#!/usr/bin/perl
# Time-stamp: <2004-06-09 11:12:38 barre>
#
# Build cross-references between classes and examples
#
# barre : Sebastien Barre <sebastien@barre.nom.fr>
#
# 0.81 (barre) :
#   - add --baselinedir d : use 'd' as baseline directory
#   - add --baselineicon f: use 'f' as icon file to report that file has
#     baseline picture
#   - add --baselinelink l: use 'l' as baseline link to picture
#   - add --baselinelinksuffix s : suffix string to append
#     to --baselinelink + filename
#   - add --datamatch s : use string s to match any usage of data files
#
# 0.8 (barre) :
#   - more robust code: comments are removed from files before parsing
#     and several identifier per lines are now supported (at last).
#   - add --datamatch s : use string s to match any usage of data files
#   - add --dataicon f : use f as icon file to report that file makes use
#     of data files
#
# 0.74 (barre) :
#   - use common build_page_doc proc
#
# 0.73 (barre) :
#   - add --linksuffix s : suffix string to append to --link + filename
#
# 0.72 (barre) :
#   - add --project name : project name, used to uniquify
#   - add slightly changed the way --link are created
#
# 0.71 (barre) :
#   - add .cpp and .cc extension to the C++ parser regexp
#
# 0.7 (barre) :
#   - update to match the new VTK 4.0 tree
#   - change default --dirs so that it can be launched from Utilities/Doxygen
#   - change default --to so that it can be launched from Utilities/Doxygen
#   - add Java support
#   - add --label str : use string as label in class page
#   - add --title str : use string as title in "Related Pages"
#   - add --dirmatch str : use string to match the directory name holding files
#   - add --unique str : use string as a unique page identifier (otherwise MD5)
#
# 0.6 (barre) :
#   - the "class to example" page is now split in different pages
#   - use --weight to increase or decrease the maximum weight of a page
#
# 0.55 (barre) :
#   - change default --to to '../vtk-doxygen' to comply with Kitware's doxyfile
#
# 0.54 (barre) :
#   - change doxygen command style from \ to @ to match javadoc, autodoc, etc.
#
# 0.53 (barre) :
#   - change default --to to '../vtk-dox'
#
# 0.52 (barre) :
#   - rename -l option to --link, add a parameter to preprend in front of
#     the example location/filename.
#
# 0.51 (barre) :
#   - fix O_TEXT flag problem
#   - switch to Unix CR/LF format
#
# 0.5 (barre) :
#   - removed version extraction feature (moved to another script)
#
# 0.4 (barre) :
#   - change (warning) default --to to '../vtk2' because I ruined my own
#     VTK distrib too many times :(
#   - change (warning) --to is now a path to a destination directory, and
#     no more a path to the destination doxygen file. The headers to update
#     will logically be searched here, while the examples to process are
#     searched in the directory where the script is launched
#
# 0.3 (barre)
#   - add '--parser' option to use specific parser only
#
# 0.2 (barre)
#   - use vtkVersion.h to fill the main page (@mainpage) with version/revision
#     => comment/remove PROJECT_NUMBER from your doxygen config file (doxyfile)
#   - add web URL to main page
#   - fix small "uninitialized variable" warning
#
# 0.1 (barre)
#   - first release (-l does not work for the moment)

use Carp;
use Data::Dumper;
use Digest::MD5  qw(md5 md5_hex md5_base64);
use Getopt::Long;
use Fcntl;
use File::Basename;
use File::Find;
use strict;

my ($VERSION, $PROGNAME, $AUTHOR) = (0.81, $0, "Sebastien Barre");
$PROGNAME =~ s/^.*[\\\/]//;
print "$PROGNAME $VERSION, by $AUTHOR\n";

# -------------------------------------------------------------------------
# Defaults (add options as you want : "verbose" => 1 for default verbose mode)

my %default =
  (
   baselineicon => "pic.gif",
   baselinelinksuffix => "",
   dataicon => "paper-clip.gif",
   dirmatch => "^Examples\$",
   dirs => ["../.."],
   label => "Examples",
   limit => 20,
   linksuffix => "",
   project => "VTK",
   store => "doc_VTK_class2examples.dox",
   title => "Class To Examples",
   to => "../../../VTK-doxygen",
   unique => "e",
   weight => 90000,
   remove_leading_slash => 0
  );

