The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.

NAME

Remote::Use - Using modules from a remote server

SYNOPSIS

  $ cat -n prime3.pl
     1  #!/usr/bin/perl -I../lib -w
     2  # The module Math::Prime::XS isn't installed in the machine
     3  # but will be downloaded from some remote server
     4  use Math::Prime::XS qw{:all};
     5
     6  @all_primes   = primes(9);
     7  print "@all_primes\n";
     8
     9  @range_primes = primes(4, 9);
    10  print "@range_primes\n";


  $ perl -I../lib -MRemote::Use=config,rsyncconfig prime3.pl

  $ cat -n rsyncconfig
     1  package rsyncconfig; # Configuration file
     2
     3  sub getarg {
     4    my ($class, $self) = @_;
     5
     6    return (
     7      host => 'orion:',
     8      prefix => '/tmp/perl5lib/',
     9      command => 'rsync -i -vaue ssh',
    10      ppmdf => '/tmp/perl5lib/.orion.installed.modules',
    11    );
    12  }
    13
    14  1;

INTRODUCTION

When I wrote this module I didn't know of the existence of PAR::Repository::Client and PAR::Repository distributions. These distributions - now embedded inside the PAR distribution - broach similar problems.

Look at them first before considering using this one. They provide a solution more robust and cover a range of problems much wider than this distribution.

Remote::Use focuses in the direct loading of modules already installed in some remote server while PAR::Repository::Client and PAR::Repository deal with the use of remote PAR repositories.

There are a few scenarios where this module can still be useful:

  • If your server and clients are homogeneous: same Perl version, very similar O.S. distributions, same external libraries installed, etc. and you want to avoid the extra work of repository creation and administration

  • If you want to use protocols like SSH, SFTP or others not currently supported by PAR::Repository::Client and PAR::Repository

  • If no binaries are involved

Remote::Use provides a way to run a Perl program when some libraries aren't availables at start time. The libraries will be downloaded from a specified server using a specified application that runs on top of some protocol. The clients must be binary compatibles with the server if binary libraries - as is the case of the module Math::Prime::XS used in the SYNOPSIS section example script - are involved. Typical downloaders are scp, rsync or wget but any other suitable alternative like lwp-mirror or Curl can be used. This means that many different protocols can be used for the transference: SSH, SFTP, HTTP, HTTPS, FTP, etc.

This way, the clients download the modules their programs use to some directory. Once the modules are downloaded they will not be downloaded again, unless the modules are removed.

To get familiar with Remote::Use start by reading Remote::Use::Tutorial

PPMDF files

A Perl Modules Descriptor File describes what Modules in the Module Server Machine will be published and what files must be downloaded for each of those modules. It is used by Remote::Use to automatically download the Perl modules need by a script being executed from a Perl Public Modules Server (See Remote::Use::Tutorial).

The file is a Perl list. For each published module Some::Module there is a key which is the associated file name Some/Module.pm and a value. The value is a hash reference that must contain at least two entries: one named dir and another named file. The second contains the list of files to be downloaded when Some::Module is used. The dir entry contains the prefix path that must be removed from the path of the source file names (at the server) to produce the target file names (at the client).

      'Some/Module.pm' => {
        dir => '/prefix/path/',
        files => [ '/auto/Some/Module/Module.so', 
                   '/Some/Module.pm', 
                   '/Some/Module.pod' ],
        bin => [ '/some/script', /another/script' ],
        man => [ '/some/man', /another/man' ],
      }

          

For each module entry additional file families can be added as illustrates the bin entry for Parse::Eyapp in the former example:

  'Parse/Eyapp.pm' => { dir => '',
    files => [ '/Parse/Eyapp.pm' ],
    bin => [ '/bin/eyapp', '/bin/treereg' ]
  }

