limiteTermes.pl

#!/usr/bin/env perl


# Declaration of pragmas
use strict;
use utf8;
use open qw/:std :utf8/;


# Call of external modules
use Encode qw(decode_utf8 encode_utf8 is_utf8);
use Getopt::Long;

# Perl modules specific for this application
use File::Copy;
use Spreadsheet::Read;
use Spreadsheet::ParseXLSX;

# Programme name
my ($programme) = $0 =~ m|^(?:.*/)?(.+)|;
my $substitute  = " " x length($programme);

my $usage = "Usage: \n" . 
            "    $programme -i raw_input_file -e Excel_file [ -o output_file ] \n" .
            "    $programme -h\n\n";

my $version     = "0.2.2";
my $changeDate  = "August 7, 2019";

# Initialising global variables 
# necessary for options
my $excel      = undef;
my $help       = undef;
my $input      = undef;
my $output     = undef;

eval    {
        $SIG{__WARN__} = sub {usage(1);};
        GetOptions(
                "excel=s"    => \$excel,
                "help"       => \$help,
                "input=s"    => \$input,
                "output=s"   => \$output,
                "xcel=s"     => \$excel,
                );
        };
$SIG{__WARN__} = sub {warn $_[0];};

if ( $help ) {
        print "\nProgramme: \n";
        print "    “$programme”, version $version ($changeDate)\n";
        print "    Modify the original “doc × term” file by deleting unwanted \n";
        print "    terms and replacing others by preferential terms in accordance \n";
        print "    with the Excel file generated by “listeTerms” and checked by \n";
        print "    an expert or at least by the user. \n";
#        print "     \n";
        print "\n";
        print $usage;
        print "\nOptions: \n";
        print "    -e  specify the name of the Excel file \n";
        print "    -h  display this help and exit \n";
        print "    -i  specify the name of the raw “doc × term” input file  \n";
        print "    -o  specify the name of the clean “doc × term” output file \n";

        exit 0;
        }

usage(2) if not $input or not $excel;

# Global variables
my %pref    = ();
my %refused = ();
my %term    = ();

readExcel();

# Opening output file ...
if ( $output ) {
        open(OUT, ">:utf8", $output) or die "$!,";
        }
# ... or sending to the standard output
else    {
        open(OUT, ">&STDOUT") or die "$!,";
        binmode(OUT, ":utf8");
        }

# Reading input file
open(INP, "<:utf8", $input) or die "$!,";
while(<INP>) {
        chomp;
        s/\r//go; # just in case ...
        my ($file, $term) = split(/\t/);
        next if $refused{$term};
        if ( $pref{$term} ) {
                # print OUT "$file\t$pref{$term}\n";
                $term{$file}{$pref{$term}} ++;
                }
        else    {
                # print OUT "$file\t$term\n";
                $term{$file}{$term} ++;
                }
        }
close INP;

foreach my $file (sort keys %term) {
        foreach my $term (sort keys %{$term{$file}}) {
                print OUT "$file\t$term\n";
                }
        }

close OUT;


exit 0;


sub usage
{
print STDERR $usage;

exit shift;
}

sub readExcel
{
my $ref = ReadData ($excel, parser => 'xlsx');

my $nb_sheets = 0;
my $term_list = 0;

if ( defined $ref->[0]{"sheets"} ) {
        $nb_sheets = $ref->[0]{"sheets"};

        for (my $number = 1 ; $number <= $nb_sheets ; $number ++ ) {
                my $sheet = $ref->[$number];
                my $title = $sheet->{"label"};
#                $title = decode_utf8($title);
                my $maxRows = $sheet->{"maxrow"};
                my $maxCols = $sheet->{"maxcol"};

                if ( $title eq "Term list" ) {
                        $term_list ++;
                        my $line = 1;
                        my ($col1, $col4, $col5) = (undef, undef, undef);
                        for ( my $col = 1 ; $col <= $maxCols ; $col ++ ) {
                                my $cell = cr2cell($col, $line);
#                                my $text = decode_utf8($sheet->{$cell});
                                my $text = $sheet->{$cell};
                                if ( $text eq '?' )  {
                                        $col1 = $col;
                                        }
                                elsif ( $text eq 'Term' )  {
                                        $col4 = $col;
                                        }
                                elsif ( $text eq 'Replace by' )  {
                                        $col5 = $col;
                                        }
                                }
                        if ( $col1 and $col4 ) {
                                for ( $line = 2 ; $line <= $maxRows ; $line ++ ) {
                                        my $cell = cr2cell($col4, $line);
#                                        my $term = decode_utf8($sheet->{$cell});
                                        my $term = $sheet->{$cell};
                                        next if not $term;
                                        $cell = cr2cell($col1, $line);
#                                        my $status = decode_utf8($sheet->{$cell});
                                        my $status = $sheet->{$cell};
                                        if ( $status eq 'x' )  {
                                                $refused{$term} ++;
                                                next;
                                                }
                                        $cell = cr2cell($col5, $line);
#                                        my $replace = decode_utf8($sheet->{$cell});
                                        my $replace = $sheet->{$cell};
                                        if ( $replace ) {
                                                $pref{$term} = $replace;
                                                }
                                        }
                                }
                        }
                }

        if ( not $term_list ) {
                print STDERR "Error: no term list in Excel file \"excel\"\n";
                exit 3;
                }
        }
else    {
        print STDERR "Erroe: Excel file \"$excel\" is empty\n";
        exit 4;
        }
}