NAME

C::Blocks - embeding a fast C compiler directly into your Perl parser

STATUS

Pre-Beta

This distribution seems to be working very well for me. However, even if the current version passes its test suite and installs on your system, it may be subtly broken and throw segmentation faults during ordinary use. Uncovering and fixing these is one of the major goals of the Alpha stage. You can be more confident in the distribution's robustness once it reaches Beta.

Currently this distribution builds and passes its test suite on at least a subset of the Linux, Mac, and Windows machines reporting on CPAN Testers, and I have implemented the functionality and many of the tests I had hoped to implement in the Alpha stage. Now that I have passing reports on the Big Three, I turn my attention to expanding the test suite.

For more on the milestones for Alpha, Beta, and v1.0, see Goals and Milestones below.

I do not yet make any promises about backwards compatibility. (But, I use this distribution in my own code, so you can anticipate that a fair bit of the API has seen use and is not likely to change much.) In particular, the compiler setup and type APIs will be ironed out, in backwards-incompatible ways, during Beta.

DESCRIPTION

This distribution contains C::Blocks, a module that adds a few lexically scoped keywords to your Perl parser. These keywords delimit C code blocks, and (in the current implementation) the code within them is compiled during Perl's bytecode compilation phase. The keywords including cblock, clex, cshare, and csub. The keyword cisa provides a simple typemap-like system.

In addition, this distribution contains C::Blocks::PerlAPI, which loads Perl's C API. I may eventually include similar modules for all standard C libraries, such as libstdio, libmath, etc. Since libperl is itself compiled with these libraries, I suspect that most of them will be redundant for most people.

It can be difficult to safely attach C structs to Perl variables. This problem was solved for XS authors with the module XS::Object::Magic. This distribution provides a port of that module called C::Blocks::Object::Magic.

This distribution presently includes C::Blocks::StretchyBuffer, which provides a rudimentary vector type. It serves an important role as a testing module. It will almost certainly be spun off into its own module by the time C::Blocks v1.0 is released, so you should not rely on its presence in the base distribution.

BUILD NOTES

On Debian systems using system Perl (local::lib or otherwise), the bad-symbol auto-detection runs much, much faster if you have libperl-dev installed. This is strongly recommended, if possible. If you use Perlbrew, it will not matter because your local Perl will have libperl available.

GOALS AND MILESTONES

I have a number of goals I would like to accomplish for this project. The basic set of milestones are (1) make it work, (2) test the heck out of it, (3) settle on an API, and (4) decide whether to throw in the kitchen sink.

Pre-Alpha

The goal during the pre-alpha stage was to get things mostly working on most platforms. Once this was accomplished, I spun the first release on CPAN. I declared victory on this stage when I got libperl caching working on all three platforms I had easy access to, even though StretchyBuffer still gave trouble on Windows.

Alpha CPAN Release, v0.01 - 0.49

The goal during the alpha stage is to achieve successful builds that can be trusted to work on the three popular "user" operating systems: Linux, Strawberry Windows, and Mac. Major work in the alpha stage will focus on OS-specific build issues and vastly expanding the test suite. The easiest and most important tests to add focus on Perl's own C API, as well as porting XS::Object::Magic's tests for C::Blocks::Object::Magic. Tests for things like scoping, memory duration, and correct line number reporting need to be expanded as well. A significant number of failures on Linux, Mac, or Strawberry Windows systems will be considered blockers to finalizing the Alpha stage.

Although originally planned as a much later addition, I have added C::Blocks::Object::Magic during alpha. This modue is a port of XS::Object::Magic. I want to keep this as a core module at least through Beta. The idea it implements is too useful to distribute it elsewhere, and it provides an excellent testing target. I hope to incorporate XS::Object::Magic's test suite as part of the test build-up for Beta.

Besides Linux, Mac, and Strawberry, I plan to officially support all operating systems for which Alien::TinyCCx successfully installs. (For the moment, that includes the addition of gnukfreebsd.) TCC itself is not guaranteed to work on Cygwin, for example, and I see no point in holding back C::Blocks due to a failure of the underlying library. (See the Beta requirements for more on Cygwin, and BSD.)

The API for tweaking the compiler will still be considered in flux, so substantial library development is discouraged during Alpha.

It is easy to produce warnings saying

warning: assignment from incompatible pointer type

This is a bug with the underlying extended symbol table handling and struct type checking. I would like this to be fixed sooner rather than later. Thankfully, it is not hard to work around this warning, and it is not a blocking issue.

Beta CPAN Release, v0.50 - v0.99

The goals after the distribution hits v0.50 include settling on a compiler setup API, finalizing the automated type handling system, cleaning up the internals, and officially supporting Cygwin and BSDs. Pre-1.0 work will focus on expanding the documentation.

The current compiler setup API involves parsing global string variables. This can certainly be done better. I want to work out a more sophisticated and programatic approach to compiler setup so that authors do not have to jump through too many hoops to set up C::Blocks when wrapping a C library. In particular, ExtUtils::Depends provides information about functions and typemaps provided by other Perl XS modules, and this API should make it easy to use this information.

The cisa keyword implements a very simple type system, introduced based on experience durinig the alpha stage. In practice, the current implementation is awkward to use. I would like to redesign the type system during Beta.