# -------------------------------------------------------------------------
# Matchers and parsers :
#
# $eliminate_matcher : regexp matching the names of the 'fake' classes
#                      that shall be eliminated/ignored,
# %parsers           : hash defining each parser by associating
#                      parser_name => [filename_matcher, func1, func2].
#  $parser_name      : language identifier (Tcl, C++, Python, etc.),
#  $filename_matcher : regexp matching the filenames belonging to the language,
# @$func1            : function called when a filename is matched,
#                      - receives the contents of the file as a reference to a
#                        string,
#                      - modifies this contents by removing the comments.
# @$func2            : function called when a filename is matched,
#                      - receives the contents of the file as a reference to a
#                        string (or the result of func1 if exists),
#                      - returns an array of unique class names that have been
#                        recognized/matched in the file contents.

my $eliminate_matcher = '^vtkCommand$';

my %parsers = (
               "Tcl" => ['\.tcl$', \&rm_tcl_comments, \&parse],
               "C++" => ['\.c(xx|pp|c)$', \&rm_cpp_comments, \&parse],
               "Java" => ['\.java$', \&rm_cpp_comments, \&parse],
               "Python" => ['\.py$', \&rm_tcl_comments, \&parse]
              );

sub rm_tcl_comments {
    my $ref = shift;
    $$ref =~ s/#.*?$//gms;
}

sub rm_cpp_comments {
    my $ref = shift;
    $$ref =~ s/\/\*.*?\*\/|\/\/.*?$|#(include|define|if).*?$//gms;
}

sub rm_no_comments {
    my $ref = shift;
}

# Parser (seems to work with all languages for the moment)

sub unique {
    my ($ref, %uniques) = (shift, ());
    foreach my $item (@$ref) {
        $uniques{$item}++;
    }
    return keys %uniques;
}

sub parse {
    my $ref = shift;
    my @matches = $$ref =~ /\b(vtk[A-Z0-9][A-Za-z0-9]+)[^"]/gms;
    return unique(\@matches);
}

# -------------------------------------------------------------------------
# Parse options

my %args;
# Getopt::Long::Configure("bundling");
GetOptions (\%args, "help", "verbose|v", "baselinedir=s", "baselineicon=s", "datamatch=s", "baselinelink=s", "baselinelinksuffix=s", "dataicon=s", "dirmatch=s", "label=s", "limit=i", "link=s", "linksuffix=s", "project=s", "store=s", "title=s", "to=s", "unique=s", "weight=i", "parser=s@", "remove_leading_slash");

my $available_parser = join(", ", keys %parsers);

if (exists $args{"help"}) {
    print <<"EOT";
by $AUTHOR
Usage : $PROGNAME [--help] [--verbose|-v] [--baselinedir path] [--baselineicon filename] [--baselinelink url] [--baselinelinksuffix url] [--datamatch string] [--dataicon filename] [--dirmatch string] [--label string] [--limit n] [--link url] [--linksuffix string] [--parser name] [--store file] [--title string] [--to path] [--weight n] [--remove_leading_slash] [directories...]
  --help          : this message
  --verbose|-v    : verbose (display filenames/classes while processing)
  --baselinedir d : use 'd' as baseline directory
  --baselineicon f: use 'f' as icon file to report that file has baseline picture
  --baselinelink l: use 'l' as baseline link to picture
  --baselinelinksuffix s : suffix string to append to --baselinelink + filename
  --datamatch s   : use string s to match any usage of data files
  --dataicon f    : use 'f  as icon file to report that file makes use of data files
  --dirmatch str  : use string to match the directory name holding files (default: $default{dirmatch})
  --label str     : use string as label in class page (default: $default{label})
  --limit n       : limit the number of examples per parser type (default: $default{limit})
  --link path     : link to example files (and prepend path)
  --linksuffix s  : suffix string to append to --link + filename
  --title str     : use string as title in "Related Pages" (default: $default{title})
  --parser name   : use specific parser only (available : $available_parser)
  --project name  : project name, used to uniquify (default: $default{project})
  --store file    : use 'file' to store xrefs (default: $default{store})
  --to path       : use 'path' as destination directory (default : $default{to})
  --unique str    : use string as a unique page identifier among "Class To..." pages (otherwise MD5) (default : $default{unique})
  --weight n      : use 'n' as an approximation of the maximum page weight (default : $default{weight})
  --remove_leading_slash: remove any leading slash in filename used after --link and before --linksuffix (default: $default{remove_leading_slash})

