#!/usr/bin/perl

# This is free software.  You may redistribute copies of it under the terms of
# the GNU General Public License <http://www.gnu.org/licenses/gpl.html>.
# There is NO WARRANTY, to the extent permitted by law.

# This script was originally written by Ken Estes Mail.com
# kestes@staff.mail.com

# a simple script to print the proper name for Perl libraries.

# It does not parse the perl grammar but instead just lex it looking for
# what we want. It takes special care to ignore comments and pod's.

# The filenames to scan are either passed on the command line or if
# that is empty they are passed via stdin.

# If there are lines in the file which match the pattern
#      (m/^\s*\$VERSION\s*=\s+/)
# then these are taken to be the version numbers of the modules.
# Special care is taken with a few known idioms for specifying version
# numbers of files under rcs/cvs control.

# If there are strings in the file which match the pattern
#     m/^\s*\$RPM_Provides\s*=\s*["'](.*)['"]/i
# then these are treated as additional names which are provided by the
# file and are printed as well.

my $perl_ns = "__PERL_NS__";

if ("@ARGV") {
  foreach (@ARGV) {
    process_file($_);
  }
} else {

  # notice we are passed a list of filenames NOT as common in unix the
  # contents of the file.

  foreach (<>) {
    process_file($_);
  }
}


foreach $module (sort keys %require) {
  if (length($require{$module}) == 0) {
    print "$perl_ns($module)\n";
  } else {

    # I am not using rpm3.0 so I do not want spaces around my
    # operators. Also I will need to change the processing of the
    # $RPM_* variable when I upgrade.

    print "$perl_ns($module) = $require{$module}\n";
  }
}

exit 0;



sub process_file {

  my ($file) = @_;
  chomp $file;

  if (!open(FILE, $file)) {
    warn("$0: Warning: Could not open file '$file' for reading: $!\n");
    return;
  }

  my ($package, $version, $incomment, $inover) = ();

  while (<FILE>) {

    # skip the here-docs "<<" blocks
    # assume that <<12 means bitwise operation
    if (((m/^\s*(?:'[^']*?'|"[^"]*?"|[^"'#]*?)*?[^"'#<@]<<[\\]?(\w+)\s*/ &&
          ($1 !~ m/^\d+$/)) ||
         m/^\s*(?:'[^']*?'|"[^"]*?"|[^"'#]*?)*?[^"'#<@]<<\s*('[^']*?'|"[^"]*?"|`[^`]*?`)\s*/
        ) &&
         ! m/q[qxwr]?\s*[{([#|!\/][^})\]#|!\/]*?<<[^<]/
       ) {
      $tag = $1;
      $tag =~ s/['"`]//g;
      while (<FILE>) {
        chomp;
        ( $_ eq $tag ) && last;
      }
      $_ = <FILE>;
    }

    # skip q{} quoted sections - just hope we don't have curly brackets
    # within the quote, nor an escaped hash mark that isn't a comment
    # marker, such as occurs right here. Draw the line somewhere.
    if ( m/^.*\Wq[qxwr]?\s*([{([#|!\/])[^})\]#|!\/]*$/ && ! m/^\s*(package)\s/ ) {
      $tag = $1;
      $tag =~ tr/{\(\[\#|!\//})]#|!\//;
      $tag = quotemeta($tag);
      while (<FILE>) {
        ( $_ =~ m/$tag/ ) && last;
      }
    }

    # skip the documentation

    # we should not need to have item in this if statement (it
    # properly belongs in the over/back section) but people do not
    # read the perldoc.

    if (/^=(head[1-4]|pod|for|item)/) {
      /^=cut/ && next while <FILE>;
    }

    if (/^=over/) {
      /^=back/ && next while <FILE>;
    }

    # skip the data section
    if (m/^__(DATA|END)__$/) {
      last;
    }

    # not everyone puts the package name of the file as the first
    # package name so we report all namespaces except some common
    # false positives as if they were provided packages (really ugly).

    if (m/^\s*package\s+([_:a-zA-Z0-9]+)\s*v?([0-9._]+)?\s*(;|{)/) {
      $package = $1;
      $version = defined($2) ? $2 : undef;
      if ($package eq 'main') {
        undef $package;
        undef $version;
      } else {
        # If $package already exists in the $require hash, it means
        # the package definition is broken up over multiple blocks.
        # In that case, don't stomp a previous $VERSION we might have
        # found.  (See BZ#214496.)
        $require{$package} = $version unless (exists $require{$package});
      }
    }

    # after we found the package name take the first assignment to
    # $VERSION as the version number. Exporter requires that the
    # variable be called VERSION so we are safe.

    # here are examples of VERSION lines from the perl distribution

    #FindBin.pm:$VERSION = $VERSION = sprintf("%d.%02d", q$Revision: 1.9 $ =~ /(\d+)\.(\d+)/);
    #ExtUtils/Install.pm:$VERSION = substr q$Revision: 1.9 $, 10;
    #CGI/Apache.pm:$VERSION = (qw$Revision: 1.9 $)[1];
    #DynaLoader.pm:$VERSION = $VERSION = "1.03";     # avoid typo warning
    #General.pm:$Config::General::VERSION = 2.33;
    #
    # or with the new "our" pragma you could (read will) see:
    #
    #    our $VERSION = '1.00'
    if ($package && m/^(?:\s*use\s+version\s*;)?\s*(?:[^#=]*=[^=~>]|)?\s*(?:our\s+)?\$(?:\Q$package\E::)?VERSION\s*=([^=~>]\s*[^;]*)/) {
      my $version_str = $1;

      # first see if the version string contains the string
      # '$Revision' this often causes bizarre strings and is the most
      # common method of non static numbering.

      if ($version_str =~ m/\$Revision: (\d+[.0-9]+)/) {
        $version = $1;
      } elsif ($version_str =~ m/\b['"]?v?(\d+(?:\.[.0-9]+)?)(_\d*|[a-zA-Z]*)?['"]?\b/) {

        # look for a static number hard coded in the script

        $version = $1;
      }
      $require{$package} = $version;
    }

    # Allow someone to have a variable that defines virtual packages
    # The variable is called $RPM_Provides.  It must be scoped with
    # "our", but not "local" or "my" (just would not make sense).
    #
    # For instance:
    #
    #     $RPM_Provides = "blah bleah"
    #
    # Will generate provides for "blah" and "bleah".
    #
    # Each keyword can appear multiple times.  Don't
    #  bother with datastructures to store these strings,
    #  if we need to print it print it now.

    if (m/^\s*(our\s+)?\$RPM_Provides\s*=\s*["'](.*)['"]/i) {
      foreach $_ (split(/\s+/, $2)) {
        print "$_\n";
      }
    }

  }

  close(FILE) ||
    die("$0: Could not close file: '$file' : $!\n");

  return;
}