx       | ␣␣␣␣x\txx       |
+-----+---------------+-----------------+

I can't use the expand command here, as it can't handle the cases when spaces and tabs are mixed. Below is an example.

user1@ubuntu$ printf "\t  \txxx" | od -t a
0000000  ht  sp  sp  ht   x   x   x
0000007
user1@ubuntu$ printf "\t  \txxx" | expand -i -t 4 | od -t a
0000000  sp  sp  sp  sp  sp  sp  sp  sp   x   x   x
0000013
user1@ubuntu$ 

As you can see, the two spaces in the original string were simply dropped. How to solve my problem? I've read other similar questions on this site, but they are not exactly the same as my problem.

Answer 1

You can use sed (I use cat -T to display the file because it shows tabs as ^I):

$ cat -T file
^I abc
^I  ^Ixde^Inot
$ sed ':x;s|^\( *\)\t|\1    |;tx' file | cat -T
     abc
          xde^Inot

Unfortunately, interpretation of \t as tab is a GNU sed extension and not POSIX specified. However, you can work around that using printf as Gilles shows here and write:

sed ":x;s|^\( *\)$(printf '\t')|\1    |;tx" file

---

What sed is doing there?

• s|^\( *\)\t|\1 |

If sed finds a tab followed by zero or more space characters anchored to the line beginning, it substitutes the tab by 4 spaces. The pair \(\) delimits the matching group (which consists of zero or more preceding space characters) that is reproduced by \1.

• tx

If a substitution was made, go to label :x. Otherwise, proceed to next line.

Answer 2

To replace all tabs that come before the first non-space character with 4 spaces, try:

perl -pe '/^(\s+)/; $k=$1; $k=~s/\t/    /g; s/^\s+/$k/'file > newfile

The script will first find all leading whitespace (spaces, tabs, and anything else) and save it as $k. It then replaces all tabs with 4 spaces in $k and finally replaces all leading whitespace with the current value of $k.

To limit to spaces and tabs only, you can do:

perl -pe '/^([ \t]+)/; $k=$1; $k=~s/\t/    /g; s/^\s+/$k/'file > newfile

Run on your example strings, these solutions produce:

$  printf '\t xxx\n   \txxx\n\t \txxx\n\tx\txx\n'  | perl -pe '/^(\s+)/; $k=$1; $k=~s/\t/    /g; s/^\s+/$k/' | od -t a
0000000  sp  sp  sp  sp  sp   x   x   x  nl  sp  sp  sp  sp  sp  sp  sp
0000020   x   x   x  nl  sp  sp  sp  sp  sp  sp  sp  sp  sp   x   x   x
0000040  nl  sp  sp  sp  sp   x  ht   x   x  nl
0000052