Example:
  $PROGNAME --verbose
  $PROGNAME --dirmatch "^Testing$" --label "Tests" --title "Class To Tests" --store "doc_class2tests.dox" --unique "t"
EOT
    exit;
}

$args{"verbose"} = 1 if exists $default{"verbose"};
$args{"baselinedir"} =~ s/[\\\/]*$// if exists $args{"baselinedir"};
$args{"baselineicon"} = $default{"baselineicon"}
  if ! exists $args{"baselineicon"};
$args{"baselinelink"} = $default{"baselinelink"}
  if ! exists $args{"baselinelink"} && exists $default{"baselinelink"};
$args{"baselinelink"} =~ s/[\\\/]*$// if exists $args{"baselinelink"};
$args{"baselinelinksuffix"} = $default{"baselinelinksuffix"}
  if ! exists $args{"baselinelinksuffix"};
$args{"dataicon"} = $default{"dataicon"} if ! exists $args{"dataicon"};
$args{"dirmatch"} = $default{"dirmatch"} if ! exists $args{"dirmatch"};
$args{"label"} = $default{"label"} if ! exists $args{"label"};
$args{"limit"} = $default{"limit"} if ! exists $args{"limit"};
$args{"link"} = $default{"link"} if ! exists $args{"link"} && exists $default{"link"};
$args{"link"} =~ s/[\\\/]*$// if exists $args{"link"};
$args{"linksuffix"} = $default{"linksuffix"} if ! exists $args{"linksuffix"};
$args{"project"} = $default{"project"} if ! exists $args{"project"};
$args{"store"} = $default{"store"} if ! exists $args{"store"};
$args{"title"} = $default{"title"} if ! exists $args{"title"};
$args{"to"} = $default{"to"} if ! exists $args{"to"};
$args{"to"} =~ s/[\\\/]*$// if exists $args{"to"};
$args{"unique"} = $default{"unique"} if ! exists $args{"unique"};
$args{"weight"} = $default{"weight"} if ! exists $args{"weight"};
$args{"remove_leading_slash"} = $default{"remove_leading_slash"} if ! exists $args{"remove_leading_slash"};

# Select parsers

my @parsers;
if (exists $args{"parser"}) {
    foreach my $parser (@{$args{"parser"}}) {
        if (exists $parsers{$parser}) {
            push @parsers, $parser;
        } else {
            carp "$PROGNAME: unknown parser : $parser\n";
        }
    }
} else {
    @parsers = keys %parsers;
}

my $os_is_win = ($^O =~ m/(MSWin32|Cygwin)/i);
my $open_file_as_text = $os_is_win ? O_TEXT : 0;
my $start_time = time();

# -------------------------------------------------------------------------
# Collect unique non-intersecting (top) examples' directories recursively
# Save current path (pwd) to avoid the ::prune bug

print "Collecting files...\n";

push @ARGV, @{$default{"dirs"}} if !@ARGV;
my (%seen, @dirs);

my $cwd = Cwd::cwd();

foreach my $file (@ARGV) {
    find sub {
        if (-d $_ &&
            $_ ne "CVS" &&
            basename($File::Find::name) =~ m/$args{"dirmatch"}/i) {
            # my ($dev, $ino) = stat $_;
            push @dirs, $File::Find::name;
            # if ! $seen{$dev, $ino}++;
            # once we are in a "examples directory", don't descend anymore
            $File::Find::prune = 1;
        }
    }, $file;
}

chdir($cwd);

# -------------------------------------------------------------------------
# Collect files within directories recursively

my @files;
foreach my $dir (@dirs) {
    find sub {
        push @files, $File::Find::name if -f $_;
    }, $dir;
}

# -------------------------------------------------------------------------
# Select unique files matching available parsers

my @parsable;
foreach my $parser (@parsers) {
    foreach my $file (@files) {
        if ($file =~ m/$parsers{$parser}->[0]/) {
            # my ($dev, $ino) = stat $file;
            push @parsable, $file;
            # if ! $seen{$dev, $ino}++;
        }
    }
}

@parsable = sort @parsable;

print " => ", scalar @parsable, " file(s) collected in ", time() - $start_time, " s.\n";

