PDL::PP-Inline - Man Page

Write PDL Subroutines inline with PDL::PP

Synopsis

  use strict; use warnings;
  use PDL;
  use Inline Pdlpp => 'DATA';

  # make data with: echo -n 'ATCGZATCG' >input.data
  # use it with aa_to_int.pl input.data

  my $file; ($file = shift, -f $file) || die "Usage: $0 filename";
  my $size = -s $file;
  my $pdl = zeroes(byte, $size);
  ${$pdl->get_dataref} = do { open my $fh, $file or die "$file: $!"; local $/; <$fh> };
  $pdl->upd_data;
  $pdl->inplace->aa_to_int;
  print $pdl, "\n";

  __DATA__
  __Pdlpp__
  pp_def('aa_to_int',
   Pars => 'i();[o] o()',
   GenericTypes => ['B'],
   Inplace => 1,
   Code => <<'EOF',
  switch($i()) {
   case 'A': $o() = 0; break;
   case 'T': $o() = 1; break;
   case 'C': $o() = 2; break;
   case 'G': $o() = 3; break;
   default: $o() = 255; break;
  }
  EOF
   Doc => "=for ref\n\nConvert amino acid names to integers.\n",
  );

Inline::Pdlpp is a module that allows you to write PDL subroutines in the PDL::PP style. The big benefit compared to plain PDL::PP is that you can write these definitions inline in any old perl script (without the normal hassle of creating Makefiles, building, etc). Since version 0.30 the Inline module supports multiple programming languages and each language has its own support module. This document describes how to use Inline with PDL::PP (or rather, it will once these docs are complete ;).

For more information on Inline in general, see Inline.

Some example scripts demonstrating Inline::Pdlpp usage can be found in the Example/InlinePdlpp directory.

Inline::Pdlpp is a subclass of Inline::C. Most Kudos goes to Brian I.

Usage

You never actually use Inline::Pdlpp directly. It is just a support module for using Inline.pm with PDL::PP. So the usage is always:

    use Inline Pdlpp => ...;

    bind Inline Pdlpp => ...;

Examples

Pending availability of full docs a few quick examples that illustrate typical usage.

A simple example

   # example script inlpp.pl
   use PDL; # must be called before (!) 'use Inline Pdlpp' calls

   use Inline Pdlpp; # the actual code is in the __Pdlpp__ block below

   $x = sequence 10;
   print $x->inc,"\n";
   print $x->inc->dummy(1,10)->tcumul,"\n";

   __DATA__

   __Pdlpp__

   pp_def('inc',
          Pars => 'i();[o] o()',
          Code => '$o() = $i() + 1;',
         );

   pp_def('tcumul',
          Pars => 'in(n);[o] mul()',
          Code => '$mul() = 1;
                   loop(n) %{
                     $mul() *= $in();
                   %}',
   );
   # end example script

If you call this script it should generate output similar to this:

   prompt> perl inlpp.pl
   Inline running PDL::PP version 2.2...
   [1 2 3 4 5 6 7 8 9 10]
   [3628800 3628800 3628800 3628800 3628800 3628800 3628800 3628800 3628800 3628800]

Usage of Inline::Pdlpp in general is similar to Inline::C. In the absence of full docs for Inline::Pdlpp you might want to compare Inline::C.

Use of PMCode

If you need to pre/post-process data around your new PDL operation, a quirk of this module is that it makes a dynamic library that can be loaded in, but does not load in the PDL-generated Perl code. You can work around that as follows:

  use Inline Pdlpp => <<'EOF';
  pp_def('collatz_pdl',
  Pars => 'beg(); end(); [o]number();  [o]highest()',
  GenericTypes => ['Q'],
  PMCode => '', # trigger the _int but keep the Perl code in main body
  Code => '
    /*...*/
  '
  );
  EOF
  sub PDL::collatz_pdl {
    PDL::_collatz_pdl_int(@_, my $n = PDL->null, my $h = PDL->null);
    # Perl scalars because is across forked processes in MCE
    map $_->sclr, $n, $h;
  }

See <https://www.perlmonks.org/?node_id=11115392> for context.