I believe that call_checker and call_parser represent a better API for the lexically-scoped keywords introduced by C::Blocks. I used the keyword API for prototyping because it was easy for me to get started. During Beta, I would like to investigate using these. If this alternative API represents an improvement over the keyword API, I want the switch to take place during Beta.

The Alpha version of C::Blocks supports (but does not enforce) single-threaded compilation of multi-threaded code. That is, it is possible and perfectly safe for a thread to compile code that is run simultaneously by many child threads. There are two problems with threads. First, the Tiny C Compiler is not itself threadsafe, so if multiple threads attempt to compile code simultaneously, things would go horribly awry. Second, the actual object code produced when C::Blocks compiles a chunk of code is stored in the parent's thread-specific storage. If the parent exits before the child does (which is not possible on Windows but is possible on POSIXy systems), the child could try to execute code from memory that has been freed. I would like to address these by protecting compilation with a mutex, and by storing object code somewhere that can live across thread boundaries. (Doing that in such a way that does not leak will be one of the more challenging parts of this particular issue.) C::Blocks can already produce thread-safe code following certain practies. The goal here is to solve the problem with code, not programming practices.

Cygwin and BSD support will be added as soon as I can get Alien::TinyCCx to build on those operating systems. Support for these operating systems is a blocker for v1.0, and so this distribution will not get out of Beta without sufficient work on Alien::TinyCCx (and probably tcc itself).

With the settling of the compiler setup API, I also hope to make it easier to write libraries that are export wrappers. The simplest example would be something that, for the moment, I am calling C::Blocks::lib. This module would make it easy to import a bunch of C::Blocks modules in one shot, i.e.

use C::Blocks::lib qw(perl gsl prima);

would be equivalent to

use C::Blocks::libperl;
use C::Blocks::libgsl;
use C::Blocks::libprima;

Of course, I've not yet decided if I want to call all libraries lib..., so this still needs some thought.

Finally, I want to create a module that helps automate the symbol table load/dump functionality and symbol verification currently under development for use with PerlAPI. During Beta, I want to create author tools for Module::Build, ExtUtils::MakeMaker, Dist::Zilla, and other distribution tools so that build-time symbol table caching is easy.

v1.0

When C::Blocks hits v1.0, it'll have a solid test suite and stable API. From v1.0 onward, it'll simply be a question of what to include in the distribution itself. Here are things I want to add (or at least think about adding) after reaching v1.0:

Simple C-based Object System?

I plan to write a module that makes it easy to write single-inheritance object oriented C code. My current plans are to steal from Prima, which does this sort of thing exceptionally well. Whether this is added to the base distribution or released as an independent one is still up in the air.

Generate Inline::C, XS, ExtUtils::Depends

Right now the C code gets compiled at Perl's parse time. Once the C code is out of the prototyping stage, it would be nice to be able to automatically extract it and generate .h and .c files that can be compiled by the system's optimizing compiler. I need to write author tools that would allow this to be done using Inline::C, or to generate XS files with the proper scoping. Of course, it would also make sense for the contents of those header files to be accessible to other XS authors via ExtUtils::Depends.

Utilize ExtUtils::Depends

C::Blocks' consumption of ExtUtils::Depends information was a Beta target. Going the other way, producing an ExtUtils::Depends-like result for C::Blocks distributions, would likely be a nice feature.

Threadsafe

The functions produced by the Tiny C Compiler are threadsafe. However, compiling code with the Tiny C Compiler itself is not threadsafe.

uses lots of global variables and is therefore not threadsafe. I would like to contribute back to the project by encapsulating all of that global state into the compiler state object, where it belongs. Others in the tcc community have expressed interest in getting this done, so it is a welcome contribution.

SEE ALSO

This module uses a special fork of the Tiny C Compiler. The fork is located at https://github.com/run4flat/tinycc, and is distributed through the Alien package provided by Alien::TinyCCx. To learn more about the Tiny C Compiler, see http://bellard.org/tcc/ and http://savannah.nongnu.org/projects/tinycc. The fork is a major extension to the compiler that provides extended symbol table support.

For other ways of compiling C code in your Perl scripts, check out Inline::C, FFI::TinyCC, C::TinyCompiler, and XS::TCC.

For mechanisms for calling C code from Perl, see FFI::Platypus and FFI::Raw.

If you just want to mess with C struct data from Perl, see Convert::Binary::C.

If you're just looking to write fast code with compact data structures, http://rperl.org/ may be just the ticket. It produces highly optmized code from a subset of the Perl language itself.

AUTHOR

David Mertens (dcmertens.perl@gmail.com)

BUGS

Please report any bugs or feature requests for the Alien bindings at the project's main github page: http://github.com/run4flat/C-Blocks/issues.

SUPPORT

You can find documentation for this module with the perldoc command.

perldoc C::Blocks

You can also look for information at:

ACKNOWLEDGEMENTS

This would not be possible without the amazing Tiny C Compiler or the Perl pluggable keyword work. My thanks goes out to developers of both of these amazing pieces of technology.

LICENSE AND COPYRIGHT

Code copyright 2013-2015 Dickinson College. Documentation copyright 2013-2015 David Mertens.

This program is free software; you can redistribute it and/or modify it under the terms of either: the GNU General Public License as published by the Free Software Foundation; or the Artistic License.

See http://dev.perl.org/licenses/ for more information.