# -------------------------------------------------------------------------
# Parse files and build xref

# %xref is a hash : xref{$class}{$parser}{$file}
# Ex: xref{"vtkPoints"}{"Tcl"}{"examplesTcl/foo.tcl"}
#     xref{"vtkPoints"}{"Tcl"}{"graphics/examplesTcl/bar.tcl"}
#     xref{"vtkNormals"}{"C++"}{"test.cxx"}

my %xref;

# %shorter_filename is a hash associating a filename to its shorter
# counterpart where any leading component matching the name of a directory
# being browsed have been removed.

my %shorter_filename;

# %use_data is a hash associating a filename to a flag (1) when the
# file makes use of data files (i.e. the contents of the file matches
# $args{"datamatch"}).

my %use_data;

# %baseline_picture is a hash associating a filename to the relative path
# to its baseline picture (relative to --baselinedir).

my %baseline_picture;

print "Parsing files...\n";

my $intermediate_time = time();

# slurp mode

my $oldslurp = $/;
undef $/;

foreach my $filename (@parsable) {

    # Read entire file into an array of lines

    sysopen(FILE, $filename, O_RDONLY|$open_file_as_text)
      or croak "$PROGNAME: unable to open $filename\n";
    my $file = <FILE>;
    close(FILE);

    # Submit the contents of the file to the corresponding parser

    foreach my $parser (@parsers) {
        if ($filename =~ m/$parsers{$parser}->[0]/) {
            $parsers{$parser}->[1]->(\$file);
            my @classes = $parsers{$parser}->[2]->(\$file);
            printf("%7s: %2d | ", $parser, scalar @classes)
              if exists $args{"verbose"};
            #print "[", join(", ", @classes), "]"  if exists $args{"verbose"};
            if (@classes) {
                $shorter_filename{$filename} = $filename;
                # print "(", join(", ", sort @classes), ") " if @classes;
                foreach my $class (@classes) {
                    $xref{$class}{$parser}{$filename}++;
                }
            }
        }
    }
    print "=> ", $filename, "\n" if exists $args{"verbose"};

    # Shorten filename if it has been added

    if (exists $shorter_filename{$filename}) {
        foreach my $dir (@ARGV) {
            last if $shorter_filename{$filename} =~ s/$dir//;
        }

        # Check for baseline picture

        if (exists $args{"baselinedir"}) {
            my ($name, $dir, $ext) =
              fileparse($shorter_filename{$filename}, '\..*');
            my @dirs = split('/', $dir);
            my $pic = $dirs[1] . "/$name.png";
            $baseline_picture{$filename} = $pic
              if -e $args{"baselinedir"} . "/$pic";
        }
    }
    $shorter_filename{$filename} =~ s/^\/// if exists $args{"remove_leading_slash"};

    # Check for data

    $use_data{$filename} = 1
      if exists $args{"datamatch"} && $file =~ m/$args{"datamatch"}/ms;
}

$/ = $oldslurp;

print " => ", scalar @parsable, " file(s) parsed in ", time() - $intermediate_time, " s.\n";

# -------------------------------------------------------------------------
# Eliminate some 'fake' classes

print "Eliminating some classes...\n";

$intermediate_time = time();
my @eliminated = ();

foreach my $class (keys %xref) {
    if ($class =~ m/$eliminate_matcher/) {
        print "   $class\n" if exists $args{"verbose"};
        delete($xref{$class});
        push @eliminated, $class;
    }
}

print " => ", scalar @eliminated, " class(es) eliminated (", join(", ", @eliminated), ") in ", time() - $intermediate_time, " s.\n";

# -------------------------------------------------------------------------
# We need to locate the C++ header of each class to add a link from
# that header to the example page. We also remove orphan classes with
# no headers.

# %headers is a hash associating a class name to the location of its header.
#   ex: $headers{"vtkPoints"} = "Common/vtkPoints.h"
# %headers_not_found is a hash associating a class name to a flag that is
#   true (1) if the location of its header has not been found.

print "Locating headers to update...\n";

$intermediate_time = time();
my (%headers, %headers_not_found);

# Let's state that no headers have been found at the moment

foreach (keys %xref) {
    $headers_not_found{$_} = 1;
}
my $headers_not_found_nb = scalar keys %headers_not_found;

$cwd = Cwd::cwd();

find sub {
    if ($headers_not_found_nb == 0) {
        # All headers have been found, let's stop descending the tree
        $File::Find::prune = 1;

