barbar's post looked like this when I approved it:

I have been struggling with sorting and merging data for some time. I managed to find a script by roboticus in Perl monks: node_id=596095 Which said:

mergefile.1
[download]

15    20    foo
22    30    bar
30    33    baz
14    22    fubar
[download]

mergefile.2
[download]

alpha    baz    17.30
gamma    foobar    22.35
gamma    bar    19.01
delta    fromish    33.03
sigma    bear    14.56
[download]

mergefile.out
bar     22      30      gamma   19.01
baz     30      33      alpha   17.30
bear    null    null    sigma   14.56
foo     15      20      null    null
foobar  null    null    gamma   22.35
fromish null    null    delta   33.03
fubar   14      22      null    null
[download]

The code used to merge mergefile.1 and mergefile.2 to create mergefile.out is below

#!/usr/bin/perl -w
use strict;
use warnings;
[download]

open F1, 'sort -k3 mergefile.1|' or die "opening file 1";
open F2, 'sort -k2 mergefile.2|' or die "opening file 2";
[download]

open OUF, '>', 'mergefile.out' or die "opening output file";
[download]

my @in1;
my @in2;
[download]

sub getrec1 {
        @in1 = ();
        if (!eof(F1)) {
                (@in1) = split /\t/, <F1>;
                chomp $in1[2];
        }
}
[download]

sub getrec2 {
        @in2 = ();
        if (!eof(F2)) {
                (@in2) = split /\t/, <F2>;
                chomp $in2[2];
        }
}
[download]

sub write1 {
        print OUF "$in1[2]\t$in1[0]\t$in1[1]\tnull\tnull\n";
        getrec1;
}
[download]

sub write2 {
        print OUF "$in2[1]\tnull\tnull\t$in2[0]\t$in2[2]\n";
        getrec2;
}
[download]

sub writeboth {
        print OUF "$in1[2]\t$in1[0]\t$in1[1]\t$in2[0]\t$in2[2]\n";
        getrec1;
        getrec2;
}
[download]

# Prime the pump
getrec1;
getrec2;
[download]

while (1) {
        last if $#in1<0 and $#in2<0;
[download]

        if ($#in1<0 or $#in2<0) {
                # Only one file is left...
                write2 if $#in1<0;
                write1 if $#in2<0;
        }
        elsif ($in1[2] eq $in2[1]) {
                # Matching records, merge & write 'em
                writeboth;
        }
        elsif ($in1[2] lt $in2[1]) {
                # unmatched item in file 1, write it & get next rec
                write1;
        }
        else {
                # unmatched item in file 2, write it & get next rec
                write2;
        }
}
[download]

My question is - how can I get this to work.??

I have these files saved in a diractory and when I try to run this from Unix command it errors: "Input file specified two times."

If I run this in Korn Shell it comes up with the warning:" sort: last character not record delimiter" and the output is:

bar    null    null    gamma    19.01
bar    22    30    null    null
baz    null    null    alpha    17.30
baz    30    33    null    null
bear    null    null    sigma    14.56
foo    15    20    null    null
foobar    null    null    gamma    22.35
fromish    null    null    delta    33.03
fubar    14    22    null    null
[download]

Which shows that the code is not working.

I altered the line "chomp $in2[2];" to read "chomp $in2[1];"
[download]

I changed the delimiters in the files to be commas, and changed the script from \t to ,

I really think this script will be able to solve my poblem of sorting and merging large files - if only I could get it to work and understand why it was not working in the first place.

Please can anyone help me by giving me any pointers?????

(I happened to have left the browser tab open and that is how I produced this copy.)

Note to barbar: Please don't go in and completely alter a node once posted. It messes with people's heads. And mine is definitely messed up enough already. Thank you.

marto:

He was referring to an old post of mine. I found a /msg from him on CB when I logged in. Lucky for me, the issue was resolved by the time I saw the question. 8^)

...roboticus

#!/usr/bin/perl -w
use strict;
use warnings;

open F1, 'sort -k3 mergefile1|' or die "opening file 1";
open F2, 'sort -k2 mergefile2|' or die "opening file 2";

open OUF, '>', 'mergefile.out' or die "opening output file";

my @in1;
my @in2;

sub getrec1 {
        @in1 = ();
        if (!eof(F1)) {
                (@in1) = split /\t/, <F1>;
                chomp $in1[2];
        }
}

sub getrec2 {
        @in2 = ();
        if (!eof(F2)) {
                (@in2) = split /\t/, <F2>;
                chomp $in2[1];
        }
}

sub write1 {
        print OUF "write1 $in1[2]\t$in1[0]\t$in1[1]\tnull\tnull\n";
        getrec1;
}

sub write2 {
        print OUF "write2 $in2[1]\tnull\tnull\t$in2[0]\t$in2[2]\n";
        getrec2;
}

sub writeboth {
        print OUF "writeboth $in1[2]\t$in1[0]\t$in1[1]\t$in2[0]\t$in2[
+2]\n";
        getrec1;
        getrec2;
}

# Prime the pump
getrec1;
getrec2;

while (1) {
        last if $#in1<0 and $#in2<0;

        if ($#in1<0 or $#in2<0) {
                # Only one file is left...
                write2 if $#in1<0;
                write1 if $#in2<0;
        }
        elsif ($in1[2] eq $in2[1]) {
                # Matching records, merge & write 'em
                writeboth;
        }
        elsif ($in1[2] lt $in2[1]) {
                # unmatched item in file 1, write it & get next rec
                write1;
        }
        else {
                # unmatched item in file 2, write it & get next rec
                write2;
        }
}
[download]

15    20    foo    
22    30    bar    
30    33    baz    
14    22    fubar
[download]

alpha    baz    17.30    
gamma    foobar    22.35    
gamma    bar    19.01    
delta    fromish    33.03    
sigma    bear    14.56
[download]

I'm not sure why this "only works on Korn shell and not in command, but there are two possibilities:

Traditionally, Unix shells perform argument wildcard expansion (globbing) on the command line, while command.com and cmd.exe, the shells on Windows do not. If you want to perform globbing yourself, use the glob version, best as follows:

# Near the top of your program
use File::Glob qw(bsd_glob); # sane whitespace handling

my @files = glob('../INPUT/*');
[download]

The other possibility is that you have two programs called sort.exe and $ENV{PATH} is set differently between your ksh and your cmd.exe. The sort.exe program that comes with Windows is incompatible with the sort.exe program that behaves like the Unixish programs do. The easiest way to make sure that the "right" sort.exe program is invoked is to give the full path, maybe C:\\programs\\cygwin\\usr\\bin\\sort.exe, instead of calling it implicitly.