205

I want to remove duplicate entries from a text file, e.g:

kavitha= Tue Feb    20 14:00 19 IST 2012  (duplicate entry) 
sree=Tue Jan  20 14:05 19 IST 2012  
divya = Tue Jan  20 14:20 19 IST 2012  
anusha=Tue Jan 20 14:45 19 IST 2012 
kavitha= Tue Feb    20 14:00 19 IST 2012 (duplicate entry)

Is there any possible way to remove the duplicate entries using a Bash script?

Desired output

kavitha= Tue Feb    20 14:00 19 IST 2012 
sree=Tue Jan  20 14:05 19 IST 2012  
divya = Tue Jan  20 14:20 19 IST 2012  
anusha=Tue Jan 20 14:45 19 IST 2012
Błażej Michalik
  • 3,727
  • 34
  • 47
divz
  • 7,597
  • 23
  • 52
  • 74

4 Answers4

466

You can sort then uniq:

$ sort -u input.txt

Or use awk:

$ awk '!a[$0]++' input.txt
Hugo
  • 25,519
  • 6
  • 77
  • 92
kev
  • 146,428
  • 41
  • 264
  • 265
  • 65
    Testing with an 18,500 line text file: `sort ...` takes about 0.57s whereas `awk ...` takes about 0.08s because `awk ...` just removes duplicates without sorting. – Hugo Oct 19 '13 at 12:38
  • 5
    @Hugo I can second that. Testing against 2,626,198 lines `awk` beats `sort`. Results show `awk` taking 5.675s and `sort` taking 5.675s. Interestingly enough the same record set took 15.1 seconds to perform a MySQL DISTINCT query on. – Tegan Snyder Feb 11 '16 at 19:13
  • @TeganSnyder You wrote the both commands took exactly same time to execute. Didn't `awk` take less time? – jarno May 17 '16 at 09:59
  • @jarno - my apologizes that was copy paste error on my part. I would need to recreate the test to see how much faster `awk` was, but it was negligible. – Tegan Snyder May 17 '16 at 14:40
  • 1
    @Hugo is there an elegant way to make this work with case insensitive? or is it better to just convert the entire doc to lowercase, then run this? – Onichan Jun 09 '16 at 02:55
  • @Onichan Try something like this: `echo -e "c\nb\nB\na" | LC_COLLATE=C -u input.txt` http://superuser.com/q/178171/83235 – Hugo Jun 09 '16 at 07:46
  • 6
    tested with 24 million rows, awk did not come to a result within 20 minutes. sort + uniq did the job in some secs. – bhelm Jul 04 '16 at 15:12
  • 6
    I downvoted this because, although poster is happy, folks could be confused by an answer that does not yield the desired output, as it sorts the input – lab419 Dec 17 '17 at 16:54
  • tried it like this: for i in $(ls .*ini); do awk '!a[$0]++' $i; done and it erased some files entirely – CIsForCookies Apr 04 '19 at 16:38
19

It deletes duplicate, consecutive lines from a file (emulates "uniq").
First line in a set of duplicate lines is kept, rest are deleted.

sed '$!N; /^\(.*\)\n\1$/!P; D'
Siva Charan
  • 17,624
  • 9
  • 58
  • 94
12

Perl one-liner similar to @kev's awk solution:

perl -ne 'print if ! $a{$_}++' input

This variation removes trailing whitespace before comparing:

perl -lne 's/\s*$//; print if ! $a{$_}++' input

This variation edits the file in-place:

perl -i -ne 'print if ! $a{$_}++' input

This variation edits the file in-place, and makes a backup input.bak

perl -i.bak -ne 'print if ! $a{$_}++' input
Chris Koknat
  • 3,082
  • 2
  • 28
  • 29
  • 1
    I like the Perl solution because it allows me to add extra conditions, e.g. only enforce uniqueness on lines matching a certain pattern. – Amos Shapira Oct 11 '18 at 04:10
0

This might work for you:

cat -n file.txt |
sort -u -k2,7 |
sort -n |
sed 's/.*\t/    /;s/\([0-9]\{4\}\).*/\1/'

or this:

 awk '{line=substr($0,1,match($0,/[0-9][0-9][0-9][0-9]/)+3);sub(/^/,"    ",line);if(!dup[line]++)print line}' file.txt
potong
  • 51,370
  • 6
  • 49
  • 80