    } elsif (-f $_ && $_ =~ /^(vtk[A-Z0-9][A-Za-z0-9]+)\.h$/) {
        # A class header has been found, let's check if it matches one pf
        # the class in xref
        my $class = $1;
        if (exists $headers_not_found{$class}) {
            print "   $class : $File::Find::name\n" if exists $args{"verbose"};
            $headers{$class} = $File::Find::name;
            $headers_not_found_nb--;
            delete($headers_not_found{$class});
        }
    }
}, $args{"to"};

chdir($cwd);

# Collect these classes that have not been associated to a header and
# remove them from xref.

my @still_not_found = keys %headers_not_found;
foreach my $not_found (@still_not_found) {
    print "   $not_found : not found (removed)\n" if exists $args{"verbose"};
    delete($xref{$not_found});
}

print " => ", scalar keys %headers, " found, ", scalar @still_not_found, " orphan class(es) removed (", join(", ", @still_not_found), ") in ", time() - $intermediate_time, " s.\n";

# -------------------------------------------------------------------------
# Build the page summary documentation

# $indent is the indentation string

my $indent = "    ";

# @words is the array of words to document

my @words = sort keys %xref;

# $header is the Doxygen string summarizing what has been documented as well
# as the credits.

my $header;
my (@summary, @credits, @legend);

push @summary,
  "  - " . scalar @words . " class(es) in " .
  scalar @parsable . " file(s) from directories matching \@c " . $args{"dirmatch"} . " on " . localtime();

push @summary,
  "  - " . scalar @parsers . " parser(s) : [" . join(", ", @parsers) . "]";

push @summary,
  "  - at most " . $args{"limit"} . " file(s) per parser";

push @credits,
  "\@version $VERSION",
  "\@author \@c $PROGNAME, by $AUTHOR";

$header = $indent . join("\n$indent", @summary) . "\n\n" .
  $indent . join("\n$indent", @credits) . "\n\n";

my $footer;

if (exists $args{"datamatch"}) {
    push @legend,
    "- <img src='" . $args{"dataicon"} . "' align='top'> : the corresponding file uses data files.";
    # hack so that the image is automatically copied to the doc dir
    $footer .= " \@image html " . $args{"dataicon"} . "\n";
}

if (exists $args{"baselinedir"}) {
    push @legend,
    "- <img src='" . $args{"baselineicon"} . "' align='top'> : the corresponding file has a baseline picture (click to display)";
    # hack so that the image is automatically copied to the doc dir
    $footer .= " \@image html " . $args{"baselineicon"} . "\n";
}

if (scalar @legend) {
    unshift @legend, "\@par Legend:";
    $header .= $indent . join("\n$indent", @legend) . "\n\n";
}

# -------------------------------------------------------------------------
# Index to class

print "Building page doc...\n";

# $prefix is a unique prefix that is appended to each link

my $prefix = "c2_" . $args{"project"} . "_";
if (exists $args{"unique"}) {
    $prefix .= $args{"unique"};
} else {
    $prefix .= md5_hex($args{"label"} . $args{"title"});
}
$prefix = lc($prefix);

# word_section_name returns the short string describing a word section

sub word_section_name {
    my ($word) = @_;
    return "\@anchor ${prefix}_$word\n$indent$word";
}

# word_section_doc returns the doxygen doc for a word

sub word_section_doc {
    my ($word) = @_;
    my @temp;
    foreach my $parser (sort keys %{$xref{$word}}) {
        push @temp, "  - $parser";
        my $count = 0;
        # @files if the set of files found for that class by that parser
        my @files = sort keys %{$xref{$word}{$parser}};
        foreach my $file (@files) {
            my $has_data;
            $has_data = ' @htmlonly <img src="' . $args{"dataicon"} . '" align="top"> @endhtmlonly' if exists $use_data{$file};
            my $has_baseline_picture;
            $has_baseline_picture = ' @htmlonly <a href="' . $args{"baselinelink"} . '/' . $baseline_picture{$file} . $args{"baselinelinksuffix"} . '"><img src="' . $args{"baselineicon"} . '" align="top" border="0"></a> @endhtmlonly' if exists $baseline_picture{$file};
            last if ++$count > $args{"limit"};
            if (exists $args{"link"}) {
                push @temp,
                '    - @htmlonly <TT><A href="' . $args{"link"} .
                  $shorter_filename{$file} . $args{"linksuffix"} . '">@endhtmlonly ' . $shorter_filename{$file} .
                    '@htmlonly</A></TT> @endhtmlonly ' .
                      $has_data . $has_baseline_picture;
            } else {
                push @temp, "    - \@c $shorter_filename{$file} $has_data $has_baseline_picture";
            }
        }
    }
    return join("\n$indent", @temp) . "\n";
}