The following example illustrates the syntax of a PPMDF file:

  pp2@nereida:~/LRemoteUse/examples$ cat /tmp/perl5lib/.orion.via.web
  (
  'Trivial.pm' => { dir => '', files => [
          '/Trivial.pm' ] },
  'Tintin/Trivial.pm' => { dir => '', files => [
          '/Tintin/Trivial.pm' ] },
  'Parse/Eyapp.pm' => { dir => '',
    files => [ '/Parse/Eyapp.pm' ],
    bin => [ '/bin/eyapp', '/bin/treereg' ]
  },
  'Parse/Eyapp/Lalr.pm' => { dir => '', files => [
          '/Parse/Eyapp/Lalr.pm' ] },
  'Parse/Eyapp/YATW.pm' => { dir => '', files => [
          '/Parse/Eyapp/YATW.pm' ] },
  'Parse/Eyapp/Treeregexp.pm' => { dir => '', files => [
          '/Parse/Eyapp/Treeregexp.pm' ] },
  'Parse/Eyapp/Parse.pm' => { dir => '', files => [
          '/Parse/Eyapp/Parse.pm' ] },
  'Parse/Eyapp/Scope.pm' => { dir => '', files => [
          '/Parse/Eyapp/Scope.pm' ] },
  'Parse/Eyapp/Options.pm' => { dir => '', files => [
          '/Parse/Eyapp/Options.pm' ] },
  'Parse/Eyapp/Output.pm' => { dir => '', files => [
          '/Parse/Eyapp/Output.pm' ] },
  'Parse/Eyapp/Node.pm' => { dir => '', files => [
          '/Parse/Eyapp/Node.pm' ] },
  'Parse/Eyapp/Grammar.pm' => { dir => '', files => [
          '/Parse/Eyapp/Grammar.pm' ] },
  'Parse/Eyapp/Driver.pm' => { dir => '', files => [
          '/Parse/Eyapp/Driver.pm' ] },
  'Parse/Eyapp/Base.pm' => { dir => '', files => [
          '/Parse/Eyapp/Base.pm' ] },
  'Parse/Eyapp/_TreeregexpSupport.pm' => { dir => '', files => [
          '/Parse/Eyapp/_TreeregexpSupport.pm' ] },
  'Math/Prime/XS.pm' => { dir => '', files => [
          '/auto/Math/Prime/XS/XS.bs',
          '/auto/Math/Prime/XS/XS.so',
          '/Math/Prime/XS.pm' ] },
  );

Here is another (summarized) example:

  ~/LRemote-Use/script$ head -23 orion.installed.modules
  (
  'CPAN/Config.pm' => { dir => '/etc/perl', files => [
          '/etc/perl/CPAN/Config.pm' ] },
  'Template.pm' => { dir => '/usr/local/lib/perl/5.8.8', files => [
          '/usr/local/lib/perl/5.8.8/auto/Template/Stash/XS/XS.so',
          '/usr/local/lib/perl/5.8.8/auto/Template/Stash/XS/XS.bs',
          '/usr/local/lib/perl/5.8.8/Template.pm' ] },
  'IO/Tty.pm' => { dir => '/usr/local/lib/perl/5.8.8', files => [
          '/usr/local/lib/perl/5.8.8/auto/IO/Tty/Tty.so',
          '/usr/local/lib/perl/5.8.8/auto/IO/Tty/Tty.bs',
          '/usr/local/lib/perl/5.8.8/IO/Tty.pm' ] },
  'IO/Pty.pm' => { dir => '/usr/local/lib/perl/5.8.8', files => [
          '/usr/local/lib/perl/5.8.8/IO/Pty.pm' ] },
  'IO/Tty/Constant.pm' => { dir => '/usr/local/lib/perl/5.8.8', files => [
          '/usr/local/lib/perl/5.8.8/IO/Tty/Constant.pm' ] },
  'Math/Prime/XS.pm' => { dir => '/usr/local/lib/perl/5.8.8', files => [
          '/usr/local/lib/perl/5.8.8/auto/Math/Prime/XS/XS.so',
          '/usr/local/lib/perl/5.8.8/auto/Math/Prime/XS/XS.bs',
          '/usr/local/lib/perl/5.8.8/Math/Prime/XS.pm' ] },
  'Template/Stash.pm' => { dir => '/usr/local/lib/perl/5.8.8', files => [
          '/usr/local/lib/perl/5.8.8/auto/Template/Stash/XS/XS.so',
          '/usr/local/lib/perl/5.8.8/auto/Template/Stash/XS/XS.bs',
          '/usr/local/lib/perl/5.8.8/Template/Stash.pm' ] },

THE CONFIGURATION PACKAGE

If the config option is set while using Remote::Use as in:

     use Remote::Use config => 'tutu/wgetconfigpm.pm';

then the configuration will be loaded from the specified configuration package file tutu/wgetconfigpm.pm.

