Retrieve unique values from column based on value from other column - arrays

Question

I have a table like this
symbol length id
A 10 id_1
A 15 id_2
A 15 id_3
B 20 id_4
B 25 id_5
... ... ...
I want to print the following in a new table
symbol length id
A 15 id_2; id_3
B 25 id_5
... ... ...
So I want to loop through the symbol column. When there are duplicate values in this column, I want to print the line where the numeric length-value is the greatest (example: symbol B). When the greatest length values are equal, I want to merge the values in the idcolumn (example: symbol A) and print this new line.
How should I do this in perl?

Answer 1

The tool in perl for coalescing duplicates is a hash. Hashes are key-value pairs, but the useful part is - the value can be an array (reference).
I'd be suggesting something like this:
#!/usr/bin/perl
use strict;
use warnings;
my %length_of;
my %ids_of;
my $heading_row = <DATA>;
while (<DATA>) {
my ( $symbol, $length, $id ) = split;
if ( not defined $length_of{$symbol} or $length_of{$symbol} < $length ) {
$length_of{$symbol} = $length;
}
push( #{ $ids_of{$symbol}{$length} }, $id );
}
print join( "\t", "symbol", "length", "ids" ), "\n";
foreach my $symbol ( sort keys %ids_of ) {
my $length = $length_of{$symbol};
print join( "\t",
$symbol,
$length,
join( "; ", #{ $ids_of{$symbol}{$length} } ) ),
"\n";
}
__DATA__
symbol length id
A 10 id_1
A 15 id_2
A 15 id_3
B 20 id_4
B 25 id_5
What this is doing is - iterating your data, and saving the highest length value (in %length_of). It's also stashing each of the ids - by symbol and length (in %ids_of). It keeps them all, so this might not be very efficient if you've a lot of data.

Answer 2

Just remember the last symbol and length and accumulate the ids:
#! /usr/bin/perl
use warnings;
use strict;
my ($last_l, $last_s, #i);
sub out {
print "$last_s\t$last_l\t", join(";", #i), "\n"
}
while (<>) {
my ($s, $l, $i) = split;
out() if $last_s and $s ne $last_s;
undef #i if $last_l < $l;
push #i, $i;
$last_s = $s;
$last_l = $l;
}
out();

Answer 3

This approach builds a hash of hashes of arrays by using the values from the symbol and length columns as keys and adding the values from the id column as array references. For the simple dataset you provided such a complex data structure is not really needed, but the approach shown below might be more flexible in cases where data is not sorted.
I use the max function from (List::Util , which is part of the core distribution) to get the maximum length value for each symbol, and Data::Dumper to help visualize things.
use Data::Dumper ;
use List::Util 'max';
use v5.16;
my (%hash, #lines) ;
while ( <DATA>) {
chomp ;
next if $. == 1 ;
push #lines, [ split ] ;
}
for (#lines) {
push #{ $hash{ $_->[0] }{ $_->[1] } }, $_->[2] ;
}
say "This is your %hash:\n", Dumper \%hash;
for my $symbol ( keys %hash ) {
my $max = max ( keys $hash{$symbol} ) ;
say "$symbol \t", "$max \t", join "; ", #{ $hash{$symbol}{$max} };
}
__DATA__
symbol length id
A 10 id_1
A 15 id_2
A 15 id_3
B 20 id_4
B 25 id_5
Output:
This is your %hash:
$VAR1 = {
'A' => {
'10' => [
'id_1'
],
'15' => [
'id_2',
'id_3'
]
},
'B' => {
'25' => [
'id_5'
],
'20' => [
'id_4'
]
}
};
A 15 id_2; id_3
B 25 id_5