# word_section_alpha returns the single alpha char corresponding to that
# word's section.

sub word_section_alpha {
    my ($word) = @_;
    $word =~ /^vtk(\w)/;
    return $1;
}

my $page_doc = build_page_doc($indent,
                              $args{"title"},
                              \@words,
                              $prefix,
                              \&word_section_name,
                              \&word_section_doc,
                              \&word_section_alpha,
                              $header,
                              $footer,
                              $args{"to"} . "/" . $args{"store"});

print join("\n", @summary), "\n => in ", time() - $intermediate_time, " s.\n";

# -------------------------------------------------------------------------
# Update class headers (add a link from the class header to the example page).
# Do not update if the section is already there.

print "Updating headers...\n";

$intermediate_time = time();
my $updated_nb = 0;

foreach my $class (@words) {

    print "   $class => " . $headers{$class} . "\n" if exists $args{"verbose"};

    sysopen(HEADER, $headers{$class}, O_RDONLY|$open_file_as_text)
      or croak "$PROGNAME: unable to open " . $headers{$class} . "\n";

    # Search for the documentation block (@class ...)

    my @dest = ();
    my $line;
    while ($line = <HEADER>) {
        push @dest, $line;
        last if $line =~ /\s\@class\s+$class/;
    }

    # Search for the end of the block (*/), and check if the xref is not
    # already there (@par      $args{"label"})

    if (defined $line) {
        while ($line = <HEADER>) {
            last if $line =~ /^\*\// || $line =~ /^\s*\@par\s+$args{"label"}:\s*$/;
            push @dest, $line;
        }

        # If the example section was not found, insert the xref to the
        # example page, read the rest of the file, and overwrite the
        # header

        if (defined $line && $line =~ /^\*\//) {
            push @dest, "\n    \@par      " . $args{"label"} . ":\n",
                    "              \@ref ${prefix}_$class \"$class (" . $args{"label"} . ")\"\n",
                        $line;
            while ($line = <HEADER>) {
                push @dest, $line;
            }
            close(HEADER);
            sysopen(HEADER, $headers{$class},
                    O_WRONLY|O_TRUNC|O_CREAT|$open_file_as_text)
              or croak "$PROGNAME: unable to open " . $headers{$class} . "\n";
            print HEADER @dest;
            $updated_nb++;
        }
    }
    close(HEADER);
}

print " => $updated_nb header(s) updated in ", time() - $intermediate_time, " s.\n";

print "Finished in ", time() - $start_time, " s.\n";


# -------------------------------------------------------------------------

sub build_page_doc {

    # $indent is the indentation string
    # $rwords is a reference to the array of words to document
    # $prefix is a unique prefix that is appended to each link
    # word_section_name returns the short string describing a word section
    # word_section_doc returns the doxygen doc for a word
    # word_section_alpha returns the single alpha char corresponding to that
    # word's section.
    # $header is the Doxygen string summarizing what has been documented as
    # well as the credits.
    # $footer is a Doxygen string appended to each the resulting page
    # $destination_file is the name of the file where this page should be
    # written to.

    my ($ident, $title, $rwords, $prefix, $rword_section_name, $rword_section_doc, $rword_section_alpha, $header, $footer, $destination_file) = @_;

    # %words_doc is a hash associating a word to its Doxygen doc (string)

    my %words_doc;

    # %sections_words is a hash associating a section (alphabetical letter) to
    # an array of words belonging to that section.
    #   Ex: $sections_words{"C"} => ("contour", "cut")
    # %sections_weight is a hash associating a section to its weight (the sum
    # of the weights of each word belonging to that section).
    # @sections is the array holding the name of all sections

    my (%sections_words, %sections_weight, @sections);

    # $navbar is the Doxygen string describing the sections' navigation bar

    my $navbar;

    my $intermediate_time = time();

    # Browse each word

    foreach my $word (@$rwords) {

        my @temp;
        push @temp, &$rword_section_name($word), &$rword_section_doc($word);
        $words_doc{$word} = $indent . join("\n$indent", @temp) . "\n";