The configuration package is a Perl package describing the connection with the Perl Public Modules Server. While the PPMDF file tell us where are the files to transfer, the configuration package says how they will be transferred. The configuration package specifies, among other things, where the PPMDF file is and what application will be used for the transference of files. The configuration package is required by the import method of Remote::Use. See an example:

  pp2@nereida:~/LRemoteUse/examples$ cat -n tutu/wgetconfigpm.pm
     1  package tutu::wgetconfigpm;
     2  use strict;
     3  use warnings;
     4
     5  sub getarg {
     6    return (
     7      command => 'wget -o /tmp/wget.log',
     8      commandoptions => '-O',
     9      host => 'http://orion.pcg.ull.es/~casiano/cpan',
    10      prefix => '/tmp/perl5lib/',
    11      ppmdf => '/tmp/perl5lib/.orion.via.web',
    12    );
    13  }
    14
    15  1;

The getarg method

The configuration file must have a subroutine named getarg. Such subroutine sets the attributes of the Remote::Use object that lead the behavior of Remote::Use during downloading. It receives as arguments the configuration package identifier and a reference to the Remote::Use object. Let us describe each of the attributes returned by getarg:

  • The command argument of getarg specifies the driver command (executable) that will be used to download the files.

                  command => 'rsync -i -vaue ssh'

    In this example we use rsync. See rsync man pages for more information. The -e ssh option tells rsync to use SSH to connect to the remote machine. The -v option increases the level of verbosity. The -u option makes rsync to skip files that are newer on the receiver. The -a option says you want recursion and want to preserve most of the attributes of the source file.

  • The host argument describes the host descriptor in terms of the application used to connect.

    Remote::Use calls the specified command (in this case rsync -i -vaue ssh) to download by asking the operating system to execute a line that can be decomposed in the following components:

     "$command $host$sourcefile $commandoptions $targetfile"

    Where $sourcefile is the file being downloaded and $targetfile is the name of the file in the target machine. The $targetfile name is deduced from the source file name and the hints given by the user in the configuration package. Usually the $command part includes the options, as in the example

                  command => 'rsync -i -vaue ssh'

    but if more options are needed after the "$host$sourcefile" prefix they can be specified using the commandoptions argument. See the configuration package wgetconfig inside the section "A CONFIGURATION FILE FOR wget" in Remote::Use::Tutorial for an example.

    For rsync connections the host attribute must be the name of the SSH connection followed by a colon:

                             host => 'orion:'

    This is because, to download using rsync a file like

      /usr/local/lib/perl/5.8.8/auto/Math/Prime/XS/XS.so

    placed at the remote server (orion) we use a command like:

      rsync -aue ssh orion:/usr/local/lib/perl/5.8.8/auto/Math/Prime/XS/XS.so XS.so

    The "$host$sourcefile" argument of the full command line can be divided into two parts: the host descriptor that includes the colon separator orion: and the file descriptor /usr/local/lib/perl/5.8.8/auto/Math/Prime/XS/XS.so.

    I usually set the multiple parameters of a connection in the ~/.ssh/config file that governs the SSH connection. As an example here is the paragraph in ~/.ssh/config that refers to the connection named 'orion':

      Host orion orion.pcg.ull.es orion.deioc.ull.es chum
      user casiano
      # The real name of the machine
      Hostname orion.pcg.ull.es
      ForwardX11 yes

    This way I don't have to write the user name and the full name of the machine each time I set a SSH connection.

    See "APPENDIX: AUTOMATIC AUTHENTICATION" and the "SEE ALSO" sections to know more about SSH configuration files.

