How to remove duplicate words from a plain text file using linux command

Question

I have a plain text file with words, which are separated by comma, for example:

word1, word2, word3, word2, word4, word5, word 3, word6, word7, word3

i want to delete the duplicates and to become:

word1, word2, word3, word4, word5, word6, word7

Any Ideas? I think, egrep can help me, but i'm not sure, how to use it exactly....

Answer 1

I'd think you'll want to replace the spaces with newlines, use the uniq command to find unique lines, then replace the newlines with spaces again.

Answer 2

Assuming that the words are one per line, and the file is already sorted:

uniq filename

If the file's not sorted:

sort filename | uniq

If they're not one per line, and you don't mind them being one per line:

tr -s [:space:] \\n < filename | sort | uniq

That doesn't remove punctuation, though, so maybe you want:

tr -s [:space:][:punct:] \\n < filename | sort | uniq

But that removes the hyphen from hyphenated words. "man tr" for more options.

Answer 3

ruby -pi.bak -e '$_.split(",").uniq.join(",")' filename ?

I'll admit the two kinds of quotations are ugly.

Answer 4

I presumed you wanted the words to be unique on a single line, rather than throughout the file. If this is the case, then the Perl script below will do the trick.

while (<DATA>)
{
    chomp;
    my %seen = ();
    my @words = split(m!,\s*!);
    @words = grep { $seen{$_} ? 0 : ($seen{$_} = 1) } @words;
    print join(", ", @words), "\n";
}

__DATA__
word1, word2, word3, word2, word4, word5, word3, word6, word7, word3

If you want uniqueness over the whole file, you can just move the %seen hash outside the while (){} loop.

Answer 5

Creating a unique list is pretty easy thanks to uniq, although most Unix commands like one entry per line instead of a comma-separated list, so we have to start by converting it to that:

$ sed 's/, /\n/g' filename | sort | uniq
word1
word2
word3
word4
word5
word6
word7

The harder part is putting this on one line again with commas as separators and not terminators. I used a perl one-liner to do this, but if someone has something more idiomatic, please edit me. :)

$ sed 's/, /\n/g' filename | sort | uniq | perl -e '@a = <>; chomp @a; print((join ", ", @a), "\n")'
word1, word2, word3, word4, word5, word6, word7

Answer 6

G'day,

And don't forget the -c option for the uniq utility if you're interested in getting a count of the words as well.

cheers,

Answer 7

Here's an awk script that will leave each line in tact, only removing the duplicate words:

BEGIN { 
     FS=", " 
} 
{ 
    for (i=1; i <= NF; i++) 
        used[$i] = 1
    for (x in used)
        printf "%s, ",x
    printf "\n"
    split("", used)
}