        # Update section(s) and section(s) weight(s)

        my $section = &$rword_section_alpha($word);
        push @{$sections_words{$section}}, $word;
        $sections_weight{$section} += length($words_doc{$word});

        print " => ", $word, "\n" if exists $args{"verbose"};
    }

    print " => ", scalar @$rwords, " words(s) documented in ", time() - $intermediate_time, " s.\n";

    @sections = sort keys %sections_words;

    # Build the navbar

    my @temp;
    foreach my $section (@sections) {
        push @temp, "\@ref ${prefix}_section_$section \"$section\"";
    }
    $navbar = "$indent\@par Navigation: \n$indent\[" .
      join(" | ", @temp) . "]\n";

    # Add the (approximate) weight of the (header + navbar) to each section

    my $total_weight = 0;
    my $header_weight = length($indent) + 24 + length($navbar);

    foreach my $section (@sections) {
        $sections_weight{$section} += $header_weight;
        $total_weight += $sections_weight{$section};
    }

    if (exists $args{"verbose"}) {
        foreach my $section (@sections) {
            printf("\t- %s : %6d\n", $section, $sections_weight{$section});
        }
    }

    print " => total weight is $total_weight in ", scalar @sections, " section(s)\n";

    print " => mean is ", int($total_weight / scalar @sections), ")\n"
      if scalar @sections;

    # Compute the alphabetical groups by joining sections depending on weights

    print "Computing alphabetical group(s)/page(s)...\n";

    # %groups is a hash associating a group id (int) to an array of sections
    # namesbelonging to that group.
    #   Ex: $groups{"0"} => ("A", "B", "C")
    # %groups_weight is a hash associating a group id to its weight (the sum
    # of the weights of each section belonging to that group).

    my (%groups, %groups_weight);

    my $groupid = 0;

    # Remove a section one by one, and put it in a group until the group if
    # full,then create a next group, etc., until the sections are exhausted.

    my @sections_temp = @sections;
    while (@sections_temp) {
        $groups_weight{$groupid} = $sections_weight{$sections_temp[0]};
        push @{$groups{$groupid}}, shift @sections_temp;
        while (@sections_temp &&
               ($groups_weight{$groupid} +$sections_weight{$sections_temp[0]})
               <= $args{"weight"}) {
            $groups_weight{$groupid} += $sections_weight{$sections_temp[0]};
            push @{$groups{$groupid}}, shift @sections_temp;
        }
        $groupid++;
    }

    if (exists $args{"verbose"}) {
        foreach my $groupid (sort {$a <=> $b} keys %groups) {
            printf("\t- %02d (weight: %7d) : %s\n", $groupid,
                   $groups_weight{$groupid}, join(", ", @{$groups{$groupid}}));
        }
    }

    print " => max weight is ", $args{"weight"}, " per group/page, but a section can not be divided\n";
    print " => ", scalar  keys %groups, " group(s) for ", scalar @sections, " section(s)\n";

    # Build documentation page
    # Browse each group, each section in this group, each word in this section

    my $page_doc;

    foreach my $groupid (sort {$a <=> $b} keys %groups) {

        my $fromto = $groups{$groupid}[0];
        $fromto .= ".." . $groups{$groupid}[scalar @{$groups{$groupid}} - 1]
          if scalar @{$groups{$groupid}} > 1;

        $page_doc .=
          "/*! \@page ${prefix}_$groupid $title ($fromto)\n\n$header";

        foreach my $section (@{$groups{$groupid}}) {
            $page_doc .=
              "\n$indent\@section ${prefix}_section_$section $section\n\n$navbar\n";
            foreach my $word (@{$sections_words{$section}}) {
                $page_doc .= $words_doc{$word}, "\n";
            }
            print "\t- $section\n" if exists $args{"verbose"};
        }
        $page_doc .= "$footer\n*/\n\n";
    }

    print "Writing documentation to ", $destination_file, "...\n";

    $intermediate_time = time();

    sysopen(DEST_FILE,
            $destination_file,
            O_WRONLY|O_TRUNC|O_CREAT|$open_file_as_text)
     or croak "$PROGNAME: unable to open destination file $destination_file\n";
    print DEST_FILE $page_doc;
    close(DEST_FILE);

    print " => written in ", time() - $intermediate_time, " s.\n";
}