  • The prefix argument describes the path in the client machine where modules will be stored. The downloaded modules will be stored below this path. Thus, the setting:

               prefix => '/tmp/perl5lib/'

    stores the files for module Math::Prime::XS

      'Math/Prime/XS.pm' => { 
         dir => '/usr/local/lib/perl/5.8.8', 
         files => [
            '/usr/local/lib/perl/5.8.8/auto/Math/Prime/XS/XS.so',
            '/usr/local/lib/perl/5.8.8/auto/Math/Prime/XS/XS.bs',
            '/usr/local/lib/perl/5.8.8/auto/Math/Prime/XS/.packlist',
            '/usr/local/lib/perl/5.8.8/Math/Prime/XS.pm' 
         ] 
      },

    respectively in:

              '/tmp/perl5lib/files/auto/Math/Prime/XS/XS.so'
              '/tmp/perl5lib/files/auto/Math/Prime/XS/XS.bs'
              '/tmp/perl5lib/files/auto/Math/Prime/XS/.packlist'
              '/tmp/perl5lib/files/Math/Prime/XS.pm' 

    That is: the target directory where the file will be finally stored is obtained by substituting in the corresponding files entry the dir prefix ('/usr/local/lib/perl/5.8.8') by the prefix attribute ('/tmp/perl5lib/') followed by the (files) word. Thus

      '/usr/local/lib/perl/5.8.8/auto/Math/Prime/XS/XS.so'

    is finally locally stored in

      '/tmp/perl5lib/files/auto/Math/Prime/XS/XS.so'

    Be sure you add that path specified in prefix to the environment variable PERL5LIB so that any Perl scripts that don't make use of Remote::Use can still have access to the downloaded modules.

  • The ppmdf option tells Remote::Use where is the PPMDF file:

        ppmdf => '/tmp/perl5lib/.orion.installed.modules',

The most important method that has to be defined inside the configuration package is getarg.

HOOKS

A PPMDF file is a Perl list. For each published module Some::Module there is a key which is the associated file name Some/Module.pm and a value. The value is a hash reference that must contain at least two entries: one named dir and another named file. The second contains the list of files to be downloaded when Some::Module is used. The dir entry contains the prefix path that must be removed from the path of the source file names (at the server) to produce the target file names (at the client).

      'Some/Module.pm' => {
        dir => '/prefix/path/',
        files => [ '/auto/Some/Module/Module.so', 
                   '/Some/Module.pm', 
                   '/Some/Module.pod' ],
        bin => [ '/some/script', /another/script' ],
        man => [ '/some/man', /another/man' ],
      }

In any entry for a module like Some/Module.pm we can add couples with the syntax

        tag => [ 'd1/f1', 'd2/f2', ... ]

to the hash entry. The tag is arbitrary and defines a family of files related with the module.

While the dir and files tags are compulsory, the others are optional. The behavior of Remote::Use for a family tag like

        tag => [ 'd1/f1', 'd2/f2', ... ]

is as follows: the family of files 'd1/f1', 'd2/f2', etc. associated with the tag will be by default downloaded to 'prefix/tag/f1', 'prefix/tag/f2', etc. Where prefix is the directory specified in the prefix option of getarg inside the configuration package.

Such behavior can be modified using hooks defined in the configuration package.

The hook pretag

If a subroutine with name pretag exists in the configuration package it will be executed for each file specified in the tag family just before the file is downloaded. The pretag subroutine receives as arguments the configuration package name, the full description of the file to download in the server (something like orion:/usr/local/bin/eyapp), the default name of the file in the client (i.e. something like /tmp/perl5lib/bin/eyapp) and a reference to the Remote::Use object. It must return the definitive full name of the file in the client (i.e. something like /home/mylogin/bin/eyapp).

The hook posttag

If a subroutine with name posttag exists in the configuration package it will be executed for each file specified in the tag family just after the file was downloaded. The posttag subroutine receives as arguments the configuration package name, the name of the downloaded file and a reference to the Remote::Use object.

