#!/usr/bin/perl
# Copyright (C) 2018 TTK Ciar
# Available for unlimited distribution and use.
# The copyright is just so someone else cannot claim ownership and sue me for use of my own code.

our $VERSION = '1.00'; # VERSION 1.00

use strict;
use warnings;
use JSON::PP;
use lib "./lib";
use Algorithm::CurveFit::Simple qw(fit);

my @DOCS;
my %OPT = (v => 1);
foreach my $arg (@ARGV) {
    if    ($arg =~ /^\-+(.+?)\=(.*)/) { $OPT{$1} = $2; }
    elsif ($arg =~ /^\-+(v+)$/      ) { $OPT{v}  = length($1) + 1; }
    elsif ($arg =~ /^\-+q$/         ) { $OPT{v}  = 0;  }
    elsif ($arg =~ /^\-+quiet$/     ) { $OPT{v}  = 0;  }
    elsif ($arg =~ /^\-+(.+)/       ) { $OPT{$1} = -1; }
    else { push (@DOCS, $arg); }
}
foreach my $k (keys %OPT) {
    next unless ($k =~ /\-/);
    my $k1 = join('_', split(/\-/, $k));
    $OPT{$k1} = $OPT{$k};
    delete $OPT{$k};
}

my $PROJECT_NAME = 'curvefit';

exit(main(\@DOCS, \%OPT));

sub main {
    my ($docs_ar, $opt_hr) = @_;
    usage() if (opt('h') || opt('help'));

    my (@xdata, @ydata);
    while(defined(my $datum = <STDIN>)) {
        next unless($datum =~ /^\s*([\-+]?[\d\.]+)[\t,]([\-+]?[\d\.]+)/);
        push @xdata, $1;
        push @ydata, $2;
    }

    my ($max_dev, $avg_dev, $src) = fit(xdata => \@xdata, ydata => \@ydata, %OPT);
    print STDERR "$max_dev\n$avg_dev\n" if (opt('v'));
    print STDERR JSON::PP::encode_json(\%Algorithm::CurveFit::Simple::STATS_H)."\n" if (opt('d') || opt('debug') || opt('profile'));
    print "$src\n";
    return 0;
}

sub opt {
    my ($name, $default_value, $alt_hr) = @_;
    return def($OPT{$name}, $alt_hr->{$name}, $default_value);
}

sub def {
    foreach my $v (@_) { return $v if (defined($v)); }
    return undef;
}

sub logger {
    return if (opt('no-log') || opt('log',1) == 0);
    my $log_ar = [localtime(), Time::HiRes::time(), $$, @_];
    my $log_rec = JSON::to_json($log_ar, {ascii => 1, space_after => 1, canonical => 1})."\n";
    print STDERR $log_rec if (opt('show-log'));
    print STDOUT $log_rec if (opt('show-log-to-stdout'));
    File::Valet::ap_f(opt('logfile',"/home/ttk/$PROJECT_NAME.log"), $log_rec) unless(opt('no-logfile'));
    return;
}

sub usage {
    print <<USAGE;
Usage: $0 [options] < data
Input must be x,y data pairs, one pair per line, separated by a comma or tab.
Options and their defaults, if any:
    --time-limit=3    Maximum number of seconds to spend calculating best fit
    --iterations=#    Maximum number of iterations to spend calculating best fit (default is to use a time limit)
    --terms=3         Number of terms in polynomial, max 10
    --inv             Invert the sense of the fit to f(y) = x
    --impl-lang=perl  Language used for output implementation: perl, C
    --impl-name=x2y   Name of function in output implementation
    --bounds-check    Implementation will check for out-of-bounds input
    --round-result    Implementation will round output to nearest integer
    --suppress-includes (C only) Do not put #include directives in output implementation
    --quiet           Do not write supplementary information to stderr
    --profile         Dump %STATS_H to stderr as json
See also: Algorithm::CurveFit::Simple
USAGE
    exit(0);
}

=head1 NAME

curvefit - Fit a polynomial to data points

=head1 SYNOPSIS

  Usage: curvefit [options] < data

  Expects x,y data pairs on STDIN, one pair per line, separated by a comma or tab.

      --time-limit=3    Maximum number of seconds to spend calculating best fit
      --iterations=#    Maximum number of iterations to spend calculating best fit (default is to use a time limit)
      --terms=3         Number of terms in polynomial, max 10
      --inv             Invert the sense of the fit to f(y) = x
      --impl-lang=perl  Language used for output implementation: perl, C
      --impl-name=x2y   Name of function in output implementation
      --bounds-check    Implementation will check for out-of-bounds input
      --round-result    Implementation will round output to nearest integer
      --suppress-includes (C only) Do not put #include directives in output implementation
      --quiet           Do not write supplementary information to stderr
      --profile         Dump %STATS_H to STDERR as json

=head1 DESCRIPTION

This is a thin wrapper around L<Algorithm::CurveFit::Simple>, which is in turn a convenience wrapper around L<Algorithm::CurveFit>.

Given a set of x,y data pairs on STDIN, it will generate a polynomial formula f(x) = y which fits that data, and write a source code implementation of that formula to STDOUT.

Additionally it will write a maximum deviation and average deviation to STDERR.  Closer to 1.0 is better.  Play with --terms=# until these deviations are as close to 1.0 as possible, and beware overfitting.  Use --quiet to suppress this information.

=head1 SEE ALSO

L<Algorithm::CurveFit::Simple>

=cut
