UNIX - Compare 2 files based on primary column

Question

I need to compare 2 files columnwise based on primary columns(it could be 1 or multiple columns as a primate key). And it should generate 3 csv files as output - differences , extra records in file1, extra records in file2

Note : tried with sdiff but it doesnt give output as desired

Example :

Here first column is a primary key

file1 :
abc 234 123
bcd 567 890
cde 678 789

file2 :
abc 234 012
bcd 532 890
cdf 678 789

Output files

differences file :
abc,234,123::012
bcd,567::532,890

extra records in file1 :
cde,678,789

extra records in file2
cdf,678,789

check the `comm` binary, a standard unix command line tool `comm --help` — knb, Mar 20 '19 at 13:39
comm will give unmatched records agreed but wont highlight the differences are in which column — Aru, Mar 22 '19 at 08:46
you can transform the data into "long format" first, check the R-tidyverse documentation for details. — knb, Mar 22 '19 at 10:28

jhnc · Answer 1 · 2019-03-24T21:01:31.380

If the files comfortably fit in memory, it's quite easy to do with hashes in Perl. For example:

#!/bin/bash

# create test data files
>cmp.d1 cat <<'EOD'
abc 234 123
bcd 567 890
cde 678 789
EOD
>cmp.d2 cat <<'EOD'
abc 234 012
bcd 532 890
cdf 678 789
EOD

# create script
>dif.pl cat <<'EOD'
#!/usr/bin/perl -w

if ( $#ARGV!=0 or ! -f "$ARGV[0]" ) {
    die "Usage: <file2 filter file1\n";
}

@KEYS = ( 0 ); # list of columns to use for primary key

# read file1 from filename given on commandline
while (<<>>) {
    chomp;
    @a1 = (split); # split line into individual fields
    $k = join "\0", @a1[ @KEYS ];

    # if $k is not unique, only final line is kept
    warn "duplicate key: $k\n" if exists $h1{$k};

    # store line in %h1 for later use
    $h1{$k} = [ @a1 ];
}

# now read file2 from stdin
# process each line as we read it
while (<<>>) {
    chomp;
    @a2 = (split); # split line into individual fields
    $k = join "\0", @a2[ @KEYS ];

    if ( exists $h1{$k} ) {
        # record exists in both files
        # calculate differences 

        @a1 = @{ $h1{$k} }; # retrieve file1 version

        # overwrite any difference fields in @a2
        map {
            $a1 = shift @a1;
            $_ = "${a1}::$_" if $a1 ne $_;
        } @a2;

        # save difference records in %hd
        $hd{$k} = [ @a2 ];

        # this will not be an extra file1 record
        delete $h1{$k};
    }
    else {
        # this record only exists in file2
        $h2{$k} = [ @a2 ];
    }
}

# format record as csv line
sub print_csv {
    print join(",", @{ $_ }), "\n";
}

print "differences file :\n";
print_csv for values %hd;
print "\n";

print "extra records in file1 :\n";
print_csv for values %h1;
print "\n";

print "extra records in file2\n";
print_csv for values %h2;

EOD

# try it out
perl dif.pl cmp.d1 <cmp.d2

Output:

differences file :
bcd,567::532,890
abc,234,123::012

extra records in file1 :
cde,678,789

extra records in file2
cdf,678,789

Note: csv output doesn't normally need to be ordered, so this code doesn't do any sorting.

score 0 · Answer 2 · answered Mar 25 '19 at 13:40

Try this command line Perl

perl -lane ' @t=@{$kv1{$F[0]}}; push(@t,$_); $kv1{$F[0]} = [@t];
if( defined($kv2{$F[0]}) ) {  $kv2{$F[0]} = "Both" } else { $kv2{$F[0]} =$ARGV; $kv3{$F[0]}=$_; }
END { 

 for my $c (keys %kv2) 
 { 
   if($kv2{$c} eq "Both") { $d1++ or print "differences file :";  

   @t=@{$kv1{$c}}; @s1=split(" ",$t[0]); @s2=split(" ",$t[1]);
   $a2= $s1[1] eq $s2[1] ? $s1[1] : $s1[1]. "::". $s2[1];
   $a3= $s1[2] eq $s2[2] ? $s1[2] : $s1[2]. "::". $s2[2];
   print $s1[0],",",$a2,",",$a3;

   }
 }

 for my $c (keys %kv2) 
 { 
   if($kv2{$c} eq "file1") { $d2++ or print "\nextra records in file1 :";  print $kv3{$c} }
 }

 for my $c (keys %kv2) 
 { 
   if($kv2{$c} eq "file2") { $d3++ or print "\nextra records in file2 :";  print $kv3{$c} }
 }

}
' file1 file2

Results:

differences file :
bcd,567::532,890
abc,234,123::012

extra records in file1 :
cde 678 789

extra records in file2 :
cdf 678 789

UNIX - Compare 2 files based on primary column

2 Answers2