in reply to Get unique fields from file

I didn't see specifically what you are having trouble with. It didn't seem like there was a specific question.

It is correct that using a hash is a good approach. I think that the uniq function is probably not where I would immediately go because of the fact that you want unique values per field, and in order to use uniq for that, you would have to hold the whole file in memory at once (even if there's a high rate of duplication within fields). Rather, I would do the unique filtering early on, line by line. That way, if there are a lot of collisions within any given field, you're only holding onto one instance, which could be a lot more memory friendly.

You're dealing with a style of CSV. It's '|' separated csv, so |sv, but I prefer using a CSV parser for that so that I don't have to deal with the intricacies of embedded / escaped separators. The Text::CSV module can grab the headers for you and can already pre-organize the data into header => value pairs for you. Here's an example of how that could look:

#!/usr/bin/env perl use strict; use warnings; use Text::CSV; my $unique_within_field = csv_to_unique_within_field(\*DATA, '|'); print "$_: ", join(', ', @{$unique_within_field->{$_}}), "\n" foreach sort keys %{$unique_within_field}; sub csv_to_unique_within_field { my($data_fh, $sep) = @_; my $csv = Text::CSV->new({}); $csv->header($data_fh, {sep_set => [$sep // ',']}); my %found; while (my $row = $csv->getline_hr($data_fh)) { $found{$_}{$row->{$_}} = undef for keys %$row; } return { map { $_ => [sort keys %{$found{$_}}] } keys %found }; } __DATA__ head1|head2|head3 val1|val2|val3 val1|val4|val5 val6|val4|val5 val2|val7|val5 val3|val7|val3

The meat here is within the csv_to_unique_within_field function. Pass into the function a filehandle, and a separator. If no separator is provided, assume comma.

The function does this:

  1. Grab the CSV header. In my example, the header identifies fields named head1, head2, and head3.
  2. For each remaining row of CSV data, populate our %found hash with hash keys for each field value within each header. I'm giving the keys 'undef' as the value to which they point. It's not important what they contain. We're using the keys as a uniqueness filter.
  3. Return a new, transformed hash where each header key points a reference to an array containing the unique values.

After this I just print the headers and all the fields they contained. Since we filtered-in only unique per header, it's just a straighforward datastructure print.