Code that uses external libraries, etc

The script below is somewhat more complicated in that it uses code from an external library (here from Numerical Recipes). All the relevant information regarding include files, libraries and boot code is specified in a config call to Inline. For more experienced Perl hackers it might be helpful to know that the format is similar to that used with ExtUtils::MakeMaker. The keywords are largely equivalent to those used with Inline::C. Please see below for further details on the usage of Inc, Libs, Auto_include and Boot.

   use PDL; # this must be called before (!) 'use Inline Pdlpp' calls

   use Inline Pdlpp => Config =>
     INC => "-I$ENV{HOME}/include",
     LIBS => "-L$ENV{HOME}/lib -lnr -lm",
     # code to be included in the generated XS
     AUTO_INCLUDE => <<'EOINC',
   #include <math.h>
   #include "nr.h"    /* for poidev */
   #include "nrutil.h"  /* for err_handler */

   static void nr_barf(char *err_txt)
   {
     fprintf(stderr,"Now calling croak...\n");
     croak("NR runtime error: %s",err_txt);
   }
   EOINC
   # install our error handler when loading the Inline::Pdlpp code
   BOOT => 'set_nr_err_handler(nr_barf);';

   use Inline Pdlpp; # the actual code is in the __Pdlpp__ block below

   $x = zeroes(10) + 30;;
   print $x->poidev(5),"\n";

   __DATA__

   __Pdlpp__

   pp_def('poidev',
           Pars => 'xm(); [o] pd()',
           GenericTypes => [L,F,D],
           OtherPars => 'long idum',
           Code => '$pd() = poidev((float) $xm(), &$COMP(idum));',
   );

Making an Installable Module

It is possible, using Inline::Module, to create an installable .pm file with inline PDL code. PDLA::IO::HDF is a working example. Here's how. You make a Perl module as usual, with a package declaration in the normal way. Then (assume your package is PDLA::IO::HDF::SD):

  package PDLA::IO::HDF::SD;
  # ...
  use FindBin;
  use Alien::HDF4::Install::Files;
  use PDLA::IO::HDF::SD::Inline Pdlapp => 'DATA',
    package => __PACKAGE__, # if you have any pp_addxs - else don't bother
    %{ Alien::HDF4::Install::Files->Inline('C') }, # EUD returns empty if !"C"
    typemaps => "$FindBin::Bin/lib/PDLA/IO/HDF/typemap.hdf",
    ;
  # ...
  1;
  __DATA__
  __Pdlapp__
  pp_addhdr(<<'EOH');
  /* ... */
  EOH
  use FindBin;
  use lib "$FindBin::Bin/../../../../../../..";
  require 'buildfunc.noinst';
  # etc

Note that for any files that you need to access for build purposes (they won't be touched during post-install runtime), FindBin is useful, albeit slightly complicated.

In the main .pm body, FindBin will find the build directory, as illustrated above. However, in the "inline" parts, FindBin will be within the Inline::Module build directory. At the time of writing, this is under .inline within the build directory, in a subdirectory named after the package. The example shown above has seven ..: two for .inline/build, and five more for PDLA/IO/HDF/SD/Inline.

The rest of the requirements are given in the Inline::Module documentation.

This technique avoids having to use PDL::Core::Dev, create a Makefile.PL, have one directory per .pd, etc. It will even build / install faster, since unlike a build of an ExtUtils::MakeMaker distribution with multiple directories, it can be built in parallel. This is because the EUMM build changes into each directory, and waits for each one to complete. This technique can run concurrently without problems.

PDLPP Configuration Options

For information on how to specify Inline configuration options, see Inline. This section describes each of the configuration options available for Pdlpp. Most of the options correspond either to MakeMaker or XS options of the same name. See ExtUtils::MakeMaker and perlxs.