DIRECT SPECIFICATION OF OPTIONS

An alternative to the use of a configuration package is to directly specify the configuration options in the use of Remote::Use as in the following example:

  pp2@nereida:~/LRemoteUse/examples$ cat -n ut1.pl
     1  #!/usr/bin/perl -w -I../lib/
     2  use strict;
     3  use Remote::Use
     4    command => 'wget -o /tmp/wget.log',
     5    commandoptions => '-O',
     6    host => 'http://orion.pcg.ull.es/~casiano/cpan',
     7    prefix => '/tmp/perl5lib/',
     8    ppmdf => '/tmp/perl5lib/.orion.via.web',
     9  ;
    10  use Trivial;
    11
    12  Trivial::hello();

The meaning of the options is at it was explained in section "The getarg method"

LIMITATIONS

  • Though not tested, more than likely, the current version of this module will only work in Unix-like systems.

  • If binary files are involved both platforms must be binary compatible. Here is an example where server and client differ in the version of the GLIBC library:

      pp2@nereida:~/LRemoteUse/examples$ reqprimetonga.pl
      receiving file list ... done
      >f+++++++++ XS.bs
    
      sent 42 bytes  received 94 bytes  272.00 bytes/sec
      total size is 0  speedup is 0.00
      receiving file list ... done
      >f+++++++++ XS.so
    
      sent 42 bytes  received 50814 bytes  101712.00 bytes/sec
      total size is 50712  speedup is 1.00
      receiving file list ... done
      >f+++++++++ XS.pm
    
      sent 42 bytes  received 5733 bytes  3850.00 bytes/sec
      total size is 5635  speedup is 0.98
      Can't load '/tmp/perl5lib//files/auto/Math/Prime/XS/XS.so' for 
      module Math::Prime::XS: /lib/tls/i686/cmov/libc.so.6: 
      version `GLIBC_2.4' not found 
      (required by /tmp/perl5lib//files/auto/Math/Prime/XS/XS.so) 
      at /usr/lib/perl/5.8/DynaLoader.pm line 225.
      at ./reqprimetonga.pl line 4
      Compilation failed in require at ./reqprimetonga.pl line 4.
      pp2@nereida:~/LRemoteUse/examples$                                 
  • Not only the libraries, the Perl interpreters must be binary compatible (which usually means they must have the same version). Here is another example in which the versions of Perl in the client (beowulf) and the server (orion) differ:

      casiano@beowulf:/tmp/Remote-Use-0.04$ export REMOTE_USE_DEVELOPER=1
      casiano@beowulf:/tmp/Remote-Use-0.04$ t/03reqprime.t
      1..1
      receiving file list ... done
      >f+++++++++ XS.so
    
      sent 42 bytes  received 15997 bytes  6415.60 bytes/sec
      total size is 15899  speedup is 0.99
      receiving file list ... done
      >f+++++++++ XS.bs
    
      sent 42 bytes  received 94 bytes  54.40 bytes/sec
      total size is 0  speedup is 0.00
      receiving file list ... done
      >f+++++++++ XS.pm
    
      sent 42 bytes  received 5733 bytes  2310.00 bytes/sec
      total size is 5635  speedup is 0.98
      /usr/bin/perl: symbol lookup error: 
      /tmp/perl5lib//files/auto/Math/Prime/XS/XS.so: 
       undefined symbol: Perl_Tstack_sp_ptr

    The version in the client is 10.0:

      casiano@beowulf:/tmp/Remote-Use-0.04$ perl -v
    
      This is perl, v5.10.0 built for i486-linux-gnu-thread-multi

    The version in the server is 5.8.8

      casiano@beowulf:/tmp/Remote-Use-0.04$ ssh orion perl -v
    
      This is perl, v5.8.8 built for i486-linux-gnu-thread-multi

ACKNOWLEDGMENTS

This work has been supported by CEE (FEDER) and the Spanish Ministry of Educacion y Ciencia through Plan Nacional I+D+I number TIN2005-08818-C04-04 (ULL::OPLINK project http://www.oplink.ull.es/). The University of La Laguna has also supported my work in many ways and for many years.

Finally, thanks to Juana, Coro and my students at La Laguna.

SEE ALSO

AUTHOR

Casiano Rodriguez Leon (casiano.rodriguez.leon at gmail dot com)

LICENCE AND COPYRIGHT

Copyright (c) 2007 Casiano Rodriguez-Leon (casiano.rodriguez.leon at gmail dot com). All rights reserved.

These modules are free software; you can redistribute it and/or modify it under the same terms as Perl itself. See perlartistic.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.