Perl 5, Python, Rakudo, C/C++, Java, Lua?

[raiph]

(I retitled and moved this content from Re^3: [perl6] Complex Attribute Validation and/or Triggers because I felt it warranted its own post.)

»»» This post is about alpha status Perl 6, not rock solid Perl 5 «««

»ö« . o { "Wouldn't it be fun to mix Rakudo Perl 6 with another lang or three, calling out to each other, passing objects back and forth, subclassing a class defined in another lang, until the whole thing dies a horrible death which you report back to #perl6?" }

• Inline::Perl5. Stefan Seifert (nine on IRC) wrote the first version of this breakthru module in September last year. It lets one write Perl 5 code in Perl 6; use P5 modules (aiui most of CPAN should work; some exceptions); treat P5 objects as if they were P6 ones; subclass a P5 class in P6; and more. It's suddenly become possible to do stuff like write a Catalyst controller in Perl 6.

• Inline::Perl6. This new CPAN module is the converse of Inline::Perl5, allowing use of Perl 6 in Perl 5 code by embedding a MoarVM based Rakudo Perl 6 in Perl 5. Also created by Stefan Seifert.

• Inline::Python for Perl 6. Also recently created by Stefan Seifert. (Note that he's the co-maintainer of Inline::Python for Perl 5.)

• p6-Inline-C by Tobias Leich (FROGGS on IRC).

• Inline::Lua for Perl 6. "Both Lua 5.1 and LuaJIT are supported." Started a few weeks ago by raydiak (author of Pray, a ray tracing engine written in Perl 6).

• The NativeCall interface. Makes it easy for non C hackers to wrap C libs without need for a C compiler or C headers on the target system. (Will become even easier when a C parser in Perl 6 starts parsing headers to automatically construct bindings.) NativeCall is being driven forward by Tobias Leich (FROGGS on IRC). His most recent work is adding support for C++ libs too.

• Use Java libs in Perl 6 code. Pepe Schwarz (psch on IRC) is leading the Java work.

[Arunbear]

Does Inline::Perl5 embed the Perl5 interpreter in the Perl6 one?

[Corion]

Yes. slides / video (no source code though, at least not on CPAN).

[Tux]

Inline::Perl5 now supports passing IO (including deep and dark IO from within XS), so this code works:

use v6;
use Slang::Tuxic;
use Inline::Perl5;

my $p5 = Inline::Perl5.new;

$p5.use ("Text::CSV_XS");

my @rows;
my $csv = $p5.invoke ("Text::CSV_XS", "new")
    or die "Cannot use CSV: ", $p5.invoke ("Text::CSV_XS", "error_diag
+");
$csv.binary (1);
$csv.auto_diag (1);

my $fh = open "/tmp/hello.csv", :r, chomp => False;

my Int $sum = 0;
while (my $r = $csv.getline ($fh)) {
    $sum += +$r;
    }
$sum.say;

To give an example of performance (seconds needed to parse 10000 lines of 5 fields each). Note that Text::CSV::Easy_* only accepts valid CSV and has no options.

Perl 5
Text::CSV::Easy_XS	0.019
Text::CSV::Easy_PP	0.017
Text::CSV_XS with bind_columns	0.033
Text::CSV_XS	0.038
Text::CSV_PP	0.517
Pegex::CSV	1.350
Perl 6
State machine (no options yet)	7.027
Inline::Perl5 + Text::CSV_XS	13.082
Inline::Perl5 + Text::CSV_XS + getline (IO)	13.142
Inline::Perl5 + Text::CSV_PP	13.623
Grammar-based (no options yet)	13.210
Regex based (no options)	39.996
Regex based, all options provided	36.075

When I started in october 2014, the Regex based solution took 260 seconds, so improvement in the language itself is made on a daily basis!

[Jenda]

Thanks for the benchmarks! Come back when, or I should rather say, if you succeed to make something that's not several hundred times slower than something that actually works. I doubt you ever will.

Ada attracted much attention from the programming community as a whole during its early days. Its backers and others predicted that it might become a dominant language for general purpose programming and not just defense-related work. Ichbiah publicly stated that within ten years, only two programming languages would remain, Ada and Lisp. Early Ada compilers struggled to implement the large, complex language, and both compile-time and run-time performance tended to be slow and tools primitive. Compiler vendors expended most of their efforts in passing the massive, language-conformance-testing, government-required "ACVC" validation suite that was required in another novel feature of the Ada language effort.

You are building another Ada, this time without the backing of DoD and, more importantly, with no chance of a Perl6 mandate.

[einhverfr]

Looks interesting. I am watching some parts of Perl 6 and this could be quite helpful. I am however taken aback by "most of CPAN should work" because I am not sure what would be the obstacles there. Are there specific features that would not work? Like maybe tail calls via goto &subname? Or is this more like "we haven't tested it entirely?

If Perl 5 modules work flawlessly in Perl 6 including language features unsupported in the latter, then that would give me additional confidence. I actually use tail calls in Perl 5 and that is one of the larger features that if it is not supported would keep me from upgrading, but if I can put it in a Perl 5 module, then no harm done.

[raiph]

I'd not be surprised if the total number of CPAN modules actually tested by everyone who's tried Inline::Perl5 is less than 100 so far. A few folk on #perl6 have mentioned using it. Tux has mentioned using Inline::Perl5 with Text::CSV_XS elsewhere in this thread. Stefan Seifert, the author of Inline::Perl5, said it was working with DBI a few days after his first commits to Inline::Perl5 and he's since demo'd using it with Catalyst.

Stefan has also briefly touched on some Perl 5 features that are not expected to work with Inline::Perl5.

Another important option when considering mixing Perl 5 and Perl 6 is Inline::Perl6.

Btw `&foo.nextwith...` is supposed to one day tailcall the `foo` sub but it's not yet been implemented in Rakudo.

[einhverfr]

Thanks for the nextwith stuff. I will keep an eye on that.

[Your Mother]

You can do tail calls in Perl5 now: 5.4.2 Creating Tail Calls in Higher Order Perl; see also Sub::Call::Tail.

The CPAN has plenty of “interesting” code that doesn’t work everywhere Perl5 does. It seems likely that a slightly smaller portion would be able work under any kind of harness, Perl6 or otherwise.

[einhverfr]

Yes you can. goto &subname also does them. I don't see anything in Perl 6 docs for this functionality however.

[Jenda]

. o { "No, it would not." }