I'm stuck in including regular expressions with a sed command.

Q: I want to replace all occurences of two spaces after the end of a sentence with just once space.

Here is what I did:

sed 's/^ $/^$/' file  

And it didn't substituted two spaces with a one space after the sentence ends.

Output I get:

This is the output.  Hello Hello

Output I want:

This is the output. Hello Hello
  • @Rahul I want the output with that contains once space after the sentence ends. Not two spaces as in first sentence. I want the second sentence output
    – Zeus
    Jul 22 '16 at 5:36
  • The answer by @Jasen is correct. BTW, the command you tried means: Substitute a line consisting of only a single whitspace by an empty line. This is why it did not do what you wanted. Jul 22 '16 at 8:35
  • 1
    Insert here: war on single vs. double space after full stop
    – gerrit
    Jul 22 '16 at 17:22
  • @gerrit That war was lost some 20 years ago. ;)
    – h4ckNinja
    Jul 23 '16 at 6:50

Your sed command 's/^ $/^$/' won't do what you want. It just replace all lines contains one space with a line contain ^$.

Depend on what characters mark end of sentence, you can do:

sed -e 's/\([.?!]\) \{2,\}/\1 /g' <file

This will replace 2 or more spaces after ., ? or ! with one space only.

  • 1
    can be shortened to sed -r 's/([.?!])\s+/\1 /g' file
    – Sundeep
    Jul 22 '16 at 5:52
  • Thanks. But would you kindly elaborate why so many round, square brackets you used? It's little confusing.
    – Zeus
    Jul 22 '16 at 6:03
  • 1
    @spasic: That's work with GNU sed only, updated with simplified version.
    – cuonglm
    Jul 22 '16 at 6:05
  • @Zeus: That's standard syntax with default BRE.
    – cuonglm
    Jul 22 '16 at 6:05
  • 1
    No, the original sed editing script will replace any line containing a single space with the literal string ^$.
    – Kusalananda
    Jul 22 '16 at 6:47
 sed 's/\.   */. /g' < file

replace dot followed by two or more spaces with dot followed by a single space.

  • Your command actually works for the whole paragraph. That's exactly what I wanted. Thanks!
    – Zeus
    Jul 22 '16 at 5:59
  • or sed -r 's/\.\s+/. /g' file with GNU sed...
    – Sundeep
    Jul 22 '16 at 6:08
  • @spasic use sed -E rather than sed -r. That works with GNU sed and *BSD sed (incl. Mac OS X) and some others. It's scheduled for inclusion as standard in POSIX "real soon now".
    – cas
    Jul 23 '16 at 7:15
  • @cas I remember sometime back, I suggested sed -E to someone on this site which didn't work but sed -r worked.. I've worked with GNU sed alone, so no idea on POSIX and other variations.. will try to include this detail in future, thanks :)
    – Sundeep
    Jul 23 '16 at 7:36
  • It must have been an old version of GNU sed. They've supported -E for quite a while now (since at least 2011, but the man and info pages still only mention -r)
    – cas
    Jul 23 '16 at 9:12

This is what you might be looking for,

tr -s " " <filename


$ echo "This is the output.  Hello Hello" | tr -s "[:blank:]"
This is the output. Hello Hello

Using sed,

$ echo "This is the output.  Hello Hello" | sed 's/\. \+/. /g'
$ echo "This is the output.  Hello Hello" | sed 's/\. \{1,\}/. /g'
This is the output. Hello Hello
  • 2
    Well, it is working thanks, but I need to include the sed command. Kindly tell something similar to above shown, like substitution, change text etc in sed.
    – Zeus
    Jul 22 '16 at 5:38
  • 2
    I wasn't aware of this tr feature,
    – Jasen
    Jul 22 '16 at 5:41
  • 3
    This approach will replace two spaces which are not end of sentence, too.
    – cuonglm
    Jul 22 '16 at 5:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.