package ResultsDataPruned;

use strict;
use warnings;
use diagnostics;

use FindBin;
use lib "$FindBin::Bin/../Modules";
use Filters;

use feature 'say';

sub query {
    my ( $experiment, $experiment_dir, @filter_config_names ) = @_;

    my $extra =
      Filters::build_filter_clause( $experiment_dir, @filter_config_names );

    # This query basically asks: How many equivalence classes
    # ended up in each possible outcome?
    # In contrast to ResultsData.pm it doesn't do any "de-pruning", e.g.
    # no expansion of the "data liveness interval length"
    # SUM(t.time2 - t.time1 + 1).
    #
    # We run with --inject-single-bit, so the bits of one class can
    # end up in different outcomes.
    # This would be counted multiple times,
    # so we'll have more results than pilots.
    #
    # The column is named "faults" so the charts still work with this.
    my $querystring = "SELECT
benchmark, resulttype, COUNT(DISTINCT g.pilot_id) AS faults
FROM variant v
JOIN trace t ON v.id = t.variant_id
JOIN fspgroup g ON g.variant_id = t.variant_id AND g.instr2 = t.instr2 AND g.data_physical_address = t.data_physical_address
JOIN result_GenericExperimentMessage r ON r.pilot_id = g.pilot_id
JOIN fsppilot p ON r.pilot_id = p.id
WHERE v.variant = '$experiment'$extra
GROUP BY v.id, resulttype
ORDER BY variant, benchmark, resulttype;";

    say $querystring;

    return $querystring;
}

sub args { return "--batch --raw"; }

sub filename {
    my @filter_config_names = grep { defined && length } @_;
    my $suffix =
      @filter_config_names ? "_" . join( "+", sort @filter_config_names ) : "";
    return "resultsdata_pruned${suffix}.csv";
}

sub postprocess { $_[0] =~ s/\t/,/g; }

1;