Auto_include

Specifies extra statements to automatically included. They will be added onto the defaults. A newline char will be automatically added. Does essentially the same as a call to pp_addhdr. For short bits of code AUTO_INCLUDE is probably syntactically nicer.

    use Inline Pdlpp => Config => AUTO_INCLUDE => '#include "yourheader.h"';

Bless

Same as pp_bless command. Specifies the package (i.e. class) to which your new pp_defed methods will be added. Defaults to PDL if omitted.

    use Inline Pdlpp => Config => BLESS => 'PDL::MyPackage';

cf "Package", equivalent for "pp_addxs" in PDL::PP.

Boot

Specifies C code to be executed in the XS BOOT section. Corresponds to the XS parameter. Does the same as the pp_add_boot command. Often used to execute code only once at load time of the module, e.g. a library initialization call.

CC

Specify which compiler to use.

Ccflags

Specify extra compiler flags.

Inc

Specifies an include path to use. Corresponds to the MakeMaker parameter.

    use Inline Pdlpp => Config => INC => '-I/inc/path';

LD

Specify which linker to use.

Lddlflags

Specify which linker flags to use.

NOTE: These flags will completely override the existing flags, instead of just adding to them. So if you need to use those too, you must respecify them here.

Libs

Specifies external libraries that should be linked into your code. Corresponds to the MakeMaker parameter.

    use Inline Pdlpp => Config => LIBS => '-lyourlib';

    use Inline Pdlpp => Config => LIBS => '-L/your/path -lyourlib';

Make

Specify the name of the 'make' utility to use.

Myextlib

Specifies a user compiled object that should be linked in. Corresponds to the MakeMaker parameter.

    use Inline Pdlpp => Config => MYEXTLIB => '/your/path/yourmodule.so';

Optimize

This controls the MakeMaker OPTIMIZE setting. By setting this value to '-g', you can turn on debugging support for your Inline extensions. This will allow you to be able to set breakpoints in your C code using a debugger like gdb.

Package

Controls into which package the created XSUBs from "pp_addxs" in PDL::PP go. E.g.:

    use Inline Pdlpp => 'DATA', => PACKAGE => 'Other::Place';

will put the created routines into Other::Place, not the calling package (which is the default). Note this differs from "Bless", which is where "pp_def" in PDL::PPs go.

Typemaps

Specifies extra typemap files to use. Corresponds to the MakeMaker parameter.

    use Inline Pdlpp => Config => TYPEMAPS => '/your/path/typemap';

Noisy

Show the output of any compilations going on behind the scenes. Turns on BUILD_NOISY in Inline::C.

Bugs

doing inline scripts

Beware that there is a problem when you use the __DATA__ keyword style of Inline definition and want to do your script containing inlined code. For example

   # myscript.pl contains inlined code
   # in the __DATA__ section
   perl -e 'do "myscript.pl";'
 One or more DATA sections were not processed by Inline.

According to Brian Ingerson (of Inline fame) the workaround is to include an Inline->init call in your script, e.g.

  use PDL;
  use Inline Pdlpp;
  Inline->init;

  # perl code

  __DATA__
  __Pdlpp__

  # pp code

PDL::NiceSlice and Inline::Pdlpp

There is currently an undesired interaction between PDL::NiceSlice and Inline::Pdlpp. Since PP code generally contains expressions of the type $var() (to access ndarrays, etc) PDL::NiceSlice recognizes those incorrectly as slice expressions and does its substitutions. For the moment (until hopefully the parser can deal with that) it is best to explicitly switch PDL::NiceSlice off before the section of inlined Pdlpp code. For example:

  use PDL::NiceSlice;
  use Inline::Pdlpp;

  $x = sequence 10;
  $x(0:3)++;
  $x->inc;

  no PDL::NiceSlice;

  __DATA__

  __C__

  ppdef (...); # your full pp definition here